Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Ingolf V. Hertel
Claus-PeterSchulz
Atoms,
Molecules and
Optical Physics 1
Atoms and Spectroscopy
Graduate Texts in Physics
Series Editors
Professor Richard Needs
Cavendish Laboratory
JJ Thomson Avenue
Cambridge CB3 0HE, UK
rn11@cam.ac.uk
To my Wife Gudrun
CPS
Preface
Atomic, Molecular and Optical physics short AMO physics is one of the canon-
ical fields of physics, a profound knowledge of which is essential for understanding
almost any other area of modern physics. And while its roots reach back over a
century and are closely connected with the early days of modern physics, current
research in AMO physics is still highly productive in respect of both, cutting edge
applications and fundamental insights as several N OBEL prizes in recent years
have documented convincingly.
Looking back at the technical development of modern industrial society which
is closely connected with modern physics one may refer (COSE, 1998) to the
20th century as that of the electron while the 21st is the century of the photon. This
interesting particle, the essential ingredient of modern optics and quantum optics,
surprises humankind since N EWTON with its wave-particle dualism. It does not only
play a key role in todays information technology but is, from a general point of
view, also the primary carrier of any information which can be obtained about the
constituents of matter and materials. Even collisions of particles with mass under the
influence of the C OULOMB force may be viewed as exchange of virtual photons.
The textbooks presented here try to give a fairly comprehensive overview on the
whole field. They cover state of the art experimental methods, and combine this
with preparing the basis for a serious, theory based understanding of key aspects in
modern AMO research. The two volumes, originally written in German language
(H ERTEL and S CHULZ, 2008), are a genuine authors translation not just an En-
glish mirror image of the original. We have rewritten much of the text, extended
it wherever appropriate, and updated a number of aspects to catch up with recent
progress in the field.
On the one hand we address advanced students of physics, chemistry and other
neighbouring fields, typically at the end of their undergraduate studies, or during
their doctoral work. On the other hand we also wish to reach young postdocs or
even mature scientists, who feel it is time they connect freshly with the topics ad-
dressed here. We consider the basics of classical geometrical optics and wave optics
as well as electrodynamics to be well known by our readers. We also expect a cer-
vii
viii Preface
the final, physically important results which we discuss and illustrate usually in
some detail. In addition, we provide several appendices for the reader interested
in more detail. We have e.g. collected a toolbox for angular momentum algebra in
atomic and molecular physics without any claim for full mathematical consistency,
but quite compact and possibly useful in practice.
Some words about formats, notation, units, typography appear in order:
Each chapter begins with a brief motto setting the tune of the chapter, followed
by short abstract guiding the reader through the text. At the end of each section a
short summary recalls what the readers should have learned from the preceding
text. All chapters build upon each other, but may be read by advanced readers
also individually: this is facilitated by intensive cross referencing of formulas and
figures, extended indices covering both volumes, a list of acronyms and important
terminology as well as references at the end of each chapter.
For clarity and homogeneity we do not reproduce original drawings or other
material from the literature. Rather, all published data have been redrawn (after
digitalization if necessary), are presented in a standard format, and all sources
used in the figures and text are properly quoted.
We consequently use the SI-System for all measurable quantities, and we empha-
size the pedagogical and practical value of a dimensional analysis for complex
physical formulas.1 On the other hand, atomic units (a.u.) facilitate the writing
of many relations in atomic and molecular physics dramatically. Hence, we use
them intensively considering, however, Eh , a0 and t0 etc. simply as abbrevia-
tions for quantities with dimensions. Phrasings such as we set , e, me , c equal
to unity are avoided, since they are highly misleading.
The finite number of letters in the Latin and Greek alphabets makes some incon-
sistencies or unusual designations unavoidable: we mention specifically, that in
order to allow the use of E for the electric field strength (an important quantity
in AMO) we use W (with appropriate indices) for energies of various types (with
the exception of the atomic unit of energy which is internationally defined as Eh ).
Occasionally we use the letter T for kinetic energy and try to avoid the neigh-
bourhood of time and temperature which are often also designated by T . Vectors
are written as r or k, unit vectors in these directions are er and ek , respectively.
We write operators as H , vector-operators as p and tensors of rank k as Ck . For
the unit operator and unit matrix we use 1. For integer numbers we mostly use
calligraphic letters such as N , while number densities are simply N to distin-
guish them from the index of refraction n which is also an often used quantity
throughout this text. Oscillations and other periodic processes are mostly char-
acterized by their angular frequencies (sometimes also by their frequencies )
and the corresponding energies are (or h).
1 We make, however, use of allowed prefixes (NIST, 2000a), such as cm1 as unit of wavenumbers
(which appears ineradicable in the literature). We also use accepted units outside the SI (NIST,
2000b), such as the enormously practical energy unit eV (electronvolt), or b (barn) as unit for cross
sections.
x Preface
Finally, we hope that these books will become a continuing source of reference
for the fastidious reader, working in or just needing to use AMO physics in her or
his special field. We ask all of you to kindly provide us with the necessary feedback.
We shall try to react to useful suggestion promptly. At the home page of the books,
http://www.mbi-berlin.de/AMO/book-homepage, we shall continuously report on
the status, list errata and possibly present additions. For additional reading and
cross referencing we have collected a few related textbooks and monographs in the
reference list below, just as typical examples without any claim for completeness.
Berlin Adlershof, Germany Ingolf V. Hertel
January 2014 Claus-Peter Schulz
References
ATKINS, P. W. and R. S. F RIEDMAN: 2010. Molecular Quantum Mechanics. Oxford: Oxford Uni-
versity Press, 2nd edn.
B ERGMANN, L. and C. S CHAEFER: 1997. Constituents of Matter Atoms, Molecules, Nuclei and
Particles. Berlin, New York: Walter der Gruyter, 902 pages.
B LUM, K.: 2012. Density Matrix Theory and Applications. Atomic, Optical, and Plasma Physics
64. Berlin, Heidelberg: Springer Verlag, 3rd edn., 343 pages.
B ORN, M. and E. W OLF: 2006. Principles of Optics. Cambridge University Press, 7th (expanded)
edn.
Preface xi
B RANSDEN, B. H. and C. J. J OACHAIN: 2003. The Physics of Atoms and Molecules. Prentice Hall
Professional.
B RINK, D. M. and G. R. S ATCHLER: 1994. Angular Momentum. Oxford: Oxford University Press,
3rd edn., 182 pages.
COSE (Committee Optical Science and Engineering): 1998. Harnessing Light: Optical Science
and Engineering for the 21st Century. Washington, D.C: National Academy Press, 360 pages.
D EMTRDER, W.: 2010. Atoms, Molecules and Photons. Berlin, Heidelberg, New York: Springer,
2nd edn.
D RAKE, G. W. F., ed.: 2006. Handbook of Atomic, Molecular and Optical Physics. Heidelberg,
New York: Springer.
E DMONDS, A. R.: 1996. Angular Momentum in Quantum Mechanics. Princeton, NJ, USA: Prince-
ton University Press, 154 pages.
H ERTEL, I. V. and C. P. S CHULZ: 2008. Atome, Molekle und optische Physik 1; Atomphysik
und Grundlagen der Spektroskopie. Springer-Lehrbuch. Berlin, Heidelberg: Springer-Verlag,
1st edn., 511 pages.
H ERTEL, I. V. and C. P. S CHULZ: 2010. Atome, Molekle und optische Physik 2; Molekle und
Photonen - Spektroskopie und Streuphysik, vol. 2 of Springer-Lehrbuch. Berlin, Heidelberg:
Springer-Verlag, 1st edn., 639 pages.
NIST: 2000a. Reference on constants, units, and uncertainties: SI prefixes, NIST. http://physics.
nist.gov/cuu/Units/prefixes.html, accessed: 8 Jan 2014.
NIST: 2000b. Reference on constants, units, and uncertainties: Units outside the SI, NIST.
http://physics.nist.gov/cuu/Units/outside.html, accessed: 8 Jan 2014.
S TEINFELD, J. I.: 2005. Molecules and Radiation, An Introduction to Modern Molecular Spec-
troscopy. Mineola, NY: Dover Edition, 2nd edn.
W EISSBLUTH, M.: 1978. Atoms and Molecules. Student Edition. New York, London, Toronto,
Syndey, San Francisco: Academic Press, 713 pages.
Acknowledgements
Over the past years, many colleagues have encouraged and stimulated us to move
forward with this work, and helped with many critical hints and suggestions. Most
importantly, we have received a lot of helpful material and state of the art data for
inclusion in these textbooks.
We would like to thank all those who have in one or the other way contributed
to close a certain gap in the standard textbook literature in this area that is at
least what we hope to have achieved. Specifically we mention Robert Bittl, Wolf-
gang Demtrder, Melanie Dornhaus, Kai Godehusen, Uwe Griebner, Hartmut Ho-
top, Marsha Lester, John P. Maier, Reinhardt Morgenstern, Hans-Hermann Ritze,
Horst Schmidt-Bcking, Ernst J. Schumacher, Gnter Steinmeyer, Joachim Ullrich,
Marc Vrakking und Roland Wester; their contributions are specifically noted in the
respective lists of references.
Of course, all other sources are also documented there which we have used for
information and which have provided the data used to generate the figures in these
books.
One of us (IVH) is particularly grateful to the Max-Born-Institute for provid-
ing the necessary resources (including computer facilities, library access, and office
space etc.) for continuing the work on this book after official retirement.
xiii
Contents of Volume 1
1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Overview, History and Magnitudes . . . . . . . . . . . . . . . . 1
1.1.1 Quantum Nature of Matter . . . . . . . . . . . . . . . . . 2
1.1.2 Orders of Magnitude . . . . . . . . . . . . . . . . . . . . 5
1.2 Special Theory of Relativity in a Nutshell . . . . . . . . . . . . . 10
1.2.1 Kinematics and Dynamics . . . . . . . . . . . . . . . . . 10
1.2.2 Time Dilation and LORENTZ Contraction . . . . . . . . . 13
1.3 Some Elementary Statistics and Applications . . . . . . . . . . . 14
1.3.1 Spontaneous Decay and Mean Lifetime . . . . . . . . . . 15
1.3.2 Absorption, LAMBERT-BEER Law . . . . . . . . . . . . 17
1.3.3 Kinetic Gas Theory . . . . . . . . . . . . . . . . . . . . 18
1.3.4 Classical and Quantum Statistics Fermions and Bosons 20
1.4 The Photon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.4.1 Photoelectric Effect and Quantization of Energy . . . . . 26
1.4.2 COMPTON Effect and Momentum of the Photon . . . . . 28
1.4.3 Pair Production . . . . . . . . . . . . . . . . . . . . . . 30
1.4.4 Angular Momentum and Mass of the Photon . . . . . . . 30
1.4.5 Electromagnetic Spectrum . . . . . . . . . . . . . . . . . 31
1.4.6 PLANCKs Radiation Law . . . . . . . . . . . . . . . . . 31
1.4.7 Solar Radiation on the Earth . . . . . . . . . . . . . . . . 34
1.4.8 Photometry Luminous Efficiency and Efficacy . . . . . 37
1.4.9 X-Ray Diffraction and Structural Analysis . . . . . . . . 40
1.5 The Four Fundamental Interactions . . . . . . . . . . . . . . . . 43
1.5.1 COULOMB and Gravitational Interaction . . . . . . . . . 44
1.5.2 The Standard Model of Fundamental Interaction . . . . . 46
1.5.3 Hadrons . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.5.4 The Electron . . . . . . . . . . . . . . . . . . . . . . . . 49
1.6 Particles in Electric and Magnetic Fields . . . . . . . . . . . . . 51
1.6.1 Charge in an Electric Field . . . . . . . . . . . . . . . . 52
1.6.2 Charge in a Magnetic Field . . . . . . . . . . . . . . . . 53
xv
xvi Contents of Volume 1
xxvii
xxviii Contents of Volume 2
xxxvii
Basics
1
Overview
Section 1.1 gives a brief survey of the canonical subject areas in physics, of
physics history and the quantum nature of atomic phenomena. An introduc-
tion to orders of magnitude of length, time and energy follows. Section 1.2
summarizes some essentials of special relativity. Section 1.3 introduces some
elements of statistical mechanics and thermodynamics. The photon, key par-
ticle in this text book, enters the scene in Sect. 1.4. Section 1.5 makes a very
short excursion into the nature of the four fundamental interactions and to
the standard model of elementary particles. Section 1.6 deals with the mun-
dane subject of how free, charged particles move under the influence of an
external electromagnetic field. Particles and waves (Sect. 1.7) and the B OHR
model of the H atom (Sect. 1.8) lead us to the foundations of modern physics.
Section 1.9 introduces one of the key concepts of quantum mechanics: space
quantization as discovered in the famous S TERN -G ERLACH experiment
which in turn led, more or less directly, to the discovery of the electron spin,
treated in Sect. 1.10.
Table 1.1 gives a compact overview of the canonical subject areas of modern
physics and connects them to the content of these textbooks: marked in italics
are those themes that are at least partially treated here. Otherwise, the table is
more or less self explaining.
At the beginning of the past century, the history of atomic, molecular and optical
(AMO) physics was almost identical to the history of physics at large. Without any
claim for completeness, we collect in Table 1.2 some milestones in the topical fields
of this textbook, and Table 1.3 highlights some key developments in theoretical
particle physics which we shall touch very briefly in Sect. 1.5. Of course, a host
of fascinating details during the development of modern physics cannot even be
mentioned here.
Table 1.2 Highlights of science history from the idea of the atom to modern physics of atoms,
molecules and quantum optics (incomplete list)
400 BC D EMOCRITOS oo (indivisible)
1808 AD DALTON Multiple proportions
1811 AVOGADRO Molecular theory of gases
1814 F RAUNHOFER First useful spectrometer
1834 FARADAY Electrolysis (FARADAY constant)
1868 M ENDELEEV Periodic table of elements
1869 H ITTORF Cathode rays
1886 G OLDSTEIN Channel rays
1895 RNTGEN X-rays
1896 B ECQUEREL Radioactivity
1897 J.J. T HOMSON e/m for electrons
1898 Marie & Pierre C URIE Polonium, radium
1898 W IEN e/m for ions
1900 P LANCK E = h
1903 RUTHERFORD Atomic nuclei
1905 E INSTEIN E = mc2
1913 B OHR Atom model
1913 M ILLIKAN e-determination
19211922 S TERN & G ERLACH Space quantization
1925 Max B ORN (N OBEL prize 1954) Fundamental research in quantum
mechanics
1926 S CHRDINGER Wave equation
1927 H EISENBERG Uncertainty relation
1947 L AMB and R ETHERFORD L AMB shift for excited H
19581966 S CHAWLOW, T OWNES, Basov, P ROKHOROV, Maser, laser and spectroscopy
M AIMAN, JAVAN, K ASTLER
1986 N OBEL prize Dudley R. H ERSCHBACH, Dynamics of chemical elementary
Yuan T. L EE and John C. P OLANYI processesa
1989 N OBEL prize Norman G. R AMSEY, R AMSEY fringes, atomic clocksb , ion
Hans D EHMELT and Wolfgang PAUL trapsa
1996 N OBEL prize R. F. C URL Jr., H. K ROTO, Discovery of fullerenesa . . . C60 etc.
R. E. S MALLEY
1997 N OBEL prize S. C HU, Methods to cool and trap atoms with laser
C. C OHEN -TANNOUDJI, W. D. P HILLIPS light
1999 N OBEL prize Ahmed Z EWAIL Femto(second) chemistrya
2001 N OBEL prize Eric A. C ORNELL, Cold atoms and
Wolfgang K ETTERLE, Carl E. W IEMAN B OSE -E INSTEIN condensationa
2002 N OBEL prize John F ENN, Koichi TANAKA Electro spray, molecular beamsa , MALDI
mass spectroscopya
2005 N OBEL prize Roy G LAUBER, Theory of optical coherencec and
John H ALL and Theodor H NSCH laser precision spectroscopya
2007 N OBEL prize Gerhard E RTL Chemical processes at surfaces
a Work from several preceding years
b Work from the 1950s
c Work from the 1960s
4 1 Basics
Table 1.3 Theory on the way from electrodynamics to the standard model of the fundamental
interactions (the dates of N OBEL prizes given here refer to discoveries and developments which
typically happened much earlier)
ca. 1850 James Clerk M AXWELL Electrodynamics
1918 N OBEL prize Max P LANCK Energy quanta
1921 N OBEL prize Albert E INSTEIN Law of the photoelectric effect
1932 N OBEL prize Werner H EISENBERG Creation of quantum mechanics
1933 N OBEL prize Erwin S CHRDINGER, Paul D IRAC Wave equations for matter
1949 N OBEL prize Hideki Y UKAWA Prediction of mesons
1954 N OBEL prize Max B ORN Statistical interpretation
of quantum mechanics
1963 N OBEL prize W IGNER, G OEPERT-M AYER, J ENSEN Structure of the nucleus
1965 N OBEL prize T OMONAGA, S CHWINGER, F EYNMAN Quantum electrodynamics
1967 N OBEL prize B ETHE Theory of nuclear reactions
1969 N OBEL prize G ELL -M ANN Quark model
1979 N OBEL prize G LASHOW, W EINBERG, S ALAM Theory of weak interaction
1982 N OBEL prize W ILSON Re-normalization, critical
phenomena
1999 N OBEL prize T H OOFT, V ELTMAN Quantum structure
of electro-weak interaction
2004 N OBEL prize G ROSS, P OLITZER, W ILCZECK Asymptotic freedom of quarks
Closely related and supplementing the above we shall also have to discuss the quan-
tization of energy. As examples we mention the
Wph = h = (1.3)
Jz = mj (1.8)
mj = j, j + 1, . . . , j. (1.9)
Here J is the absolute value of the angular momentum, Jz its projection on a given
axis (here z) in space. One distinguishes orbital angular momentum (integer quan-
tum number) and intrinsic angular momentum or spin (integer for so called bosons,
half integer for so called fermions, see Sect. 1.5). To characterize these properties
of a given orbit or particle we shall refer to its angular momentum (or spin) as being
0, 1/2, 1, etc.
Before going into details, we want to give an orientation about orders of magnitude
for some relevant physical observables with which we shall have to deal. Orders
of magnitudes are always important in physics and the student is well advised to
roughly memorize some of these data. This will turn out to be very helpful when
trying to find her or his bearings later on when indulging into real, own measure-
ments. (A list of the most important fundamental constants of nature is provided in
Appendix A.)
diameter of
Planck length
our galaxy
bacteria
atomic nuclei
next star
universe
insects
human
atoms
virus
cells
1035 1015 1010 105 100 105 1010 1015 1020
dominant forces: length / m
strong interaction gravitation (among quasi neutral objects)
electroweak interaction
electromagnetic interaction
Fig. 1.1 Length scales in the universe and the range of dominance of the four fundamental forces:
strong, electro-weak, electromagnetic and gravitational interaction (see Sect. 1.5)
1018 m quark 1 am
electron
? ? (lepton)
1.1 Overview, History and Magnitudes 7
homo sapiens
Planck time
universe
lifetime
0 meson
human life
one year
neutron
lifetime
earth
see enlarged scale
10-44 10-16 10-12 10-8 10-4 100 104 108 1012 1016
time / s
Fig. 1.3 Time scales in the universe. The range marked with the fat double arrow is shown in
Fig. 1.4 on an enlarged scale
pointed out that the finite extension of the atomic nuclei play a role (proton radius
ca. 0.881015 m) for ultra-high precision spectroscopy (see Chap. 9). On the other
end of the atomic length scale we have to be aware that the wavelength of electro-
magnetic radiation (spectral range of the visible, VIS, from 380 nm to 760 nm, see
Sect. 1.4.5) is of high relevance in all spectroscopic investigations.
primary
processes fast
of photo- digital electronics
synthesis life time of excited
atomic states fast camera shutter
ultrafast physics
1018 1015 1012 109 106 103 1
atto femto pico nano micro milli s
characteristic time scales / seconds (s)
Fig. 1.4 Enlarged time section from seconds to attoseconds (1018 s) from Fig. 1.3. A presently
very active area of research in AMO is ultrafast and attosecond physics
electronic
W & Z boson
rest mass p+
excitation in atoms,
molecular molecules,
molecular vibrations solids
rotation
1 nK 1 K 1 mK 1K 1000 K 1 MK 1000 MK
LHC lead
room
protons
Fig. 1.5 Energy and equivalent temperature of quantum systems: the scale starts today at one
or several 100 pK, referring to the coldest B OSE -E INSTEIN condensates (BEC) and cosmic mi-
crowave background radiation (CMB(R)) at 2.725 K on the one end of the scale. On the other end
we find collisions artificially generated at the large hadron collider (LHC) with 14 TeV for protons
and more than 1000 TeV for lead nuclei. AMO is interested mainly in the red marked energy range
1.1 Overview, History and Magnitudes 9
Table 1.4 Orders of magnitude of atomic interactions here for typical examples such as H, alkali
metal atoms and He; the structure of volume 1 of these textbooks essentially follows this energetic
scheme
Interaction Order of magnitude See
cm1 eV kHz K
Pure C OULOMB Z/r 30000 4 1015 43000 Chap. 2
Exchange (in He n = 2) 1000 to 6000 0.12 to 0.7 3 1010 1400 Chap. 7
to 1.81011 to 8600
C OULOMB screening 3000 0.4 1014 4300 Chap. 3
Fine structure (FS) 1 to 1000 104 to 0.1 3 1010 1.4 Chap. 6
to 3 1013 to 1400
External el. mag. fields 1 104 3 1010 1.4 Chap. 8
Hyperfine structure 103 to 1 107 to 104 3 107 1.4 103 Chap. 9
to 3 1010 to 1.4
which should be memorized. We shall often come across this very important con-
stant, typically indicating some influence of, or connection with special relativity.
The currently best value of , measured with very high accuracy, is reported in
Appendix A, based on 2010 CODATA (NIST 2010). There, one also finds precise
conversion factors between different units of energy.
We finally mention that one may, from fundamental constants, also construct a so
called P LANCK energy WP = c2 (c/G)1/2 = 1.221 1019 GeV. This is an energy
of cosmic magnitude which may be related to the first moments after the big bang.
There are indications for unification of the fundamental forces at such energy, i.e. at
least three coupling constants might become equal at this energy.
Section summary
The history of AMO physics is identical to the early history of physics at
large. We have recalled some basic observations documenting the quantum
nature of submicroscopic matter.
The essence of (any) physics is to obtain a quantitative description of nature.
Numbers and orders of magnitude are thus essential, and a good feeling for
the scales of lengths, time and energy is important for working in physics.
Section 1.1.2 gives a summary.
Specifically, we memorize P LANCKs constant h 6.63 1034 J s and the
dimensionless fine structure constant 1/137 which will be steady com-
panions on our voyage through AMO physics.
10 1 Basics
We cannot give here a serious introduction to the theory of relativity and assume
the reader to be somewhat familiar with E INSTEINs special theory of relativity.
Generally speaking, throughout this textbook relativistic effects will be treated as
a kind of afterthought where necessary (e.g. in Chap. 6). It is, however, useful to
recall here some formulas for later use.
We first remember that N EWTONs equations, specifically his second axiom, re-
main fully valid under relativistic conditions. Thus,
dp
=F (1.11)
dt
at high velocities v, has just to be applied to the relativistic momentum:
mv
p= = mv and p = mc = mc 2 1. (1.12)
1 2
Here m is the rest mass of a particle (its intrinsic mass in its rest frame, also called
invariant mass) and
1 v
= with = (1.13)
1 2 c
is the so called L ORENTZ factor. One may also rewrite (1.13) as
2 2 2 = 1 (1.14)
which is invariant under L ORENTZ transformation (it is a constant). For highly rel-
ativistic particles 1 one obtains
1 1/ 2 2 . (1.15)
Next we recall the equivalence of mass and energy. In the particles rest frame
the famous E INSTEIN relation reads
2 m2 c4 2 2 m2 c4 = m2 c4 ,
2 m2 c4 p 2 c2 = m2 c4
W 2 p 2 c2 = m2 c4 . (1.17)
1.2 Special Theory of Relativity in a Nutshell 11
holds with energies Wi according to (1.19) to which potential energies may have
to be added due to internal or external fields. In addition
relativistic momentum conservation p i = const
Note that
P 1 P 2 = W1 W2 /c2 p1 p 2
is the inner product of two four vectors P 1 and P 2 . The length of a four-vector is
invariant under L ORENTZ transformation (M INKOWSKI Norm):
respectively. In modern theoretical literature the introduction of this relativistic mass is, however,
usually omitted to avoid confusion: it is the energy that changes with velocity, while the rest mass
m is L ORENTZ invariant.
12 1 Basics
W2
P2 = p2
c2
W 2 p 2 c2 = W 2 p 2 c2 = const. (1.20)
This expression describes a system of particles with total energy and total momen-
tum W , p in one frame, and W , p in an other frame of reference, respectively.
This is completely equivalent to (1.17), which was referring to a single particle and
its rest frame on the right hand of the equation. As an example, for two (or more)
particles the transformation from the laboratory system (lab) to the centre of mass
system (cm) would read
2
Wlab plab
2 2
c = Wcm
2
, (1.21)
where in the laboratory frame Wlab and p lab are the sums of relativistic energies and
momenta of all particles, respectively, while Wcm is the sum of all particle energies
in the centre of mass system in which per definition the sum of all momenta is
p cm = 0.
The kinetic energy T of a relativistic particle is defined by
1 + wk /2 3
= 2wk 2wk 1 wk + . (1.24)
1 + wk 4
The approximate equality is an expansion for small kinetic energies. The relation
between the magnitude of the momentum p and the kinetic energy T is derived by
inserting (1.22) into (1.19) and squaring, so that
For small kinetic energies one recovers the classic relation T = p 2 /2m.
A nonrelativistic treatment is only possible if T mc2 . For an electron this is a
rather limited range of energies since
Wrest e = me c2 = 0.511 MeV. (1.26)
1.2 Special Theory of Relativity in a Nutshell 13
We want to transform times and positions that are known in one frame of reference,
say the rest frame of a particle with coordinates x , y , z , t , moving at a velocity
v in respect of another frame, say the laboratory system with coordinates x, y, z, t.
A distance
x in the moving system is seen from the laboratory system at rest as
much shorter (L ORENTZ contraction):
x = x / . (1.27)
Conversely, the time difference
t of two events measured in the moving system
will appear to be stretched in a measurement in the laboratory system (so called time
dilation):
t =
t . (1.28)
We recall the so called twin paradox: the twin brother travelling in a spacecraft at
nearly the speed of light returns only a little bit older, while his twin who remained
on earth has already become an old man.
Closely related is the relativistic D OPPLER shift. Assume, as indicated in
Fig. 1.6, a source moving with velocity v to emit radiation of angular frequency
(wave vector k with k = /c, wavelength = c/ ). In the laboratory sys-
tem at rest one registers an angular frequency (wave vector k) at an angle with
respect to v:
k 1 1
= = = = = . (1.29)
k (1 cos ) 2 1 cos
= 2 . (1.31)
14 1 Basics
We shall come back to this remarkable fact in the context of synchrotron radiation
in Sect. 10.6.2.
In the limit of very small velocities v one expands (1.29) in powers of = v/c
to recover the classical D OPPLER shift:
= k v. (1.33)
For later use we also give an expansion in terms of the kinetic energy. For wk =
T /(mc2 ) = 1 1 we obtain from (1.30) (at = 0 and ):
wk
= wk 2wk + wk 2wk 1
2 + . (1.34)
2
Section summary
Most of the relations from special relativity communicated here will be used
quite often throughout this textbook. The reader may want to memorize at
least (1.12), (1.13), (1.17), (1.20), (1.22), and (1.26)(1.28).
Thermodynamics and statistics are together a big and important theme in physics
and physical chemistry. Many substantial textbooks exist on the subject and a vari-
ety of important aspects may be found well presented by online scripts in the Inter-
net. Here we present only a collection of topics and formulas from this wide field
with particular relevance to atomic and molecular physics. We start with some rather
elementary remarks about exponential probability distributions. They will be exem-
plified for spontaneous decay of unstable (excited) states of quantum systems and
for absorption of radiation. We then present a collection of formulas from kinetic
gas theory and end with probability distributions for classical particles, fermions
and bosons. But before going into specific examples we define some general termi-
nology.
A probability distribution w(x) describes the probability w(x)dx to find a ran-
dom variable between the values x and x + dx. Properly normalized the probability
to find the system with any value of x must be unity:
w(x)dx = 1. (1.35)
0
1. The decay constant A does not depend on the number of particles that are in-
cidentally present in the experiment but only on the properties of the object(s)
studied.
2. It does also not depend on the time at which the particle or state decays: the decay
occurs at some arbitrary, a priori unknown time and we cannot predict what that
time will be; we only know the probability Adt that a decay may happen within
a time interval dt.
No predictions can be made about the destiny of a specific particle or state. How-
ever a quite accurate prediction can be made for a large ensemble of, say N , parti-
cles or objects in unstable states.2
Here and in the following N gives the number of particles in a volume V of interest while
2 Note:
N = N /V refers to the number density (or particle density) measured in units [N] = m3 .
16 1 Basics
dN = N Adt. (1.39)
The minus sign here indicates reduction. If we start at time t = 0 with N0 particles
in one particular initial state, the number N (t) of particles that at time t are still
found in their initial state is obtained by integration:
t
dN
= Adt ln N (t) ln N0 = At
N 0
where we have introduced the mean lifetime = 1/A and the so called half lifetime
1/2 = ln 2/A = ln 2 = 0.692 . The latter is the time during which half of the ini-
tial particles have decayed. Often one is interested in the number of decay processes
per unit time in the sample (at time t):3
dN N (t) ln 2 N0 t/
A= = AN (t) = = N (t) = e . (1.41)
dt 1/2
The exponential decay law is of fundamental nature and describes the statistical
(also called spontaneous or natural) decay of excited atomic or molecular states or
of electron hole pairs (excitons) in a solid as well as the decay of radioactive nuclei
(and thus the decrease of radioactive radiation with time) or of elementary particles
(barions, mesons, etc.) such as the decay of the neutron which we shall describe
in Sect. 1.5.3.
The exponential distribution or decay law (1.40) is displayed in Fig. 1.7. One
directly recognizes the significance of the half lifetime 1/2 . Alternatively to a linear
display shown in Fig. 1.7(a) one often uses a logarithmic scale for the probability.
As seen in Fig. 1.7(b) the exponential decay then gives a straight line. This is often
used to recognize exponential decays and to estimate the half lifetime.
Finally, we may recast (1.40) into a probability w(t)dt for any of the initial N0
particles to decay between a time t and t + dt:
1 dN 1
w(t) = = et/ . (1.42)
N0 dt
We note that this probability distribution is normalized according to (1.35): any
given particle will definitely decay at some time between t = 0 and . Conversely,
the probability that a particle has not yet decayed before a time t is exp(t/ ).
And with (1.37) we verify indeed that the mean lifetime (also mean or average
decay time) is t
= = 1/A as introduced in (1.40). The variance (1.38) for the
exponential decay function is 2 .
3 In nuclear physics this is called activity which must not be confused with the decay constant (or
0.8 1/2
1/4
0.6
1/2 1/8
0.1
0.4
1/4
0.2
1/8
0.0 0.01
0 1 2 3 4 5 6 t / 1/2 0 1 2 3 4 5 6 t / 1/2
Fig. 1.7 Exponential decay law (a) in linear, (b) in logarithmic display. Note that the time axis is
scaled here in units of the half lifetime 1/2 so that for t = 1 and t = 2 and t = 31/2 the probability
decreases to 1/2, 1/4 and 1/8, respectively as indicated by the dashed lines
Exponential distribution functions of the kind just discussed do not only describe
probabilities as a function of time. Another important application is the absorption
of fast moving particles or electromagnetic radiation (i.e. photons, including visible
light as well as -rays) when passing through matter.
We discuss prototypically the absorption of visible light of intensity I (z) when
passing over a distance z through a medium. The intensity is defined as the total
energy transported in the light beam per unit time and area. It is measured in [I ] =
J s2 m2 = W m2 .
Alternatively, in the particle picture, we may describe the photon flux (z) =
I (z)/(h) with [] = particles s1 m2 as a function of position z (each photon
having an energy h). As indicated in Fig. 1.8, light with intensity I (z) at position
z is absorbed on the way through a medium. In analogy to (1.39) and (1.40), the
intensity dI absorbed over a short distances dz is proportional to that distance and
z
18 1 Basics
at position z. Thus, the result is again an exponential decay law here it is the light
intensity which decays with the optical path z travelled.
The proportionality constant is called absorption coefficient and is measured
in units [] = m1 . It may be understood on an atomistic, statistical basis: the ab-
sorber medium may e.g. be an atomic gas in its electronic ground state, it may also
be glass or a liquid. In any case it consists of many absorbing particles (atoms,
molecules, defects in a solid), each of which has a characteristic absorption cross
section ([ ] = m2 ) for the photons which pass through the medium. The absorp-
tion coefficient is then the product of cross section and particle density (of the
absorbers), = N .
For a medium of finite thickness d the transmitted intensity is thus given by the
so called L AMBERT-B EER law
The probability w(z)dz that any of the photons arriving in the initial flux 0 at z = 0
is absorbed between z and z + dz is given by the probability distribution
1 d 1
w(z) = = ez/ l with l = 1/( N ). (1.46)
0 dz l
In analogy to the notation used in Sect. 1.3.1, here l is called mean free path length
of the photon. It is the distance a photon can on average move freely in the
medium without being absorbed. And exp(z/ l) is the probability that a photon
has survived up to distance z.
The statistical interpretation of the properties of ideal and real gases by the kinetic
theory of gases has played an important role in the history of atomic and molecular
physics. Here we just communicate some basic terminology and results without
derivation.
4 Inthe chemical literature one often writes = ln 10C 2.303C, with C being the con-
centration of the absorbent (e.g. in dilute liquids or gases), measured in [C] = mol L1 . The
molar absorption (or extinction) coefficient thus has the dimension [] = L mol1 cm1 =
1000 mol1 cm2 . The so called absorption (or extinction) is then log(I0 /I (x)) = Cx.
1.3 Some Elementary Statistics and Applications 19
One mol of a gas (an SI unit) is defined as the amount of a substance that
contains as many elementary entities (e.g. atoms, molecules, ions, electrons) as there
are atoms in 12 g of the carbon isotope 12 C. The corresponding number of particles
is the AVOGADRO constant:
The relative atomic (or molecular) mass Mr , formerly called standard atomic
(molecular) weight, of an isotope (substance) X is defined as Mr (X) = m(X)/
[m(12 C)/12] where m(X) is the mass of the corresponding atom and m(12 C) that
of 12 C. Note, that one typically finds Mr given in [Mr ] = g mol1 so that the mass
of one atom (molecule) of this substance is m = Mr /NA . Formally, however, Mr is
a dimensionless number and m = (Mr /NA ) g mol1 .
The classical equipartition theorem states that in a system at thermodynamic
equilibrium each degree of freedom that enters quadratically into the total energy
of the system contributes equally to it with kB T /2. The total internal energy of the
complete system is then given by5
kB T RT
U = f NA =f (1.48)
2 2
where f is the number of degrees of freedom per particle, T the absolute tempera-
ture of the gas, kB the B OLTZMANN constant (1.83) and R the molar gas constant
R = N A kB .
In an ideal (atomic) gas only the kinetic energy is of relevance and f = 3. Tem-
perature then corresponds to the average internal kinetic energy u of the particles
according to
1 3
u = mv 2 = kB T . (1.49)
2 2
These particles
move in the gas with an average velocity v 2 , their average momen-
2
tum being m v . From this, one may calculate the pressure p of an ideal gas as the
momentum transfer per unit time and area by elastic collisions and back reflections
with and from the walls containing the gas. Since 1/6 of all atoms move into one
direction we obtains with the particle density N
p = N mv 2 /3 = N kB T . (1.50)
pV = NA kB T = RT (1.51)
5 We use the traditional letter U for this energy/ mol and u for its average per particle.
20 1 Basics
In a real gas particles do collide not only with the walls but also with each
other. With the particle velocity v, the particle flux (i.e. the number of particles
passing through a unit area per unit time) is vN . If is the gas kinetic cross section,
measured in [ ] = m2 , the time tcol and the distance l a particle moves on average
freely between two collisions are
1
tcol = and l = v
tcol , (1.52)
vN
respectively. The bracket
indicates that one has to average over the velocity dis-
tribution N(v) of the gas particles. Somewhat more precisely, in the denominator v
is the relative velocity between the colliding particles.
If only one particle species
is of interest their average relative velocity is 2v
. Thus, the so called mean free
path becomes
1
l= , (1.53)
2 N
where is an average gas kinetic cross section. For typical elastic collisions between
atoms and molecules it is on the order of 1019 m2 . Relations similar to (1.53) may
also be written for the absorption of ions and nucleons or of light, X- or -rays, see
(1.46).
Finally, we come back to the equipartition theorem. For molecules one has to
add 1 to the number of degrees of freedom f for each accessible rotation, and 2
for each vibrational mode (for kinetic and potential energy). Thus, in a solid f =
6 per atom, in a diatomic molecular gas f = 7 per molecule (three translational
coordinates, two rotational axes, one vibrational mode), while for linear and bent
triatomic molecules we have f = 9 and 10, respectively and so on. However, in
these cases one has to account for quantization of rotational and vibrational energies:
they become inaccessible (frozen) at very low temperature. Thus, the equipartition
theorem in its simple form (1.48) holds only at sufficiently high temperature.
When speaking of mean energy, velocity, lifetime or free pathway of particles (pho-
tons, atoms, molecules, ions, electrons, nuclei etc.) one implies that these quantities
are described by a statistical distribution. Classically, B OLTZMANN statistics pro-
vides the statistical distribution of energies in all fields of physics. It refers to the
probability for finding a certain energy u per particle in an ensemble. This energy
may be kinetic energy or internal excitation of the particles (e.g. electronic, vibra-
tional and rotational excitation in a molecule). Quantum physics requires certain
modifications, but for low enough particle densities and sufficiently high tempera-
tures these are very small, as we shall see in a moment.
In view of the possible quantization of energy we have to distinguish discrete
and continuous energy states. Energies of the system may be realized by different
quantum mechanical states. Different states i with identical energies ui are denoted
1.3 Some Elementary Statistics and Applications 21
Ni gi
= exp(ui /kB T ) (1.54)
N Z(T )
or dN g(u) exp(u/kB T )du (1.55)
with the total number density N of the particles and the so called partition function
Z(T ) = gi exp(ui /kB T ) which ensures that summation over all states i on the
right hand side of (1.54) gives 1.
Normalization in the continuum case needs some more detailed consideration.
We exemplify this by the distribution of velocities vx , vy , vz in an ideal gas. With
the particle mass m the kinetic energy is u = m(vx2 + vx2 + vx2 )/2 and g(vx , vy , vz ) =
const (since for < vx,y,z < no velocity vector is a priori more probable than
another) one may write the velocity distribution
3/2
which is normalized such that integration over all velocities gives 1. If, on the other
hand, one is interested in the probability of finding a particle with a certain mag-
nitude of velocity in a range v and v + dv one has to integrate over all angles so
that dvx dvy dvz = 4v 2 dv. This leads to the well known M AXWELL -B OLTZMANN
velocity distribution
dN 2 m 3/2 2 mv 2
= v exp dv, (1.57)
N kB T 2kB T
which is again properly normalized so thatthe integral over all velocities 0 v <
is 1. The most probable velocity is vm = 2kB T /m. One may rewrite (1.57) as an
energy distribution by substituting u = mv 2 /2:
3/2
dN 2 1 u
= u exp du. (1.58)
N kB T kB T
state 1 2 3 1 2 3 1 2 3
Fig. 1.9 How two particles may be distributed onto three states; this illustrates the key difference
between the statistics according to B OLTZMANN (distinguishable classical particles, here red and
black), B OSE -E INSTEIN (indistinguishable particles, no further limitations), and F ERMI -D IRAC
(indistinguishable particles, PAULI principle)
and thus fully recovers the classical equipartition theorem for three degrees of free-
dom in the
form (1.49). It also confirms the expression (1.49) for the averaged square
velocity v 2
= 3kB T /m, while v
= 8T kB /m.
So much about classical statistics. When reconsidering this from a quantum me-
chanical point of view one has to account for phase space quantization as well as for
the indistinguishability of identical particles. The former aspect implies that even
the continuum is not completely continuous. Rather, the 6 dimensional phase space
(3 position and 3 momentum coordinates) has a finite cell size h3 . From this and
with the degeneracy ge = 2s + 1 due to the spin s of the particles under consid-
eration (see Eq. (1.8)) follows the density of states for a gas of non-interacting
particles in the continuum:
gs (2m)3/2 4 2m3/2
g(u) = u = gs u. (1.60)
4 2 3 h3
Its dimension is Enrg1 L3 . We shall give a derivation of g(u) for the model of a
free electron gas in Sect. 2.4.3 and discuss the quantization of electromagnetic radi-
ation in Sect. 2.2.2, Vol. 2. Here we refrain from elaborating on the derivation of the
statistical distributions, refer to the standard textbooks on statistical thermodynam-
ics and quantum statistics, and just summarize some key results.
In respect of indistinguishability, quantum mechanics knows two kinds of par-
ticles that behave differently: Bosons and fermions which we have introduced al-
ready in Sect. 1.1.1 as particles with integer and half integer spin s, respectively.
For fermions (e.g. e , e+ , p, 3 He, etc.) the PAULI exclusion principle (N OBEL prize
1945) holds so that each (discrete or continuum) quantum state can only be occu-
pied by one particle at most. In contrast, bosons (e.g. photons, 2 H = D, 4 He, 12 C,
etc.) are not restricted by the PAULI principle, i.e. each state may be occupied by
many particles. But again, identical bosons are indistinguishable in contrast to clas-
sical theory. Figure 1.9 illustrates the fundamental differences between the three
1.3 Some Elementary Statistics and Applications 23
statistics for the most simple example: how can two particles be distributed onto
three states?
Detailed considerations on the probabilities for populating (many) energy levels
for many particles lead to the different statistics: the B OLTZMANN distribution for
classical particles, the B OSE -E INSTEIN distribution6 for bosons, and the F ERMI -
D IRAC distribution7 for fermions. To compare all three statistics (concentrating on
a quasi continuum of states) we also rewrite the B OLTZMANN distribution (1.55)
suitably:
1
B OLTZMANN dN = g(u)du (1.61)
exp[(u )/(kB T )]
1
F ERMI -D IRAC dN = g(u)du (1.62)
exp[(u )/(kB T )] + 1
1
B OSE -E INSTEIN dN = g(u)du. (1.63)
exp[(u )/(kB T )] 1
6 B OSE first applied this to photons while E INSTEIN generalized it to any bosons.
7 F ERMI and D IRAC developed it independently from each other in 1926, F ERMI somewhat earlier
than D IRAC.
8 Inthermodynamics, the chemical potential is defined as the partial derivative G/ N of the free
enthalpy G (G IBBS potential) with respect to the particle number N at constant temperature and
pressure. Thus, gives the amount of energy that is necessary to change the number of particles
of a system (by 1) without disturbing the equilibrium of the system.
24 1 Basics
For a gas of free, noninteracting particles, we may insert the density of states
g(u) from (1.60) explicitly, introduce a quantity
4 2m3/2
A = gs (1.64)
N h3
which emphasizes the quantum nature of these statistics, and obtain:
dN udu
B OLTZMANN =A (1.65)
N exp( u
kB T )
dN udu
F ERMI -D IRAC =A u (1.66)
N exp( kB T ) + 1
dN udu
B OSE -E INSTEIN =A u . (1.67)
N exp( kB T ) 1
gs (2mkB T )2/3
exp( ) = (1.69)
N h3
gs (2mkB T )3/2
or = = ln . (1.70)
kB T N h3
At high temperatures and not too high densities the chemical-potential is thus neg-
ative. Just to obtain some feeling for typical values of this quantity we note that at
normal conditions N = NL = 2.687 1025 m3 and 273 K for He with m 4 u,
s = 0 and gs = 1 one finds exp( ) 252106 compared to which the additive
constant = 1 in the denominators of (1.66) and (1.67), respectively, is irrele-
vant. This is characteristic for gases under standard conditions where no difference
among the three statistics is observable, and 0.293 eV is virtually identical
for all three statistics.
However, at low temperatures and/or high particle densities significant differ-
ences exist. For F ERMI -D IRAC and B OSE -E INSTEIN statistics has to be deter-
mined from a numerical integration of (1.68) with = 1, respectively.
It is important to note, that for B OSE -E INSTEIN gases must not be positive to
avoid singularities in (1.67) (see, however, Sect. 2.2.2, Vol. 2). For = 0 the integral
on the left side of (1.68) reaches its maximum and can be evaluated in closed form.
1.3 Some Elementary Statistics and Applications 25
(a) (b)
Tc = 0.18K Tc = 0.32K
u
0 1 2 3 4 5 6 0 1 2 3 4 5 6
u / k BT
Fig. 1.10 Comparison of the three statistical energy distributions for an atomic mass 4 u
at 100 mbar and two different temperatures (a) 2.5 K and (b) 1 K. The red lines refer to
B OSE -E INSTEIN statistics as applicable to He atoms, the grey line illustrates the classical
M AXWELL-B OLTZMANN distribution and the black line represents the energy distribution ob-
tained from F ERMI -D IRAC statistics
Section summary
Statistical distributions are important in many areas of classical and quantum
physics. They describe the probability to find an observable at a certain posi-
tion in space or time or with a particular energy, frequency etc. Characteristic
are the mean value (1.37) and the variance (1.38).
Most common is the exponential distribution which we have introduced in
the context of spontaneous decay of excited states (1.40). It may be charac-
terized by the half-lifetime of the excited states (the time after which half of
the excited states have decayed) which is t1/2 = ln 2/A = ln 2, with the
mean lifetime and A the transition probability. Analogue relations hold for
the absorption of radiation though matter, described by the L AMBERT-B EER
absorption law (1.44).
We have reviewed a few basic concepts from kinetic gas theory: the average
free energy per atom and degree of freedom is kB T /2; the mean free path
between two collisions in a gas is l = 1/( 2 N ), with 1015 m2 the gas
kinetic cross section and N the particle density.
We have also discussed the three relevant statistics (1.61)(1.63) for energy
distribution in an ensemble of particles: the classical B OLTZMANN statistics,
F ERMI statistics (valid for fermions, i.e. particles with half integer spin) and
the B OSE -E INSTEIN statistics (valid for bosons, particles with integer spin
quantum number). At temperatures a few degrees above the critical tempera-
ture (1.71) for B OSE -E INSTEIN condensation all three distributions are almost
identical.
From classical wave optics we know that light can be described as electromagnetic
waves: diffraction and interference are the experimental observations onto which
this viewpoint is based. In terms of geometrical optics, light propagation may even
be described simply by so called light rays or beams a notion which can be based
on wave optics as a special case (more in Chap. 1, Vol. 2).
However, light has also particle properties. The key observations documenting
this aspect are summarized in this section. Quantum mechanics, to be addressed in
some detail in the next chapter, reconciles both points of view or rather: it provides
a set of rules for a consistent interpretation of the experimental observations.
One of the fundamental observations on the quantum nature of light is the photo-
electric effect. With light of a wavelength (frequency = c/) one illuminates a
metal surface from which electrons emerge. One measures the kinetic energy T of
1.4 The Photon 27
potential energy
e
T
h
WA
electron
sea x
Fig. 1.11 Simple potential well model to explain the photoelectric effect. The photon of energy
h rises an electron from the electron sea in the metal (bound) into the continuum (free). A hole
is left in the sea
these emitted electrons and makes some quite remarkable observations (astonishing
at least at the time early in the 20th century when this was discovered):
T (max) = h WA , (1.72)
where h is the P LANCK constant (1.2) and WA the so called work function or
electron affinity of the metal surface from which electrons emerge (for experi-
ments in the gas phase WA has to be replaced by the ionization potential WI of
the atoms or molecules studied).
E INSTEIN (1905) in his annus mirabilis presented the interpretation for the photo-
electric effect one of the key steps in the early days of modern physics for which
he received the N OBEL prize in physics9 in 1921: Light energy exists only in well
defined energy packets of
Wph = h = . (1.73)
This energy packet is the elementary quantum of light, called photon. Light obvi-
ously has both: wave and particle character.
To get some numerical feeling, let us consider yellow light (from the sun or
from a sodium street lamp) at a wavelength of = 589 nm. With c = we have
= 5.09 1014 Hz. Thus, Wph = h = 3.37 1019 J = 2.10 eV is the energy of
the photon!
One may visualize the photoelectric effect by a very simple potential well model
for quasi free electrons in a metal. The energetic relations between T , WA and h
are illustrated in Fig. 1.11. If the photon has an energy h > WA it may eject an
electron from the electron sea (where it is bound but may freely move within the
metal) into the continuum (where it is unbound). If the electron originates from the
9 Note: not for his at least equally important theory of special relativity.
28 1 Basics
Fig. 1.12 (a) Experimental scheme to study the C OMPTON effect. (b) Kinematics of the scattering
process. Prior to collision energy (momentum) are W (p) and We (pe ) for photon and electron,
respectively; the dashed quantities refer to the situation after the process
The C OMPTON effect (N OBEL prize in 1927) may be observed with an experi-
mental setup as sketched schematically in Fig. 1.12(a). Highly energetic photons
( -radiation) are scatted from quasi free metal electrons. The momentum of the
photon is
= 0 incoming radiation
(dashed) = 90
= 45 intensity of the
scattered = 135
- radiation
= 0.710 = 0.710 ' = 0.751
tered -radiation is shifted in respect of the incoming . The shift depends on the
scattering angle :
= C (1 cos ). (1.75)
10 Often the reduced C OMPTON wavelength /me c = a0 = 3.8110 1012 m is used. In rel-
ativistic quantum mechanics one typically measures lengths in units of the reduced C OMPTON
wavelength /me c, and atomic energies in units of me c2 .
30 1 Basics
The photoelectric effect and the C OMPTON effect are two major mechanisms by
which high energy photons interact with matter. For completeness we also mention
pair production: in the vicinity of a nucleus a photon can be converted into an elec-
tron and a positron. This process is symbolically written as e +e+ and has the
energy balance
h = 2me c2 + Te + Te+ , (1.77)
so that this process becomes possible if and only if the energy of the photon
h > 2me c2 1.022 MeV, the rest mass of one electron and one positron. The ex-
cess energy is converted (essentially) into kinetic energy Te+ +Te of the two emerg-
ing particles. For momentum conservation (at threshold the electron and positron
momenta are very small) the process can only occur in the presence of a nucleus
which by C OULOMB interaction carries away the surplus momentum of the
photon. Pair production may also be viewed as excitation of an electron into the
world from the D IRAC sea in which a whole is created (the positron).
All three processes, photoelectric effect, C OMPTON effect and pair production,
are the key mechanisms for absorption of high energy photons in by atoms (specifi-
cally in the solid state). We shall come back to this in Sect. 10.5.3.
For completeness we mention that the exactly inverse process is extremely im-
probable due to phase space considerations. In contrast, the generation of two pho-
tons by positron-electron annihilation e +e+ 2 is a well known process, in
which energy and momentum conservation is easily realized. The two photons are
emitted in exactly opposite direction. This process is exploited in positron emission
tomography (PET), today a widely used medical technique for precise tumour imag-
ing. The positron in this case originates from an artificial isotope attached to a drug
which is accumulated specifically in tumour cells. Detecting the two photons after
e e+ annihilation in coincidence allows one to localize their origin in the human
body.
For completeness we mention already here that the particle photon also has an
intrinsic angular momentum , called the photon spin with a quantum number s = 1.
We shall learn more about the experimental evidence in Sect. 4.1.4. The photon
spin will play an important role in various contexts, and a quantum mechanical
description of the photon will be presented in Sect. 2.2, Vol. 2.
We may even attribute a mass to the photon, the equivalent of its energy:
mP h = h/c2 . (1.78)
Note, however, that the rest mass of the photon is zero: it exists only as a particle
moving with the speed of light. We shall later on discuss that this has serious conse-
quences for space quantization of its angular momentum. Briefly, the massless parti-
1.4 The Photon 31
cle photon has only two substates with sz = jz = , while according to (1.9) a par-
ticle with rest mass and angular momentum j = s = 1 has three possible substates.
Electromagnetic radiation is the key for most spectroscopic studies in atomic and
molecular physics. The relevant radiation ranges from radio frequency photons
whose energy is in the eV energy range up to the hard X-ray region with energies
up to MeV. Figure 1.14 gives an extended overview on the whole electromagnetic
spectrum of relevance. Note that slightly different definitions are used for the spec-
tral ranges shown in Fig. 1.14, depending on the field of application. We follow the
specifications of ISO 21348 (2007).11 Different units are used in different spectral
regions: Frequencies in the very low energy range, wavelengths in the infrared
(IR), in the visible (VIS), ultraviolet (UV) and vacuum-ultraviolet (VUV) spectral
range.12 For still shorter wavelengths, i.e. in the extreme ultraviolet (XUV), in the
soft and hard X-ray region as well as for -rays one uses energy units ( in eV,
keV, MeV). In spectroscopy a convenient measure is the reciprocal wavelength, the
so called wavenumber
= 1/ (1.79)
that is proportional to the photon energy
The SI unit of wavenumbers is m1 but commonly one still uses [ ] = cm1 , often
literally called wavenumber. Up to date energy conversion factors are found e.g.
at NIST (2011).
Quantum mechanics, one may say, was triggered by understanding the photoelec-
tric effect due to E INSTEIN (1905) (N OBEL prize 1921). But before that (1900),
P LANCKs law had already revolutionized the world of physics (N OBEL prize 1918)
giving an accurate interpretation of black body radiation for which the depen-
dence on wavelengths and absolute temperature T had been measured with very
high precision. The characteristic behaviour is illustrated for several temperatures
in Fig. 1.15. The interpretation of this fundamental distribution of radiation forced
Cs atomic clock
9192631770 Hz
FIR
soft
RF EUV X-ray
microwave IR visible UV hard X-ray -ray
XUV
A BC VUV
(c) 1 km 1m 1 mm 1 m 1 nm 1pm 1 fm
Fig. 1.14 Spectrum of electromagnetic waves (for acronyms see text and p. 81ff.). (a) Specific sources and properties, (b) terminology, (c) wavelength scale, (d)
frequency scale, (e) energy scale, (f) examples for applications, (g) examples for methods of generation. Note the narrow range of visible radiation. Only a few
special sources of radiation are specifically emphasized such as CMB(R) and H Lyman-. The ultraviolet (UV) regions UVA, UVB, and UVC are abbreviated
1 Basics
here by A, B, and C
1.4 The Photon 33
1.0
1.0 5772 K (sun)
1073 K
3400 K 288 K (earth)
0.5 4000
10 106
u( ) / J m3 m-1
2856 K
0.5 10 0 5 10 15 m
1873 K
100
0.0
0 1 2 3 4 5
visible spectrum (390 to 750) nm wavelength / m
Fig. 1.15 P LANCKs law at different temperatures: 5772 K effective black body temperature of
our sun, 3400 K special purpose, short-lived incandescent bulbs, 2856 K CIE standard illumi-
nant A (about 100 W tungsten incandescent bulb), 1873 K blast furnace at the discharge aperture,
1073 K dark red glow (kitchen stove, grill fire), 288 K earth surface temperature (average)
8h 3 d
u()d = , (1.81)
c 3 exp(h/kB T ) 1
3 d
u()d = (1.82)
2 c3 exp(/kB T ) 1
Alternatively one plots as done in Fig. 1.15 the distribution as a function of the
wavelength so that [u()]
= J m4 = 103 mJ m3 m1 and
8hc d
u()d = . (1.84)
exp(hc/kB T ) 1
5
34 1 Basics
Equations (1.81) and (1.84) are the standard form of P LANCKs law for the black
body radiation density as originally published by P LANCK (1900). Often one is
also interested in the intensity at a given point, direction and wavelength per unit of
projected area, solid angle and frequency interval with [L ] = W m2 sr1 Hz1
(or [L ] . . . per wavelength interval), also called spectral radiance. Since black
body radiation is intrinsically isotropic, the corresponding expressions are obtained
by simply multiplying (1.81) and (1.84) with c/4 , i.e. the prefactor 8hc is re-
placed by 2hc2 . Typically measured is the spectral distribution of the intensity,13
i.e. of the radiation power emitted (per area) into the forward hemisphere, with
[I()] = W m2 nm1 . Integration over the cos angular distribution (projection
of the surface area onto the direction of emission) gives a factor of so that
2hc2 d
I()d = . (1.85)
5 exp(hc/kB T ) 1
The wavelength max at which this spectral distribution of radiation reaches its
maximum, decreases with temperature as seen in Fig. 1.15. Explicitly one finds
(from dI()/d = 0 for the maximum) the so called W IEN wavelength displace-
ment law:
max T = b with b = 2.8977721(26) 106 nm K. (1.86)
Finally, the total power emitted per unit area from the surface of a black body
is obtained by integrating (1.85) over all wavelengths. This gives the important
S TEFAN -B OLTZMANN law for the intensity of black body radiation:
2 5 kB4 4
I (T ) = I()d = T = B T 4 . (1.87)
0 15 h3 c2
Thus, the (spectrally integrated) intensity depends on the fourth (!) power of
the absolute temperature T . The proportionality factor B = 5.670373(21)
108 W m2 K4 is called S TEFAN -B OLTZMANN constant.
At this point, a few words are in order on the radiation which we receive every day
from our sun. Some relevant parameters are summarized in Table 1.5. The spectral
distribution of the solar radiation at the top of the earth atmosphere is shown in
Fig. 1.16 (wiggly red line). Since about 2004 it is constantly monitored by satellite.
The solar constant S is the integral over this spectrum. The daily results can be
obtained online from SORCE (2012) (the data reported there are renormalized to
13 In radiometry one uses the terms irradiance or radiant flux (see also Sect. 1.4.8). For consistency
with the general custom in AMO physics we usually call this quantity intensity of the radiation,
measured in [I ] = W m2 .
1.4 The Photon 35
Table 1.5 Some properties of sun and earth in the context of solar radiation (data from SSE 2012,
unless otherwise specified)
Mean radius sun R 6.9551 105 km
Mean radius earth RE 6371.0 km
Mean distance sun-eartha RSE 149.60 106 km 1 uab
Solar constantc S 1360.8(5) W m2 (from KOPP and L EAN 2011)
Radiation power emitted by sun 384 109 PW (into 4 sr)
Radiation power received by earthd 173.5 PW (from the sun)
Effective temperature TS 5772 K (for the above value of S)
Sun surface temperature Tph 4400 K to 6600 K (top to bottom photosphere)
Albedo (B OND) a 0.306 (fraction of radiation reflected)
Earth temperaturee TE 254 K (effective black body)
TEa 288 K (average at surface)
a Dueto the ellipticity of the orbit the distance varies between parhelion (minimum) and aphelion
(maximum) by about 6.9 % between 4th of January and 4th of July
b Per definition the astronomical unit of length is 1 ua = 149597870700 m; 1 ua is almost identical
O2 PLANCK at 5772 K
I () / W m2 nm-1
1.0 0.05
H 2O
solar H2O 20
H2O
~
O3 spectrum
at sea H2O 20
H2O,
level CO2
0.0
0 1000 2000 3000 4000
UV vis IR wavelength / nm
Fig. 1.16 Spectral intensity distribution I() of the solar radiation: as measured by SORCE
(2012) above the earths atmosphere (for > 2400 nm from ASTM 2008); at sea level (AM1.5
global tilt spectrum from ASTM 2008, see text); black body radiator at 5772 K. Also indicated
are the main absorbing molecules in the earths atmosphere
36 1 Basics
Fig. 1.17 Definition of air mass coefficient AM for solar radiation standards. Table on the right:
solar radiation (integrated from 280 to 4000 nm) at normal incidence for different zenith angles
a the astronomical unit of length 1 ua, see Table 1.5). The spectral distribution is
remarkably stable over time, the average changes being less than the width of the
red line in Fig. 1.16. As illustrated, the distribution is approximated reasonably well
by a P LANCK distribution from a black body at 5772 K (smooth black line). The
latter is obtained from (1.85) by multiplying it with (R /1 ua)2 , where R is the
solar radius. No further scaling is needed to obtain this fit.
As we see, the fit is not perfect, but considering the gigantic nuclear fusion re-
actor which our sun actually is, and considering its complicated photosphere, the
relative similarity with a black body radiator is quite remarkable and so is the sta-
bility of the distribution. A variety of values for the sun temperature may be found
in the literature. From the recent, highly accurate measurement of the solar constant,
S = 1360.8 W m2 , by KOPP and L EAN (2011) (see also SORCE 2012) and with
appropriate scaling of the S TEFAN -B OLTZMANN law (1.87), we obtain the effective
black body temperature of the sun as TS = (1 ua /R )1/2 (S/B )1/4 = 5772 K.
Also shown in Fig. 1.16 is the spectral distribution of radiation which reaches the
ground level after partial absorption and scattering of the incoming solar radiation
by atmospheric gases. As sketched in Fig. 1.17, this obviously depends on the zenith
angle of the sun ( = 90 latitude angle). For not too large the optical path
length through the atmosphere is given by hx = h0 / cos , where h0 7.7 km is the
effective vertical height of the earths atmosphere (i.e. the height where the pressure
has dropped to 1/e of its value on ground). The ratio
hx 1
AM = (1.88)
h0 cos
is called air mass coefficient. An approximate empirical formula (see e.g. H ONS -
BERG and B OWDEN 2012) for the intensity reaching the earth surface is
0.678
I = 1.1 I0 0.7AM , (1.89)
1.4 The Photon 37
where I0 is the radiation arriving on top of the atmosphere (essentially the solar
constant S, given in Table 1.5). The prefactor 1.1 accounts for radiation scattered
from the air and retroreflected from the ground, the exponentials reflect somehow
the L AMBERT-B EER absorption law. With reference to Fig. 1.17 we also note that
the intensity depends in addition on the angle of incidence ( ) onto the receiving
surface, Isur = I cos( ), with (1.89) referring to normal incidence.
As a standard value, AM = 1.5 (short AM1.5) has been adopted, considered rep-
resentative for most industrialized countries at noon time. Based on measurements
and modelling, one also has defined two standards spectral distributions arriving on
the earth surface, both for AM1.5: (i) direct normal incidence of radiation and (ii)
hemispherical (global) incidence on a 37 degree tilted surface, which also includes
scattered and retroreflected radiation (ASTM 2008, G173-3).14 The latter spectrum
is shown in Fig. 1.16 and referenced the table with Fig. 1.17.
We note at this point that this spectrum has its maximum at ca. 500 nm, i.e. near
the maximum of the spectral sensitivity of the human eye at 555 nm. The evolution-
ary context behind this coincidence is evident.
In contrast, the surface of our earth with an average temperature of ca. 288 K
emits in the IR spectral range with a maximum at about 10 m where the green-
house gas CO2 absorbs (we shall come back to this in Sect. 5.3.1, Vol. 2) fortu-
nately we have to say: without it, the temperature on the earth surface would cor-
respond to the effective black body temperature of the earth TE = 254 K and life
would not be possible.
Black body radiators are very inefficient when used to generate visible light. Even
the sun emits only a fraction of its total radiation energy into the visible (VIS) spec-
tral region: only 46 % of the total radiation power are emitted between 380 and
760 nm, as one finds from integrating (1.87). Incandescent light bulbs at typical
temperatures convert much less energy into the visible spectral range (e.g. 6.5 %
for a 100 W bulb with its tungsten wire at 2856 K). Worldwide it is understood that
intense efforts are needed to save energy. Primary energy resources must be used
in a most efficient way and energy efficient lighting is an import potential. As a
consequence, the good old light bulb, which humankind has gotten so used to since
over 100 years, is approaching its last days very rapidly.
The overall efficiency in creating luminosity as registered by the human eye is
even much worse, since u() has to be multiplied by the physiological sensitiv-
ity V () of the eye. This so called photopic luminous efficiency function peaks at
555 nm,15 where the human eye has its maximum sensitivity. It is standardized by
14 See also ISO 60904-3 (2008) or DIN EN 60904-3. Note that for AM1.5 normal incidence and
Table 1.6 Relation between photometric and radiometric quantities after O HNO (2010)
Photometric Unit Relation to lm Radiometric Unit
Luminous flux lm (lumen) Radiant flux W (Watt)
Luminous intensity cd (candela) lm sr1 Radiant intensity W1 sr1
Illuminance lx (lux) lm m2 Irradiance W m2
Luminance cd m2 lm sr1 m2 Radiance W sr1 m2
Luminous exitance lm m2 Radiant exitance W m2
Luminous exposure lx s Radiant exposure J m2
Luminous energy lm s Radiant energy J (joule)
Color temperature K (Kelvin) Radiance temperature K
the Commission international de lclairage (CIE) and forms the basis for all lu-
minosity determinations. For back of an envelope calculation it may roughly be
approximated by a Gaussian, centred at 560 nm with a FWHM 104 nm (from a
least square fit). But for precise work tabulated values from the literature have to be
used, e.g. from CIE (or as plotted by D ICK LYON 2008, with detailed references).
For a brief and concise introduction into photometry we refer to O HNO (2010)
from whom we also have adopted Table 1.6. It compares (physiologically weighted)
photometric quantities with (directly energy related, physical) radiometric quanti-
ties.16 The relevant photometric SI unit is the candela, cd, defined as . . . the lumi-
nous intensity, in a given direction, of a source that emits monochromatic radiation
of frequency17 540 1012 Hz and that has a radiant intensity in that direction of
(1/683) W per sr.
However, the physically more fundamental unit is the lumen (lm), the unit for the
luminous flux. It measures the photometric equivalent of the total radiation power
emitted by a given light source. Thus, the hypothetical, ideal light source driven by
1 W electric power emits 683 lm at 555 nm. It is said to have an overall luminous
efficacy of 683 lm W1 and its luminous efficiency is defined as 100 % it emits the
maximum possible amount of visible light.
All other light sources have (much) lower luminous efficacy and efficiency. One
may compute the
V ()I()d
luminous efficiency = 0 (1.90)
I()d
0
16 Note that this radiometric terminology somewhat confusingly differs from the standards used
elsewhere in optical and laser physics. A laser beam e.g. is typically characterized by its intensity
measured in units [I ] = W m2 , while here in radiometry the corresponding quantities in Table 1.6
are called irradiance or radiant exitance.
17 That is at a wavelength of 555 nm where the human eye has its maximum sensitivity.
1.4 The Photon 39
ted per unit area. This luminous efficiency is a function of T and has a maximum of
about 14.5 % at T = 7000 K. For our sun it is ca. 14 %, while for the above men-
tioned CIE standard Illuminant A the efficiency is only about 2.5 %, correspond-
ing to a luminous efficacy of ca. 17 lm W1 . A typical standard 100 W incandescent
bulb emits a total luminous flux of 1360 lm and thus its efficacy is 13.6 lm W1 . Of
all the electric energy used to heat the tungsten wire of the bulb, 98 % is lost as IR
radiation and eventually dissipated as heat! So called halogen lamps are somewhat
more efficient since a special chemical process reduces evaporation of the cathode
material and the temperature of the tungsten wire can be significantly higher. Still,
the general efficiency problem of incandescent lamps remains.
Thus, the challenge is, to exploit more efficient ways for converting electrical
energy into visible light. One direction that currently is followed uses miniaturized
fluorescent lamps. Typically, in a mercury gas discharge UV light is generated at
253.7 and 185 nm. It is then converted into visible light of different wavelengths
by fluorescing materials, so called phosphors. Typically three to five different phos-
phors with reasonable efficiency are used, whose fluorescence bands by additive
colour mixing appear more or less as white light. Much current development is
focused onto these materials. By judicious choice of the fluorescent bands one may
obtain a more or less continuous coverage of the visible range thus imitating the
visible part of the black body radiation without loosing energy into other spectral
regions.
The colour temperature of such a source is defined as the temperature of a black
body radiator that best matches that spectrum in the visible. Presently, the lumi-
nous efficacy of compact fluorescent lamps (commercial energy saving lamps) is
on the order of (50 to 60) lm W1 (7.3 % to 8.7 % efficiency) but up to more
than 100 lm W1 (14.6 % efficiency) can be achieved with long tubular fluorescent
lamps. For street lighting also high and low pressure sodium discharge lamps are
used, easily recognized by their yellow-orange light (around 590 nm). Low pres-
sure Na discharges have the highest luminous efficacy in state-of-the-art lighting
technology of up to 200 lm W1 (29 % efficiency).
Light emitting diodes (LED) could, in principle, be able to supersede these val-
ues: they convert electric current so to say more or less directly into light. Again,
additive colour mixing of at least three LEDs is required for generating white light
in practical devices. The luminous efficiency is presently of a similar order of magni-
tude as for fluorescent lamps, somewhere between 15 and 25 %. However, strategies
are discussed to achieve much higher efficiencies (e.g. B RETSCHNEIDER 2007), and
the price for such high-tech products typically decreases exponentially with years
of experience and mass production.
Occasionally the question comes up about a theoretical maximum of luminous
efficiency. It is difficult to answer: consider a (very hypothetical) light source con-
verting electric power to a 100 % into a spectrum, which in the visible (380 and
760 nm) corresponds exactly to the ideal white light of our sun, and is zero else-
where. According to (1.90) this would correspond to a luminous efficiency of 38 %.
Hence, conversion of 100 % electric energy into visible light is per definition
a completely unrealistic goal: it would mean totally green illumination. Thus, any
40 1 Basics
d lattice planes
s
d
claims for a luminous efficiency higher than 3438 % have to make massive com-
promises on the effective colour temperature. We shall certainly witness an exciting
development in the lighting industry over the coming years.
Electromagnetic radiation in all spectral ranges light in the widest sense is today
one of the most important tools to reveal the structure and dynamics of matter. In
later chapters we shall learn a lot about the various spectroscopic techniques that are
used in this context.
At this point we want to mention, at least briefly, one of the key methods for struc-
tural analysis: X-ray diffraction, i.e. scattering and interference of electromagnetic
radiation of very short wavelength from crystalline matter. It is treated systemati-
cally in textbooks and monographs on solid state physics (see also Sect. 1.7.2).
The basis for a variety of such methods is multiple beam interference from the
crystal lattice structure studied. As sketched in Fig. 1.18 a multitude of parallel, so
called lattice planes may be thought to reflect the X-rays. According to Fig. 1.18
the optical path difference between two rays reflected from neighbouring planes is
2s = 2d sin , where d is the distance of two lattice planes in the crystal and the
so called B RAGG angle.
Thus, the reflected X-ray beams (shown in Fig. 1.18 are only two of them) inter-
fere constructively if and only if B RAGGs law (also called B RAGG condition)
holds. Here is the wavelength of the scatted X-ray radiation and z is an integer.
For reference we also report some further quantitative relations relevant to X-ray
scattering. One defines a reciprocal lattice vector
G = hg 1 + kg 2 + lg 3 (1.92)
1.4 The Photon 41
constructed with the basis vectors of the unit cell g 1 , g 2 , g 3 in the reciprocal lattice,
using the so called M ILLER indices h, k, l that characterize the lattice planes in the
crystal. Without entering into details of crystal lattice theory we just note that these
basis vectors in the reciprocal lattice relate to the ordinary basis vectors a 1 , a 2 , a 3
in the crystal lattice by
g i a j = 2ij , (1.93)
and in Fig. 1.18 we identify d 2/|G|. With wave vectors k and k of incoming
and outgoing X-ray radiation B RAGGs law (1.91) is now written
k = k k = G. (1.94)
Normally one investigates only elastic scattering where |k| = |k | and one may
rewrite (1.94) as (k G)2 = k 2 or G2 2kG = 0. We may finally write the condi-
tion for diffraction:
2
G G
2kG = G 2
or k = . (1.95)
2 2
In the latter form the diffracted wave vectors describe a plane bisecting the recip-
rocal lattice vector G and being perpendicular to it. One may construct such planes
for all elementary reciprocal lattice vectors and combine them to a closed surface
in reciprocal lattice space. They form the so called B RILLOUIN zone (BZ) that rep-
resents all wave vectors of radiation that can be B RAGG reflected by the crystal.
BZs are a very important concept in solid state physics, specifically relevant for the
theory of band structure as we shall briefly discuss in Sect. 2.8. Since several dif-
ferent lattice vectors with different alignment in space exist (typically more than 3,
see Eq. (1.92)), BZs may be rather complex surfaces. Corresponding surfaces may
also be constructed from 2G, 3G etc. and one distinguishes the 1st, 2nd, 3rd, etc.
B RILLOUIN zone.
The intensity of the diffracted X-ray radiation depends on the (electron) charge
density distribution (r) in the unit cell of the crystal and is proportional to the
absolute square of the so called structure factor
F(hkl) = d3 r(r) exp(iG r) (1.96)
cell
= Fj (G) exp i2(xj h + yj k + zj l) .
j
The summation has to be carried out over all atoms in the unit cell of the crystal,
each of which is characterized by its atomic form factor18
Fj (q) = d3 rN (j ) (r) exp(iq r). (1.97)
atom
(a) (b)
Fig. 1.19 (a) X-ray diffraction image from a large single crystal (80 100 50 m) of the human
enzyme prolidase, recorded with h = 13.05 keV at the beamline BL14.1 of the Free University
Berlin at BESSY.The maximum resolution of the diffraction image corresponds a lattice plane
distance of the crystal lattice of 0.25 nm. In the squares magnified sections of the image are shown.
(b) Secondary structure mode of the enzyme in dimer form with unbound Mn2+ (red spheres).
With kind permission from M UELLER et al. (2007) and private communications
These form factors have to be determined for each atom by integration over the par-
ticle density N (j ) (r) of all its electrons (the dimension of N being L3 ). Without
entering into the details we note here that the imaginary part of the atomic form fac-
tor is related to the photo-absorption cross section, while the real part characterizes
the elastic (coherent) photon scattering. For radially symmetric charge distributions
it is
(j )
N (r) sin(qr) 2
Re Fj (q) = 4 r dr, (1.98)
0 qr
Section summary
Photons have (i) a well defined energy W = h = (with = c/ and
= ck) as determined by the photoelectric effect, (ii) a momentum p = k
documented by the C OMPTON effect, and (iii) an angular momentum . They
have no rest mass and exist only moving with the speed of light.
The spectrum of electromagnetic radiation (we call it light in a general
sense) ranges from radio frequencies ( km, h 109 eV) to -rays
(105 nm, h 108 eV). The visible spectrum is only a very small part of
it (the wavelength range from 380 and 760 nm).
P LANCKs radiation law (1.81) was a corner stone in the development of quan-
tum physics. It describes the spectrum of a black body such as our sun or
incandescent light bulbs. The maximum of the spectrum shifts with tempera-
ture according to W IENs displacement law max T 2.9 106 nm K, the total
intensity (per unit area) follows the S TEFAN -B OLTZMANN law I (T ) = B T 4 .
A hypothetical, ideal light source with a luminous efficiency of 100 %, driven
by 1 W electric power emits 683 lm (lumen) at 555 nm. Incandescent light
bulbs have a luminous efficiency of only 23 %.
X-ray radiation is a very powerful tool for structural analysis. Constructive
interference occurs at angles of incidence (in respect of the lattice plane)
according to the B RAGG law 2d sin = z. Other formulations of the B RAGG
law make use of the reciprocal lattice vector (1.92) and M ILLER indices. The
structure factor (1.96) and the atomic form factor (1.97) describe the intensi-
ties in the diffraction pattern.
We cannot give here an even brief introduction into particle physics and the un-
derlying theory. The so called standard model (SM) of quantum chromodynamics
(QCD) is a sophisticated theory which cannot be treated in passing. However, it is
important to know a few basics when discussing the interactions relevant in atomic,
molecular and optical physics. The four fundamental interactions
1. Gravitation
2. Electromagnetic interaction
3. Weak interaction
4. Strong interaction
may be seen as what holds the world together in its innermost folds as far as we
understand it today.
In our daily life we are confronted almost exclusively with the first two of these
forces in particular gravitation plays a key role in our everyday experience, while
electromagnetic interactions are perceived commonly in a more indirect manner:
through the action of various machines and equipment, via lighting gears (or occa-
44 1 Basics
Table 1.7 The four fundamental interactions, exchange bosons and coupling constants
Interaction Fermion Exchange Mass Couplingb Range Dependence
(e.g.) boson / GeV c2 /m on distance
Gravitation e , p, n Gravitona 0 5.9 1039 1/r 2
Electromagn. e , p Photon 0 7.30 103 1/r 2
Weak e , W -boson 80.4 105 1018 1/r 5
Z0 -boson 91.2 107 1/r 7
Strong p, n -meson 135.139 1 1015 1/r 7
Quarks Gluons 0 0.119
a Hypothetical, not yet found
b Values for gravitation and electromagnetic interaction refer to a pair of protons
The interaction of two particles depends in general in a complex manner from their
distance r. Only the well known force laws for gravitation
1.5 The Four Fundamental Interactions 45
m1 m2 r
F g = G (1.99)
r2 r
1 q1 q2 r
Fe = (1.100)
40 r 2 r
can be written in this simple, closed form, with masses m1 , m2 and charges q1 , q2 of
the interacting particles, respectively. They may be derived from a scalar potential
V (r) 1/r by
19 We note here that for very high energies the electromagnetic coupling constant em changes
(essentially ln W ); at energies 90 GeV (roughly corresponding to the mass of the Z boson)
experiments have determined 1/em 1/128.
46 1 Basics
The other two forces, strong interaction and weak interaction, play an important role
on a subatomic length scale only: as indicated in Table 1.7, their range of interaction
is finite in complete contrast to C OULOMBs law and gravitation. Strong interac-
tion is mainly responsible for the cohesion of nuclear matter, i.e. it takes care of
the stability of nuclei which otherwise under the influence of electrostatic repulsion
would simply explode. Weak interaction on the other hand plays a decisive role in
-decay, i.e. when an electron is emitted from an atomic nucleus.
Electromagnetic, strong and weak interaction are described today consistently
and convincingly by the standard model of QCD only gravitation resists so far
a unified interpretation. At the core of the theory is a set of objects that may be
considered point like20 (<1018 m) real elementary particles without any internal
structure but with some well defined properties in terms of mass, charge and spin
(intrinsic angular momentum). They are summarized in Fig. 1.20.
The standard model distinguishes two varieties of elementary particles:
1. Matter particles
6 quarks are subject to strong interaction and carry one of three colour charges
red, green or blue; they also carry a charge +2/3e or 1/3e and are
subject to electromagnetic as well as to weak interaction.
6 leptons are all subject to weak but not to strong interaction i.e. they are
colourless. The electron, muon and tauon carry also an electric charge e
and are subject to electromagnetic interaction. In contrast, the corresponding
neutrinos do not have an electric charge. The have, however, a (very small)
mass.
All of these are fermions with spin 1/2. They are complemented (not shown
in Fig. 1.20) by the corresponding antiquarks (in the colours antired, antigreen,
antiblue) and antileptons, their charges are opposite.
2. Exchange particles (or gauge bosons) are considered to be exchanged between
the matter particles during interaction. They are all bosons with spin quantum
number s = 1 and may be seen as mediators of the forces. They are formed for a
very short time only, i.e. virtually, and are unobservable directly.21
0 0 0 electroweak
colourless
Fig. 1.20 Elementary particles according to the standard model: matter particles (fermions, spin
1/2) and exchange particles (gauge bosons, spin 1). The masses and charges are indicated:
masses are given in terms of their energy equivalent Wm = mc2 . *) According to CERN (2013)
On 4 July 2012, the ATLAS and CMS experiments at CERNs Large Hadron Collider announced
they had each observed a new particle in the mass region around 126 GeV /c2 . . . consistent with
the H IGGS boson but it will take further work . . . . This is based on theoretical work for which
E NGLERT and H IGGS (2013) received the N OBEL prize
We have discussed already in Sect. 1.4 the photon. It is the exchange particle
for electromagnetic interaction by which all charged particles interact (in addition
to weak or strong interaction and to gravitation). Strong interaction is characteristic
for quarks as well as for antiquarks (not shown in Fig. 1.20) with charge 2/3e
and +1/3e. They are never observed as free particles and occur only as compound
systems (confinement). On the other hand, at very close distance their interaction
tends to zero (so called asymptotic freedom).
In analogy to electric charge one attributes to the quarks a colour charge by
which they interact with each other. The colour comes in three varieties red, green
or blue (in our graphs coloured red, grey or black). In each composite particle the
colours have to add in such a manner that their sum leads to white (in analogy to
the colours observed from the visible spectrum of light). Antiquarks are coloured
with complementary colours antired, antigreen or antiblue. The exchange bosons
of the strong interaction are called gluons (reminding us of glue). They come in
8 different varieties encompassing one colour and one anticolour each (while the
48 1 Basics
Proton Mass
mp = 1836.15 me
u
Charge = 1e0 = 1.67261027 kg
u Spin = 1/2 = 938.27231 MeV/c 2
d stable = 1.007 276 47 u
Neutron
Charge = 0 mn = 1838.68 me
u Spin = 1/2 = 1.6749 1027 kg
d unstable = 939.5656 MeV/c 2
d t1/2 = 10 min 14s = 1.008664 92 u
Fig. 1.21 Structure of the two nucleons: proton and neutron. They are built from up and down
quarks (u and d, respectively). The special choice of colours is here without direct relevance and is
supposed to just illustrate the principle of colours in quantum chromodynamics: each of the three
fundamental colours (commonly called red, green, blue) has to be present exactly once. The wiggly
lines indicate exchange of gluons (strong interaction)
photon does not contain any charge). Both, gluons and the photon are massless
particles while the three exchange bosons of weak interaction Z0 , W+ and W are
again colourless but have each a (different) mass.
1.5.3 Hadrons
u u
n d d p
d u
W _
e
Fig. 1.22 F EYNMAN diagram for the neutron decay by weak interaction (half lifetime
1/2 = 10 min +14 s): by emission of an electron e and an antielectron neutrino e a down quark
d (charge 1/3e) is changed into an up quark u (charge +2/3e). The interaction occurs via a
virtual W boson
In contrast to the barions the so called mesons (medium heavy particles) are built
from a quark and an antiquark with complementary colours (colour and anticolour),
which makes them also white. Barions and mesons together are called hadrons.
Hadrons are all subject to strong interaction, to weak interaction and as far as
electrically charged also subject to electromagnetic interaction.
The leptons electron, muon and tauon and their respective neutrinos are, in contrast
to the hadrons genuine elementary particles, at least as far as we know today. They
interact among each other as well as with quarks, protons and other barions and
atomic nuclei by electro-weak and (as far as charged) by electromagnetic interac-
tion.
In atomic, molecular and optical physics we are almost exclusively concerned
about electromagnetic interactions and by far the most important particle is the elec-
tron. Let us have a somewhat closer look at it. Its rest mass is
Generally, free and isolated particles are only observed with an integer multiple
of e (positive, zero or negative). We point out again, however, that the bound quarks
and antiquarks have charges 2/3e or 1/3e.
Directly related to the elementary charge is the energy unit electronvolt
Section summary
The four fundamental interaction responsible for keeping the world at its
innermost together are gravitation, electromagnetic interaction, weak inter-
action, and strong interaction. Gravitation and electromagnetic interaction en-
ergies are 1/r, i.e. have in principle infinite range, and are dominant in
macroscopic physics and atomic physics, respectively. Weak and strong inter-
action are of short range on the order 1018 and 1015 m, respectively, and
from an atomic physics point of view refer to forces inside the nuclei.
The standard model of elementary particle physics describes electromagnetic,
weak and strong interaction in a unified scheme. It comprises 6 quarks (matter
particles, fermions), 6 leptons (including the electron, all fermions, subject
to weak and electromagnetic interaction, if charged) and 4 types of exchange
bosons, including the photon which mediates electromagnetic interaction. The
quarks have charges 2e/3, e/3, spin 1/2 and come with three kinds of
colour charges. They are subject to weak, strong and electromagnetic forces.
They are building blocks of all other massive particles, including the nucleons
(neutron and proton) which in turn are constituents of atomic nuclei.
Leptons are a different kind of elementary particles, subject only to weak
and electromagnetic interaction (if charged). Of particular interest in atomic
and molecular physics is the electron. We memorize its electric charge e
1.6 1019 C. To move it through an electric field requires energy or sets
energy free. The atomic unit of energy is thus 1 eV 1.6 1019 J.
dp
F= = q(E + v B). (1.105)
dt
In this form the equation is correct even in the relativistic limit. However, for sim-
plicity we shall restrict the following discussion to the nonrelativistic limit unless
pointed out differently.
Geometry (b and ) as well as the voltage U applied are known. The ratio (and only
the ratio) of kinetic energy T to charge q may be determined according to (1.106) by
the electric deflection method! This is valid in general, completely independent of
how complicated the setup may be. In practical applications one often uses special
capacitor designs e.g. cylindrical or hemispherical capacitors (the latter is described
in some detail in Appendix B.3, Vol. 2) as well as segments of these for determining
the particle energy if its charge is known. In addition to a measurement of the energy,
for extended, divergent particle beams such setups allow to focus the trajectories
from the entrance aperture onto the exit aperture of an energy analyzer.
To also determine the mass of a particle one needs an additional measurable quan-
tity. In a purely magnetic field B the movement of the charge q is determined by
dp
F= = qv B
dt
with the force F being perpendicular to v and B (see Fig. 1.25). The change in
energy
dW
= vF = qv v B 0
dt
is thus identical to zero, and v = const. This holds also relativistically so that = 0
and = const.
The relativistic equation of motion thus reads
dp dv
= m = qv B,
dt dt
with the rest mass m of the particle and p = mv. The particle moves on a circle
that is determined by balancing the centrifugal force Fc and the magnetic force
F = qvB:
dp 2 2
= Fc = mv = mv = qvB, (1.107)
dt r r
p
so that mv = p = qBr or = rB (1.108)
q
mv p
and r = = . (1.109)
qB qB
If one wants to use this orbital radius r for mass selection, again only one parameter
can be extracted: with a magnetic field the ratio of momentum and charge p/q can
be measured.
The expressions given above also hold in the nonrelativistic case where 1.
For highly relativistic energies as they are e.g. realized in electron storage rings built
for the generation of synchrotron radiation (see Sect. 10.6.2) one has more or less
v
54 1 Basics
If one combines the two types of fields judiciously one may achieve with T =
p 2 /2m a determination of m/q:
m 1 p 2 /q 2
= . (1.112)
q 2 T /q
In the classical, so called double focussing mass spectrometers, ions pass succes-
sively electric and magnetic fields. Such devices are still used today but have lost
much of their importance in relation to the FT-ICR mass spectrometers discussed in
Sect. 1.6.3.
Another type of mass spectrometers, today widely used, is the so called
quadrupole mass spectrometer (QMS). It exploits the mass dependence of the dy-
namic stability regions for ion trajectories in a rapidly oscillating electric quadrupole
field (see e.g. D EHMELT and PAUL 1989): for specific ratios of DC to AC fields only
ions with a specific mass to charge ratio m/q may pass on a stable trajectory through
the setup while all other ions undergo increasing oscillations and do not reach the
exit slit. The interesting mathematics follows M ATHIEUSs differential equation (see
e.g. W IKIPEDIA CONTRIBUTORS 2013).
In time of flight mass spectrometers one accelerates ion in electric fields to select
m/q through their time of flight to the detector. In state-of-the-art setups of this type
one exploits a variety of tricks to focus ions of the same mass but from different
starting positions at the ion source (W ILEY and M C L AREN 1955) as well as those
with different initial kinetic energies (M AMYRIN 1994) at the same time on the
detector. With present commercial solutions high mass to charge ratios m/q (up
to 4000 u) may be reached and high mass resolution (m/
m 60 000) can be
achieved.
For special applications two other, classical setups are still useful and deserve to
be mentioned. We already mentioned J.J. T HOMSONs method with parallel electric
(E) and magnetic (B) fields which nowadays is used to analyze particle emission
with high kinetic energies T . Assume the fields to point into +y-direction, the ions
enter the field perpendicular in +z-direction (velocity vz ) and pass through them on
a length l. In this case the L ORENTZ force (1.105) leads to acceleration both in y-
as well as in x-direction. After a time t1 = l/vz the particles have passed the fields
and acquired velocities
Behind the fields the ions are detected on a position sensitive detector: J.J. T HOM -
SON used a photographic plate, today sophisticated electronic devices are available
as we shall describe in Appendix B.1, Vol. 2. If the detector is positioned at z = s
far behind the field, the ions hit the detector at a time t2 = s/vz and their deflections
are
qEls qEls qBls qBls
y = v y t2 = 2
= and x = vx t2 = = . (1.114)
mvz 2Wkin mvz 2mWkin
56 1 Basics
v = E/B
the charged particles just fly on straight line trajectories. If the mass is known one
may thus determine the momentum and hence the kinetic energy.
Electrons in plasmas but also in clusters and condensed matter (in particular quasi
free electrons in metals and semiconductors) may carry out collective oscillations,
so called plasma oscillations with a characteristic frequency that plays an impor-
tant role in many areas of physics. The simplest model starts with is a quasi neutral
plasma of electrons e and ions with a density N of quasi free electrons, equal to
that of the ions. Assume we displace all electrons in respect of the ions by a distance
x as shown in Fig. 1.26.
This leads to a surplus of surface charge density = eN x that causes an elec-
tric field between electrons and ions E = /r 0 . The equation of motion for each
electron in this field is thus
me x = eE = e/r 0 = e2 N/r 0 x.
This differential equation is simply that of a harmonic oscillator with a force con-
stant kP = e2 N/r 0 . The angular frequency of the oscillations
kP N e2
p = = (1.116)
me me r 0
1.7 Particles and Waves 57
p2
r () = 1 , (1.117)
2
which for < p becomes negative.
To interpret this result, we consider a plane wave exp[i(kx t)] propagat-
ing in +x-direction, with k = n/c being the magnitude of the wave vector in
the medium and n = r the index of refraction. According to (1.117), below the
plasma frequency n becomes imaginary and we write k = i. Correspondingly the
wave, exp(x) exp(it), is damped. Thus, the medium absorbs at all frequen-
cies < p and is transparent for > p .
We finally note that in isolated particles, the situation is somewhat more compli-
cated, e.g. so called plasmon resonance phenomena will be observed at p .
Section summary
In an electromagnetic field E, B particles with a charge q are subject to the
L ORENTZ force F = q(E + v B), where v is the particles velocity.
This can be used to manipulate and guide the charged particles (typically elec-
trons and ions) and to select their energy and/or mass/charge ratio.
We note two interesting frequencies: (1) In a homogeneous magnetic field B
a free electron moves on a circle, perpendicular to the field (or spirales) with
the cyclotron
frequency c = v/r = eB/( me ). (2) The plasma frequency
p = Ne 2 /(m e r 0 ) is characteristic for oscillations of quasi free electrons
in clusters and metals (particle density N ).
screen
cathode +U
aperture
Wkin low Wkin high
Fig. 1.27 D EBYE -S CHERRER diffraction schematically: (a) setup and generation of the diffrac-
tion cones; (b) typical diffraction patterns for a poly-crystalline target: top view onto the screen (a)
as seen from the right
h 1.23 nm
dB = = . (1.120)
2me T T / eV
At a kinetic energy 100 eV the wavelength of an electron is thus on the order of the
size of an atom 0.1 nm. Note, however, that (1.118) also holds in the relativistic
case, with p given by (1.25). Specifically, for electrons that have been accelerated
by a voltage U one has to use
eU
p= 2me eU 1 + . (1.121)
2me c2
With the rest energy me c2 = 0.511 MeV for electrons this leads already at a moder-
ate kinetic energy of 50 keV to a reduction of the wavelength of 2.5 % in comparison
to (1.120). In a precision experiment this clearly has to be accounted for.
elastically
diffracted electrons
Fig. 1.28 Low energy electron diffraction (LEED): (a) experimental setup; (b) typical diffraction
pattern from a single crystalline ordered surface
The diffraction structures are generated by many single reflexes from the
micro-crystals and appear according to (1.91) under the B RAGG angles 2 =
arcsin(z/2d) with z = 0, 1, 2 . . . The various distances d of the crystals lattice
planes contain information about the crystal structure. Since the crystals are aligned
at random into all directions of space, all electrons diffracted into an angle 2 will
be emitted into a cone around the incoming electron beam. For each m and each
d such a cone exists. For each cone a circle is formed at its intersection with the
plane of observation (photographic plate, or today more likely a CCD camera). In
Fig. 1.27(b) such diffraction patterns for electrons of low and higher energy are
sketched schematically.
10
6 4 2 0 2 4 6
scattering angle / mrad
even be diffracted by quasi macroscopic optical elements, and they interfere. For
this nice experiment a transmission grating was manufactured by state-of-the-art
nano-technology. The diffraction pattern may be completely understood in terms
of K IRCHHOFFs diffraction theory which has been developed more than 150 years
ago.
One may of course ask how far one can push such kind of wave optics with
particles and particle beams. This is an interesting theme which may even be po-
tentially relevant for technical applications. Atom lithography and atom optics with
very cold atoms is subject to current research. We have already mentioned briefly in
Sect. 1.3.4 research on B OSE -E INSTEIN condensation (BEC). One particular fasci-
nating aspect is that the wavelength of the matter waves increasing with decreasing
temperature according to (1.118). It eventually reaches values comparable to the av-
erage distance between atoms (particle density N 3 ). With (1.118) and (1.50)
this happens when
h2
T N 2/3 . (1.122)
3mkB
Comparison with (1.71) shows that this estimate agrees with the critical temperature
for BEC apart from a numerical prefactor and the missing spin degeneracy. This
is a nice, plausible result.
One may also push the wave concepts into the opposite direction and ask about
the macroscopic limits of the wave-particle
duality: at constant kinetic energy the
DE B ROGLIE wavelength decreases m, i.e. with the square root of the mass.
Diffraction experiments with larger molecules are thus increasingly more difficult
and much more challenging.
Z EILINGER and his associates (see e.g. A RNDT et al. 1999) were able to show
that even such big objects as Fullerene molecules, C60 , are subject to wave mechan-
ics. In a single slit diffraction experiment one observes quite normal diffraction
patterns, as documented in Fig. 1.30 and as one actually would expect. The ul-
timate conundrum in this context is about really large objects as sketched in the
cartoon Fig. 1.31 the question is, what will happen? We leave this as a little mental
1.7 Particles and Waves 61
exercise for the reader who may want to estimate orders of magnitude for a pertinent
experiment.
px x (1.123)
22 Note that measurements of microscopic observables are based on probability distributions. Thus,
the exact value on the right hand side of the uncertainty relation (, h/2, h etc.) depends on the na-
ture of that distribution and on the exact definition of uncertainty, e.g. FWHM, foot width, position
of the first minimum, etc.
62 1 Basics
Fig. 1.32 Classical Gedankenexperiment according to H EISENBERG why an electron may not
be localized by an optical microscope better than allowed by the uncertainty relation
x
an electron beam by an
aperture leads to diffraction
px
and hence to uncertainties in e-
the determination of the p
electrons lateral momentum
x
be small and the opening angle of the microscope has to be kept large, in order
to obtain high resolution. According to A BB the smallest structure
x one may
resolve in an optical microscope is determined by its numerical aperture n sin and
the wavelength
h
x = , (1.124)
n sin
with n = 1 for the optical index of refraction (in vacuum). Small and large sin
imply, however, an intrusion into the experimental situation, since the photon will
thereby lose or change its momentum p by
h
px = p sin = sin ,
h
and transfer it to the electron. Together with the A BB relation (1.124) we obtain
px
x h and thus have derived the uncertainty relation (1.123). Any other
method trying to localize the electron leads eventually to the same result.
Assume e.g. we try to localize an electron beam by a slit this leads to diffrac-
tion as indicated in Fig. 1.33. To obtain an estimate for the uncertainty in the x-
component of the electron momentum
px = p sin we may use the angle min at
which the first minimum of the diffraction pattern occurs. The latter is well known
from the corresponding formula in wave optics:
sin min = .
x
Using this one reads from Fig. 1.33
h/p
px = p sin = p =p =
px
x = h,
x
x
which again is equivalent to the uncertainty relation (1.123).
1.7 Particles and Waves 63
W t . (1.125)
The big dilemma of physics at the beginning of the 20th century was the inability
to understand stable electron orbits around the atomic nuclei. Electrons on circu-
lar orbits are accelerated charged particles which according to classical physics
should continuously radiate electromagnetic waves and hence loose energy. Thus,
they should get slower and slower with time and being negatively charged should
eventually fall into the positively charged nucleus!
Alas! Why do the electrons remain on stable orbits? As we shall see in Sect. 1.8,
B OHR simply postulated that stable electron orbits have angular momenta equal to
integer multiples of . Among theses stationary states according to B OHR there
are also excited ones which decay spontaneously by emitting radiation according to
an exponential decay law exp(At). But still, the ground state is completely stable
even in the B OHR model!
A quantitative solution of this problem is given by quantum mechanics. Inter-
estingly, the few basics about the wave nature of the electron which we have dis-
cussed so far, already allow us to answer this puzzle in principle. We simply use
the H EISENBERG uncertainty relation to obtain an estimate for the minimum en-
ergy without adopting a specific model for the atom: Let a be the mean radius of
the atom. With a high probability electrons will be found within that radius. This
immediately leads to an estimate of the uncertainty
p in a determination of the
electrons momentum by writing (1.123) now
p2 2
T= .
2me 2me a 2
e2
V =
40 a
this gives a total energy W T + V . Assuming the equality to hold one obtains
2 e2
W= . (1.127)
2me a 2 40 a
64 1 Basics
W obviously depends on the atomic diameter a. So let us now find the lowest pos-
sible energy! We just apply the standard rule of calculus to find this minimum:
dW 1 42 0 e2 me a !
= = 0,
da 4 me a 3 0
from which we obtain for the radius of the atomic ground state:
42 0 0 h2
a0 = = . (1.128)
e2 me e2 me
Putting this back into (1.127) the total energy of the atomic ground state becomes
1 me e4 me e4
Wmin = 2 2
= 2 . (1.129)
32 2 0 80 h2
This is exactly the value of the B OHR model for the ground state of the hydrogen
atom! Big surprise with such a simple estimate? Well, the idea is obviously right.
One should, however, not overestimate the quantitative assertion: the correct numer-
ical result is a direct consequence of the specific form of (1.126) with on the right
hand side, rather than h/2 or something else (see footnote 22). And from this we
have derived the specific estimate for the minimum product of position and momen-
tum uncertainties.
Section summary
We have discussed here some elements of quantum mechanics. While the pho-
ton was recognized as the particle associated with electromagnetic waves,
we now associate matter waves with the well known particles. The DE
B ROGLIE wavelength of particles dB = h/p also holds for relativistic ve-
locities if the correct expression for p is used.
In the nonrelativistic limit, for
slow electrons, we memorize dB 1.2 nm / T / eV. Convincing experimen-
tal evidence comes from diffraction experiments with particles.
The uncertainty relation may be said to express the particle wave duality in
terms of minimal uncertainties in a measurement of two canonic conjugate
variable, e.g. for momentum and position
p
x or for energy and time
W
t . The precise value of the lower limit (, h/2 or the like) depends
on the specific problem and on the accuracy which appears tolerable.
The stability of the atomic ground state can be viewed as a direct consequence
of the uncertainty relation. (Why do negative electrons not fall into the positive
nucleus?)
In 1913, during the big time of quantum mechanical discoveries, Niels B OHR
(N OBEL prize in physics 1922) worked as a young postdoc with Ernest RUTHER -
1.8 BOHR Model of the Atom 65
FORD in England. RUTHERFORD had developed an atomic model based on his scat-
tering experiments of alpha particles by atoms according to which . . . the atoms
consist of a positively charged nucleus surrounded by a system of electrons kept
together by attractive forces from the nucleus; the total negative charge of the elec-
trons is equal to the positive charge of the nucleus. Further, the nucleus is assumed
to be the seat of the essential part of the mass of the atom, and to have linear di-
mensions exceedingly small compared with the linear dimensions of the whole atom
. . . . J.J. T HOMSON expanded this by assuming that . . . the atom consists of a
sphere of uniform positive electrification, inside which the electrons move in circu-
lar orbits (both quotes from B OHR 1913).
As discussed above, the big challenge in physics was then to explain the highly
problematic fact that these atoms do not collapse: why do electrons circle on sta-
ble atomic orbits without loosing energy? Remember, this was 10 years before
DE B ROGLIE postulated his matter waves as discussed in Sect. 1.7.1! B OHR , so
to say, just cut the Gordian knot and developed the fundamental ideas for a the-
ory of atomic structure. He knew from BALMERs spectroscopic work, that the
energy levels of atoms follow the phenomenological equation W n2 with in-
teger numbers n. These numbers became B OHRs quantum numbers. Since accord-
ing to classical mechanics the energy of a particle on a circular orbit scales in-
versely to the square of its orbital angular momentum, B OHR suggested that the
angular momenta of stable orbits be directly proportional to these quantum num-
bers. And since the P LANCK constant h, then already known from P LANCKs ra-
diation law and E INSTEINs explanation of the photoelectric effect, has the unit of
an angular momentum, he postulated that angular momenta be quantized in units
of h/2 = . From a classical view point, this quantization could neither be proven
nor even be justified, but B OHR demonstrated that with these assumptions the ex-
perimentally observed spectrum of hydrogen could be explained with unique preci-
sion.
Many versions and numbers of B OHRs postulates may be found in the literature
and in a multitude of Web pages on the subject. In his original paper B OHR (1913)
used the following two principle assumptions:
1. The dynamical equilibrium of the systems in the stationary states can be dis-
cussed by help of the ordinary mechanics, while the passing of the systems be-
tween different stationary states cannot be treated on that basis.
2. The latter process is followed by the emission of a homogeneous radiation, for
which the relation between the frequency and the amount of energy emitted is
the one given by P LANCKs theory.
electron round the nucleus in a stationary state of the system is equal to an entire
multiple of a universal value which he identified as h/2 , today written as . In
concise form, the angular momentum L = r p is quantized and its magnitude on
a stationary circular orbit is
h
L = rme ve = n = n, (1.130)
2
with the velocity ve , the electron mass me and the radius r of the orbit.
The second of B OHRs principle assumptions is about the photon energies emit-
ted in a transition from one stationary orbital n to another one n . In mathematical
form it reads:
h = Wn Wn . (1.131)
Thus, we have derived a classical expression for the radius, while the kinetic energy
may be written in terms of the potential V (r)
me ve2 Ze2 1
T= = = V (r).
2 80 r 2
(Note that the latter equality is directly given by the C OULOMB law but also corre-
sponds to the classical virial theorem.) Thus, the total energy of the system is
Ze2
W = T + V = T = < 0, (1.133)
80 r
with the minus sign characterizing bound states: one defines zero total energy W = 0
for an electron that is just no longer bound (no kinetic energy at r ).
Finally, the quantization condition (1.130) is introduced by rewriting the kinetic
energy, again using classical mechanics:
L2 (n)2
W = T = = . (1.134)
2me r 2 2me r 2
1.8 BOHR Model of the Atom 67
One may now obtain independent expressions r(n) = an and W (n) = Wn for any
given n = 1, 2, . . . < by solving the set of Eqs. (1.133) and (1.134) accordingly.
One finds B OHRs orbital radii
n2 0 h2 n2
an = = a0 (1.135)
Z e2 me Z
for the stationary states of hydrogen (Z = 1) and hydrogen like ions (Z > 1), with
n called principle quantum number and a0 B OHRs radius:
0 h2
a0 = = 5.2918 1011 m. (1.136)
e2 me
The latter is used as unit of length in atomic physics (we memorize a0 0.05 nm).
The total energy (or term energy) of the stationary states is
Z 2 me e4 Z2 Z2
Wn = = E h = R hc. (1.137)
n2 802 h2 2n2 n2
me e4 e2
Eh = 2R hc = = = 4.359 1018 J =
27.211 eV, (1.138)
402 h2 40 a0
a0 a2 2 2 h3
t0 = = 2me 0 = 0 4 = 2.4189 1017 s. (1.140)
v0 h me e
23 Notto be confused with the rest energy of the electron Wrest = me c2 , also called natural unit of
energy (see e.g. NIST 2010).
68 1 Basics
Fig. 1.34 Orbital radii rn = a0 n2 /Z and energies Wn = Eh Z 2 /(2n2 ) for H (Z = 1, red) and He+
(Z = 2, black) according to the B OHR model
The atomic unit of mass is the rest mass of the free electron:
Note, however, that this must not be confused with the unified atomic mass unit
which is defined as 1/12 of the mass of an unbound atom of the nuclide 12 C at rest
and in its ground state. It is a unit outside the SI but widely used and recommended
by the Comit International des Poids et Mesures.
Ions with only one electron, i.e. He+ , Li++ , . . . , U91+ are called hydrogen like. For
the examples H and He+ Fig. 1.34 illustrates the magnitudes of orbital radii rn =
n2 a0 /Z and energies Wn = (Z 2 /2n2 )Eh .
So far we have always assumed that the nuclear mass is at rest (mn = ). However,
correctly we should describe the electron motion in the centre of mass system.
For a two particle system this kinematic correction is done by simply replacing
the electron mass me with the reduced mass m e:
me mn
me m
e = . (1.143)
me + mn
1.8 BOHR Model of the Atom 69
Z2 Z2
Wn = E h = RH hc, (1.145)
2n2 n2
with Eh and the RYDBERG constant R (1.138) being replaced by
e
m e
m
E h = Eh and RH = R , respectively. (1.146)
me me
From the term energies (1.137) one derives with (1.131) the famous RYDBERG
(RYDBERG -R ITZ) formula for the spectra of hydrogen and hydrogen like ions H,
D, He+ , Li++ , Be+++ , . . . , U91+ :
1 1
Wn1 n2 = = h = Wn1 Wn2 = Z R 2 22
(1.147)
n1 n2
1 1 1
= = Z R 2 2
2
in wavenumbers (1.148)
n1 n2
1 1
= Z 2 R c 2 2 as transition frequency. (1.149)
n1 n2
State-of-the-art spectroscopy of the hydrogen atom contributes today decisively to
the most precise measurements of fundamental constants. We shall discuss their so-
phisticated methods and their amazing precision in Sect. 6.5.4. In order to compare
these experimental results with various theoretical predictions one has to correct
them of course according to (1.146) with the different kinematic correction factors
for each nuclear mass.
The B OHR model works surprisingly well for H and H like ions. B OHRs term
energies (1.137) in these cases are identical to those derived from nonrelativistic
quantum mechanics. Unfortunately, the model fails for all other atoms in spite of
some serious efforts made in the early days of quantum mechanics. It also fails in
the relativistic case when ve /c = (e2 /20 hc)(Z/n) = Z/n. Here the fine struc-
ture constant 1/137 according to (1.10) appears again. Relativistic effects ob-
viously become important for large Z and low n. We may neglect them as long as
Z/n 1.
70 1 Basics
So far we have also completely ignored the spin of the electron which is also of
relativistic origin. The electron spin leads to an extension of the concept of angu-
lar momentum and one of its practical consequences is fine structure splitting (see
Sect. 6), which the B OHR model of course does not consider.
We have to be aware that the concept of electron orbitals with well defined radii
will have to be reconsidered very critically: strictly speaking, from todays under-
standing of quantum mechanics, the B OHR model is simply not correct as we shall
discuss in detail in Sect. 2.6.11. Still, it serves in many cases as a helpful first step
towards formulating and understanding the correct quantum mechanical theory.
Section summary
The B OHR model of hydrogen (and H-like atoms), historically one of the cor-
nerstones in the development of modern physics, makes surprisingly correct
predictions for energy levels, Wn = Eh Z/(2n2 ) and transition frequencies
h = Wn Wn . . . though just based on heuristic postulates. Even the B OHR
radii rn = a0 n2 /Z allow a reasonable interpretation as we shall see in the next
chapter. Thus, the B OHR model remains a reference for quick cross checks
and estimates. Also, its educational value should not be underestimated.
For higher precision, some small correction for finite mass of the nucleus are
needed as summarized in Sect. 1.8.5.
Using the results of the B OHR model, we have introduced the system of
atomic units (a.u.) with energies measured in Eh (twice the binding energy
of the H atoms ground state), lengths in a0 (the radius of the first B OHR or-
bit), time in t0 (the time an electrons needs to circle once around that orbit),
and the mass in me (the rest mass of the electron).
One of the key experiments for quantum mechanics has been performed in 1922 by
S TERN (N OBEL prize 1943) and G ERLACH. It showed with then unsurpassed clarity
that classical mechanics and electrodynamics are unable to explain the observations
on the level of atomic dimensions.
Before trying to understand this experiment and its consequences, we have to recall
some basics from mechanics and electrodynamics. According to the B OHR model
the electron circles around the atomic nucleus, a motion that in a classical picture
may be viewed as a spinning top.
As indicated in Fig. 1.35 this orbit is connected with an electric current
e ev
I= = .
t 2r
1.9 S TERN -G ERLACH Experiment 71
+I
of
This current, enclosing an area A (A = r 2 ), generates a magnetic moment M
the magnitude
ev evr eL
M = IA = r 2 = = .
2r 2 2me
It is proportional to the magnitude of L and oriented into the opposite direction. The
magnetic moment of an orbital angular momentum is thus
e L
M= L = B . (1.150)
2me
Note: the gyromagnetic ratio derived here, M/L = e/2me = ge B is indepen-
dent of the specific geometry of the motion. It does not depend on whether a point
charge or an extended charge distribution orbits around the nucleus. However, this
above classical derivation is valid only for orbital angular momenta. It will have
to be modified in view of the results of the S TERN -G ERLACH experiment.
In the B OHR model angular momenta (1.130) are quantized and appear only
in units of . Thus, L/ is an integer number in B OHRs theory, and the unit of
magnetic moment is the so called B OHR magneton
e 1
B = = 927.400915(23) 1026 J T . (1.151)
2me
T = M B. (1.152)
Its potential energy T d is thus given by
VB = M B = MB cos , (1.153)
where is the angle between M and B and zero energy has been assumed for
M B. Minimum potential energy corresponds to M parallel to B, and L an-
tiparallel to it.
72 1 Basics
F = grad V (r).
Otto S TERN had invented the so called molecular beam method24 and already ap-
plied it to determine the M AXWELL -B OLTZMANN velocity distribution in gases
(see Sect. 1.3.4). In his famous paper Ein Weg zur experimentellen Prfung der
24 In this and later context, an atomic or molecular beam source may be thought to just consists of
a reservoir with the species to be investigated (possibly heated, or cooled, possibly mixed with an
inert carrier gas). Through a small orifice in it, the atoms or molecules diffuse or stream into the
vacuum, and their divergence angle is collimated by one or more apertures along the beam axis.
1.9 S TERN -G ERLACH Experiment 73
z (a) z (b)
atom
oven magnet B on S
S B off
N x y
magnet B on N
slit photo plate profile
Fig. 1.37 Schematic of the S TERN -G ERLACH experiment. (a) The atomic beam from the oven
is collimated through a slit, may then be deflected in an inhomogeneous magnetic field, and will
finally hit a photographic plate. (b) Side view of magnet poles and atom beam
(a) z
y
z
(b) y
Fig. 1.38 Two wire field: it corresponds approximately to the field created by two wires through
which a current flows. The corresponding magnetic field lines are illustrated in (a). In an experi-
ment the field is realized by a pair of suitably shaped permanent magnets as indicated in (b). The
red line parallel to the y-direction indicates the position of the atomic beam entering the space be-
tween the magnets from the back. It will be deflected by the inhomogeneous field in z-direction
z B
Fz = M . (1.157)
z
z
(a)
x
z
(b) x
(c) x
Fig. 1.39 Classical expectation about the S TERN -G ERLACH experiment: (a) View onto the
atomic beam coming out of the yz plane without magnetic field. (b) Alignment of the magnetic
moments prior to entering the magnet, distributed at random. (c) Atomic beam profile after deflec-
tion by the magnetic field. The distribution would arise from B/z being strongest for x = 0 and
decreasing for |x| > 0
What Do We Expect?
Let us assume the profile of the atomic beam without magnetic field arriving in the
detector plane is shaped as indicated in Fig. 1.39(a). Classically, the magnetic mo-
ments M are expected to be randomly distributed in all directions as indicated in
Fig. 1.39(b), their projections M z onto the +z-direction ranging from B L/ to
+B L/. The atomic beam enters the inhomogeneous magnetic field and its atoms
are deflected corresponding to their distribution of Mz . Behind the magnet the re-
sulting spatial distribution is registered by a suitable detector. With the magnetic
field switched on, the classical expectation is to observe a broadening of the beam
in z-direction proportional to the field and the average magnetic moments reflect-
ing the randomly distributed Mz components. The deflection on the z-axis (x = 0)
will be largest since there the field gradient B/z is highest.
25 A quaint note aside: the silver traces only became visible after Otto S TERN blew the smoke of
his (sulfur containing) cigar onto the photographic plate an early, unintended contribution to the
photochemistry and catalysis of the photographic development process.
1.9 S TERN -G ERLACH Experiment 75
atom signal
magnetic field bears out the dramatic splitting into two components. This observa-
tion, which is also made for many other atoms, is a direct consequence of space
quantization as we shall discuss in a moment.
How can the results of the S TERN -G ERLACH experiment be interpreted? We re-
member, the B OHR model postulates that angular momenta are quantized according
to (1.130): Thus, we expect L = with to be a positive integer number or zero.
Let us assume now, that a similar relation holds for the components of angular mo-
menta, say for Lz in z-direction. Then Lz would assume 2 + 1 integer multiples
of :
Lz = m with m = , + 1, . . . , . (1.158)
76 1 Basics
Fig. 1.42 Schematic of a L ANGMUIR -TAYLOR detector: (a) Potential for an electron at an isolated
tungsten surface, (b) potential and energy of the valence electron in a K atom, (c) potential felt by
a K atom when hitting a tungsten surface, (d) simple detector setup
This quantization is called space quantization (in German language somewhat more
precisely: Richtungsquantisierung). The number m is often called magnetic quan-
tum number since it was originally observed in a magnetic field. We prefer, how-
ever, the more general term projection quantum number.
In contrast to a classical statistical distribution of angular momenta, as illustrated
in Fig. 1.39(b), there are only 2 + 1 allowed projections of L = onto the z-
axis. The number 2 + 1 is called multiplicity of the state (in Sect. 1.3.4 we have
introduced it as degeneracy a term used in the context of energies). According to
(1.150) the magnetic moment is M = B L/, thus space quantization of L also
implies quantization of M. The component Mz may thus assume values from |M|
to |M| which correspond to the components |L| to |L| of the orbital angular
momentum.
A closer quantum mechanical analysis (see Chap. 2) shows that the conjectures
made above are nearly correct except for the fact that the magnitude of the angular
momentum is given by
L = ( + 1), (1.159)
J
magnetic moment for J : MJ = gJ B (1.160)
and its projection onto a given axis, say z, is given by
MJ z = gj B mj with mj = j, j + 1, . . . , j. (1.161)
The so called L AND g-factor is gL = 1 for pure orbital angular momenta and =2
for electron spin states. We shall derive general expressions gj in Chap. 8. A mul-
titude experiments confirms that the magnetic moment of atoms and molecules is
indeed proportional to the total angular momentum J and oriented opposite to it.
Equation (1.156) has also to be modified:
e
L ARMOR frequency for J : j = gj B = gj L . (1.162)
2me
Section summary
The S TERN -G ERLACH experiment has been another corner stones in the de-
velopment of quantum mechanics. It proofs what we call space quantization:
angular momenta J are oriented in space so that their projection on a given
axis is Jz = mj with mj = j, j + 1, . . . , j where j is the angular mo-
mentum quantum number.
Orbital angular momenta have integer quantum numbers, intrinsic quantum
numbers of particles may be half integer (fermions) or integer (bosons).
78 1 Basics
We emphasize that this value of the electron g factor cannot be explained by any
kind of classical charge distribution on an orbit, for which always (1.150) holds,
i.e. gL = 1.
In contrast, the relativistically correct D IRAC theory leads exactly to ge = 2 for
the electron spin. Thus, according to (1.161) the component of the magnetic moment
of the electron in respect of a given axis, say z, becomes indeed Mz = ge B /2 =
B as experimentally observed.
It is interesting to note that the L ARMOR frequency (1.162) for an electron with
spin s and ge = 2 becomes
e e
s = ge L = ge B B, (1.166)
2me me
and is, according to (1.111), (nearly) identical to the cyclotron frequency c =
(e/me )B of an electron in a magnetic field (in the nonrelativistic limit).
We point out, however, that high precision measurements show a small, but signif-
icant difference between L and c , i.e. document a deviation from (1.165), in fact
one finds ge 2.0023 . . . . A theoretical understanding of this so called anomalous
magnetic moment of the electron is provided by quantum electrodynamics (QED)
for which T OMONAGA, S CHWINGER and F EYNMAN received the N OBEL prize in
1965 as we shall explicate in Sect. 6.6.
The so called E INSTEIN - DE -H AAS effect gives one further impressive piece of ev-
idence for the non-classical nature of the magnetic moment of the electron based
80 1 Basics
on a macroscopic measurement. In this experiment one uses the fact (not to be dis-
cussed here in detail) that ferromagnetism is caused by many electron spins which
are oriented in parallel.
One determines the torque exerted onto a (soft) ferromagnetic iron core (cylin-
der) due to a change of magnetization: the magnetic field of the solenoid tends to
orient the magnetic moments of the electron spins in parallel to this magnetic field
and hence there will also be a preferential orientation of the electron spins. These
spins all together constitute a total inherent angular momentum of the cylinder. In
order to change it, a torque has to act on this cylinder which in the setup shown
in Fig. 1.44 arises from the external magnetic field. The torque is measured by the
twisting angle of a thin spring wire (silica).
In a quantitative experiment one starts with a completely demagnetized probe
and allows a well defined current pulse to flow through the solenoid. This leads to
a changing magnetic field Hsolenoid (t) as sketched in Fig. 1.45. During this pulse a
torque acts on the system which was originally at rest.
After the magnetic field has decayed, the remanent magnetization Mrem is as-
sumed to be N times the z-component of the magnetic moment Mz of the individ-
ual electrons. With (1.161) one obtains
e e
Mz = ge B ms = ge Sz Mrem = Brem = N ge Sz , (1.167)
2me 2me
where N is the number of electrons that after the pulse remain oriented (magnetized)
in the direction of the applied magnetic field (remanence). In turn, Brem may easily
Hsolenoid
Bmagnet
Brem
t
Fig. 1.45 Temporal evolution of the magnetic field generated by the solenoid in Hsolenoid after
starting to discharge the capacitor C along with the magnetic induction Bmagnet resulting from it
in the ferromagnetic cylinder leading to a remanent part Brem
Acronyms and Terminology 81
with Irod = mR 2 /2, the moment of inertia of the ferromagnetic cylinder, m being its
mass and R its radius. The angular frequency of the cylinder rod after application
of the magnetic field may be determined from the maximum twist angle max of the
silica wire onto which the cylinder is mounted: the initial kinetic energy has then
been transformed into potential energy:
Irod 2 kr 2
= max . (1.169)
2 rod 2
From the ratio of Brem according to (1.167) and the angular momentum Irod rod
according to (1.168) we obtain
Brem N ge S z e e
= = ge .
Irod rod N Sz 2me 2me
Section summary
The spin related properties of the electron have again be summarized in
(1.163)(1.165).
The E INSTEIN - DE H AAS experiment demonstrates the electron spin on a
macroscopic level using in a ferromagnetic rod. When its magnetization is
reversed in external magnetic field, the magnetic moments of N the electrons
have to reversed from +/2 to /2. Thus, the total angular momentum of
all electrons in the rod changes by N . Since N is a large number, this in-
duces a macroscopic rotation of the rod, which can be measured, confirming
the microscopic measurements of ge .
CCD: Charge coupled device, semiconductor device typically used for digital
imaging (e.g. in electronic cameras).
chemical-potential: In statistical thermodynamics defined as the amount of energy
or work that is necessary to change the number of particles of a system (by 1)
without disturbing the equilibrium of the system, see in Sect. 1.3.4.
CIE: Commission international de lclairage, International Commission on Illu-
mination, provides e.g. colorometric tables (http://files.cie.co.at/204.xls).
CMB(R): Cosmic microwave background, radiation at 2.725 K from the origin of
the universe.
DC: Direct current, unidirectional electric voltage and current.
ESS: European Spallation Source, large scale facility for generating neutrons for
structural research, see http://europeanspallationsource.se/.
EUV: Extreme ultraviolet, part of the UV spectral range. Wavelengths between
10 nm and 121 nm according to ISO 21348 (2007).
FIR: Far infrared, spectral range of electromagnetic radiation. Wavelengths be-
tween 3 m and 1 mm according to ISO 21348 (2007).
FT: F OURIER transform, see Appendix I.
FWHM: Full width at half maximum.
HERA: Hadron-Elektron-Ring-Anlage, for collision experiments between elec-
trons of 30 GeV and protons of 820 GeV operated at DESY very sucessfully
until 2007.
HF: High frequency, part of the RF spectrum. Wavelengths from 10 m to 100 m
or frequencies from 3 MHz to 30 MHz according to ISO 21348 (2007).
ICR: Ion cyclotron resonance, spectroscopy (specifically mass spectroscopy)
based on irradiation with a radio frequency in resonance with the cyclotron fre-
quency of an ion in magnetic field (see Sect. 1.6.3).
IR: Infrared, spectral range of electromagnetic radiation. Wavelengths between
760 nm and 1 mm according to ISO 21348 (2007).
LED: Light emitting diode.
LEED: Low energy electron diffraction, see Sect. 1.7.2.
LF: Low frequency, part of the RF spectrum from 30 kHz up to 300 kHz.
LHC: Large hadron collider (not to be confused with left hand circularly polarized
light), high energy physics facility at CERN (Geneva) providing particles with
collision energies up to 14 TeV for protons and up to 1 PeV for heavy ions.
MF: Medium frequency, part of the RF spectrum from 300 kHz up to 3 MHz.
MIR: Middle infrared, spectral range of electromagnetic radiation. Wavelengths
between 1.4 m and 3 m according to ISO 21348 (2007).
MW: Microwave, range of the electromagnetic spectrum. In spectroscopy MW
usually refers to wavelengths from 1 mm to 1 m corresponding to frequencies
between 0.3 GHz and 300 GHz; ISO 21348 (2007) defines it as the wavelength
range between 1 mm and 15 mm.
NIR: Near infrared, spectral range of electromagnetic radiation. Wavelengths be-
tween 760 nm and 1.4 m according to ISO 21348 (2007).
NIST: National institute of standards and technology, located at Gaithersburg
(MD) and Boulder (CO), USA. http://www.nist.gov/index.html.
Acronyms and Terminology 83
References
A RNDT , M., O. NAIRZ, J. VOS -A NDREAE, C. K ELLER, G. VAN DER Z OUW and A. Z EILINGER:
1999. Wave-particle duality of C60 molecules. Nature, 401, 680682.
ASTM: 2008. G173-03 Reference Spectra Derived from SMARTS v. 2.9.2, American Soci-
ety for Testing and Materials (ASTM). http://rredc.nrel.gov/solar/spectra/am1.5/ASTMG173/
ASTMG173.html, accessed: 7 Jan 2014.
B OHR , N.: 1913. On the constitution of atoms and molecules. Philos. Mag., 6, 26, 125.
B OHR , N. H. D.: 1922. The N OBEL prize in physics: for his services in the investigation of the
structure of atoms and of the radiation emanating from them, Stockholm. http://nobelprize.org/
nobel_prizes/physics/laureates/1922/.
B RETSCHNEIDER , E.: 2007. Efficacy limits for solid-state white light sources, Laurin Publishing
Co., Inc., Pittsfield MA, USA. http://www.photonics.com/Article.aspx?AID=28677, accessed:
7 Jan 2014.
DE B ROGLIE , L.: 1929. The N OBEL prize in physics: for his discovery of the wave nature of
electrons, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1929/.
CERN: 2013. The search for the H IGGS boson, Geneva. http://home.web.cern.ch/about/
physics/search-higgs-boson, accessed: 8 Jan 2013.
C HANTLER , C. T., K. O LSEN, R. A. D RAGOSET, J. C HANG, A. R. K ISHORE, S. A. KO -
TOCHIGOVA and D. S. Z UCKER : 2005. X-ray form factor, attenuation, and scattering tables
(version 2.1), NIST. http://physics.nist.gov/ffast, accessed: 7 Jan 2014.
C OMPTON , A. H.: 1927. The N OBEL prize in physics: for his discovery of the effect named after
him, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1927/.
C ORNELL , E. A., W. K ETTERLE and C. E. W IEMAN: 2001. The N OBEL prize in physics: for
the achievement of Bose-Einstein condensation in dilute gases of alkali atoms, and for early
fundamental studies of the properties of the condensates, Stockholm. http://nobelprize.org/
nobel_prizes/physics/laureates/2001/.
D EHMELT , H. G. and W. PAUL: 1989. The N OBEL prize in physics: for the development of the
ion trap technique, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1989/.
D ICK LYON: 2008. Luminosity function, Wikimedia Commons. http://commons.wikimedia.
org/wiki/File:Luminosity.png, accessed: 7 Jan 2014.
E INSTEIN , A.: 1905. ber einen die Erzeugung und Verwandlung des Lichtes betreffenden
heuristischen Gesichtspunkt. Ann. Phys., 17, 132.
E INSTEIN , A.: 1921. The N OBEL prize in physics: for his services to theoretical physics, and
especially for his discovery of the law of the photoelectric effect, Stockholm. http://nobelprize.
org/nobel_prizes/physics/laureates/1921/.
E NGLERT , F. and P. W. H IGGS: 2013. The N OBEL prize in physics: for the theoretical discovery
of a mechanism that contributes to our understanding of the origin of mass of subatomic par-
ticles, and which recently was confirmed through the discovery of the predicted fundamental
particle, by the ATLAS and CMS experiments at CERNs large hadron collider, Stockholm.
http://nobelprize.org/nobel_prizes/physics/laureates/2013/.
H ERZOG , R.: 1935. Berechnung des Streufeldes eines Kondensators, dessen Feld durch eine
Blende begrenzt ist. Arch. Elektrotech., 29, 790802.
H ONSBERG , C. and S. B OWDEN: 2012. PVCDROM Air Mass, UNSW and Solar Power
Labs at ASU, Australia. http://www.pveducation.org/pvcdrom/properties-of-sunlight/air-mass,
accessed: 7 Jan 2014.
ISO 21348: 2007. Space environment (natural and artificial) process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
KOPP , G. and J. L. L EAN: 2011. A new, lower value of total solar irradiance: evidence and
climate significance. Geophys. Res. Lett., 38, L01706.
M AMYRIN , B. A.: 1994. Laser-assisted reflectron time-of-flight mass-spectrometry. Int. J. Mass
Spectrom. Ion Process., 131, 119.
References 85
M ATHER , J. C. and G. F. S MOOT: 2006. The N OBEL prize in physics: for their discovery of the
blackbody form and anisotropy of the cosmic microwave background radiation, Stockholm.
http://nobelprize.org/nobel_prizes/physics/laureates/2006/.
M ILLIKAN , R. A.: 1923. The N OBEL prize in physics: for his work on the elementary charge
of electricity and on the photoelectric effect, Stockholm. http://nobelprize.org/nobel_prizes/
physics/laureates/1923/.
M UELLER , U., F. H. N IESEN, Y. ROSKE, F. G OETZ, J. B EHLKE, K. B UESSOW and U. H EINE -
MANN : 2007. Crystal structure of human prolidase: the molecular basis of PD disease,
Hinxton, UK: PDB entry 2okn. The European Molecular Biology Laboratory (EMBL-EBI).
http://www.ebi.ac.uk/pdbe-srv/view/entry/2okn/summary.html, accessed: 7 Jan 2014.
NIST: 2010. Reference on constants, units, and uncertainties, NIST. http://physics.nist.gov/cuu/
Constants/, accessed: 7 Jan 2014.
NIST: 2011. Conversion factors for energy equivalents, NIST. http://physics.nist.gov/cuu/
Constants/energy.html, accessed: 7 Jan 2014.
O HNO , Y.: 2010. Radiometry and photometry for vision optics. In: M. BASS, ed., Handbook of
Optics, vol. II, 37.1. New York: McGraw-Hill.
PAULI , W.: 1945. The N OBEL prize in physics: for the discovery of the exclusion princi-
ple, also called the Pauli principle, Stockholm. http://nobelprize.org/nobel_prizes/physics/
laureates/1945/.
P LANCK , M.: 1900. Zur Theorie des Gesetzes der Energieverteilung im Normalenspektrum.
Verh. Dtsch. Phys. Ges., 2, 235245.
P LANCK , M. K. E. L.: 1918. The N OBEL prize in physics: in recognition of the services
he rendered to the advancement of physics by his discovery of energy quanta, Stockholm.
http://www.nobelprize.org/nobel_prizes/physics/laureates/1918/.
S CHLLKOPF , W. and J. P. T OENNIES: 1996. The nondestructive detection of the helium dimer
and trimer. J. Chem. Phys., 104, 11551158.
SORCE: 2012. SORCE Solar Spectral Irradiance, Boulder, Co.: Laboratory for Atmo-
spheric and Space Physics, University of Colorado and NASA. http://lasp.colorado.edu/
lisird/sorce/sorce_ssi/, accessed: 7 Jan 2014.
SSE: 2012. Solar System Exploration Our Solar System, NASA. http://solarsystem.nasa.gov/
planets/profile.cfm?Display=Facts&Object=Sun, accessed: 7 Jan 2014.
S TERN , O.: 1921. Ein Weg zur experimentellen Prfung der Richtungsquantelung im Magnet-
feld. Z. Phys., VII, 249253. Nachdruck: Z. Phys. D, Atoms, Molecules and Clusters 10, 111
116 (1988).
S TERN , O.: 1943. The N OBEL prize in physics: for his contribution to the development of the
molecular ray method and his discovery of the magnetic moment of the proton, Stockholm.
http://nobelprize.org/nobel_prizes/physics/laureates/1943/.
T HOMSON , J. J.: 1906. The N OBEL prize in physics: in recognition of the great merits of his the-
oretical and experimental investigations on the conduction of electricity by gases, Stockholm.
http://www.nobelprize.org/nobel_prizes/physics/laureates/1906/.
T OMONAGA , S.-I., J. S CHWINGER and R. P. F EYNMAN: 1965. The N OBEL prize in physics:
for fundamental work in quantum electrodynamics, with deep-ploughing consequences for
the physics of elementary particles, Stockholm. http://nobelprize.org/nobel_prizes/physics/
laureates/1965/.
W IKIPEDIA CONTRIBUTORS: 2013. Mathieu function, Wikipedia, The Free Encyclopedia.
http://en.wikipedia.org/wiki/Mathieu_function, accessed: 7 Jan 2014.
W ILEY , W. C. and I. H. M C L AREN: 1955. Time-of-flight mass spectrometer with improved
resolution. Rev. Sci. Instrum., 26, 11501157.
Elements of Quantum Mechanics
and the H Atom 2
Overview
If the reader is already familiar with quantum mechanics he may just want
to browse this chapter and return later if necessary. However, readers who
have experienced quantum mechanics up to now only as compulsory math-
ematical exercise may perhaps read this chapter with advantage and find it
helpful to approach the indispensable instruments without big formal hurdles.
In Sects. 2.12.3 we summarize a minimum of formalism. Section 2.4 treats
as a first example the well known particle in a box and the free electron gas
which in atomic and solid state physics is an important elementary model.
Section 2.5 gives an overview of how to treat angular momenta, needed in
all following text, specified in Sect. 2.5.4 for the electron spin s = 1/2. Sec-
tion 2.6 offers a crash course in nonrelativistic quantum mechanics of the
H atom essential knowledge for all the following chapters. We refrain here
from formal derivations in favour of a plausible, possibly somewhat hand wav-
ing introduction. Finally, Sect. 2.7 presents a first, elementary approach to in-
teractions of atomic electrons with external fields, which will be extended and
deepened in Chap. 8. Clearly, such a brief introduction into quantum mechan-
ics cannot substitute a profound study of the formal theory. It should, however,
facilitate the access to it and make the reader fit for the following chapters.
The classical picture of a well defined trajectory with known coordinate x(t) and
momentum p(t) looses its validity in quantum mechanics, as sketched in the phase
diagram Fig. 2.1: position and momentum cannot be measured simultaneously. They
may only be determined with an accuracy in accordance with the uncertainty rela-
tion
pi
xi h/2 . Quantum mechanics only makes statements about probability
amplitudes (r, t). These may assume the form of so called Wave-packets which
localize particles in space and time as well as the uncertainty relation allows it. One
finds a particle at position r and time t with the probability
2
w(r, t) = (r, t) . (2.1)
One may easily visualize the concept of a probability amplitude for the example
of photons in a YOUNGs double slit experiment. The probability to find a photon
at position r and time t is proportional to the intensity I (r, t) of the light, and the
latter is proportional to the square of the field amplitude. Let us consider only one
component of polarization, say Ex . The dependence of the intensity on position may
then be written as
2 2
I (r) Ex (r) = (r) = w(r).
The last two equalities can help us to get used to quantum mechanical terminology:
we call the quantity (r) now the (position dependent) probability amplitude or
wave function. In the case of light, it is simply represented by the electric field
component Ex . One determines it according to the laws of optics as solution of the
corresponding wave equation
We may see this expression as a reinterpretation of the classical YOUNG double slit
experiment in terms of quantum mechanical probability amplitudes.
In an actual experiment one may now reduce the light intensity such that only
one single photon at a time is near the double slits and participates to the observed
interference pattern. One may easily verify such a setup with a particle counter
detecting single photons one may even hear the individual photons click.1 If a
sufficient number of such single photon events are recorded and added up, one finds
quite counter intuitively that the classical diffraction pattern known from optics
is recovered! Thus, the probability distribution of each individual photon behind the
double slit is determined by the wave amplitudes = 1 + 2 and it is impossible
to say through which slit the particle has passed (see Fig. 1.31). One also says that
one photon always interferes only with itself. We shall discuss this statement in some
detail in Chap. 2, Vol. 2 and quantify it statistically.
From the perspective just discussed let us have a look at the matter waves intro-
duced in Sect. 1.7.1. We have already familiarized ourselves with the relation be-
tween momentum and wavelength as postulated by DE B ROGLIE. Matter waves are
1 For the experts: One may invest a lot of thought and substantial experimental efforts to make
absolutely sure that only isolated photons reach the double slit setup (for a recent review on single
photon sources see E ISAMAN et al. 2011). In the present context we shall be content if the average
time tav between each interference event (photon counted) is long compared to the coherence time
of the photon source c = 1/
tav , with
being the bandwidth of the source (see Sect. 2.1.4,
Vol. 2).
90 2 Elements of Quantum Mechanics and the H Atom
And as in optics, diffraction and interference occurs, e.g. at a double slit for which
(2.4) describes the experimentally observable signal of particles.
However, in contrast to electromagnetic radiation, where was identified with
the electric (or magnetic) field, for matter waves (r) cannot be attributed to any
directly measurable physical quantity: the observable physics is described by the
probability w(r). So we call (r) simply the probability amplitude for finding a
particle at position r.
Otherwise, analogue considerations are valid for photons and particles of matter.
If e.g. we try to localize them on their pathways, we loose the interference patterns!
One important general rule holds: interference phenomena are observed if differ-
ent but indistinguishable pathways exist on which the particles may proceed toward
the detector. In contrast, no interference is observed if the two pathways may in
principle be distinguished even when no distinction is made in the actual experi-
ment.
Section summary
In classical mechanics well defined trajectories in position r(t) and mo-
mentum p(t) space describe the motion of particles. In contrast, quantum
mechanics describes probability amplitudes (r) and probabilities dw(r) =
|(r)|2 d3 r for finding a particle at a certain position r in a volume element
d3 r.
Diffraction and interference of matter waves can be described by concepts in
analogy to those used in wave optics. The big difference is that for photons
the probability amplitude is also a directly measurable quantity (electric or
magnetic field) while for matter waves (r) cannot be measured directly
only probabilities |(r)|2 refer to the real world.
p2 2m(W V (r))
(r) + (r) =
(r) + (r) = 0.
2 2
Rewritten, this is already the stationary S CHRDINGER equation
2
(r) + V (r)(r) = W (r), (2.6)
2m
or somewhat more compact
(r) = W (r)
H (2.7)
with the eigenfunction (r) and the eigenenergy W . The H AMILTON operator H
represents the total energy (briefly just Hamiltonian):
=
+ V (r).
2
H (2.8)
2m
In the often encountered one-dimensional case the S CHRDINGER equation (2.6) is
further simplified:
2 d2 (x)
+ V (x)(x) = W (x). (2.9)
2m dx 2
= 2 + V (r) =
p2
2
H + V (r). (2.10)
2m 2m
Here we have introduced the momentum operator (a vector operator)
92 2 Elements of Quantum Mechanics and the H Atom
/x
p = i = i /y
(2.11)
/z
so that
2 2 2
p2 =
p
p = 2 2 = 2
= 2 + + . (2.12)
x 2 y 2 z2
Thus, (2.10) is the operator form of the classical energy conservation law
p2
W = Wkin + V = + V (r).
2m
(r, t) = i (r, t)
H (2.13)
t
2 (r, t)
or more explicitly:
(r, t) + V (r) (r, t) = i .
2m t
It cannot be derived, and we just communicate it here as it was found by Erwin
S CHRDINGER in the beginning of 1926 by the way: during a winter ski holiday
in the Swiss Alps. We point out some key aspects:
The S CHRDINGER equation is a linear PDE of 2nd order in space and 1st order
in time! As a consequence, time dependence of stationary states is truly complex.
The linear superposition principle may be applied to the solutions.
The statistical interpretation of quantum mechanics understands solutions
(r, t) of this PDE as probability amplitudes for finding a particle at a posi-
tion r in space at time t according to (2.1).
Although the S CHRDINGER equation cannot be derived, it is able to describe
a wealth of atomic phenomena and to predict observables quantitatively (under
nonrelativistic conditions) with excellent accuracy: we emphasize that is this very
fact which defines the validity of a physical theory.
Equivalently, formal (algebraic) quantum mechanics can be deduced from a con-
sistent set of axioms, which are, however, also heuristically assumed.
2.2 SCHRDINGER Equation 93
Consistent alternatives of wave equations for matter waves are the D IRAC equa-
tion for fermions, a multi component spinor equation and the K LEIN -G ORDON
equation, a single component PDE of 2nd order in time, which turns out to be
valid for bosons both are relativistically correct.
(r, t) = H
If the Hamiltonian itself does not depend on time, H (r), the time depen-
dence of the wave function may be factored by a product ansatz:
(r, t) = i (r, t)
H (r)(t) = i (r)(t)
H
t t
H (r) i (t)
= W.
(r) (t) t
The last identity (with the constant W which has to be determined) must hold so
that the former equality can be valid for all values of r and t. One thus has to
solve id(t)/dt = W (t) and H (r) = W (r). In this case, the time dependent
equation leads to the trivial solution
W
(t) exp i t . (2.15)
The position dependent part is nothing but the stationary S CHRDINGER equation
(2.6) and the parameter W introduced is the total energy of the system. The overall
wave function is given by
W
(r, t) = (r) exp i t . (2.16)
Note: The time dependence is truly complex and the imaginary unit i in the prefactor
= H
is needed for the solution! In the present case, however, with H (t) the time
dependence is trivial in the sense that only
2 2
w(r, t) = (r, t) = (r) (2.17)
can be measured, which contains only information about the stationary state.
In order to find these stationary atomic states for a given potential V (r)
which are the equivalent to B OHRs stationary orbits one has to solve the station-
ary S CHRDINGER equation (2.6) under suitably chosen boundary conditions. For
bound states this typically leads to a whole series of discrete total energies W < 0
with a set of quantum numbers characterizing the states. The corresponding series
of wave functions for individual electrons called orbitals describe the probability
of finding the particles at a given position in space.
94 2 Elements of Quantum Mechanics and the H Atom
k
(r, t) = C exp i(t kr) = C exp i t kr (2.19)
2m
W pr
= C exp i t .
With this we obtain the important dispersion relation for free matter waves:
W p2 k 2 2 2
= = = = (k) or W = k . (2.20)
2m 2m 2m
Note: The probability to find this particle, w(r, t) = | (r, t)|2 = |C|2 , is indepen-
dent of position and time as expected for an infinitely extended plane wave. In
other words, a particle with well defined momentum may not be localized at all as
expected according to the uncertainty relation (1.123).
Section summary
The stationary S CHRDINGER equation (2.6) may be gleaned from the classi-
cal wave equation combined with a free interpretation of the DE B ROGLIE
wavelength in a conservative potential. It is validated by the excellent
agreement of its predictions with experimentally observed data in the sub-
microscopic world at nonrelativistic energies.
In its most compact form it reads H = W , where the H AMILTON operator
(2.10), also called Hamiltonian, is constructed in full analogy to its classical
counterpart, just replacing the momentum by the quantum mechanical mo-
mentum operator p = i.
The time dependence of the wave function is described by the time dependent
S CHRDINGER equation (2.13). In the case that the Hamiltonian itself is not
time dependent, the (stationary) solutions are given by the product of the solu-
tions (r) of the stationary equation (2.6) and a simple exponential function
exp(i(W/)t).
The most simple solution of the time dependent S CHRDINGER equation is a
plane wave exp[i((W/)t kr)], with the energy W = k 2 /(2m).
2.3 Basics and Denitions of Quantum Mechanics 95
Here we summarize briefly the fundamental axioms of quantum mechanics and re-
call some terminology and rules which we shall use later on:
Note: the sum includes all bound (discrete) and free (continuum) states.
One calls a basis orthonormal if
fk |fi = ki , (2.22)
where |
is the scalar product of two state vectors |
and |
. With this, the
projection of the state |
onto the basis vector |fk
is
fk |
= ci fi |fk
= ck . (2.23)
i=1
Wave functions, the most commonly used representation of states, are formally
obtained by expanding |
in a continuous position basis {|r
} where r extends
over all points in 3D position space. We write (2.21) as ket
|
= d r r r
3
or as bra | = d 3 r r r . (2.24)
and obtain from (2.24) the definition of a wave function in position space,
r|
= |r
= d3 r r r r = (r), (2.26)
96 2 Elements of Quantum Mechanics and the H Atom
and (r) = |r
. In practice, wave functions are determined by solving the sta-
tionary S CHRDINGER equation. With (2.24) and (2.25) we derive (after one 3D
integration) the scalar product in terms of wave functions:2
|
= (r)(r)d3 r = |
. (2.27)
We finally note that, equivalent to the state expansion in position space, one may
define wave functions (p) in momentum space by
|
= d3 p p p and
p|
= d3 p p pp = (p).
Operators
Linear operators play a key role in quantum mechanics: in general an operator
changes an object onto which it acts (e.g. a state vector, a wave function, another
and let it act on a ket
operator) into something different. Let us call the operator A
vector |
:
A|
= |
.
Linearity implies that for a superposition of states |
= c1 |1
+ c2 |2
+
c1 |1
+ c2 |2
+ = c1 A|
A 1
+ c2 A|
2
+ .
B
The product of two operators A is defined by
B)|
(A B|
=A . (2.29)
B
Operator multiplication is distributive, i.e. A C
= (A
B)
C = A(
B C)
but not neces-
sarily commutative. In general AB = B A and one defines a commutator
B]
[A, =A
B B
A (2.30)
2 We shall use the bra| and |ket symbols rather loosely. In particular, we shall often write wave
which only in special cases may vanish (see Sect. 2.3.3). With |
and |
being
two state vectors or wave functions one defines as matrix element of A between
states |
and |
(which may or may not be basis states of an operator A:
= (A)d
A = |A
3 r. (2.31)
Without going into details we define the so called adjoint (or Hermitian conjugate)
by
operator A
= |A
A or A = |A
(2.32)
3
or d r = (A)d
A 3 r. (2.33)
Of particular importance are the so called Hermitian operators, let us call them O.
They are self-adjoint:
O
O. (2.34)
By this definition and with (2.31)(2.33) the matrix elements
O|
= (O)
= |O
d3 r = (O)d 3r (2.35)
= |O
= |O
(2.36)
Observables
Observables are all physical quantities which can in principle be measured (ob-
served). Every quantity which is observable in classical physics is represented quan-
tum mechanically by a linear Hermitian operator, let us call it O. A quantum sys-
tems can be characterized by a set of eigenstates (eigenvectors) |fk
of an observable
O which it may in principle assume. From the eigenvalue equation
k
= k |fk
O|f (2.37)
one determines the eigenvalues k of the observable O for the eigenvectors |fk
. In
any individual physical measurement of the observable O only one of its eigenvalues
k can be observed.
From (2.37) and the orthonormality of the |fk
basis one sees that the matrix
in the basis of its eigenstates between |fi
and |fk
are
elements of O
k
= k ik ,
Oik = fi |Of (2.38)
3 To verify these relations one simply expands |
and |
in a basis of eigenvectors (eigenfunc-
tions) of O.
98 2 Elements of Quantum Mechanics and the H Atom
The latter relation follows directly with (2.22). If we measure now the observable
many times (as one does in a real experiment), the result of each individual mea-
O
The probability to detect this particular
surement is one of the eigenvalues i of O.
eigenvalue i is determined by the probability amplitude ci :
O|
=O ci |fi
= ci i |fi
. (2.40)
The average value measured for this observable, i.e. the result of many individ-
ual measurements applied to the same state |
, is called expectation value of the
observable:
O
|ci |2 i = |O
. (2.41)
()
Somewhat more general, if the state of the system is |
= i ci |gi
, i.e. it is
its expectation
given in an arbitrary basis {|gi
} and |gi
are not eigenstates of O,
value is
()
()
= |O|
O
= ci
gi |O|
()
ck |gk
= ci k
c() . (2.42)
gi |O|g k
i k ik
Unit Operator
We note in passing a nice mathematical trick by rewriting (2.39) as
|
= |fi
ci = |fi
fi |
= |fi
fi | |
.
i i i
2.3 Basics and Denitions of Quantum Mechanics 99
From this we deduce the often very useful fact that the quantum mechanical unit
operator4 (which does not change the operand) may be written as
1= |fi
fi |, (2.43)
i
as long as the states |fi
represent a complete orthogonal basis and the summation
goes over all states of this basis.
Quantization
Experimentally, when an observable O is determined in an individual measurement
one always finds one of its eigenvalues. With the measurement one prepares also
One
the corresponding eigenfunction (eigenstate, eigenvector) of the observable O.
may say that by a measurement one projects the eigenvector out of the state |
under investigation.
|n
= Wn |n
.
H
Sz |sms
= ms |sms
.
2.3.2 Representations
We have already made use of different ways to describe quantum systems and their
changes. The standard terminology speaks about representations of states and oper-
ators.
SCHRDINGER Representation
In the S CHRDINGER representation (or picture) the operators are differential oper-
ators. The states are represented by wave functions. The scalar product is an integral
according to (2.27) and the orthogonality of basis states is defined by (2.28). The
matrix elements of an operator A in respect of an arbitrary basis set |fk
, |fi
, . . .
are
(in general not eigenfunctions of A)
k
= fi (r)Af
Aik fi |Af k (r)d3 r = fk A
fi = A , (2.44)
ki
where we have used the definition of adjoint operators (2.32). If the operator is
Hermitian, i.e. represents and observable, this relation reads
Oik fi |Ofk
= fi (r) Of k (r) d3 r (2.45)
= k (r) fi (r)d3 r
Of i
= Oki
= fk |Of
. (2.46)
HEISENBERG Representation
In the H EISENBERG representation (or picture) the operators (we mention in this
context the N OBEL prize for H EISENBERG 1932) A are matrices, which are de-
termined by their matrix elements Aik . The states, say |
or |
, are vectors in
H ILBERT space5 which we write
|
= = b1 f 1 + b2 f 2 + b3 f 3 +
|
= = c1 f 1 + c2 f 2 + c3 f 3 +
i
= i |i
and B|
A| i
= i |i
.
Thus, also the following relations must be valid for simultaneous measurability of
and B:
the operators A
B|
A i
= A
i |i
= i A| A|
i
= i i |i
= B i
.
B
Hence, the operators must commute, A =! B
A.
Equivalently:
and B
Simultaneous measurement of two observables A is possible if and only
if their commutator vanishes:
B
A B
A = [A,
B]
= 0. (2.48)
From the above follow some simple recipes of how to translate classical quantities
into quantum mechanical operators. One simply has to substitute:
r r and pi i =p
i
xi
(2.49)
or p i , i , i = i =
p.
x y z
All other substitutions are derived from these rules. In particular, the classical
Hamiltonian total energy
p2
Hclass = + V (r) = Wkin + V with p 2 = p p
2m
becomes the H AMILTON operator:
We shall now find the eigenfunctions and eigenvalues of the momentum in the
S CHRDINGER picture, starting with the one dimensional case:
d(x)
x (x) = px (x)
p i = px (x).
dx
One easily verifies that (x) = exp(ipx x/) = exp(ikx x) are solutions of this eigen-
value problem. Each value of px (with < px < ) is an eigenvalue of the mo-
mentum operator p x in x-direction: the eigenfunction is thus a plane wave with a
continuum of eigenvalues.
This is easily extended into 3D space. The eigenvalue equation
p (r) = i(r) = k(r) is solved by
a plane wave (r) = C exp(ik r).
with p = k and the angle between e and k just as one would guess.
Section summary
In the S CHRDINGER picture, states (bra |
and ket |) of quantum sys-
tems are represented by wave functions (r) and (r), respectively. The
H EISENBERG picture uses state vectors in H ILBERT space.
Quantum states may be expressed as a linear superposition of states (2.21)
from a complete, orthonormal basis with fi |fk
= ik . The unit operator
may be written i |fi
fi |.
The S CHRDINGER representation uses differential operators. Classical the-
ory is translated by replacing r r and p p = i.
Matrix elements of an operator A k
= f (r)Af
are Aik fi |Af k (r)d3 r
i
in the S CHRDINGER picture. In the Heisenberg representation operators are
defined by the corresponding matrices.
The adjoint A of an operator A is defined by A
|
= |A
.
Hermitian
operators are self-adjoint A = A, and Aik = Aki holds.
2.4 Particles in a Box And the Free Electron Gas 103
Eigenstates of the momentum operator p are plane waves exp(ik r), with
p = k.
2 d2 n (x)
= Wn n (x).
2m dx 2
Solutions can in principle be (x) = sin(kx) or cos(kx). However, since the wave
function cannot penetrate into the wall, it must vanish on the walls which at x = 0 is
only possible for the sin(kx) solution. For continuity on the other wall sin(kL) = 0
and thus k = n/L must hold, with n = 1, 2, 3, . . . . In summary, the solutions
(eigenfunctions) are standing waves with nodes on both walls of the box. The
eigenenergies assume discrete values Wn :
2 nx 2 k 2 h2 n2
n (x) = sin with Wn = = . (2.52)
L L 2m 8mL2
The expectation value of the momentum p x is derived from
x n
= n (x)
px
= n p px n (x)dx
2 nxL d sin nx
= i
sin L
dx
L0 L dx
i2n L nx nx
= sin cos dx 0.
L2 0 L L
This corresponds to the fact that the particle in the box moves with equal probability
back and forth. In contrast, the square of the momentum,
2 2 L nx 2 d2 sin nx
x = n p
p x2 n = n p x2 n dx = sin L
dx
L 0 L dx 2
104 2 Elements of Quantum Mechanics and the H Atom
2(n)2 L nx 1 h2 2
= sin2 dx = n ,
L3 0 L 4 L2
=
is not zero. With this and H px2
/2m we verify Wn in (2.52).
A next step towards reality is the extension into the 3D space. The movement of a
particle is now restricted to a large but finite 3-dimensional box, with rigid walls
beyond which the probability to find a particle is zero. For simplicity we assume
the box to be a cube with edge length L as illustrated in Fig. 2.2(a). Inside the box
particles move freely. Stationary solutions are plane waves (2.18) which we write in
the box as real functions now as product in three dimensions:
To be continuous, the wave function must vanish on the walls of the cube:
sin(kj L) = 0 kj = n j for j = x, y, z. (2.54)
L
With these boundary conditions, and in analogy to the 1D case (2.52),
2 2 2 k 2 2 2 2
W= kx + ky2 + kz2 = = n , (2.55)
2m 2m 2mL2
now with three integer quantum numbers nx , ny , nz and n2 = n2x + n2y + n2z .
One may view this in k or n space as indicated in Fig. 2.2(b). Equation (2.54) says
that precisely one solution exists for each lattice point with integer nx , ny and nz .
Thus, one reads from the figure that the total number of states with quantum num-
bers 1 to nx , 1 to ny and 1 to nz , i.e. the number of states with n n2x + n2y + n2z is
given by NZ (n) = 1/8 4/3 n3 . Expressing n by the energy W , the total number
of states with energies W is
1 (2mW )3/2 3
NZ (W ) = L .
6 2 3
If the particle has a spin s, we also have to account for the energy degenerate ge =
2s + 1 possible orientations in space. Dividing by the volume of the box L3 , we
obtain the total number of available states per unit volume:
ge (2mW )3/2
NZ (W ) = . (2.56)
6 2 3
From this we derive the number of states in an energy interval from W to W + dW ,
called density of states (DOS), here per unit volume:
dNZ (W ) ge (2m)3/2 4 2m3/2
g(W ) = = W = ge W. (2.57)
dW 4 2 3 h3
We note in passing, that assuming the phase space to be quantized in unit cells of
size h3 leads to exactly the same result.6 For later use we also give the density of
states in respect of a specific element of solid angle d and express the energy by k,
the magnitude of the wave vector:
d2 NZ (W ) d ge mk
dg = = d. (2.58)
dW 4 (2)3 2
w(W ) / eV -1 (a) W
w (W ) / eV -1 (b)
W
0.2 0.2
T = 0K
T = 50 K
0.1 0.1 T = 500K
Ne Ne
T = 293 K
0 0
0 2 4 6 8 6.8 6.9 7.0 7.1 7.2
F W / eV F
Fig. 2.3 F ERMI -D IRAC probability distribution according to (2.61) as a function of energy for
electrons in a metal with a F ERMI energy F = 7 eV at different temperatures T : (a) energy range
from 0 eV to 9 eV, (b) expanded scale around F ; at T = 0 the electron density Ne extends up to
F (grey shading up to the dashed vertical red line); the full red line indicates the density of states
W
total number of electrons per volume with kinetic energies between 0 and F is
thus
1 2me F 3/2
Ne = 2NZ (F ) = . (2.59)
3 2 2
We may invert this to obtain the so called F ERMI energy F . At absolute zero tem-
perature F is identical to the chemical-potential introduced in Sect. 1.3.4, and
corresponds to the maximum electron energy. One state after the other is filled, each
by two of electrons, up to the F ERMI energy:
2 2 2/3
F = 3 Ne . (2.60)
2me
Typical F ERMI energies for metals range from 1.6 eV (Cs) to 14.3 eV (Be).
Expressing the prefactor A given by (1.64) with ge = 2 in terms of the F ERMI
energy F one may write the probability distribution for finding a given energy W
in the electron gas in a box:
dNe 3 1 3/2 W
w(W )dw = = dW (2.61)
Ne 2 F exp[(W )/(kB T )] + 1
with F F as long as kB T F .
Section summary
The particle in a box model allows a most simple description of electrons
moving freely in a metal. The wave functions (2.53) in the 3D box (vol-
ume L3 ) have nodes at theboundaries of the box. The energies are W =
2 2 n2 /(2mL2 ) with n = n2x + n2y + n2z and ni representing positive val-
ues in 3D integer number space.
From this the DOS (2.57) is derived, which according to (2.57) is
ge m3/2 W , where m is the particles mass and ge = 2s + 1 the degen-
eracy due to its spin.
In the case of electrons (Fermions) each state can be filled with up to 2 elec-
trons. Then, at T = 0 K the highest energy with occupied states is the F ERMI
energy F = 2 (3 2 Ne )2/3 /(2me ).
At temperatures T > 0 the boundary between occupied and unoccupied states
smears out according to (2.61). The width of the boundary layer is on the
order of kB T .
Angular momenta play a central role in atomic and molecular physics and Ap-
pendix B gives a summary on the essentials: their abstract definitions, properties,
combinations and the relevant algebra. Here we introduce the S CHRDINGER pic-
ture of orbital angular momenta as used in the quantum mechanics of the H atom.
The orbital picture emerging from this concept is directly accessible to physical
imagination and visualization. At the end of this section we shall generalize the
basic concepts, including electron spin.
Quantum mechanical problems may be treated with advantage in polar rather than in
Cartesian coordinates if they have a symmetry centre, e.g. if the potential depends
only on the distance r from origin V (r) = V (r) as in the C OULOMB case. The
108 2 Elements of Quantum Mechanics and the H Atom
y
x
transformations between Cartesian (x, y, z) and polar coordinates (r, , ) are read
from Fig. 2.4,
x = r sin cos
y = r sin sin (2.62)
z = r cos ,
2 1 1 2
L = 2
sin + 2 . (2.65)
sin sin 2
In full analogy to classical mechanics, the form of (2.64) suggests two components
of kinetic energy:
2 1 2
radial energy Hr = r and (2.66)
2m r 2 r r
2
rotational energy rot = L .
H (2.67)
2mr 2
L = r
p. (2.68)
This has to be expressed in polar coordinates. We just show this for one coordinate
by way of example:
z = i x y .
L (2.69)
y x
With (2.62) one transforms the expression in brackets into polar coordinates:
x y z
= + + = r sin sin + r sin cos +0
x y z x y
= y +x =x y .
x y y x
z = i .
L (2.70)
We are now ready to discuss eigenvalues and eigenstates of angular momentum op-
erators. They will be used in more or less all of the following chapters. In the spirit
of this textbook, again we only sketch the basic concepts, present a collection of nec-
essary tools, and refer the reader to the standard textbooks on quantum mechanics
and angular momentum algebra for details.
1
with eigenfunctions m = exp(im), (2.74)
2
and eigenvalues m where m = 0, 1, 2, . . . .
x , L
[L y ] = L
x L y L
y L z ,
x = iL
(2.75)
[L z ] = iL
y , L x z , L
and [L x ] = iL
y .
2
L Y (, ) = L2 Y (, ) (2.76)
with the eigenvalue L2 and use a the product ansatz towards its solution:
Y (, ) = ( )(). (2.77)
z and substitute ( )m ()
As azimuthal part we try the eigenfunctions (2.74) of L
into (2.65). This leads to
1 ( ) 1 2
2
sin m () + 2 ( ) 2 m ()
sin sin
= L2 ( )m () = L2 Y (, )
1 m2
2 sin 2 = L2 . (2.78)
sin sin
Thus, only one ODE remains to be solved. Several procedures lead to the cor-
rect solutions. One may directly use the associated L EGENDRE polynomials, known
from the mathematics of ODEs, or (perhaps more elegantly) by exploiting the prop-
erties of the angular momentum operators to find appropriate recursion formulas.
In any case, one has to demand physically reasonable boundary conditions. In anal-
ogy to (2.72) for the z-component, the wave functions must be finite and unique for
0 . Without proof we communicate here that such physically meaningful
solutions exist for which the following relations hold (for completeness we include
again the z-component):
x -2
z only acts
The validity of (2.81) follows directly from (2.74) and (2.77), since L
onto the component of Y m (, ). This implies
2
2
L L z
z=L L
2
or z = 0.
L ,L (2.84)
Vector Diagram
With (2.79) one may write the magnitude of the angular momentum as:
|
L| = ( + 1). (2.85)
For a given set of quantum numbers m the exact direction of the total angular
momentum is undefined. Precisely defined is only the magnitude and component
Lz = m in respect of the z-axis. One visualizes these relations with the help of
a vector diagram shown in Fig. 2.5. ofthe electron spin (see Fig. 1.43). Figure 2.5
illustrates the example = 2, |
L|/ = 6 2.45 with L z / = m = 2, 1, 0, 1, 2.
One may consider the vector arrows statistically
distributed around the z-axis, i.e.
on cones of height m with a side length ( + 1).
Spherical Harmonics
The eigenfunctions of
2
L and Lz are called spherical harmonics Y m (, ). General
formulas and properties are summarized in Appendix B.1.2, and specific expres-
sions up to = 3 are tabulated in Table B.1. A graphical illustration of the angular
dependence is shown in Fig. 2.6.
The Y m (, ) are orthonormalized:
2
d Y m (, )Y m (, ) sin d = m m . (2.86)
0 0
2.5 Angular Momentum 113
x y
Fig. 2.6 3D-plot of the s, p and d spherical harmonics; plotted are the squared moduli as a func-
tion of angles, the shading colours indicate the sign of Y m (, 0). For an alternative representation
in the real basis see Appendix D.3 and in particular Fig. D.1
Y m (, ) = (1)m Y m (, ), (2.87)
Y m ( , + ) = (1) Y m (, ), (2.88)
In the following text we shall, for compactness, usually write the spherical harmon-
ics in bra and ket form, substituting
Y m (, ) | m
and Y m (, ) m|. (2.89)
In this notion, the orthogonality relations (2.86) and the matrix elements of an oper-
ator A are written as
m m = mm m .
and A m, m = m|A (2.90)
As shown by the S TERN -G ERLACH experiment, the electron has in addition to mass
and charge one further property which we have identified in Sect. 1.10 as an intrinsic
angular momentum, called spin. The spin is characterized by the spin quantum num-
ber s = 1/2. Its magnitude is |S| = s(s + 1), and two orientations with angular
momentum components /2 and /2 are possible. Closely related to the spin, the
electron has also a magnetic moment with a g factor as defined by (1.162) close
to ge 2.
One simply transfers the formal rules which we have introduced in Sects. 2.5.2
2.5.3 onto the properties of the spin. Clearly, the spin may not be imaged in position
space. However, we may define quite formally in analogy to the orbital angular
momentum L a new vector operator S with a square magnitude operator S 2 and a
component Sz in z-direction for which the general commutation rules for angular
momenta (2.75) and (2.84) are valid:
2
[
Sx ,
Sy ] = i
Sz , [
Sy ,
Sz ] = i
Sx , [
Sz ,
Sx ] = i
S ,
Sy ,Sz = 0.
(2.91)
They imply, as in the general case, that the components of the spin cannot be mea-
sured simultaneously. However, it is possible to determine its magnitude together
with one of its components is (e.g. the z-component
Sz ).
We now introduce spin states |sms
. In analogy to (2.79)(2.83) (for the orbital
angular momentum)
for which the spin points into +z- and z-direction, respectively. We have intro-
duced here three equivalent notations which are commonly used for compactness.
One also finds in the literature the notation spin function and without bra or ket.
Alternatively, one speaks about spin up () and spin down () states. In any
case, the relations
2 |
= 3 2 |
S 2 |
= 3 2 |
S
4 4
(2.94)
Sz |
= |
Sz |
= |
2 2
hold, together with the orthonormality relations
|
= |
= 0 and |
= |
= 1. (2.95)
2.5 Angular Momentum 115
From the commutation rules (2.91) one may derive how the other components of
S
act onto the basis (here without proof):
i i
Sx |
= |
Sx |
= |
Sy |
= |
Sy |
= |
. (2.96)
2 2 2 2
In this basis each arbitrary spin state of an electron may be expressed as
| = + | + | , (2.97)
which are normalized by definition. As an exercise, the reader may show with the
aid of (2.94) and (2.96), that the parameters and give the polar and azimuthal
angles, respectively, at which the so defined spin state is oriented in space.
The above provides a toolbox which is fully sufficient to describe the properties
of the spin states. Nevertheless one often writes perhaps for historical reasons or
better visualization operators in the form of matrices and eigenstates as vectors,
the so called spinors
+
= and = + (2.99)
1 0
with the basis = and = .
0 1
116 2 Elements of Quantum Mechanics and the H Atom
S= , (2.100)
2
with
being a vector operator, the so called PAULI vector, which is built from the
PAULI matrices:
0 1 0 i 1 0
x = y = and z = (2.101)
1 0 i 0 0 1
3 1 0
S = Sx2 +
Sy2 +
2
Sz2 = 2 . (2.102)
4 0 1
For later use we note here that the PAULI matrices anti-commutate
i j + j i = 2ij , (2.103)
and thus
x2 = y2 = z2 =
1 and x y = y x = iz . (2.104)
Section summary
Orbital angular momentum operators may be derived from L=r p , with
p = i, or be constructed from the commutation rules (2.75).
From the spatial representations of the operators z according to
2
L and L
(2.65) and (2.70), respectively, one obtains eigenvalues and wave functions
as summarized in (2.79)(2.83), and schematically illustrated in the vector
diagram Fig. 2.5.
The shape of the orbitals with lowest angular momentum s, p and d shown
in Fig. 2.6 should be memorized. General formulas and properties are sum-
marized in Appendix B.1.2, specific expressions are tabulated in Table B.1.
The electron spin obeys the same commutation rules as orbital angular mo-
menta. Its intrinsic angular momentum, the spin, is however s = 1/2 and the
projection quantum number ms = 1/2. A frequently used representation of
the spin operators are the PAULI matrices (2.101) and (2.102) which act on the
so called spinors two component representations of the spin eigenfunctions.
2.6 One Electron Systems and the Hydrogen Atom 117
2
=
H
p2
+ V (r) = Hr + L + V (r) (2.106)
2me 2me r 2
2 1 2
with Hr = r , (2.107)
2me r 2 r r
r + L
H + V (r) n m (r, , ) = Wn m n m (r, , ). (2.108)
2me r 2
2 1 d 2 d 2 ( + 1)
r + + V (r) Rn (r) = W Rn (r). (2.110)
2me r 2 dr dr 2me r 2
2 ( + 1)
Veff (r) = V (r) + . (2.111)
2me r 2
With this and the substitution
2 d2 un
2
+ Wn Veff (r) un (r) = 0. (2.113)
2me dr
Note: The total energy does not depend on the projection quantum number m and is
thus be written W = Wn . Zero energy is usually set for the completely unbound
system, i.e. for electron and nucleus at infinite distance with no kinetic energy.
Bound electrons have negative energies Wn < 0, while free electrons have total
energies W > 0. Extending (2.89) to the full electron wave function one often ab-
breviates
Rn (r)Y m (, ) |n m
. (2.114)
More specifically, one even writes these atomic orbitals shorthand as n = 1s, 2s,
2p, 3s, 3p, 3d, etc. in the notation introduced in Sect. 2.5.3.
We recall here the concept of atomic units (a.u.), introduced in Sect. 1.8.3:
energy Eh = me e4 02 h2 /4
length a0 = 0 h2 e2 m1
e / = / me Eh (2.115)
time t0 = 202 h3 e4 m1
e /.
Numerical values of these quantities are given in Appendix A, and the most recent,
accurate updates can be found at NIST (2010). We use these definitions here to
rewrite the radial S CHRDINGER equation 2
(2.113). We multiply (2.113) by me /
and a0 , and apply the identity a0 = / me Eh to obtain in dimensionless form:
2
1 d2 un ( + 1) V (r/a0 )
+ W n /E h + un (r) = 0. (2.116)
2 d(r/a0 )2 2(r/a0 )2 Eh
1 d2 un
2
+ Wn Veff (r) un (r) = 0 (2.117)
2 dr
( + 1)
with Veff (r) = V (r) + and for the H atom V (r) = Z/r.
2r 2
Following this scheme one may rewrite all atomic equations in a dimensionless,
rather clean looking form. Theoreticians, in particular, like this kind of equation very
much and even give the recipe to just set = e = me = 1 which really oversimpli-
fies what has to be done. The procedure has one serious disadvantage: a dimensional
analysis is no longer possible which is often highly commendable to check com-
plex calculations. Thus, we typically try to use equations in a form indicated by
(2.116), and carry the a.u. a0 , Eh , and t0 explicitly along. Sometimes one may even
be able to combine elementaryconstants to truly dimensionless quantities, such as
the fine structure constant = Eh /me c2 according to (1.10).
Up to now we have treated the problem as if the electron would orbit around a space
fixed centre. As the nuclear mass is much larger than the electron mass me in the
case of the H atom with a proton as nucleus mp 1840me the centre of mass
is indeed very close to r = 0. For more demanding precision one has, however, to
correct for the difference. As in classical mechanics (see corrections to the B OHR
model described in Sect. 1.8.5), in quantum mechanics too, one transforms the two
body problem into an effective one particle problem by replacing the electron mass
me with the reduced mass m e (1.143) of the system. All a.u. have, in principle, to
be replaced correspondingly (kinematic correction):
me e
m e
m
a0 a 0 = a0 , Eh E h = Eh , and t0 t0 = t0 . (2.118)
e
m me me
For simplicity we shall, however, in the following text continue to use me and the
units a0 , Eh and t0 , and refer to the exact calculations if relevant.
While we know already the angular part of the hydrogenic wave functions, we
are still left with the task to find physically meaningful solutions to the radial
S CHRDINGER equation (2.117). They have to behave reasonably at r = 0 (see
below) and must not diverge for r . From this follows necessarily that only a
particular set of discrete total energies Wn < 0 leads to such solutions. To find these
is the task at hand.
Before applying mathematics we want to obtain a qualitative picture to support
our physical intuition. Figure 2.7 illustrates this for the = 0 case in a C OULOMB
potential V (r) = Z/r. We derive the kinetic energy of the electron at different
120 2 Elements of Quantum Mechanics and the H Atom
W,V(r )
0
Wn r
Wkin
rcl
V
un (r ) large classically for-
bidden region:
= h /p exponential
small decrease of
wavefunction r
Fig. 2.7 Bound state radial wave functions for s states, schematic. Top: C OULOMB potential
V (r) 1/r (red) and total energy Wn < 0 (black) determine the classical turning point rcl
(onset of the classically forbidden region). Bottom: The characteristic behaviour of radial wave
functions un is explained by changes of the kinetic energy Wkin in different regions of the poten-
tial (see text)
positions in the potential from Wkin = Wn V (r) and take the corresponding DE
B ROGLIE wavelength = h/p = h/(2me Wkin )1/2 as an indication for changes in
the radial wave function un (r). It obviously will change more rapidly for small
r (large Wkin ) than in the neighbourhood of the classical turning point rcl , where
Wkin = 0. In the classically forbidden area with r > rk (negative kinetic energies)
we expect exponential damping of the wave function, as illustrated in Fig. 2.7.
In a next step we explore the limits for very large and very small r. For the
limiting case r we may neglect the potential altogether and (2.117) becomes
a simple oscillator equation:
1 d2 un
+ Wn un (r) = 0.
2 dr 2
The classical solution is un (r) exp(i 2Wn r). Since for bound states Wn < 0,
we note for large r
lim un (r) r n exp 2|Wn |r . (2.119)
r
In the opposite limit r 0 the centrifugal term ( +1)/2r 2 dominates the potential
in (2.117) and
1 d2 un ( + 1)
un (r) = 0
2 dr 2 2r 2
r0
has the solution un (r) = Ar +1 as one easily verifies. Thus, we note for small r
lim Rn (r) = lim un (r)/r r . (2.120)
r0 r0
2.6 One Electron Systems and the Hydrogen Atom 121
We now specialize to the H atom. In the present chapter we neglect the size of the
atomic nucleus (positively charged with +Ze) since nuclear radii rnuc are much
smaller than atomic radii, typically ratom 105 rnuc . Thus, we are dealing with a
pure C OULOMB potential
1 Ze2
V (r) = , (2.121)
40 r
apart from small and very small effects which will be treated in Chaps. 6 and 9,
respectively. The general solution of the radial S CHRDINGER equation (2.110) is
found by using a power series of the type
...
Rn (r) = exp 2|Wn |r Ak r k ,
k=
which includes the limiting cases just discussed. Well known results from mathemat-
ics are used and we summarize here the results, again without proof. For hydrogen
like systems i.e. for one electron in the C OULOMB potentialof a Z fold charged
nucleus the radial function is
n 1
[(n + )!]2 k
2 +1
Ln+ () = (1)k+1 . (2.123)
(n 1 k)!(2 + 1 + k)! k!
k=0
We introduce here finally the often used term good quantum numbers: They char-
acterize the eigenvalues of such observables that may be measured simultaneously
with the H AMILTON operator, i.e. their operators commutate with the Hamiltonian.
We already know and m as typical examples: they are part of the set of quantum
numbers characterizing the total energy of the system; z are simultaneously
2
L and L
measurable with H .
122 2 Elements of Quantum Mechanics and the H Atom
Wn / eV n
=0 =1 =2 =3 =4
0
8
6
5s 5p 5d 5f 5g
- 0.85 4
4s 4p 4d 4f
- 1.51 3
3s 3p 3d
- 3.40 2
2s 2p Z2
Wn = E h
2n 2
independent of
-13.6 1
1s
Fig. 2.8 Term energies of the hydrogen atom (Z = 1) for different n and
Table 2.1 The lowest atomic levels, their energies in the H atom and the degeneracy of the states
(with Eh = 27.2 eV)
Shell Orbital n m Wn Degeneracy
Without spin With spin
sum in shell sum in shell
K 1s 1 0 0 Eh /2 1 1 2 2
L 2s 2 0 0 Eh /8 1 2
2p 2 1 0, 1 3 4 6 8
M 3s 3 0 0 Eh /18 1 2
3p 3 1 0, 1 3 6
3d 3 2 0, 1, 2 5 9 10 18
N 4s, p, d, f 4 Eh /32 16 32
These solutions of the S CHRDINGER equation are the quantum mechanical equiv-
alent to B OHRs stationary orbits. Substituting unl (r) according to (2.122) into the
radial equation (2.113) one finds the eigenenergies Wn for the H atom. Remark-
ably, they are identical to the energies (1.137) from the B OHR model.8 These results
are summarized in Fig. 2.8 and Table 2.1.
We recall that the states are characterized by the principle quantum number
n = 1, 2, 3, . . . , the angular momentum quantum number (0 n 1) and
the projection quantum number m ( m ). Each set n m of quantum num-
bers refers to a different wave function (atomic orbital). In addition, we also have to
consider the spin of the electron, with projections ms = 1/2.
8 The corrections for finite mass of the atomic nucleus (Sect. 1.8.5) also apply.
2.6 One Electron Systems and the Hydrogen Atom 123
Veff (r ) / E h
=1 =2
0.2
5 10 15 20
0
n=2 n=3
r /a 0
COULOMB potential
- 0.5 n =1 = Veff for = 0
Fig. 2.9 Illustration of degeneracy: Shown are the pure C OULOMB potential (red line) and the
effective potentials (black) for the H atom. The term energies Wn are indicated by horizontal
lines in their respective effective potentials: for = 0 (dotted red), for = 1 (dashed grey) and for
= 2 (heavy, full red). As indicated, the C OULOMB potential leads to a characteristic degeneracy
of states with equal n but different
However, for the H atom the energies Wn depend only on the principle quantum
number n. Thus, the energy levels are degenerate, the total degeneracy in a shell
n being gn = 2n2 . This is summarized in Table 2.1 for n = 1 to 4, also showing
the assignment of orbitals with equal principle quantum number n (i.e. with equal
energies and similar orbital radii) to specific shells (K, L, M, N . . . corresponding
to n = 1, 2, 3, 4, . . .).
It is important to note that degeneracy is a special property of the pure
C OULOMB potential, while m degeneracy occurs for all atoms if no external field
is present. Figure 2.9 illustrates degeneracy in the potential energy diagram, also
showing the effective potentials (2.111) for = 1 and = 2 (for = 0 effective and
C OULOMB potential are identical).
The radial wave functions have very specific shapes, which may be understood from
the properties of the effective potentials. Figure 2.10 illustrates this schematically by
way of example for the n = 3 level and orbital angular momenta = 0 and 1. Shown
are C OULOMB potential, centrifugal potential and effective potential for = 1. The
classically forbidden areas (Wkin < 0) are grey shaded, the classical turning points.
While the radial functions for = 0 start with a finite value at r = 0, for = 1
the probability there is = 0 (since ( + 1)/(2r 2 ) ). Oscillations of the radial
functions are expected only in between the classically turning points.
Table 2.2 presents the radial wave functions Rn (r) for the six energetically low-
est states (n 3) of atomic hydrogen (Z = 1) and H-like atoms (Z > 1) in closed
form.
From the radial wave functions one derives the probability distributions
2
w(r)dr = Rn (r) r 2 dr (2.125)
124 2 Elements of Quantum Mechanics and the H Atom
1 R21 (r) = 1
( Z )3/2 e/2
2 6 a 0
3 0 R30 (r) = 1
( Z )3/2 (6 6 + 2 )e/2
9 3 a 0
1 R31 (r) = 1
( Z )3/2 (4 )e/2
9 6 a 0
for the electron to be found between r and r + dr. For comparison with the classical
picture of an orbiting electron one has to consult these probability distributions as
a function of distance from the nucleus. A graphical illustration of the radial wave
functions Rn (r) and the radial probability distributions w(r) is shown in Fig. 2.11.
Closer inspection of the probability distributions shows, that their maxima match ex-
actly the radii of the respective B OHR orbits for the largest = n 1 at any given n!
The square of the wave function |n m (r, , )|2 represents the probability distribu-
tion for finding electrons per volume element, or equivalently the electron density
within the atom as a function of position in space. Since these atomic orbitals form
2.6 One Electron Systems and the Hydrogen Atom 125
Fig. 2.11 Radial wave functions of the H atoms Rn (r) and probability distributions r 2 Rn 2 (r) for
the K, L and M shell. The dashed, vertical lines in the probability plots for the highest at a given
n indicate the corresponding B OHR radii. Note the drawing for the 1s orbital on a blown up r-scale
3
z 1s 0 2p 1 3s0 3p 1
5 10 10
0 0 0 0
-5 -10 -10
-3
-3 0 3 -5 0 5 -10 0 10 -10 0 10
3d 1 3d 2 20 4 s 0 40 5s 0
10 10
0 0 0 0
-10 -10
-20 -40
-10 0 10 -10 0 10 -20 0 20 -40 0 40
x
Fig. 2.12 Density plots for some characteristic H atom wave functions n m . Plotted are equiden-
sity lines (red high, grey low probability density). The distances are given in a0
a base for solving many key problems in atomic and molecular physics, we recom-
mend our readers to visualize and memorize their general shape intensively. One
finds a host of instructive Internet-sites, e.g. with Java applets to generate the H or-
bitals in a variety of displays. Thus, we present in Fig. 2.12 only a small selection of
cuts through the 3D density distributions. Plotted are the contour lines of the den-
sity |n m (x, y = 0, z)|2 = |Rn (r)Y m (, )|2 in the zx plane on a linear scale (in
contrast to many presentations in the WWW, where one finds the density plotted on
a log-scale). For clearness we have indicated the highest densities by red areas.
126 2 Elements of Quantum Mechanics and the H Atom
We finally note again, that only the ns orbitals have a finite density at the origin,
|ns (0)| > 0, while
n m (0, , ) 0 for > 0. Specifically, from (2.122) and with
Y00 (, ) = 1/ 4 we record for later use:
n00 (0)2 = Z 3 / a 3 n3 . (2.127)
0
according to n (with n < n). Clearly, due to the m and (for the H atom) degener-
acy, the angular momentum quantum numbers do not influence the position of the
spectral lines as one also verifies by a glance at Fig. 2.8. This holds at least in 1st
order approximation. Finer effects will be discussed in Chap. 6.
A lot of scientific detective work of the early pioneers went into discovering the
connection between spectra and term energies, and the series are named after those
who originally discovered them. Most prominent are the LYMAN (n = 1 in the
VUV), BALMER (n = 2 in the VIS and near UV) and PASCHEN series (n = 3 in
the near IR). Within the series, the BALMER lines are historically referred to as H-
alpha, H-beta, H-gamma etc., more generally the lines of the different series
are designated as Ly-, and Ly-, Ly- , . . . , Ba-, Ba-, Ba- , . . . and so on, with
= n + 2, = n + 3, etc.
One often needs to know the expectation values of the electron distance from origin
r to a certain power k. In principle, they could be determined by a large number of
individual measurements of this value in an suitably designed experiment. Quantum
mechanics provides:
k
r = n |r k |n
= Rn (r)r k Rn (r)r 2 dr = 2
Rn (r)r 2+k dr (2.129)
0 0
with 2
Rn (r)r 2 dr = 1.
0
These integrations are trivial but somewhat tedious. Using (2.122) and (2.123) one
obtains (for later reference):9
n2 1 ( + 1)
r
n = a0 1+ 1
Z 2 n2
"
#
2 n4 3 ( + 1) 1/3
r n = a02 2 1 + 1
Z 2 n2
1 1 Z
= (2.130)
r n a0 n2
1 1 Z2
=
r 2 n a02 n3 ( + 1/2)
1 1 Z3
= .
r 3 n a03 n3 ( + 1/2)( + 1)
One often hears the verdict that the B OHR model is basically wrong albeit predict-
ing the correct term energies Wn and spectra (1.131) for the H atom in agreement
with quantum mechanics and thus should be abolished when teaching modern
atomic physics. We do not adhere to such dogmatic view.
Of course, B OHRs orbits have to be replaced by atomic orbitals, i.e. by den-
sity distributions of the electrons within the atoms, and B OHRs quantization condi-
tion (1.130), L = n, has to be confronted with its quantum mechanical analogue,
z () = m(). Obviously, m n 1 is not equal to n as postulated in the
L
B OHR model but it is a good first guess.
On the other hand, the concept of angular momentum quantization, of stationary
states, and of radiation emitted upon transition between them with h =
Wn was
a brilliant, daring and instrumental step on the way to understand the quantum nature
of microscopic phenomena even though the fact that states without orbital angular
momentum ( = 0) do exist, does not fit at all into the B OHR model. However, the
higher the angular momenta, the closer the atomic orbitals correspond to B OHRs
orbits. And in general, the correspondence principle holds:
However, the maxima of the probability distributions are of course not directly
observable; rather one may compare expectation values such as n |r|n
= r
. For
the largest angular momenta = n 1 one obtains from (2.130)
1 2 a0 a0
r
= lim n + n = n2 , (2.131)
n 1 2 Z Z
which obviously agrees in the limit with B OHRs prediction (1.135). Conversely,
for the smallest values of = 0, the average radius r
is distinctively larger than
predicted by the B OHR model, namely (3/2)n2 a0 /Z.
In summary, we do not feel that the B OHR model should be completely forgotten.
Apart from its outstanding historical impact, we shall come across a number of spe-
cific aspects in modern AMO physics for which useful, simple models or concepts
have been stimulated by images of electrons moving on classical trajectories com-
bined with appropriate quantization rules. We may think of B ORN -O PPENHEIMER
approximation for molecular physics and of semiclassical trajectory calculations in
atomic scattering theory, to mention just two important, and very successful exam-
ples or, as we shall see in the following section, the derivation of the magnetic
moment of an electron associated with its orbital angular momentum.
Section summary
The eigenenergies of the H atom, Wn = Z 2 /(2n2 )Eh , depend only on the
principle quantum number n . . . a specific consequence of the pure C OULOMB
potential. For precision measurements this has to be corrected by m e /me
where m e is the reduced mass of the electron.
Electron wave functions for the H atom can be expressed in analytical form
n m (r, , ) = Rn (r)Y m (, ) with the spherical harmonics Y m (, ) and
the radial function Rn (r) being proportional to the L AGUERRE polynomials.
The asymptotic behaviour of the wave functions should be memorized:
limr Rn (r) exp( 2|Wn |r) and limr0 Rn (r) r .
good quantum numbers characterize the eigenvalues of such observables that
may be measured simultaneously with the H AMILTON operator.
The so called normal Z EEMAN effect is actually not normal at all. It is observed
only in special cases. It concerns atoms in an external magnetic field. The word
normal simply alludes to the classical interpretation of such a situation ignoring
the electron spin. We shall treat atoms in external fields in great detail in Chap. 8.
Thus, the subject is touched here just briefly, and only since it confronts us for the
first time with the removal of a specific energy degeneracy, the m degeneracy in this
case.
2.7 Normal ZEEMAN Effect 129
with
W being the change of total energy in respect of the unperturbed state. Here
and in the following we use state vectors instead of wave functions n m |n m
for compact writing.
The present problem is a particular simple case for the so called perturbation the-
ory to which we shall come later in some detail. At this point we just remember that
according to (2.79) and (2.81) the eigenfunctions n m (r, , ) = Rn (r)Yn (, )
of the H atom, or their eigenstates |n m
, respectively, are also eigenfunctions
(eigenstates) of z with
2
L and L
z |n m
= m|n m
.
L (2.136)
130 2 Elements of Quantum Mechanics and the H Atom
m=0
B
m =1
The latter relation obviously implies that in a magnetic field the energy degeneracy
for different m is removed. The originally identical energies of the 2 + 1 states
|n m
which correspond to a given value of n and , now split into 2 + 1 different
energy sub-levels. According to (2.137) the splitting is proportional to m and B.
The origin of this splitting is that the magnetic field breaks the spherical symmetry
characteristic for the unperturbed H atom.
We illustrate this for the example of an np state |npm
with = 1 and sublevels
m = 1, 0, and 1. Figure 2.14 shows the energy changes
Wm as a function of the
magnetic field B.
One may observe this splitting e.g. in optical emission spectra. Figure 2.15 com-
pares the emerging spectra for (a) a p s and (b) a d p transition. The individ-
ual transitions are indicated by black, downward arrows. The selection rules applied
in this plot are
= 1 and
m = 0, 1. They will be derived and discussed in
Sect. 4.4.
As the degeneracy and hence the number of split levels is 2 + 1, a manifold
of transitions may occur if states with > 1 are involved. This is illustrated in
Fig. 2.15(b) where the upper levels correspond to a d state. Since, however, for
the normal Z EEMAN effect the splitting between neighbouring levels is always
B B according to (2.137), independent of and m, and since for all transitions
m = 0, 1, one nevertheless sees only a line triplet in all cases.
As mentioned at the beginning of this section: in reality this kind of Z EEMAN
effect is seen only in special situations (see Sect. 8.1.2), since usually the spin of
electrons plays an important role and complicates the observations.
We note here an important message from a situation which one typically en-
counters in quantum systems with two or more degenerate states |1
, |2
, |3
: The
degeneracy is removed as soon as some additional, perturbing interaction V1 has to
be considered for which the matrix elements i|V1 |j
between some of these states
do not vanish.
2.8 Dispersion Relations 131
m=+2
(a) (b) +1
+1
BB 0
p 0
d 1
1 BB
2
+1
BB
p 0
s m = 1 0 +1
1
m = 1 0 +1
Fig. 2.15 Normal Z EEMAN effect for (a) p s and (b) d p transitions. In case of equal
splitting in the excited and ground states one observes in each case a line triplet in spite of the
5-fold splitting in the d state
Section summary
A magnetic field B removes the central symmetry and hence the m degener-
acy. For the so called normal Z EEMAN effect, theory predicts a level splitting
Wm = B mB.
The selection rules for optical transitions are
= 1 and
m = 0, 1.
We make a brief detour here, illuminating the borderline between atomic, molecular,
optical and solid state physics. Traditionally, dispersion relations are used in optics
and characterize an important material property: the dependence of the wavelength
(or wavenumber k = 2/) of electromagnetic radiation on its frequency . From
a quantum mechanical point of view one may generalize this to describe the relation
between the energy of a photon W = (or in fact the energy of any other particle)
and its wave vector k.
For the massless particle photon in vacuo, with c = = /k, the dispersion
relation
W = = c|k| (2.138)
is obviously a linear relation between energy and wave vector. In contrast, the en-
ergy of a freely moving, nonrelativistic electron (whose mass is me ) is
me 2 p2 2 k 2
W (k) = WP + v = WP + = WP + , (2.139)
2 2me 2me
where we have used the DE B ROGLIE relation (1.118). WP allows for arbitrary en-
ergy calibration and may e.g. account for a potential energy or the rest mass energy.
Thus, in this case the dispersion relation is quadratic. Both cases are illustrated
graphically in Fig. 2.16.
132 2 Elements of Quantum Mechanics and the H Atom
k k
It should be noted that the quadratic relation (2.139) for particles with mass is
also in accord with the quantum mechanical calculation for a free particle in a 1D-
or 3D-box according to (2.52) or (2.55). Providing the box is large enough, the
energies may be considered continuous, i.e. they generally follow (2.139). However,
this is only the most simple model for the electronic band structure in a solid. The
particle in a box model does not account for the fact that the electrons in a solid
do not really move freely: rather, the electrons move in the lattice of atomic ions
and experience a periodic potential with strong attraction close to the ionic cores.
Elsewhere the potential is partially screened by bound and other free electrons in
the solid. We mention here two crucial consequences arising from this situation.
First, the dispersion relation (2.139) will have to be modified, which in principle
requires a serious band structure calculation. Many phenomena may, however, be
explained by the so called parabolic approximation. It parameterizes the band en-
ergy by introducing an effective mass me of the electron which may even depend on
the direction into which the electron moves:
2
kx ky2 kz2
W (k) = WP 2
+ + . (2.140)
2mx 2my 2mz
The sign allows to extend the concept also to electron holes (positive charges in
the latice) and both these quasi-particles may have different masses. This mass
enters into all further calculations on the dynamics and statistics of electrons and
holes in a solid, e.g. into the F ERMI -D IRAC statistics discussed in Sect. 2.4.3.
Second, one has to account for the periodicity of the motion in the lattice. The
solutions for this problem are so called B LOCH waves
which are the product of a plane wave exp(ikr) and a periodic function uk (r). The
latter has to obey the periodic boundary condition
uk (r + T ) = uk (r), (2.142)
where T is any translation from one elementary cell of the lattice into another. Over-
all we still expect the dispersion relation (2.139) for electron energies to be more or
less valid. As detailed studies show, one may have to replace the electron mass me
by a (different) effective mass meff but the quadratic dependence of the energy W
2.8 Dispersion Relations 133
on the wave vector typically holds. The electron motion in general averages over
the periodic potential. However, the energies and eigenfunctions will be influenced
strongly by the lattice potential if the electrons are particularly close to the ions or
particularly far away from them. This is specifically relevant when the electron wave
vector k = k BZ is at (or close to) the B RILLOUIN zone (BZ), i.e. if the correspond-
ing wave functions are constructively interfering (see Sect. 1.4.9). Such electrons
feel the periodic potential strongly, all others experience only an average.
We cannot go into detail of these concepts which are fundamental for the the-
ory of band structure in solid states. We just emphasize some aspects which are
also of importance in molecular physics. Thus, let us discuss the particularly simple
situation of the 1D case, which e.g. describes a chain of atoms with a distance a
from each other (also, with slight modifications, a ring like molecule). According to
(1.95) and (1.93) the B RAGG condition for the ntn BZ then reads
k = kBZ = n . (2.143)
a
The corresponding wave functions are essentially exp(ikBZ x) and exp(ikBZ x), and
for free travelling electrons the energy in that situation would be 2 k 2BZ /2me in both
cases. We have two energetically degenerate solutions. However, the most general
solutions are linear superpositions of both, i.e. standing waves, with the two physi-
cally reasonable cases:
These correspond to cos kBZ x and sin kBZ x. The probability |(x)|2 to find the elec-
tron close to the lattice ions is maximal in the first case, minimal in the latter. The
consequence of the perturbing periodic potential is removal of the degeneracy, sim-
ilar to Z EEMAN splitting discussed in the last section. And since the perturbing
potential is highly attractive, we expect the energy to be lowered in the first case,
and to be risen in the second case. In effect, at the boundary of the BZ we expect
the energies to split into two, and a gap to arise between two bands of the otherwise
continuous 2 k 2 /2me distribution.
This is illustrated in Fig. 2.17(a). Since the periodicity of the system does not
favour any particular origin in the reciprocal lattice, one projects all the possible
energies onto the first BZ as shown in Fig. 2.17(b). Electrons in the system may as-
sume all energies W in the grey shaded energy bands. The energy regions marked
band gap are energetically forbidden.
In summary, in the solid state case continuous energy bands, with gaps in be-
tween them, replace the discrete energy levels which we have discussed for the
atomic case (specifically for the H atom in the present chapter). One must, how-
ever, be aware of the fact that Fig. 2.17 shows only a particularly simple situation
(1D case, one valence electron only). In general, the band structure of solid states
is much more complicated, and depending on how many electrons are available to
fill the bands, it provides the basis for such different materials as metals, isolators or
semiconductors.
134 2 Elements of Quantum Mechanics and the H Atom
(a) W W (b)
3rd 2nd 1st 1st 2nd 3rd band gap
band gap
band gap
k k
- 3 a
2 - 2 3
a -
a a 0
a a a
Fig. 2.17 Emergence of energy bands in a periodic system (lattice constant a) and band gaps.
(a) Allowed energies ( ) with distortions at the boundaries of the 1st, 2nd, and 3rd B RILLOUIN
zone in comparison with the free particle ( ) as a function of k. (b) Projection onto the first
B RILLOUIN zone
Section summary
Dispersion relations describe how the energy W of the system depends on
the wave vector k. The most simple cases are (a) the photon (a massless
particle) W = ck, and (b) an electron (a particle with mass) moving freely
W = 2 k 2 /(2me ).
The influence of an average potential can be accounted for by an effective
mass, replacing me which even may depend on the direction of the electrons
momentum.
B LOCH waves, (r) exp(ikr)uk (r), are constructed to include the period-
icity of the lattice in the function uk (r) = uk (r + T ).
References
B ORN , M.: 1927. Das Adiabatenprinzip in der Quantenmechanik. Z. Phys., 40, 167192.
B ORN , M.: 1954. The N OBEL prize in physics: for his fundamental research in quantum
mechanics, especially for his statistical interpretation of the wave function, Stockholm.
http://nobelprize.org/nobel_prizes/physics/laureates/1954/.
E ISAMAN , M. D., J. FAN, A. M IGDALL and S. V. P OLYAKOV: 2011. Invited review article:
single-photon sources and detectors. Rev. Sci. Instrum., 82, 071101.
H EISENBERG , W. K.: 1932. The N OBEL prize in physics: in recognition of the great merits
of his theoretical and experimental investigations on the conduction of electricity by gases,
Stockholm. http://www.nobelprize.org/nobel_prizes/physics/laureates/1932/.
ISO 21348: 2007. Space environment (natural and artificial) Process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
NIST: 2010. Reference on constants, units, and uncertainties, NIST. http://physics.nist.gov/cuu/
Constants/, accessed: 7 Jan 2014.
S CHRDINGER , E. and P. A. M. D IRAC: 1933. The N OBEL prize in physics: for the discov-
ery of new productive forms of atomic theory, Stockholm. http://nobelprize.org/nobel_prizes/
physics/laureates/1933/.
Periodic System and Removal of Degeneracy
3
Overview
This is a quite compact and important chapter. The attentive reader should
be able to work through it rather quickly after having refreshed her or his
basic knowledge in the previous two chapters. We shall make here the first
steps to generalize the methods successfully applied for the H atom, and
allow for an interaction potential which is no longer strictly proportional
to 1/r. This forms the basis for understanding the physics behind the peri-
odic system of the elements which is summarized in Sect. 3.1. The most sim-
ple multi-electron systems are the alkali atoms. Their energy levels are dis-
cussed in Sect. 3.2 phenomenologically and analyzed qualitatively and quan-
titatively, briefly explaining quantum defect theory. In Sect. 3.3 we introduce
time independent perturbation theory a tool that will be used later on quite
frequently and illustrate it by way of example for the energies of alkali
atoms.
The periodic system provides the structural basis for our understanding of atoms
and molecules. Thus, albeit fairly well known, it appears worthwhile to summarize
here the underlying concepts, observations and definitions.
The theory of the hydrogen atom as a prototype atom contains already all ingredi-
ents needed for understanding of how the more complex atoms are built up. The
periodic system of the elements follows naturally from the so called Aufbau princi-
ple (from German Aufbau = build-up): to a first order approximation one treats
the N electrons of an atom (with a nuclear charge Z) as independent of each other
and assumes that their respective wave functions look very similar to those of the
electron in a hydrogen atom. It turns out that this approach serves astonishingly well
for a first guess. Of course one has to modify the potential that each individual elec-
tron sees with all the other electrons around it clearly cannot simply be Z/r:
the nuclear charge will partially be screened off by (N 1) electrons.
We shall discuss this screening and its consequences in detail later. Here we
simply note, that each electron (numbered as i = 1, 2, . . . , N ) is characterized by a
set of quantum numbers:
They correspond to the quantum numbers of the electron in an H atom. The entirety
of quantum numbers for all electrons of an atom in a particular state is called its
In its best known form the PAULI exclusion principle (short PAULI principle,
N OBEL prize 1945) states that no two electrons in the same atom can have the
same four quantum numbers, briefly
More generally, no two identical fermions (i.e. particles with half integer spin
s = 1/2, 3/2, etc.) can be in the same quantum state. This is an empirically con-
firmed property of fermions. In Sect. 7.3.1 we shall discuss and use the quantum
mechanical formulation: The total wave function of identical fermionsis antisym-
metric in respect of the exchange of two particles. We shall see there that both for-
mulations are completely equivalent.
3.1 Shell Structure of Atoms and the Periodic System 139
One defines so called electron shells which are occupied by all electrons of an atom
with the same principle quantum number n. The shells are denoted by the letters K,
L, M, N, . . . , as summarized in Table 3.1. As already noted in Sect. 2.5.3, the letters
s, p, d, f, g stand for the angular momentum quantum numbers = 0, 1, 2, 3, 4. The
number of states in a shell with the principle quantum number n is 2n2 (including
spin states with ms = 1/2) all of which are degenerate in the H atom case.
According to the PAULI principle, each state may be occupied with no more than
one electron. The electron configuration of an atom is written in compact form as
illustrated in Table 3.2 for the ground state of some light atoms. For larger atoms
(e.g. Na) one summarizes the inner shells by the symbol of the next smaller rare gas
atom in brackets [ ].
The Aufbau principle of the periodic system assumes that these shells are sub-
sequently filled as the number of electrons increases. Figure 3.1 illustrates the shell
filling scheme up to neon graphically.
140 3 Periodic System and Removal of Degeneracy
s p s p
L L
K H K He s p s p
L L L L
K Li K Be K B K C
L L L L
K N K O K F K Ne
Fig. 3.1 How the K and L shell of the periodic system of elements are filled with electrons. Arrows
indicate the spin projection (1/2) of the electrons
Table 3.3 gives a complete overview of the periodic table of elements, indicating by
colour shadings how the electron shells are filled. The elements within one group
(vertical columns) have equivalent outer shell electrons (differing only by n) and
typically show corresponding similarities in their chemical and physical properties.
Each period (horizontal row) corresponds to the filling of one particular shell and
comprises elements with different electron configuration and usually different prop-
erties. Exceptions are the d block of transition elements, the Lanthanides and the
Actinides, for which deferred inner electron shell filling occurs. The radioactive el-
ements of the 7th period have been (except for Fr and Ra) generated artificially
at heavy ion storage rings (in very low concentration). They typically have mean
lifetimes of only seconds to minutes, and little is known about their electron config-
uration or even about their physical and chemical properties. Nevertheless, they are
interesting objects for fundamental studies.
We refer the reader to numerous, well linked presentations in the internet, such
as W IKIPEDIA CONTRIBUTORS (2014). The source par excellence for further infor-
mation is, however, NIST (2011) from where a host of tabulated properties of the
elements including all available spectroscopic data can be retrieved. Quite instruc-
tive is also the animation from the U NIVERSITY OF C OLORADO (2000).
Table 3.3 Periodic system of the elements. Left on the top of the elements the atomic number is given, below each element the configuration of the last
3.1
built in electron is shown. The filling of different shell is marked by colour shadings: s electrons and p electrons determine the main groups, adding
d electrons leads to the subgroups in the middle of the periodic table that contain all transition metals. The deferred addition of 4f and 5f electrons
occurs for Lanthanides and Actinides
Shell Structure of Atoms and the Periodic System
141
142 3 Periodic System and Removal of Degeneracy
Be Mg Zn Cd Yb Hg
25
3d 104s 24p 6
[Ar]3d 104s 2
[Kr]4d 105s 2
[Xe]4f 146s 2
20
3s 2
WEA and WI / eV
15
10
0
Z= 0 20 40 60 80
He Ne Ar Kr Xe Rn
Fig. 3.2 Ionization potentials WI (red) and electron affinities WEA (grey) of the atoms as a func-
tion of nuclear charge Z. The full vertical lines indicate shell closures, the dashed lines the closure
of subshells, corresponding to the electron configuration given
The essential basis for the experimentally well confirmed shell structure of the
elements is (beyond the PAULI principle) the astonishingly good model of quasi-
independent electrons in the so called central field approximation. We shall explain
this in detail in Chap. 10 where a profound understanding of the periodic system
will emerge.
The ionization potentials (IP) WI for all natural elements of the periodic system
are presented in Fig. 3.2. The diagram illustrates impressively the shell nature of
atomic structure: The IPs assume very pronounced maxima for the rare gas atoms
(He, Ne, Ar, Kr, Xe, Rn) that are characterized by closed shell. They are chemically
particularly inert. Smaller maxima are also in between, whenever a subshell for one
particular is completely filled.
A complementary behaviour is seen in the electron affinities WEA (grey line),
i.e. the negative of the binding energies, which may be gained by the formation of
a negative ion (anions) energies. They correspond to the negative binding energies
of the electrons in the anion (WB = WEA ) and are particularly high, if the corre-
sponding neutral atom has just one vacancy in the outer shell for the electron to be
accepted which holds for the H atom and the halogens (F, Cl, . . .) and becomes
zero for the rare gases with their closed shells: no stable rare gas anions are known
to exist (see also Sect. 6.4.5, Vol. 2).
The atomic radii also support the shell structure of atoms very clearly. They are
plotted in Figs. 3.3 and 3.4 as functions of nuclear charge Z. Of course, the term
3.1 Shell Structure of Atoms and the Periodic System 143
r / nm
methods: W IGNER -S EITZ
radii (red) and VAN DER
WAALS radii (black) 0.2
0.1
He Ne Ar Kr Xe
Z
0.2
0.1
0
He Ne Ar Kr Xe
Z
atomic radius is not very well defined as we have learned in Chap. 2 the size of
an atom is characterized by the probability to find its electrons in a certain distance
of the nucleus, and a limit cannot be defined uniquely. One may e.g. use the so called
W IGNER -S EITZ radius, rWS . That is the radius of a sphere of the same volume that
the atom occupies on average in the condensed phase if such a phase exists for that
particular atom. From particle density N or mass density , relative atomic mass Mr
and AVOGADRO number NA one finds
3 3 3Mr
rWS = = 3 . (3.4)
4N 4NA
A similar quantity is the VAN DER WAALS radius, that gives the distance up to
which chemically not bound atoms can approach each other (for a more precise def-
inition see Sect. 3.2.6, Vol. 2). Both quantities are shown in Fig. 3.3 for the elements
H to Ba.
Alternatively, Fig. 3.4 shows calculated atomic radii. They are extracted e.g. from
quantum mechanical expectation values, as discussed in Sect. 2.6.11 relying on
good computations of the respective atomic wave functions. An empirical deter-
mination is based on so called covalent radii that are derived from a set of well
known binding lengths of molecules, preferably diameters. By comparison with
144 3 Periodic System and Removal of Degeneracy
other molecules one may improve these estimates without big efforts as also shown
in Fig. 3.4 (asterisks). In spite of all ambiguities in defining the term atomic radius
one recognizes very clearly the general trends as a function of atomic number. Rare
gases typically have the smallest radii, alkali metals the largest: electrons in closed
shells all see essentially the same, high charge, while in the outermost shell of an
alkali atom this charge is strongly screened for the one valence electron due to the
core electrons.
We shall discuss this in detail in the following sections of the present chapter.
Before doing so we mention one more little difference between Fig. 3.3 and Fig. 3.4:
while the minima of the atomic radii for rare gases are quite pronounced in the
latter case, such clear cut minima cannot be identified for the VAN DER WAALS and
W IGNER -S EITZ radii. Obviously, in addition to the electron density the influence
of polarizability plays here an important role.
Section summary
According to the PAULI principle each identical fermion in a quantum system
must differ by at least one quantum number.
The periodic system of the elements is based on the PAULI principle. The
PAULI principle prevents electrons from falling all into the energetically low-
est level and thus ensures the variety of the chemical elements. One may say
that it is behind the mystery of life.
The atomic shells are denominated by K, L, M, N, . . . according to their prin-
ciple quantum number n = 1, 2, 3, 4, . . . , respectively. Each shell may contain
up to 2n2 electrons. The Aufbau principle (building up principle) of the peri-
odic system says that electrons fill states with quantum numbers n mms one
after the other (essentially) in numerical sequence.
Chemical and physical properties of the elements depend crucially on the
number of electrons in the outermost shell. Completely filled shells corre-
spond to the rare gases; within each main period they have the highest ion-
ization potential and the smallest radius (as represented e.g. by the W IGNER -
S EIZ radius Eq. (3.4)).
The most simple cases of multi-electron systems are those where one or several
inner shells are completely filled with electrons (closed shells) and only one electron
is found in the outermost shell. A look at the periodic system in Table 3.3 identifies
these as the elements of the 1st group, the alkali atoms and the alkali like ions. Their
electronic ground state configuration is {[Rg]ns}, where [Rg] stands for a rare gas
configuration of the atomic core, e.g. Li: {[He]2s} Na: {[Ne]3s} K: {[Ar]4s}, etc.
In the following we shall focus on the active electron in the respective new shell,
filled only with this one electron of particular interest. It is called valence electron
(in the original German literature also Leuchtelektron = lighting electron).
3.2 Quasi-One-Electron System 145
497
transitions (double arrows) 3p 3d
.2
are indicated with
.2
wavelengths given in nm. The 3s
460
3
81
.3
diagram has been generated
2.6
610
from the NIST data bank. The
term energies are related to
2
2
the respective (negative)
323.
2p
binding energies Wn of the
valence electron by
0.8
Wn W2s , with WI = W2s 1
67
being the IP of the system
0
2s angular momentum
The most detailed and precise information comes again from spectroscopy. The
spectra and hence the term schemes of the alkalis turn out to be similar to those
of the H atom with its 1s ground state electron with two important modifications:
(i) the active valence electron is now initially in an ns state with n > 1 and the inner
shells are filled (thus no VUV radiation is observed) and (ii) the degeneracy is
now removed (which makes the spectra richer, i.e. more complicated to interpret,
than that of the H atom).
As a characteristic example Fig. 3.5 shows the term diagram of Li(1s)2 n as
derived from a host of spectroscopic information collected by generations of spec-
troscopists. This type of term scheme indicating the possible transitions is usually
called G ROTRIAN diagram.
The general relation between the energies Wn of stationary states and spectral
lines observed (in wavenumbers) is again
1 1
n n = = (Wn Wn ), (3.5)
hc
now, however, with different energies for each .1
The basics of emission, absorption and fluorescence spectroscopy will be intro-
duced briefly in Sect. 4.2.2 and a variety of more refined modern methods will be
described later, e.g. in Sect. 6.1. Here we only give a survey over the collected re-
1 Strictly speaking, one should even introduce an additional quantum number j for the total angular
momentum, which becomes increasingly relevant as Z gets larger. We have already familiarized
ourselves with it in the context of the S TERN -G ERLACH experiment in Sect. 1.9.5 and shall come
back to it in detail in Sect. 6.2.5.
146 3 Periodic System and Removal of Degeneracy
Wn / E h H Li Na K Rb
ionization continuum
0
n=8
- 0.020
- 0.031 n=4 6p 4f 5f
4s 4p 4d 4f
4s 4p
4d 4f 5s
5p 4d 4f 6s 4d 7s 7p 5d
- 0.056 n=3 4p 3d 5p 6p
3s 3p 3d 3p 3d 6s
3s 4s 5s
4p 5p
- 0.125 n=2 3p
2s 2p 2p
4s 5s
2s 3s
- 0.500 n=1
1s
Fig. 3.6 Overview on the term energies of the alkali atoms in comparison to the H atom. Charac-
teristic is the removal of degeneracy and the decrease of the energies in respect of the H atom for
small angular momenta
sults for the alkalis and refer the interested reader to the spectroscopic data bank of
K RAMIDA et al. (2013) which we have already mentioned.
A comparison of the energies of the H atom with those of all alkali atoms (va-
lence electron) shows Fig. 3.6. The characteristic removal of degeneracy lets the
term energies of the alkali valence electrons always lie below those of the H atom,
Wn < Eh /(2n2 ), owing to the higher nuclear charge Z. However, this lowering of
energy gets smaller as the orbital angular momentum gets larger. Figure 3.6 clearly
documents that Wns < Wnp < Wnd < Wnf . For the nf terms the energies of the al-
kali atoms are practically identical to those in atomic hydrogen. We shall understand
in a moment why that is so.
Eh
Wn = with n = n . (3.6)
2n2
The parameter is called quantum defect and n the effective quantum number. For
the moment this is simply a parameter, determined experimentally, which allows
one to order the spectroscopic data in a systematic manner.
A comparison of experimental data shows that strongly depends on the orbital
angular momentum and upon closer inspection also slightly on the principle
3.2 Quasi-One-Electron System 147
(n, )
ns
function of binding energy 1.36
Wn for orbital angular
momenta = 0, 1, 2, 3 and 4. 1.35
Full black circles are 0.89
experimental data taken from np
0.87
K RAMIDA et al. (2013), open 0.85
circles are denoted there as - 4 - 2 - 0.8 - 0.4 0
determined by interpolation
or extrapolation . . . or by 0.014
semi-empirical
calculations. . . . Full lines nd
represent linear least square 0.012
fits to the measured data.
Note the break in energy 0.010
scale for the ns and np series, 0.0015
and the different scale for the nf
nd, nf and ng series 0.0010
0.0004 ng
0.0000
- 1.6 - 1.2 - 0.8 - 0.4 0
Wn / eV
quantum number n. Figure 3.7 illustrates this for Na as example. Obviously, for
large n the quantum defect approaches a constant value that depends only on . We
shall come back to the theoretical interpretation of in the framework of the so
called quantum defect theory (QDT) in Sect. 3.2.6. For a quantitative comparison
one fits the experimentally determined term energies, i.e. the excitation energies
in respect of the ground state, by an extended RYDBERG -R ITZ formula (see e.g.
W EBER and S ANSONETTI 1987):
Eh
term energy Wn = Wn0 0 (3.7)
2(n (n, ))2
with (n, ) = + B/(n )2 + C/(n )4 + (3.8)
or (n, ) + DWn . (3.9)
Here Wn0 0 is the ground state energy, which is related to the IP by WI = Wn0 0 .
The quality of present day spectroscopic data (K RAMIDA et al. 2013) allows a very
precise determination of the parameters. The thus derived quantum defects of all
alkali atoms for large n are summarized in Table 3.4 and Fig. 3.8.
One sees very clearly that the quantum defect increases strongly with the nuclear
charge Z and decreases with increasing orbital angular momentum. We also want
to emphasize that for large n (small |Wn |) decreases indeed linearly with Wn
(not shown here) according to (3.9) which allows for (i) a determination of WI with
spectroscopic accuracy and (ii) analytic continuation of (n, ) (W, ) into the
continuum where W > 0. In the following we shall try to understand these findings
148 3 Periodic System and Removal of Degeneracy
(Z-1)e0
qualitatively and quantitatively by looking at the potentials and wave functions, and
finally we shall introduce QDT which explains that the quantum defect is not just
some kind of fudge parameter.
A quite reasonable model for the alkali atoms assumes that the Z 1 core elec-
trons (which fill the inner shells of the core ion completely) do influence the energy
terms of the one outermost electron (the valence electron) only by screening the
pure C OULOMB potential of the nucleus. As illustrated in Fig. 3.9, we consider an
N electron problem (nuclear charge Z = N ). At large distances, the valence elec-
3.2 Quasi-One-Electron System 149
2 4 6
0
-1/r VS (r) r / a0
-2
VS (r ) / E h
-4
-6 - Z/r
-8
-10
Fig. 3.10 Effective potential VS (r) in which the valence electron of an alkali atom moves (red),
compared to a fully screened C OULOMB potential 1/r and the unscreened potential Z/r that
the electron experiences close to the nuclear charge Z (here with Z = 11 representing the Na
atom). The grey shaded area indicates the radius of the ionic core
tron sees only the screened nuclear charge 1e and experiences something about
the nucleus only if it dives into the ionic core. One may treat this simplified problem
with nearly the same methods as the H atom except that we no longer have a pure
1/r potential.
Rather we have to solve the problem for a screened Z/r potential (in a.u.)
Z/r r 0
VS (r) = 1/r + VC (r) in between (3.10)
1/r r
where the ionic core is accounted for by a suitable, smooth potential, in the simplest
case e.g.
VC (r) = (Z 1) exp(r/rS )/r (3.11)
as illustrated schematically in Fig. 3.10.
n m (r) = Rn (r)Y m (, ).
Again we substitute un (r) = rRn (r) and obtain in complete analogy to (2.113) the
radial differential equation (in a.u.)
150 3 Periodic System and Removal of Degeneracy
d2 un ( + 1)
+ 2 W n VS (r) + un (r) = 0 (3.13)
dr 2 2r 2
with the important difference that VS (r) is no longer the pure C OULOMB potential
but rather a suitably screened one as described by (3.10).
The task at hand now is to identify from the infinite manifold of solutions for
(3.13) those which are physically meaningful. They have to reproduce the correct
asymptotic behaviour both for large and small distances r. Since we have derived
the corresponding relations in the case of the H atoms without any reference to
the interaction potential, (2.119) and (2.120) must be valid in the present case too.
r the radial wave function has to be damped exponentially, un
Thus, for
exp( 2|Wn |r), and Wn is the sought-after energy of the stationary states of the
system. On the other hand, for r 0 the radial function has to follow un r +1 .
This very behaviour allows us to explain the experimentally observed depen-
dence of the quantum defect (Table 3.4): according to (2.125) the probability to find
an electron at a distance r from the nucleus is w(r) = 4r 2 Rn 2 (r) = 4u2 (r).
n
Specifically for small distances it is thus r 2 +2 . Hence, the higher its angular mo-
mentum, the less the electron is influenced by the ionic core. For very large the
wave function will be essentially hydrogen like, and the quantum defect correspond-
ingly small, just as documented in Sect. 3.2.2.
For a truly quantitative analysis of wave functions and eigenenergies we have
to integrate (3.13) explicitly. Even though the radial S CHRDINGER equation for
a screened potential of the type shown in Fig. 3.10 cannot be solved analytically,
numerical integration of (3.13) may be done today on any PC without problems.
A number of robust and simple procedures are available, e.g. the often used RUNGE -
K UTTA method. Typically one integrates for a given trial energy Wn from the inside
outwards as well as from the outside inwards, starting with the asymptotic forms just
discussed. One accepts only such solutions that can be matched continuously and
differentiable and by variation of Wn one thus obtains the discrete eigenstates and
eigenenergies of the system.
Of course, the result of such calculations can only be as good as the potential
VS (r) which so far we have only described qualitatively. There are a number of
useful approximations to estimate VS (r). In the most simple approach one guesses a
parametric form which is fitted such that the experimentally observed energies of a
few states are reproduced exactly. With the thus determined empirical core potential
one may derive the wave functions for any state of the system and compute other
quantities, such as the energy terms, optical transition probabilities, polarizabilities
and so on.
0 3s
4 r 2Rn (r) / a 0
0.2
1 2s 3p
2
0
0.1
5 3d
1s
0 0.0
0 1 2 0 5 10 15 r / a0
Fig. 3.11 Radial electron probability distributions w(R) in Na for (a) the core electron orbitals
and (b) orbitals of the valence electron (alternatively in the 3s ground state and in the 3p and 3d
excited states). Also shown is the cumulated radial electron probability in the Na+ core (Z = 11,
closed [Ne] shell), down-scaled by a factor of 40. Grey shaded is the radius of the ion core (1.8a0 ).
The distributions shown by red lines have been calculated ab initio using a finite difference atomic
structure code, FDAlin, conveniently provided by S CHUMACHER (2011). The black dashed and
dash dotted lines are derived from QDT and are not expected to be valid inside the ionic core (see
Sect. 3.2.6). Note in particular the excellent agreement for 3s
2 This holds at least for the smaller alkali atoms where relativistic effects play a minor role and
quantitatively that the valence electron stays predominantly outside the core. This
holds a fortiori for excited states, here very clearly seen for the 3d state the orbital:
its radial wave function (R3d (r) r 2 for small r) leads to very little probability for
finding the electron inside the core as Fig. 3.11 illustrates.
This explains again convincingly why the quantum defect decreases so rapidly
with as illustrated in Fig. 3.8. It also clarifies the overall striking similarity of the
alkali spectra with those of the hydrogen atom which we have seen in Fig. 3.6. The
nf electrons, for which at the origin R4f (r) r 3 and thus w(r) r 8 , practically
never come close to the nucleus and the quantum defect can essentially be neglected
as for all higher values of (see Fig. 3.8).
We recall again: degeneracy observed in H and H like ions is just a very specific
consequence of the pure C OULOMB potential. Deviations from the C OULOMB po-
tential lead to different energies for different , i.e. to a removal of this degeneracy.
The deviations are the larger the more the electron feels of the ionic core. For very
large the situation is practically identical to the pure 1/r potential in the H atom
case.
Even though in principle the full quantum mechanical problem of bound states in
small alkali atoms may be treated numerically, it is useful to return once more to the
model of a single electron in an effective potential. So far, we have not yet under-
stood why the extended RYDBERG -R ITZ formula works so well with essentially
constant quantum defect (3.8) for large n as documented by Fig. 3.7. This remark-
able observation hints at some genuine physical background, even though the ef-
fective quantum number n and the quantum defect were introduced in (3.6) just
as useful empiric parameters. It turns out that understanding this physical origin of
the quantum defect allows one to extract useful concepts and further quantitative
information.
The roots of the so called quantum defect theory (QDT) go back to the early
days of quantum mechanics and H ARTREE (1928), while between 1950 and 1990
S EATON (1983), FANO (1986), J UNGEN (1996) and their students have developed
it into a powerful theoretical framework and extended it to multichannel problems
(MQDT). It has been applied successfully to calculate oscillator strengths for atomic
and molecular transitions including high lying RYDBERG states, to determine accu-
rate ionization potentials and photoionization cross sections (see also Sect. 5.5), to
understand autoionizing series and perturbations in multi-electron spectra, as well
as in electron ion scattering theory and is used even in solid state structure the-
ory.
The key to QDT is that for large r far outside the atomic core electronic wave
functions evolve in a pure C OULOMB potential and may be derived as analytic func-
tions also for non-integer quantum numbers with the correct damping at large r
(in a.u.)
3.2 Quasi-One-Electron System 153
un (r) r n exp 2|Wn |r = r n exp r/n , (3.14)
r
3 The r n factor used here in contrast to (2.119) can improve the convergence.
4 This is a rather crude choice. It leads, however, to qualitatively correct wave functions. The radius
of the ionic core for Na+ is typically given in the literature as 0.095 nm = 1.8a0 . At this distance
this screened potential VS (r) is about 1.1/r.
154 3 Periodic System and Removal of Degeneracy
(a) (c)
0.6 0.2
H 18s
r Rn / arb. un.
0 0
H 20s
-0.2
0 10 20
50 H 18s H 20s
-1.0
Na+ core
(b) (d) Na 18s
0.6 0.2
Na 20s
0 0
-0.2
0 10 20
50 Na 18s Na 20s
-1.0 r / a0
0 500 1000
Fig. 3.12 Illustration of the quantum defect as phase shift = for Na in the 18s and
20s states: (a) The radial wave functions rRn (r) for a pure C OULOMB potential (H atom) are
compared (d) to those in an Na+ pseudopotential VS (r) (see Fig. 3.10). All wave functions are
normalized to their minimum, which is set to 1. The blow ups of the r scale (c) and (d), respec-
tively, illustrate the strong influence of the Na+ ionic core (grey shaded area) onto the Na wave
functions. It leads to a shift of the rRn maxima in respect of the H atom; this shift is directly
related to the quantum defect (n, )
deeper attractive potential VS . While the Na 18s and 20s wave functions are nearly
identical in this small r range (apart from their magnitude owing to normalization),
they differ dramatically from their counterparts in the H atom. Outside the core the
potential is purely C OULOMBic for both Na and H hence the two pairs of wave
function differ essentially only by this phase shift .
QDT treats the outside wave functions as a linear superposition of two pure
C OULOMB functions. The mathematics involved is not completely trivial and
we can only outline here just a few basics, following the excellent review of
S EATON (1983) and his concise mathematical summary from 2002. The radial
S CHRDINGER equation (3.13) is solved in two parts:
"
FI (r) for 0 r < r0 with VS (r)
u(r) = (3.15)
FII (r) for r0 r with 1/r.
(If alkali like ions are discussed with a remaining charge ZC of the ionic core, the
1/r potential has to be replaced by ZC /r.)
3.2 Quasi-One-Electron System 155
Since this is an ODE of 2nd order it has two linearly independent sets of C OULOMB
functions as solutions for which one finds various forms in the literature. For QDT
S EATON (1983) defines s(, ; r) as the regular solution, and c(, ; r) as the irreg-
ular solution. When treating the H atom in Sect. 2.6.5 we have considered only the
regular solution un s(n , ; r) for n = 1/n2 (n being an integer); asymptoti-
cally they are r +1 for r 0 and r n exp(r/n) for r . It is important to
note that for all other values of the regular solution s(n , ; r) diverges at large r
hence the n give the eigenvalues of the H problem.
The irregular solutions c(, ; r) are r at the origin so that any radial
wave function R(r) rc(, ; r) would diverge hence they were considered non-
physical for the H problem. It is interesting to note, however, that for large r and
n = 1/(n + 1/2)2 (and only for these) they are damped r n exp(r/n).
Hence, it is plausible as shown already by H ARTREE (1928) that at arbitrary
energies = 1/n2 a physically meaningful solution of the problem for r > r0 is:
FII (r) = cos n s(, ; r) + sin n c(, ; r)
(3.18)
= (1)n cos()s(, ; r) + sin()c(, ; r) .
Bound States
Specifically for bound states = 1/(n )2 < 0, S EATON shows
sin(n ) n3 1/2
s(, ; r) = (1) cos n K
(2)1/2 K 2
(3.19)
n3 1/2
cos(n )
c(, ; r) = (1) + sin n K
(2)1/2 K 2
Since n = (n+1 n ) 2/n3 in the limit of large n, the bound state functions
s(n , ; r) are said to be normalized per unit energy in this limit.
Since the component of the solutions (3.19) diverges for large r it is not suit-
able for the description of any realistic physical wave function. And indeed, when
inserting (3.19) with (3.21) into the general solution (3.18) the terms cancel and
one obtains for r > r0 simply:
3
1/2
n 2r
FII n , ; r = (1) K n , Wn , +1/2 . (3.24)
2 n
This is the sought-after general wave function in a pure C OULOMB potential outside
the atomic core for arbitrary energy! The W HITTAKER functions are included in
advanced mathematical programmes, e.g. in Mathematica, and we have computed
these functions for the 18s and 20s case of Na with n = n 1.3848. The result
is completely identical to the wave functions shown in Fig. 3.12(b) and (d) within
the limits one could see in the graphs except inside the ionic core where it is not
expected to be valid.
3.2 Quasi-One-Electron System 157
Continuum States
We have already indicated that the behaviour of continuum wave functions may be
gleaned from analytic continuation of the quantum defect (n, ) (, ) for =
k 2 > 0. What we have vaguely addressed as phase shift between the oscillations
of the bound state wave functions will then become indeed a genuine phase shift
between a pure outgoing C OULOMB wave and the continuum wave function dis-
torted by the screened Z/r potential of the ionic core. Here too, (3.18) applies for
the outside part of the wave function, FII , and at a suitable r0 one has to join FI
and FII . Again various forms for the regular and irregular pure C OULOMB contin-
uum wave functions are used. According to S EATON (2002), Eqs. (86)(90) and
(113)(118)
' (
s(, ; r) Im exp i( /2) Wi/km +1/2 (2ikr)
' ( (3.27)
c(, ; r) Re exp i( /2) Wi/km +1/2 (2ikr)
158 3 Periodic System and Removal of Degeneracy
0.4
s (, ;r)
0.0
c( , ;r)
-0.4
0 5 10
r
Fig. 3.13 C OULOMB continuum s waves ( = 0) at = 4 for attractive potential: regular solution
(red line, zero at origin), irregular (black line, finite at origin) adapted from S EATON (2002). The
asymptotic behaviour (3.30) (pink and grey dashed lines, respectively) gives a nearly perfect match
to the exact solutions already at r 3
where Wi/km +1/2 (2ikr) is again a W HITTAKER W function, this time to a com-
plex argument, and for the presently considered attractive C OULOMB potential we
have r > 0. The C OULOMB phase shift
decreases rapidly with k and corresponds to the phase difference of the C OULOMB
in respect of a free spherical wave (not to be confused with the phase shift () due
to the ionic core potential VS being different from purely C OULOMB).
We have suppressed here a discussion on the important question of normalization
and simply note that
s(1 , ; r)s(2 , ; r)dr = (1 2 ), (3.29)
0
i.e. s(1 , ; r) is normalized to scale (see Appendix J). Again, the regular solution
goes to 0 at the origin while the irregular solution stays finite or diverges. For large r
these wave functions are essentially outgoing spherical waves apart from a slowly
varying logarithmic phase shift characteristic for the C OULOMB potential (see (J.8))
i.e. they behave essentially like sine and cosine functions:
2 2
lim s(, ; r) = sin and lim c(, ; r) = cos (3.30)
r k r k
1
with = kr + ln(2kr) /2 + (k, ). (3.31)
k
Figure 3.13 illustrates these continuum functions for = 0 at a moderate energy =
4 or k = 2 where 0 0.078 102 . The exact solutions (adapted from S EATON
(2002), Fig. 1) are apparently matched rather perfectly by their asymptotic form
(3.30) already at surprisingly small values of r 3.
We now come back to the general QDT theme of constructing a wave function
for r > r0 , this time in the continuum, using the analytic continuation () de-
rived from bound state quantum defects. Being interested in large r we insert the
3.2 Quasi-One-Electron System 159
Z 2 (Z qs )2
Wn = 2
Eh = Eh , (3.33)
2n 2n2
or in wavenumbers
(Z qs )2 (Z qs )2
n = E h /2 = R
hcn2 n2
2|Wn | n Z Z qs
= = = . (3.34)
Eh R n n
Plots of the square roots of the energies as a function Z are called M OSLEY di-
agram and should give according to (3.34) a straight
line. The screening parameter
qs is then obtained from the axis intercept of 2|Wn |/Eh where Z qs = 0. Full
screening of the C OULOMB potential would imply qs Z 1. By way of exam-
ple we discuss the energy terms for Na and Na-like ions with data taken from the
K RAMIDA et al. (2013) data bank. The series starts with Z = 11 (Na) and continues
with Mg, Al, Si, P, S, Cl, etc.
Figure 3.34 presents the data for the ns, np and nd terms with n = 3, 4 and 5.
The graph documents that indeed the square root of the energies follows to a very
good approximation a straight line as a function of the nuclear charge Z as predicted
by (3.34). Admittedly we still need two parameters, qs and n n to fit a whole
series well. As expected, for states with the highest orbital angular momentum the
160 3 Periodic System and Removal of Degeneracy
4
3s
3p
Z-10.03
3d
3 2.91
4s
2Wn / E h 2 4p
Z-10.02 4d 5s
3.90 5p
1 5d
Z-10.02
4.97
0
10 11 12 13 14 15 16 17 18 19 20 Z
Na I Al III P V Cl VII Ca X
Fig. 3.14 M OSLEY diagram for Na like ions and different n states. Following spectroscopic
traditions we designate the degree of ionization with roman numbers: I for neutral atoms, II for
singly ionized, III for doubly ionized atoms, etc. Symbols correspond to the experimental data
according to K RAMIDA et al. (2013), the lines are fits following essentially (3.34). Nearly perfect
screening (qs = 10) is shown for the d states
energies agree best with the straight lines as validated in Fig. 3.14 particularly well
for the 3d, 4d and 5d series. The screening in all three cases is nearly perfect (qs =
10) and die slope of the lines is almost ideally 1/n. For the p and s states the
screening is obviously not perfect: in particular for the 3s states the screening is
only just above qs = 9, i.e. the electron still sees effectively nearly two charges of
the nucleus. In any case, the surprisingly good agreement of experimental data with
the prediction (3.34) supports very impressively the model of a quasi-one-electron
system over a whole isoelectronic series.
For higher Z the lines typically fall into the X-ray region. M OSLEY diagrams are
therefore also used successfully to characterize X-ray spectra from inner shells (see
Sect. 10.5.2).
Section summary
The spectra of alkali atoms resemble those of the H atom in the VIS except
that now degeneracy is removed and consequently several series of lines are
observed for each principle quantum number n.
Thus, a good concept to understand these spectra is a quasi-one-electron
model where all but one active electron (the valence electron) constitute the
closed shell of the ionic core. They (partially) screen the charge of the nucleus
but otherwise do not participate in standard spectra.
The energy terms of the alkali atoms (and alkali like ions) depend thus
on n and of the valence electron and may be described by Wn =
Eh /[2(n )2 ]. The quantum defect (n, ) depends strongly on and
weakly on n. For large n it becomes nearly independent it. The higher , the
smaller is , i.e. the closer the energy levels are to those of the H atom (and
H like ions, respectively).
3.3 Perturbation Theory for Stationary Problems 161
These findings are explained by the low probability of the valance electron to
be found close to the nucleus, being r 2 +2 for r 0.
Quantum defect theory (QDT) gives a very useful analytic foundation to the
observed behaviour of quantum defects: it turns out that the quantum defect
relates to the phase shift = between a pure C OULOMB wave function
(H atom) and the wave function for the alkalies.
M OSLEY diagrams summarize the spectral lines of a whole series of ions with
equal electron configuration as a function of the nuclear charge Z. Plotting
2|Wn |/Eh vs. Z leads to essentially straight lines.
is assumed to be solved. Let the Hamiltonian for the new problem be given by
=H
H (r, p
0 + U ), (3.36)
162 3 Periodic System and Removal of Degeneracy
with the perturbation operator U (in the most simple case just a scalar potential).
For perturbation theory to be applicable we have to assume that for the averaged per-
| |H
turbation |U 0
| holds. We thus write the S CHRDINGER equation (3.12)
in the form
0 + U
(H )k = Wk k , (3.37)
which has to be solved by the perturbation approach. One then expands the ener-
gies and wave functions into a series in terms of small quantities Wk and k
assuming that U is also small on the order of . The formally correct procedure is
to compare quantities of the same power in the smallness parameter and thus to
obtain correction terms of the order , 2 , etc. These are the corrections in 1st, 2nd,
etc. order perturbation theory. The key quantities characterizing the perturbation are
the matrix elements:
(0) (0)
Uj k = j U k = j(0) (r)U (r, p)k(0) (r)d3 r. (3.38)
We abbreviate the procedure just outlined somewhat. Since the eigenfunctions i(0)
of the unperturbed Hamiltonian H 0 , derived from (3.35), form a complete, orthonor-
mal basis set, any function may be constructed as linear superposition from them
including the solution of (3.37). Thus, we use the perturbation ansatz
(0)
k = ai i with |ak | 1 and |ai | 1 for i = k. (3.39)
i
The conditions |ak | 1 and |ai | 1 are crucial; they establish the essence of the
perturbative approach.
Introducing (3.39) into the S CHRDINGER equation (3.37) we obtain
0 + U
[H Wk ] ai (0) = 0
i
i
0
H
(0) ai (0) Wk ai (0) = 0
ai i + U i i
i i i
(0) (0)
ai (0) = 0.
ai W i W k i + U (3.40)
i
i i
In the last step we have made use of 0th order solution (3.35). Next, (3.40) is multi-
(0) (0) (0)
plied from the left with k and integrated over all space. With k |i
= ki
one obtains
(0) (0) (0)
ak Wk(0) Wk + ak k(0)
U k + ai k U i = 0.
i=k
3.3 Perturbation Theory for Stationary Problems 163
Since |ai | 1 for i = k, and since the matrix elements of the perturbation operator
are small too, one can neglect to 1st order the whole sum over i = k in this
U
expression. With this ak may be factored out, and one arrives immediately at the 1st
order correction for the energy
(0) (0)
W = Wk Wk = Ukk = k U k ,
(0)
(3.41)
Inserting the 1st order solution (3.41) for the energy correction this becomes
(0) (0) (0) (0) (0)
0 = aj Wj Wk k U k ai j U i .
(0)
+
i
If again one neglects all terms that are quadratically small, the third term can be
dropped and of the sum only the term with i = k remains (since ak 1), so that
(0) (0) (0) (0)
0 = aj Wj Wk + j U k ,
| (0)
j(0) |U Uj k
(0) k (0) (0) (0)
k = k + (0) (0)
j = k + (0) (0)
j . (3.42)
j =k Wk Wj j =k Wk Wj
For the next step we insert the results of 1st order perturbation theory again into
the S CHRDINGER equation (3.37) and repeat the whole procedure. The 2nd order
correction for the energy is thus derived as follows:
| (0)
j(0) |U
k(0) H0 (0) + k (0)
Wk = +U k (0) (0)
j
j =k Wk Wj
164 3 Periodic System and Removal of Degeneracy
| (0)
j(0) |U
(0) (0) (0) (0)
k k
(0) k
= Wk + U k + (0) (0)
U j
j =k Wk Wj
(0)
|Uj k |2
Wk = Wk + Ukk + (0) (0)
. (3.43)
j =k Wk Wj
1. We see that many states (in principle an infinite number) may contribute to the
determination of wave functions as well as of energies. Practical considerations
will limit the efforts, and one has to choose judiciously which states are included
in a perturbation expansion.
2. The contribution of individual states j depends both on the magnitude of the
(0) (0)
matrix elements |Uj k | and on the resonance denominators 1/(Wk Wj ). The
closer a the energy of a perturbing state j to that of state k which is perturbed,
the stronger the modification.
(0) (0) (0) (0)
3. If several states 1 , 2 , 3 are degenerate, i.e. have the same energy W1 =
(0) (0)
W2 = W2 , one has to be careful with the perturbation approach because of
the resonance denominators. Only if the non-diagonal matrix elements of the
perturbation disappear, i.e. only if U12 = U13 = U23 0, one may follow the
perturbation procedure just described.
In other words, we have now to solve a set of g linear algebraic equations, which
can lead to a set of g eigenvalues W and state vectors. Generally speaking, one
has to diagonalize the Hamiltonian matrix by a unitary transformation. The matrix
elements of the diagonalized matrix are the eigenvalues of the system.
As well known, a solution of (3.45) only exists if the determinant of the matrix
vanishes. Thus, one first has to find the solutions of
W
det(H 1) = 0. (3.46)
To illustrate how to use 1st order perturbation theory, we briefly return to the alkali
atoms as a simple example. We emphasize the pedagogical aspect of this subsection
that is not meant to generate accurate numbers for energy levels. With the interaction
potential (3.10) discussed in Sect. 3.2.3 the Hamiltonian of the perturbed problem
(in a.u.) is written
= 1
2 + VS (r) = H
H 0 + VC (r) with VS (r) = 1/r + VC (r), (3.47)
2
and we recognize
0 =
2 /2 1/r
H
being the unperturbed Hamiltonian of the H atom.
The spherically symmetric perturbation potential U (r) = VC (r) originates from
the interaction of the valence electron with the full nuclear charge Z at small r and
screening at larger distances. For demonstration we use again the particularly simple
model potential VC introduced in (3.11). Figure 3.15 illustrates this potential for Na
(Z = 11). The screening radius has now been calibrated, as described below, to fit
the experimentally determined ground state binding energy W3s .
The matrix elements of the perturbation (3.11) are computed with the hydrogen
(0)
eigenfunctions Rn (r)Y m (, ) (0th order solutions), which have been treated in
detail in Sect. 2.6.1. Since the perturbation potential is spherically symmetric and
166 3 Periodic System and Removal of Degeneracy
-4
VC (r) =
-6 (Z 1) -r/rS
e
-8 r
acts only on the radial part of the wave function, the matrix elements simply are
(0) (0)
VC n m,n m = mm VC (r)Rn (r)Rn (r)r 2 dr = VC n ,n . (3.48)
0
In 1st order perturbation theory only the diagonal matrix elements are needed. We
(0)
make explicit use of the radial wave functions Rn (r) of the H atom as given in
(2.122), using the series expansion (2.123) for the L AGUERRE polynomials. The
numerical integration of (3.48) can be done with standard desktop computing pro-
grammes within seconds.
(0)
Inserting the binding energies Wn = 1/2n2 of the hydrogen atom into (3.41)
and noticing that with VC (r) < 0 the matrix elements are also negative, the energies
of the alkali atoms in 1st order perturbation theory finally are (in a.u.)
(0) 1
Wn = Wn + VC n ,n = |VC n ,n |. (3.49)
2n2
All energy terms are indeed lowered in respect of the H atom as experimentally
observed.
In Table 3.5 we communicate some numbers that we have derived in this man-
ner for Na and compare them with the experimental data from the NIST data bank
(K RAMIDA et al. 2013). For comparison, H atom binding energies are also shown.
Acronyms and Terminology 167
The data illustrate quite clearly what may be achieved with such a simple approxi-
mation. Of course, the results can only be as good as the perturbation potential that
describes the additional interaction and one cannot expect miracles from (3.11) with
only one free parameter a rather crude guess of the perturbation potential. An ad-
ditional problem is, that the wave functions experience a strong phase shift in the
ionic core, as we have seen in Sect. 3.2.6. This is of course not taken into account
in 1st order perturbation theory where the energy is computed from 0th order wave
functions for the H atom.
In view of these difficulties, the results shown in Table 3.5 may be considered
reasonable. But it becomes also obvious that 1st order perturbation theory should
only be used for a first orientation in such problems with several competing influ-
ences. If precision is demanded one has to resort at least to the methods outlined in
Sect. 3.2.
Section summary
Perturbation theory can often give important qualitative and semi- quantitative
insights into the physics expected from exposing atoms or molecules to spe-
cific interactions as long as the averaged perturbation is small compared to
| |H
the Hamiltonian defining the basic interaction in the system, |U 0
|.
In 1st order perturbation theory the change of energy (of non-degenerated
(0) (0)
states) is
W = Ukk = k |U |k
, while the 1st order wave function is
given by (3.42).
Higher order approaches are needed if first order (diagonal matrix elements)
vanishes.
For degenerate states a different approach has to be taken. One expresses the
perturbed Hamiltonian in matrix form (3.44), using as many and as good basis
functions as appropriate. One then has to diagonalize this matrix, using the
standard procedures of linear algebra. The procedure can be extended to quite
general problems, providing suitable basis functions are available.
References
FANO , U. and A. R. P. R AU: 1986. Atomic Collisions and Spectra. Orlando: Academic Press Inc.,
409 pages.
F REEMAN , R. R. and D. K LEPPNER: 1976. Core polarization and quantum defects in high
angular-momentum states of alkali atoms. Phys. Rev. A, 14, 16141619.
H ARTREE , D. R.: 1928. The wave mechanics of an atom with a non-Coulomb central field. Part I.
Theory and methods. Proc. Camb. Phil. Soc., 24, 89110.
I DZIASZEK , Z. and P. S. J ULIENNE: 2010. Universal rate constants for reactive collisions of
ultracold molecules. Phys. Rev. Lett., 104, 113202.
ISO 21348: 2007. Space environment (natural and artificial) Process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
J UNGEN , C.: 1996. Molecular Applications of Quantum Defect Theory. New York, London: Taylor
& Francis, 664 pages.
K RAMIDA , A. E., Y. R ALCHENKO, J. R EADER and NIST ASD T EAM: 2013. NIST Atomic
Spectra Database (version 5.1), NIST. http://physics.nist.gov/asd, accessed: 7 Jan 2014.
M ARGENAU , H.: 1984. The Miracle of Existence. Woodbridge, CT, USA: Ox Bow Press.
NIST: 2011. NIST physics laboratory holdings by element, NIST. http://physics.nist.gov/
PhysRefData/Elements/per_noframes.html, accessed: 7 Jan 2014.
O SPELKAUS , S. et al.: 2010. Quantum-state controlled chemical reactions of ultracold potassium-
rubidium molecules. Science, 327, 853857.
PAULI , W.: 1945. The N OBEL prize in physics: for the discovery of the exclusion princi-
ple, also called the pauli principle, Stockholm. http://nobelprize.org/nobel_prizes/physics/
laureates/1945/.
S CHMIDT , K. and M. A. L EE: 1998. Visual Schrdinger: A visualizer-solver. http://fermi.la.asu.
edu/Schroedinger/, accessed: 7 Jan 2014.
S CHUMACHER , E.: 2011. FDAlin programme, computation of atomic orbitals (Windows and
Linux), Chemsoft, Bern. http://www.chemsoft.ch/qc/fda.htm, accessed: 5 Jan 2014.
S EATON , M. J.: 1983. Quantum defect theory. Rep. Prog. Phys., 46, 167257.
S EATON , M. J.: 2002. Coulomb functions for attractive and repulsive potentials and for positive
and negative energies. Comput. Phys. Commun., 146, 225249.
U NIVERSITY OF C OLORADO: 2000. Davids wizzy periodic table, Physics 2000. http://www.
colorado.edu/physics/2000/applets/a2.html, accessed: 7 Jan 2014.
W EBER , K. H. and C. J. S ANSONETTI: 1987. Accurate energies of ns, np, nd, nf, and ng levels
of neutral cesium. Phys. Rev. A, 35, 46504660.
W IKIPEDIA CONTRIBUTORS: 2014. Periodic table, Wikipedia, The Free Encyclopedia. http://en.
wikipedia.org/wiki/Periodic_table, accessed: 7 Jan 2014.
Non-stationary Problems: Dipole Excitation
with One Photon 4
Overview
The present chapter concentrates on electric dipole transitions (E1), while
Sect. 5.4 will also treat electric quadrupole (E2) and magnetic dipole (M1)
transitions. After some basics and terminology on electromagnetic radiation,
polarization, and photon spin (Sect. 4.1), the essentials of spectroscopy are in-
troduced in Sect. 4.2, the E INSTEIN A and B coefficients are defined, and the
classical model of a radiating oscillator is reviewed. The advanced reader may
jump over this section and ignore also Sect. 4.3.14.3.5, where the elements
of time dependent perturbation theory are summarized. However, Sect. 4.3.6
with terminology and some key results as well Sect. 4.4 with essentials on
selection rules for dipole (E1) transitions are needed in the following sections
and should be read carefully. The same holds for Sect. 4.5 where the angular
characteristics of dipole radiation are presented. Section 4.6 may be used by
the expert reader just as a source of reference with details on the evaluation
of matrix elements and E INSTEIN coefficients. In Sect. 4.7 photoinduced lin-
ear combinations of states are discussed a theme of broad relevance. In this
context we also introduce quantum beats and indicate some spectroscopic per-
spectives. Finally, we ask the very fundamental, almost philosophical question
whether electrons may really jump from one stationary state into another
and present experiments illuminating this profoundly.
Before discussing radiation induced transitions we define the terminology for de-
scribing electromagnetic waves used here and in most of the following text. For
brevity, we shall speak about electromagnetic waves, electromagnetic radiation
and light more or less synonymously, although the latter is often used more specif-
ically for the visible part of the spectrum.
For the present purpose it is sufficient to consider only plane, monochromatic waves
extending over all space, and to concentrate on the electric field component of the
wave which is responsible for E1 transitions.1
The electric field vector E(r, t) is an observable in the real world depending on
position vector r and time t. We thus write it as2
i
E(r, t) = E0 eei(krt) e ei(krt) (4.1)
2
with the (real) field amplitude E0 , the unit polarization vector e, and the wave vector
k with |k| = 2/ = /c.
Even though it is sometimes convenient to use a complex representation (and take
its real part as observable after all calculations are done), we emphasize strongly that
it is important to write E(r, t) as a real quantity in order to be able to describe all
observed physical phenomena as done here neither of the two summands in (4.1)
can be ignored as we shall see shortly!
From classical electrodynamics we know that the field amplitude E0 is related to
the (time averaged) intensity I of the electromagnetic wave by
E0 = 2I /(0 c) = 2I Z0 = 27.45 I "1/2 (4.2)
with the electric constant 0 , the speed of light c and Z0 the characteristic vacuum
impedance. For practical purposes we communicate a handy numerical expression:
I
E0 = 2745 V m1 . (4.3)
W cm2
1 We find this approach (leading to correct results for E1 transitions) conceptually more accessible
than the general, rigorous treatment of all transition types based on the vector potential. The latter
is outlined in Appendix H, while in Chaps. 1 and 2, Vol. 2, we shall generalize (4.1) and learn how
to treat spatial distributions and quasi-monochromasy of real light beams adequately.
2 Theoverall phase angle 0 does not play a role in the present discussion and will generally be
ignored. However, we shall have to come back to it in Chaps. 1 and 2, Vol. 2.
4.1 Electromagnetic Waves: Electric Field, Intensity, Polarization 171
In reality, one usually has to deal with light of a certain bandwidth, i.e. with an
intensity distribution I() (intensity per unit angular frequency). Then one has to
replace I I()d and to integrate expressions derived for transition probabilities
over the whole available spectral range.
In this basis, the polarization of light which propagates in the +z-direction (e0 ) may
again be expressed by only two basis vectors (e+1 , e1 ). One easily verifies again
the orthonormality
eq eq = qq with eq = (1)q eq . (4.6)
For atomic problems the helicity basis is often better adapted, since atoms too are
usually described in a spherical coordinate system. For reference, we express the
Cartesian basis in terms of the helicity basis:
1 i
ex = (e+1 e1 ) and ey = (e+1 + e1 ). (4.7)
2 2
For later use we also give the unit polarization vectors for linearly polarized light in
45 and 135 direction in respect of the x-axis:
1 1
e 45 = (ex + ey ) and e 135 = (ex ey ). (4.8)
2 2
172 4 Non-stationary Problems: Dipole Excitation with One Photon
The unit vector e = e+1 describes left hand circularly polarized light (LHC), also
called + light. Inserting (4.5) into (4.1) gives
1
E +1 (r, t) = E0 sin(kr t)ex + cos(kr t)ey , (4.12)
2
and e1 stands for right hand circularly polarized light (RHC) or light:
1
E 1 (r, t) = E0 sin(kr t)ex cos(kr t)ey . (4.13)
2
An illustration of + light gives Fig. 4.1. Shown is the E vector at a fixed time
t = /2 along the z-axis. As indicated E rotates at a fixed position clockwise
around the wave vector k of the light, i.e. with positive helicity. One verifies this
most directly in (4.12) with kr = 0, or correspondingly in Fig. 4.1, if one considers
the time propagating.3
The most general unit polarization vector for light propagating into an arbitrary
direction can be written in either Cartesian or spherical basis:
1
1
e = ax e x + ay e y + az e z = aq e q with |aq |2 = 1. (4.14)
q=1 q=1
In the spherical basis it may be specialized for any elliptically polarized light prop-
agating parallel to the +z axis (i.e. into e0 direction) by
3 Methods for generating and detecting polarized light will be discussed in Chap. 1, Vol. 2.
4.1 Electromagnetic Waves: Electric Field, Intensity, Polarization 173
z
+ light (LHC)
k
+t
+t 0
k y
+t
Fig. 4.1 Schematic illustration of left hand circularly polarized light + (LHC). The grey arrows
indicate the direction of the electric field perpendicular to the z k axis at a fixed time, correspond-
ing to t = /2. For larger times the sense of the rotation is indicated. The somewhat surprising
definition for left hand circular polarized light is of historical origin: in times prior to the laser,
people used to look into the oncoming light beams, hence they considered + light as rotating
anti-clockwise, i.e. LHC polarized
with the component a0 = 0 in this coordinate system). The ellipticity angle de-
scribes the degree of ellipticity, the alignment4 angle gives the direction of the
ellipse in respect of ex . Converting (4.15) into Cartesian coordinates with (4.5) and
inserting into (4.1) gives an explicit, parameterized expression for the elliptic field
vector in terms of and :
E el (r, t) = (E0 / 2)
'
cos sin(kr t ) + sin sin(kr t + ) ex (4.16)
(
+ cos cos(kr t ) sin cos(kr t + ) ey .
This describes the standard form of an ellipse on which the E vector rotates if (for
fixed r) the phase kr t + varies between 0 and 2 . As sketched in Fig. 4.2,
this ellipse is inclined by an angle in respect of the x-axis, and for 0 /2
its major and minor axes a and b, respectively, are:
a = E0 (cos + sin )/ 2 = E0 sin( + /4)
(4.17)
b = E0 (cos sin )/ 2 = E0 cos( + /4).
4 Unfortunately the terms alignment and orientation are mixed up time and again in the liter-
ature: Alignment refers to the direction of a polar vector (e.g. the E vector in the case of linearly
polarized light). In contrast, orientation specifies the sense of rotation of an axial vector (e.g. the
angular momentum of left and right circularly polarized light).
174 4 Non-stationary Problems: Dipole Excitation with One Photon
Note that a 2 + b2 = E02 . In the literature instead of the ellipticity angle often a
(somewhat ambiguously defined) so called ellipticity is used:
The main part of this chapter will be concerned with absorption and emission of
electromagnetic radiation from atoms. In general, one has to distinguish two dif-
ferent coordinate systems: one in which the atom is best described we call it the
atomic frame (at) while the photon may possibly be better described in another
frame we call it the photon frame (ph).
These coordinate systems may differ from each other. In the spirit of Sect. 4.1.2
a convenient choice for the photon frame is illustrated in Fig. 4.3. We shall use this
henceforth unless otherwise stated. The atomic frame may e.g. be defined by an ex-
ternal electric or magnetic field in respect of which the projection quantum numbers
m are defined. The photon frame is defined as shown in Fig. 4.3 by the direction of
light propagation (wave vector k), with the axes z(ph) k and y (ph) z(at) .
(at)
The unit vectors eq in the atomic frame (helicity basis) indicated in Fig. 4.3
by heavy red arrows may in a classical picture be seen to represent three different
classical dipole oscillators. This will be discussed in Sect. 4.5.1.
4.1 Electromagnetic Waves: Electric Field, Intensity, Polarization 175
m = -1 k
x (at)
At this point we want to extend our knowledge about the properties of photons sum-
marized in Sect. 1.4. In addition to momentum and energy the photon also possesses
an intrinsic angular momentum, the so called photon spin with a quantum number
sph = 1. Projected onto the wave vector k LHC and RHC polarized light have angu-
lar momenta ms with ms = 1 and 1, respectively. Owing to the transverse nature
of electromagnetic waves the third component with ms = 0 does not exist.
In a particularly beautiful and fundamental experiment this has been observed for
the first time by B ETH (1936). The setup and the results are shown schematically
in Fig. 4.4. LHC photons are prepared by passing linearly polarized light through
a /4 plate. These photons then pass through a /2 plate where they are converted
into RHC. This implies an angular momentum change of 2 per photon which is
transferred to the /2 plate in order to conserve the total angular momentum of the
system. For more efficiency the light beam is retroreflected behind another /4 and
/2 plate
RHC RHC 0
/4 plate
x y xy-polarization plane -5
of linear polarization
linearly polarized light in = -135 -90 -45 0 45
Fig. 4.4 Experiment of B ETH (1936) on mechanical detection and measurement of angular mo-
mentum of light
176 4 Non-stationary Problems: Dipole Excitation with One Photon
Section summary
We describe electromagnetic radiation by a monochromatic plane wave. As
a physical observable it is represented by a real valued function (4.1). Both
exponential terms will turn out to be relevant when describing the interaction
of the electromagnetic field with quantum systems.
The helicity basis (4.5) is convenient for describing orthonormal (4.6) unit ba-
sis vectors for problems in atomic physics. The most general unit polarization
vector for light propagating into +z-direction is given by
at physically reasonable boundary conditions (in the following we shall often ab-
breviate all quantum numbers by one letter, e.g. n j = j, a or b). Energy zero is
defined such that for bound states Wj < 0 while the free electron has a continuous
spectrum of energies Wkin = W 0. As discussed in the previous two chapters, the
asymptotic behaviour of the radial wave function of the bound states is given by
lim Rj (r) exp 2|Wj |r
r
while the continuum states are described essentially by C OULOMB wave functions.
4.2 Introduction to Absorption and Emission 177
W2 | 2
WI
W1 | 1
Here ba = 2ba is the angular frequency of the radiation, ba = c/ba its vacuum
wavelength and ba = 1/ba = ba /c its wavenumber. Figure 4.6 illustrates three
basic types of setups for spectroscopic investigations. Specialties and refinements
will be discussed in detail later, but essentially all spectroscopic measurements fol-
low in principle one of these three concepts:
hot gas,
plasma
emission spectroscopy
cold gas
white light absorption spectroscopy
lamp
fluorescence spectroscopy,
las
Fig. 4.6 3 basic methods of spectroscopy very schematic. Left: experimental setups, middle: type
of observed spectra, right: term schemes and transitions
Absorption
As derived in Sect. 1.3.2 the L AMBERT-B EER absorption law (1.44)
dI
= Na I I (z) = I0 e Na z = I0 ez (4.21)
dz
describes the exponential reduction of the intensity I (z) of electromagnetic radia-
tion when passing through an absorbing medium of particle density Na with [Na ] =
number of absorbing particles m3 . Here Na is assumed to be constant over dis-
tance z and time t.
We now want to understand the process on a microscopic level and derive quan-
titative expressions for the absorption cross section ([ ] = m2 ) and the absorp-
tion coefficient = Na ([] = m1 ). Since the energy of a photon is we can
4.2 Introduction to Absorption and Emission 179
n (n-1)
|a Na
translate the intensity of the radiation field I ([I ] = W m2 ) into its photon density
Nph ([Nph ] = number of photons m3 ) or energy density u = Nph ([u] = J m3 )
by
I = cNph = cu. (4.22)
As schematically indicated in Fig. 4.7, on an atomistic level the absorption process
corresponds to a loss of photons from the field, i.e. a reduction of Nph by dNph
which is identical to a loss dNa of particle density Na in the initial state |a
and gain
dNb of density Nb in the final excited state |b
.
In contrast to the situation described by (4.21), we are now interested in the
change of N with time t at fixed position z for a constant intensity I . Since a change
over a distance dz corresponds to a change over a time dt = dz/c we may rewrite
(4.21) with (4.22) as
1 dNa I
R ba = = ba = ba (4.24)
Na dt
for the transition into the state |b
from state |a
. We have introduced here also the
photon flux = I /.
Note that the rate R ba has the dimension T1 ; it is the probability for absorption
of photons by an atom per unit of time. To obtain the total number of events per unit
time one has to multiply it by the number of atoms in the volume under observation.
In the above discussion we have tacitly assumed that the radiation to be absorbed
is strictly monochromatic and tuned into resonance with the absorption line. In real-
ity, however, the intensity has a spectral distribution, which we indicate by I() with
[I()] = W m2 s1 . Thus, only that fraction of intensity can be absorbed which is
in resonance with the angular frequency ba of the transition. Usually one refers the
transition probability to the spectral radiation density u(), i.e. the energy density
5 Forclarity, throughout this chapter we shall use R ba , B ba etc. for rates and coefficients averaged
over initial and summed over final substates, since in general we have to account for degenerate
energy levels. In contrast, Rab (e) refers to transitions between specific substates b and a with
polarization e.
180 4 Non-stationary Problems: Dipole Excitation with One Photon
Table 4.1 Definitions in the context of absorption and emission of electromagnetic radiation
Symbol Eq. Term Unit Remarks
absorption cross section m2 effective absorbing area
N particle density m3
=N absorption coefficient m1 I = I0 exp(x)
= amplification coefficient m1 if < 0
I = c Nph light intensity W m2
Nph = I /(c ) photon density m3
I = E02 /(2Z0 ) intensity E0 field amplitude
Z0 = 1/(c0 ) impendence in free space 376.73 also vacuum impedance
Nph = r 0 2
2 E0 energy density J m3 radiation field
Nph
u() =
spectral radiation density J m3 Hz1
0 E02
= energy
Vol
= I
c
= 2
= I ()
2c W s2 m3
u() = 2u()
bandwidth (frequency) Hz
. . . (angular frequency) s1
du per spectral range d, which is related to the spectral intensity distribution I()
by
I() du d du u()
u() = = = = , (4.25)
c d d d 2
with the spectral radiation density given alternatively per unit angular frequency
= 2 or per unit frequency of the electromagnetic radiation. (We shall use the
former unless mentioned otherwise.)
Relevant for the transition probability is the intensity at the transition frequency.
Consequently, we now rewrite (4.24) as:
1 dNa I(ba )
R ba = = B ba = B ba u(
ba ). (4.26)
Na dt c
The thus defined constant B ba is called E INSTEIN coefficient for absorption with
[Bba ] = m3 s2 J1 . Note that rates, transition probabilities and matrix elements
from state |a
to state |b
are typically indexed from right to left.
Table 4.1 summarizes the terms and definitions most frequently used for the
quantitative description of radiation induced dipole transitions. Most of the discus-
sion to follow in this chapter will focus on the E INSTEIN coefficients. They contain
the essence of the transitions in quantum systems. Polarization and frequency de-
pendence play an important role.
Stimulated Emission
Up to now we have assumed that only one dominantly populated ground state level
contributes to the absorption process. And the absorption as such was thought to be
weak enough so that it does not change the population significantly. We now con-
4.2 Introduction to Absorption and Emission 181
n (n+1)
|a Na
sider the inverse process, the so called stimulated emission which we have schemat-
ically illustrated in Fig. 4.8. For this to happen, the excited state must be populated.
For simplicity we assume for the moment that all atoms are initially found in the
excited state |b
as indicated in Fig. 4.8. In complete analogy to (4.24) and (4.26)
the corresponding emission rate is
1 dNb I(ba )
R ab = = B ab = B ab u(
ba ). (4.27)
Nb dt c
We thus find that more photons may come out of the system than going into it if
the excited state has a higher population than the ground state so that this process
dominates. B ab is called E INSTEIN coefficient for stimulated emission.
In general, both processes absorption and stimulated emission happen in a
quantum system. According to (1.54) the ratio of the population densities are given
by Nb /Na = (gb /ga ) exp[(Wb Wa )/(kB T )] in thermodynamic equilibrium. At
room temperature kB T 25 meV is small, while electronic excitation in atoms typ-
ically requires some eV; hence Wb Wa kB T and Nb Na . In contrast, if the
energy gaps are much smaller, as it is the case e.g. for molecular vibrational and ro-
tational states, one has to consider induced emission also in standard spectroscopic
absorption experiments.
A completely different situation arises when skillful schemes are applied to
achieve a significant population of some specific excited state |b
such that Nb > Na
(for some state |a
, not necessarily the ground state). Then, stimulated emission can
be stronger than absorption and laser action becomes possible. We shall discuss this
in some detail in Chap. 1, Vol. 2.
From experimental experience we know that excited atoms may also decay sponta-
neously. Quantum electrodynamics (QED) explains this by the interaction of quan-
tum systems with the vacuum field: In a fully quantized description the electromag-
netic field is represented by harmonic oscillators, their energy being (Nj + 1/2)j .
Here Nj is a (typically very large) integer number proportional to the intensity of
the radiation at the frequency j , and j /2 is the so called zero-point energy. The
latter represents the eigenenergy of the ground states of the field oscillators at zero
intensity of the radiation. This vacuum field is essentially a consequence of the
uncertainty relation. One may say, somewhat figuratively, that it is this (isotropic)
182 4 Non-stationary Problems: Dipole Excitation with One Photon
zero point oscillation of the quantized radiation field, which forces an excited quan-
tum system to decay spontaneously. We shall present a quantitative description of
spontaneous emission based on this concept in Chap. 2, Vol. 2.
Here we consider a heuristic, classical approach in which the atomic electron is
treated as an accelerated charge with a dipole moment D(t) = er(t). This clas-
sical picture does not allow to extract quantitative results. It leads, however, to an
intuitive understanding. According to classical electrodynamics such a dipole emits
electromagnetic waves. In the nonrelativistic limit the electric field at large distances
R from the source is given by
E(R, t) =
1 k
k
t = er (t ) ,
D (4.28)
2
40 c R k k 40 c R 2
with t = t R/c being the retarded time, and r representing the components of
the dipole acceleration perpendicular to the wave vector k. We identify this with
projecting r onto the x (ph) y (ph) plane of the photon frame introduced in Fig. 4.3.
For a harmonic dipole oscillator r(t) = r (at) exp(iba t) with amplitude r (at) in the
atomic frame and angular frequency ba the field amplitudes are given by
2
eba (ph)
Ex (R, t) = r (at) ex ei(kRba t) (4.29)
40 c2 R
2
eba (ph)
Ey (R, t) = r (at) ey ei(kRba t) . (4.30)
40 c2 R
In Sect. 4.5.1 these expressions will be used for a semiclassical derivation of angular
and polarization characteristic of atomic dipole radiation. Here we are just interested
in the overall power emitted. From (4.28) we obtain with (4.2) the intensity I =
0 c|E(R, t)|2 , and on average6 an energy
dW
|D(t)| 2
d = I R 2 d = sin2 k dk (4.31)
dt (4)2 0 c3
is emitted per unit time into a solid angle dk = sin k dk dk . Here k is the angle
between D and k. Integration of sin2 k over all solid angles gives a factor 8/3 so
that the total radiation power emitted is
2
dW 1 e2 dp
P= = | 2=
D| (4.32)
dt 60 c 3 60 me c dt
2 3
dW 1 e2 |r (at) /2|2 ba
4
P= = . (4.33)
dt 6 0 c3
Due to this radiation the excited state of the atom decays. In a classical picture
the oscillation amplitude (i.e. the orbital radius of the electron) continuously de-
creases contrary to the spectroscopic observation of discrete, stationary states:
the atom (or any other quantum system) is either in the excited or in the ground
state.
Thus, quantum mechanically (4.33) is interpreted as a probability statement:
The probability that one excited atom emits a photon of energy during the time
interval dt is
dW 1 e2 |r (at) /2|2 ba
3
R ab dt = A ab dt
(spont)
dt = = (4.34)
3 0 c 3
with 1/137 being the fine structure constant (1.10). The exponential decay law
ensuing form (4.34) has already been discussed in Sect. 1.3.1. For the excited state
density Nb we obtain according to (1.39) and (1.40) in the absence of other pro-
cesses
1 dNb 1
= A ab = Nb (t) = Nb0 eAab t = Nb0 et/ab
Nb dt ab
with the average excited state lifetime ab = 1/A ab , the half lifetime 1/2 =
(ln 2)ab , and Nb0 the excited state density at time t = 0.
We note at this point, that (4.35) is almost identical to the exact expression
for the so called E INSTEIN A-coefficient for spontaneous emission, which will be
discussed in Sect. 4.6.2. The specific properties of the emitting atom, here repre-
sented by |r (at) /2|, require of course a stringent quantum mechanical interpreta-
tion. For a prominent case, the 3p 3s transition in sodium (the so called Na
D line at = 589 nm), we try an intelligent guess by setting r (at) 0.190 nm
(the atomic radius, see Fig. 3.4). From (4.35) we obtain A 3s3p 3.2 107 s1 or
= 1/A ab 31 ns which is at least on the same order of magnitude as the cor-
rect, experimentally determined value 16.2 ns. Of course, we do not expect exact
results from this classical model.
184 4 Non-stationary Problems: Dipole Excitation with One Photon
|a Na
initial final initial final initial final
I I I
ab ab
characteristics
of spectra ab
Fig. 4.9 Absorption, induced and spontaneous emission schematically. In the top row the atom-
istic view point is indicated, below typical spectroscopic patterns are illustrated: the detected in-
tensity as a function of angular frequency of the radiation
Figure 4.9 summarizes the above considerations and findings. In a real experiment
one has to consider all three processes simultaneously for describing an atomic sys-
tem in the presence of an electromagnetic field correctly. A special case is a system
of atoms, molecules or condensed matter oscillators in thermodynamic equilibrium
with its own radiation field.
E INSTEIN has taken this thermodynamic equilibrium as a starting point for a very
elegant and simple derivation of P LANCKs radiation law Fig. 1.15. It is based on
the so called principle of detailed balance: under stationary conditions each process
must be in equilibrium with its inverse. One treats the problem as a representative
two level system sketched in Fig. 4.10. The kinetics of this system is described by so
called rate equations. The population of excited and ground state may change due
to the three processes as follows:
dNb dNa
= A ab Nb Nb B ab u(
ba ) + Na B ba u(
ba ) = . (4.36)
dt dt
Stationarity means that the particle densities do not change:
dNb,a
= 0 = A ab Nb Nb B ab u(
ba ) + Na B ba u(
ba )
dt
A ab Nb A ab
=1 = u(
ba ).
(Na B ba Nb B ab )u(
ba ) B ba (Na /Nb ) B ab
4.2 Introduction to Absorption and Emission 185
A ab A ab
ba ) =
u( = .
Bba (Na /Nb ) Bab Bba (ga /gb )eba /(kB T ) B ab
3 1
u() = /(k
,
2 3
c e BT ) 1
3
A ab = R ab = 2 3 B ab and gb B ab = ga B ba .
(spont)
(4.38)
c
Note the 3 factor interestingly, it agrees with the classical prediction (4.35).
We point out that the B coefficients used here refer to spectral radiation densities
Section summary
We have introduced three prototype methods for obtaining information from
quantum systems by interaction with an electromagnetic field: emission, ab-
sorption and fluorescence spectroscopy.
Induced processes (absorption and induced emission) are introduced phe-
nomenological. The rates for both processes are proportional to the spectral
ba ) = I(ba )/c at the transition frequency studied.
radiation intensity u(
A classical interpretation of spontaneous emission encounters severe prob-
lems which have led to B OHRs second theorem and to the probabilistic
interpretation of the emission process. Nevertheless, it provides an informa-
tive first guess on the spontaneous emission probability and useful information
(4.32) for later use.
E INSTEINs has derived P LANCKs radiation law from rate equations for a
system of quantum oscillators in thermal equilibrium with the radiation field.
This leads immediately to quantitative relations between the E INSTEIN coef-
ficients for spontaneous and induced emission.
The readers should memorize the famous factor 3 (or 3 ) between them:
A ab 3 B ab = 3 (ga /gb )Bba it will play an important role on several oc-
casions.
7 In
the literature often u() = 2 u()
per unit frequency is used. Then one has to replace B ba
B ba
()
= B ba /2 in all relevant equations.
186 4 Non-stationary Problems: Dipole Excitation with One Photon
where cj (t) is the time dependent probability amplitude of state j (r) and the prob-
ability to find a certain final state |j
at time t is given by
2
wj (t) = cj (t) . (4.42)
If the summation (4.41) is carried out over sufficiently many states and the basis |j
is complete, it represents a solution as close to exact as desired.
Using this ansatz and the time dependent Hamiltonian (4.40), the time dependent
S CHRDINGER equation (2.13) is written as
| (t)
H0 + U p , r, t) (t) = i
( (4.43)
t
[cj (t)eij t ]|j
cj (t)eij t H0 + U
(p , r, t) |j
= i .
t
j j
Inserting H0 |j
according to (4.39) on the left hand side of this equation, and using
the rules of product differentiation on the right hand side leads to
(
cj (t)eij t Wj + U p , r, t) |j
j
dcj (t)
ij t ij t
=i cj (t)(ij )e + e |j
.
dt
j
With i(ij ) = Wj the first terms in the sums on either side cancel. One multiplies
from the left with b|eib t and applies b|j
= bj . With Wb = b and the time
dependent perturbation matrix elements
Uij (t) = i|U ( (
p , r, t)|j
= i (r)U p , r, t)j (r)d3 r (4.44)
Note that this is a general and still exact formulation for atoms or molecules ex-
posed to a time dependent interaction. In particular, it is not restricted to the dipole
approximation nor even to monochromatic plane waves.
In practice one can only sum over a finite number of terms. A perturbation approach
(
is possible if |U 0
|, i.e. if the averaged perturbation is small com-
p , r, t)
| |H
pared to the averaged Hamiltonian of the unperturbed system (H atom, alkali atom
188 4 Non-stationary Problems: Dipole Excitation with One Photon
etc.). Then the initial conditions will change only slightly with time. To obtain the
probability amplitude cb (t) for the final state b in 1st order time dependent per-
turbation theory one starts in 0th order by setting on the right hand side of (4.45)
ca(0) (t) 1 (initial state a) and cj(0) (t) 0 for all j = a. From this one obtains for
b = a in 1st order:
dcb (t) i
= Uba (t)eiba t . (4.46)
dt
If the interaction is switched on at time t = 0, integration over time leads to the time
dependent transition amplitude in 1st order perturbation theory:
i t
cb (t) = Uba t eiba t dt . (4.47)
0
One may iterate the procedure by inserting (4.47) into (4.45) to obtain the 2nd order
solution and so on.
Often one is interested only in how the probability amplitude cb (t) evolves after
many oscillation of the field, i.e. for t 1/ba . In the limit t the 1st order
probability amplitude (4.47)
i
)ba (ba )
cb () = Uba t ei(ba )t dt = U (4.49)
0
is apart from numerical prefactors nothing but the F OURIER transform (I.2)
of the perturbation potential at the frequency ba of the transition. In conclusion
atomic or molecular transitions can only be excited by an interaction potential
which contains F OURIER components U )ba () which are in resonance with the tran-
sition frequency ba = .
8A somewhat heuristic derivation of eE0 and p, r) for electric dipole transitions will be pre-
D(
sented below in Sect. 4.3.4, while a more rigorous, general derivation and specialization is found
in Appendix H.
4.3 Time Dependent Perturbation Theory 189
i it
(
Uba (t) = b|U p , r, t)|a
= b| eE0 De D eit |a
2 (4.50)
i
= eE0 Dba eit
Dba eit with
Dba = b|D|a
.
2
The transition amplitude (4.47) now becomes
eE0 t
Dba ei(ba )t
Dba ei(ba +)t dt
cb (t) = (4.51)
2 0
eE0 Dba ei(ba )t D ei(ba +)t
= ba . (4.52)
2 i(ba ) i(ba + )
We note here that both terms, exp(it) and exp(+it), are relevant. And one
sees already in (4.51) that significant contributions are expected only for stationary
phases, i.e. for ba = 0, respectively. Otherwise the contributions to the integral
oscillate rapidly and disappear in the limit t . With the angular frequency of
the radiation > 0 (per definition), and a referring to the initial, b to the final state,
the first term in (4.52) describes absorption, the second induced emission:
absorption b a: ba > 0
D exp(it) term relevant
(4.53)
emission a b: ba < 0
D exp(+it) term relevant.
We now apply this general formalism to the so called electric dipole transitions
(E1 transitions) and present here a slightly heuristic introduction of the interaction
potential. It leads to the same result as the more stringent (but also less evident)
derivation given in Appendix H.1.6. In both cases the key simplification exploits the
fact that the wavelengths of interest are typically large compared to atomic dimen-
sions ( a0 ). Hence, the electromagnetic field E(r, t) described by (4.1) may be
expanded in powers of r/ or k r 1. To 1st order exp(ik r) reduces to 1, and
only the temporal change of the electric field E(t) needs to be considered.
The force on an electron from this oscillating field is eE(t), and for the
electron-nucleus pair the
i
U (r, t) = D E(t) = er E(t) = E0 er eeit e eit . (4.55)
2
190 4 Non-stationary Problems: Dipole Excitation with One Photon
It depends on the electron position r and on time t. The field amplitude E0 is related
to the radiation intensity I via (4.2), and e describes the polarization of the field.
Hence, by comparing (4.55) with (4.50) one finds that the dipole transition operator
for absorption is given by
D = r e = D e/e and eE0 = e 2I /(c0 ), (4.56)
In the following we evaluate explicitly the probability amplitude for absorption, i.e.
we focus for the moment only on the first term in (4.52), with ba > 0 (Wa < Wb ).
If the electromagnetic field is switched on at t = 0, we have cb (t) 0 for t < 0
while for t > 0 the transition amplitude evolves as
eE0 ei(ba )t 1
cb (t) =
Dba . (4.58)
2 i(ba )
4.3 Time Dependent Perturbation Theory 191
Fig. 4.11
g( ) / (t/2 ) 1
g() = sin2 [(ba )t/2]/(ba )2
as a function of ba in
units of 2/t . For t it
becomes proportional to the
delta function
-3 -2 -1 0 1 2 3
( ab - ) / (2 /t)
Figure 4.11 illustrates the characteristic frequency dependence g(). For large times
t 1/(ba ) it becomes arbitrarily narrow, and at the same time arbitrarily high
(t/2 at = ba ). This line profile is normalized such that g()d = 1, and
is given by
2 sin2 ( ba2 t) t
g() = (ba ). (4.60)
t (ba )2
Thus, g() becomes a representation of the D IRAC delta function in the limit
t .
(abs)
According to (4.59) wba (t) grows linearly with time. The approximation is of
course only valid for not too large fields and not too large times so that the general
assumption of perturbation time |cb (t)|2 1 is fulfilled for all b = a. At the same
time we are interested in the stationary state, i.e. in times t 1/(ba ).9
(abs) (abs)
We now divide wba (t) by time t to obtain the transition rate Rba = wba (t)/t,
i.e. the transition probability per unit time which in this approximation obviously
becomes independent of time. This is indeed the quantity one determines in an ab-
sorption experiment.
Finally, we have to account for the fact that in practice strictly monochromatic
electromagnetic waves do not exist. Each radiation has a certain bandwidth
=
2
. We introduce this frequency dependence by replacing the intensity I in a
spectral range from to + d by I()d, where I() = cu() is the spectral
9 We shall see in Chap. 5 that these somewhat contradictory requirements can already be overcome
by accounting for a finite lifetime of excited states, and a full treatment will be given in Chap. 10,
Vol. 2 in the framework of the optical B LOCH equations.
192 4 Non-stationary Problems: Dipole Excitation with One Photon
intensity distribution. Thus, in (4.59) the square of the perturbation amplitude e2 E02
according to (4.56) has to be replaced by
2
e2 E02 e2 I ()d. (4.61)
c0
(abs)
The absorption rate dRba = wba (t)d/t for radiation of angular frequency be-
tween and + d is then given by
e2 I()
dRba = I ()|
Dba |2 g()d = 4 2 |
Dba |2 g()d, (4.62)
0 c 2
which is identical to the excitation probability of state |b
from |a
per unit time,
with the fine structure constant 1/137 according to (1.10).
So far we have tacitly assumed that the absorption occurs between two isolated
states |a
and |b
with sharp energies Wb and Wa , i.e. at a well defined sharp line of
angular frequency ba = (Wb Wa )/. The total transition rate Rba is then obtained
by integrating (4.62) over all frequencies:
I() 4 2
Rba = dRba = 4 2 |
Dba |2 g()d = Dba |2 I(ba ). (4.63)
|
shows that Rba is indeed a rate, i.e. gives the probability for an absorption process
per atom per unit of time. In the present perturbative approach we have assumed
that Rba tobs 1 during the relevant observation time tobs .
We see now explicitly that the excitation probability is proportional to the
F OURIER component of the spectral radiation density u() = I()/c at the tran-
sition frequency ba . The integration (4.63) is of course based on the assumption
that I() is constant over the absorption line. For typical classical radiation sources
this is trivially a correct assumption. In laser spectroscopy, however, this is not nec-
essarily true as we shall explain in Sect. 5.2.3 and more profoundly in Chap. 10,
Vol. 2.
For the integration in (4.63) we have assumed that the electromagnetic wave
interacts for a sufficiently long time with the system, so that the limit 1/ba t
in (4.60) is well approximated. With typical periods of the light field on the order
of femtoseconds (fs) this approximation may even be applied to excitation by ns
pulses often used in laser spectroscopy. But (4.62) is still valid for much shorter
interaction times: according to (4.60) this does, however, imply a broadening of the
line. We shall come back to this aspect in Sect. 6.1.7.
4.3 Time Dependent Perturbation Theory 193
To derive (4.63) for the absorption process we have evaluated the first exponent
in (4.51). The second becomes only relevant if ba < 0, i.e. for induced emission
(Wa > Wb ). This leads in complete analogy to the rate Rab | Dab |2 = |r ab e |2 for
induced emission, with identical prefactors. Because of hermicity r ab = r ab holds
and thus Rba = Rab . The rates for induced emission and absorption for a specific
transition between the states |b
|a
are identical.
In contrast to the general discussion in Sects. 4.2.34.2.5 we refer here to indi-
vidual substates described by a set of quantum numbers e.g. j m and not to energy
levels which may consist of several such (degenerate) sublevels. The key quantity
which characterizes an electric dipole transition is the dipole transition operator, ac-
cording to (4.56) Dab = r ba e, i.e. essentially the scalar product between dipole
matrix element D ba = er ba and polarization vector e with the dipole transition
amplitude
r ba = b|r|a
= b (r)ra (r)d3 r = a|r|b
= r ab . (4.64)
Note that r ba is a vector, specific for each system and each transition,10 and the
transition probabilities also depend critically on the polarization e of the radiation.
Before evaluating this in detail, we briefly summarize the results for electric dipole
transitions obtained so far.
We rewrite the probability (4.63) for inducing E1 transitions per unit time be-
tween the substates |a
= |ja ma
and |b
= |jb mb
with polarization e:
4 2
Rba = I (ba )|r ba e|2 = Rab (4.65)
4 2 c
= |r ba e|2 u(
ba ) = B(jb mb ; ja ma ; e)u(
ba ).
In full analogy to (4.26) and (4.27) we have introduced the substate and polarization
specific E INSTEIN B coefficients:
4 2 c
B(jb mb ; ja ma ; e) = |r ba e|2 = B(ja ma ; jb mb ; e). (4.66)
Again, these B coefficients are defined in respect of a spectral intensity distribution
I(ba ) = cu(
ba ) per unit angular frequency at the transition frequency ba =
|Wb Wa |/.
10 More precisely, er
ba is called dipole length matrix element. According to (H.25) the correspond-
ing dipole velocity matrix element is
p|a
= iba me eb|r|a
.
eb|
Both formulations lead to identical results if exact wave functions are used. For approximate solu-
tions (i.e. quite generally) significant differences may occur.
194 4 Non-stationary Problems: Dipole Excitation with One Photon
(spont) 3 d
dRab = B(ja ma ; jb mb ; e) = C|r ba e|2 d
2 c3 8
(4.67)
2 3
ba e2 ba3
= C r ab e d with C = = .
2c2 8 2 0 c3
A discussion of the corresponding E INSTEIN A coefficients will be postponed un-
til we have derived the angular distributions of the radiation patterns in Sect. 4.5.
A clean derivation of (4.67) will be given in Chap. 2, Vol. 2.
In summary, absorption and emission of electromagnetic radiation due to E1
transitions is characterized by
According to (4.57) the matrix elements for transitions |ja ma
|jb mb
are
Dba =
r ba e and Dab = r ab e = Dba for absorption and emission, respectively. They
determine
whether the transition may occur at all (so called selection rules),
its dependence on polarization and angular distribution,
and its overall strength.
In the following sections these aspects will be discussed in detail, in particular the
key quantity r ba e in (4.68) will be evaluated for different polarizations and ge-
ometries. As we shall see, this may be somewhat involved since in general, r ba and
e are described most conveniently in different coordinate frames, e.g. in the atomic
(at) and photon (ph) frame, respectively, as introduced in Sect. 4.1.3.
However, we do not want to end this section without treating at least one par-
ticularly simple example. We choose excitation (and induced emission) by linearly
polarized light and exploit the fact that r ba = (xba , yba , zba ) is a well defined vec-
tor which may be represented by its Cartesian components. Let (r) be the angle
4.3 Time Dependent Perturbation Theory 195
I(ba )
Rba = Rab = 4 2 |r ba |2 cos2 (r) (4.69)
with |r ba |2 = (xba
2 + y 2 + z2 ). This expression may be somewhat misleading in as
ba ba
far as depending on the alignment of the polarization vector not all components of
r ba contribute equally; e.g. for the simplest geometries where the linear polarization
is chosen parallel to any of the axes (4.69) becomes
I( ) |xba |2 for e x
ba
Rba = Rab = 4
2
|yba |2 for e y (4.70)
|zba |2 for e z.
Section summary
The semiclassical approach describes radiation induced processes by a classi-
cal, time dependent electromagnetic field and treats the interacting atomic or
molecular system quantum mechanically.
Time dependent perturbation theory is usually a good approximation for low
enough radiation intensities. Time dependent transition amplitudes (4.47) are
computed from the matrix elements of the perturbation operator U (t).
In the limit t , the transition amplitude (4.49) between initial (a) and
final state (b) is essentially the F OURIER component of Uba (t) at the transition
frequency ba .
The dipole approximation describes so called E1 transitions and is usually an
excellent approximation for a quantitative description of transition probabili-
ties as long as the wavelength of the radiation is a0 . It neglects the change
of the field with position over atomic distances. In the dipole length formu-
lation the interaction operator is E0 D where E0 is the field amplitude and
Dba = r ba e, with e being the unit polarization vector of the radiation.
Both exponential terms of the real field amplitude (4.1) matter: the
D exp(it) term is responsible for absorption, the
D exp(it) term for in-
duced emission. While the semiclassical treatment gives quantitatively cor-
rect results for absorption and induced emission, spontaneous decay requires,
strictly speaking, the use of QED to be discussed in Chap. 2, Vol. 2.
196 4 Non-stationary Problems: Dipole Excitation with One Photon
As we have learned in Sect. 4.1.4 the photon has a spin sph = 1. But only two
projections ms q = 1 in respect of the direction of propagation, |sph q
= |1 +1
and |1 1
are realized, describing LHC and RHC polarized light, respectively. One
speaks of positive and negative helicity. The fact that the third component q = 0 is
missing, reflects the transverse nature of free electromagnetic waves. According to
relativistic quantum field theory this is a consequence of the photon being a massless
particle.
Quite down-to-earth, the angular momentum properties of the photon allow us
to derive selection rules for E1 transitions without any computational efforts: we
simply apply the classical conservation law for angular momenta. For the system
atom + photon the total angular momentum and its projection onto a particular
direction in space is conserved during absorption or emission of a photon. Thus,
transitions may only be observed if the total angular momentum of the whole system
J is conserved. Denoting the total angular momentum operators of the atom with Ja
and J b in the lower and the upper state, respectively, the following rules in vectorial
form hold:
for absorption
J =Ja +
S ph Jb
(4.71)
for emission
J = J b J a + S ph .
absorption
(bottom, Dab = r ab e ) of
|a
circularly polarized light ma ma
(left: LHC + ) and y (at) y (at)
(right: RHC ). The k || z k || z
selection rules
m = 1 are
summarized in (4.74). Here x (at) in x (at) in
the atomic at and photon (ph) +(LHC) light -(RHC) light
coordinate frames are
identical, with the photon k mb k mb
wave vector k z(at) |b
emission
|a
out ma = m b - 1 out ma=mb +1
y (at) y (at)
x (at) x (at)
j = 0, 1 with ja , jb 0, but 0 0
abbreviated: (ja jb 1) = 1, (4.72)
while (ja jb 1) = 0 indicates that the triangular relation does not hold.
Clearly, the sum of the projections of the angular momenta before and after the
process have to be conserved as well. Hence, for dipole transitions the corresponding
selection rule is
q = m = mb ma = 0, 1. (4.73)
Schematically this is illustrated in Fig. 4.13 for absorption and emission of pure
circularly polarized + (LHC) and light (RHC), propagating in +z(at) direction.
Figure 4.14 shows the situation for linearly polarized light.
Note that due to the transverse nature of electromagnetic waves the selection rule
m = 0 refers to a coordinate system perpendicular to that in the two other cases:
for linearly polarized light it is convenient to choose the photon frame (ph) in respect
198 4 Non-stationary Problems: Dipole Excitation with One Photon
mb
z (at) |b
|a
ma= mb
y (at)
emission
x (at) k
e 0 || z (at)
of the atomic coordinate frame (at) with the electric field vector E z(ph) z(at) and
the light travelling in the x (at) y (at) plane, as illustrated in Fig. 4.14.11
The selection rules just outlined are necessary but not sufficient conditions for E1
transitions. To obtain quantitative expressions for the transition probabilities sum-
marized in (4.68) one has to evaluate | Dba |2 = |r ba e|2 explicitly. To do so, a spe-
cific coordinate system has to be chosen. For the unit polarization vector e we have
discussed suitable basis sets in Sect. 4.1.2. In principle, one may use Cartesian (xyz)
or polar coordinates (r ). Depending on the geometry of the experiment one or the
other choice may be advantageous. For linear polarization we have shown by way
of example that a Cartesian basis with e parallel to one of the axes leads directly
to the particularly simple result (4.70). However, since the angular part of atomic
wave functions is usually presented in polar coordinates, for a general description
the corresponding helicity basis (introduced in Sect. 4.1.2) is more appropriate for
expressing the polarization vector.
We introduce spherical components rq of the position vector r. With x =
r sin cos , y = r sin sin and Table B.1 one easily verifies
1
1
r = r = r = xex + yey + zez =
rq eq =r C1q (, )eq (4.76)
q=1 q=1
1
1
1
r C1q (, )eq = r (1)q C1q (, )eq = (1)q rq eq = r .
q=1 q=1 q=1
1
r ba e = b|r|a
e = b jb mb |r|a ja ma
e = mb |rq |ma
eq e
q=1
1
= b |r|a
jb mb |C1q |ja ma
eq e. (4.79)
q=1
12 More details for evaluating these matrix elements are made available in Appendix C.
200 4 Non-stationary Problems: Dipole Excitation with One Photon
Note that emission differs from absorption as described by (4.57). The polarization
vector e (creation of a photon) replaces e (annihilation of a photon):
1
r ab e = rab
ja ma |C1q |jb mb
eq e = (r ba e)
q=1
1
= rab (1)q ja ma |C1q |jb mb
eq e . (4.80)
q=1
We now discuss the simplest case, assuming that the light polarization is given by a
basis vector in the atomic frame, e := eq . All transition rates (4.68) are now propor-
tional to |
Dba |2 = |r ba eq |2 . With (4.79) this implies
2 2
b |r|a
jb mb |C1q |ja ma
.
(abs) (ind) (spont)
Rba = Rab dRab (4.81)
The selection rules for any E1 transition are essentially contained in the angular
part of the matrix elements. With (4.77) the angular part of the transition probability
(4.81) is
2
jb 1 ja
jb mb |C1q |ja ma
2 = (2jb + 1) jb C1 ja
2 . (4.82)
mb q ma
4.4 Selection Rules for Dipole Transitions 201
Whether a dipole transition is possible at all between two levels with total angular
momenta ja and jb depends on reduced matrix element jb C1 ja
. It does not
vanish if and only if the triangular relation (4.72) holds, as already gleaned from
angular momentum conservation.
Whether transitions can occur between specific sublevels depends on the 3j sym-
bol in (4.82). It can only be finite if the sum of its projection quantum numbers in
the second row vanishes (see Appendix B.2). Hence, (4.73) is confirmed as addi-
tional selection rule. The underlying conservation of total angular momentum has
been illustrated in Fig. 4.13.
Note: We point out again the differences between absorption and emission as a
consequence of replacing e in (4.79) by e to obtain (4.80) with r = r (real valued
vector). E.g., as spelled out in (4.74), + light absorption increases the projection
quantum number in the nal state by q = 1: mb = ma + 1; in contrast, emission
of + decreases of m in the nal state: ma = mb 1. But in both cases the se-
lection rule (4.73) holds per denition. This is a consequence of Dab = r ab e =
(r ba e) =
Dba : the transition probabilities depend only on the absolute squares of
the matrix elements. A corresponding consideration holds for light with q = 1.
For linearly polarized light, according to (4.74) and Fig. 4.14, the unit polar-
ization vector becomes e = e0 so that with (4.81) and (4.82) the transition rate is
simply proportional to
2
jb 1 ja
for E z(at) mb = ma . (4.83)
mb 0 ma
Thus, with such linearly polarized light one observes transitions with
m = 0, often
called transitions. In contrast to (4.74) the E vector of the photon and not its wave
vector k points into the z(at) direction. The propagation angle within the x (at) y (at)
plane does not play a role in this case.
For reference we also note that
1
|r ba |2 = mb |rq |ma
2 = |xba |2 + |yba |2 + |zba |2 . (4.84)
q=1
The explicit evaluation of the matrix elements in (4.81) may become quite com-
plicated, depending on the angular momentum coupling schemes in which the states
|ja ma
and |jb mb
are described. In later chapters we shall have to discuss this in
more detail for various examples.
For the moment we just evaluate the most simple case for single electron transitions
between uncoupled, pure orbital angular momentum states, |na a ma
and |nb b mb
,
ignoring the electron spin. Inserting (C.28) into (4.81) one obtains
202 4 Non-stationary Problems: Dipole Excitation with One Photon
Wn / eV
n =0 =1 =2 =3 =4
0
8
6
5s 5p 5d 5f 5g
-0.85 4 4s 4p 4d 4f
-1.51 3
3s 3p 3d
-3.40 2
2s 2p
-13.6 1 1s
Fig. 4.15 G ROTRIAN diagram for the H atom illustrating the selection rule
= 1 in E1 tran-
sitions (shown are only emission lines, the electron spin is neglected)
2 2
Rba na |r|nb
a ma |C1q | b mb
2
= na |r|nb
(2 a + 1)(2 b + 1)
2
2
a 1 b a 1 b
ma mb +q ( a 1 b ) . (4.85)
ma q mb 0 0 0
The first 3j symbol confirms again the selection rules (4.72) and (4.73). The second
3j symbol, however, adds an additional selection rule: since its projection quantum
numbers all vanish, a + 1 + b must be even, lest it disappears according to (B.49).
Thus with the constraint ( a 1 b ) = 1 a new selection rule holds:
a = b 1. (4.86)
The physical origin of this rule is parity conservation which is discussed in more
detail in Appendix D. Briefly, the overall parity of the system has to be conserved,
i.e. its symmetry in respect of inversion at the origin has to be the same before and
after absorption or emission of the photon. The angular part of the wave functions
for such pure states is completely described by spherical harmonics, their parity
being (1) . The photon which is represented by C1 has negative parity. The parity
of the whole system before absorption of the photon is thus (1) a +1 while it is
(1) b after the photon has been absorbed. Thus, for parity conservation a + 1 + b
must indeed be even.
Up to now, we have used pure states for the hydrogen and alkali atoms. As an
example, Fig. 4.15 shows the term scheme of the H atom with some E1 transitions.
Specifically, for an s p transition three processes are allowed: |0ms
|1mp
with
ms = 0and mp = 0 or 1. According to (C.44) all three angular matrix elements
are 1/ 3, independent of the substates, so that
|ns |r|np
|2
for an = 0 = 1 transition: Rab . (4.87)
3
4.5 Angular Dependence of Dipole Radiation 203
The electron spin, so fare neglected, will modify details of the term scheme
Fig. 4.15. The corresponding selection rules will be discussed in Chap. 6. In Ap-
pendix C one finds compact formulas for the evaluation of (4.82) in the general case.
In Chap. 7 we shall see, however, that transitions between states with angular mo-
mentum quantum numbers 0 and 1 are characteristic for closed shell multi-electron
systems. They occur e.g. among so called singlet states where the spins of several
electrons compensate each other. E1 transitions may then indeed be described by
the orbital angular momenta of the optically active electron, e.g. p s. In the
remainder of the present chapter we shall identify jb = 1 ja = 0 processes as
1 P 1 S transitions within a singlet system, indicated by the superscript, while
1 0
the subscript stands for j .
To actually calculate the transition probabilities, in addition to the angular part
of the matrix elements, which we have so far discussed in detail, one also has to
compute the radial transition matrix elements. For the H atom Appendix C.5 sum-
marizes the necessary formulas. A rather comprehensive collection of spontaneous
transition probabilities for many atoms of the periodic system is found at K RAMIDA
et al. (2013).
Section summary
Selection rules for light induced E1 transitions may directly be inferred from
angular momentum conservation.
For the total angular moments the triangular relation (ja jb 1) = 1 must hold,
the photon having a spin 1.
For the projection quantum numbers q =
m = mb ma = 0, 1 reflects the
photon polarization q = 1 (with respect th z(at) k) and q = 0 (with respect
th z(at) E).
Quantitatively, the transition probabilities are proportional to
2
jb 1 ja
2
rab (2jb + 1) jb C1 ja
2 .
mb q ma
We now turn to the angular characteristics for absorption and emission of electro-
magnetic radiation in dipole transitions. We make explicit use of the coordinate
frames introduced in Sect. 4.1.3: the atomic frame (at) for describing the atom and
the photon frame (ph) in which the light is best described. The latter is character-
ized by the polar (k ) and azimuthal angle (k ) of the wave vector k as shown in
Fig. 4.3. Section 4.5.1 presents a first somewhat heuristical semiclassical ap-
proach by interpreting the classical oscillator as dipole transition matrix element
(at)
r ab . In Sect. 4.5.2 we shall show that this may be cast into an exact quantum me-
chanical formulation with essentially identical results, now including all necessary
numerical prefactors.
204 4 Non-stationary Problems: Dipole Excitation with One Photon
The three components directly reflect the three types of classical oscillators illus-
trated graphically in Fig. 4.3. They correspond to the three types of transitions which
we have already summarized in (4.74):
For q = 0 ( light) the electron oscillates linearly in z(at) direction with an am-
plitude ma |r0 |mb
. From the classical picture we expect zero radiation in z(at)
direction and maximum intensity emitted into the x (at) y (at) plane with linear po-
larization z(at) .
The terms with q = +1 and q = 1 represent an electron on a circular orbit in the
x (at) y (at) plane. The emitted light may propagate into the whole solid angle 4 .
Maximum intensity and fully circularly polarized light is intuitively expected
for propagation into +z(at) direction, while light detected in the x (at) y (at) plane
will be linearly polarized z(at) axis.
More quantitatively, one may glean electric field amplitudes from (4.29) and (4.30)
(at)
by projecting r ab onto the x (ph) y (ph) plane. For light emitted from a linear oscil-
(at)
lator with amplitude ma |rq |mb
e0 one reads from Fig. 4.3
(ph) (ph)
Ey = 0 and Ex ma |r0 |mb
sin k (4.89)
z (at) k
k= 90 k< 90
E,
k
x (at) y (at)
E, k
Fig. 4.16 Angular distribution of dipole radiation from a linear oscillator ( light)
I (k , ) sin2 k . For two different directions of the wave vector k the alignment of the polar-
ization vector is indicated (red, full and dashed vector arrows, respectively)
z (at) k
E, +
k
y (at )
k
x (at) E,
k
Fig. 4.17 Angular distribution of dipole radiation from a circular oscillator ( light)
I (k , ) (1 + cos2 k )/2. The polarization vector is indicated for two different directions of
observation: at 0 < k < /2 (full red ellipse with vector arrow), corresponding to elliptically
polarized light, and for k = /2 (dashed red arrow) corresponding to linear polarization
In Fig. 4.18 the standard geometry for observation of the different transition types
is shown, as e.g. realized for observing the so called normal Z EEMAN effect (see
Sect. 2.7). The three cases q = 0, 1 may in principle be observed independently
since the corresponding atomic lines are split in the magnetic field, with frequencies
0 and 1 = 0 L /2 for transitions with
m = q = 0 and 1, respectively.
206 4 Non-stationary Problems: Dipole Excitation with One Photon
(a) (b) - +
(c)
-
z (at ) +
-1 0 1
+1 spectrum
1P
1 0
mb = -1
-1 0 1
Fig. 4.18 Normal Z EEMAN effect as realized in a 1 S0 1 P1 transition; (a) energy levels in an
external magnetic field B ex z(at) , (b) experimental setup, (c) spectrum from light emitted at
k = 0, (d) spectrum from and light emitted at k = /2
We emphasize again that the level scheme Fig. 4.18(a) should not be misinter-
preted: the arrows indicating transitions with 1 and 0 refer to different coordinate
systems. The external magnetic field defines here the direction of z(at) B ex ; emis-
sion parallel to that axis (i.e. k z(at) B ex ) implies transitions with
m = q = 1
( light) only; in contrast,
m = q = 0 transitions imply light, linearly po-
larized with the electric field E B ex , emitted into any direction except in z(at)
direction.
What has been deduced above from a more or less classical picture also follows from
a rigorous quantum mechanical treatment. The angular and polarization character-
istic of emission and absorption rates are obtained by inserting (4.80) and (4.79),
respectively, into the general expression (4.68). Typically, the matrix element r ba
and the polarization vector e are described in different coordinate frames, most con-
veniently in the atomic (at) and photon frame (ph) as illustrated in Fig. 4.3. The
(spont) (ind) (abs)
transition rates dRab , Rab , and Rba between two angular momentum basis
states |jb mb
|ja ma
are then proportional to
1 2
(at) (ph) 2 (at) (ph) 2
r e = r e = mb |rq |ma
eq(at) e(ph)
ba ab
q=1
1 2
(ph)
= b |r|a
jb mb |C1q |ja ma
eq e .
(at)
(4.93)
q=1
4.5 Angular Dependence of Dipole Radiation 207
We emphasize that the unit vector e(ph) represents the polarization of the
Note that irrespective of the direction of the above arrows (4.93) holds as writ-
ten (see also the note on p. 201). To obtain practical relations for the angular dis-
(at)
tributions, such as (4.90) and (4.92), we have to express eq and e(ph) in the same
coordinate system. Which of these frames is more convenient depends on the exper-
imental situation. For illustration we carry this out explicitly as a kind of exercise
on coordinate rotation, using the rules explained in Appendix E.
m = q = 0 ( component)
m = q = 1 ( + component)
k (ph) k (ph)
e1(at) =e ik
cos2 e1 + sin2 e1 . (4.96)
2 2
m = q = 1 ( component)
k (ph) k (ph)
e1 = eik sin2 e1 + cos2 e1 .
(at)
(4.97)
2 2
208 4 Non-stationary Problems: Dipole Excitation with One Photon
Note that in the photon frame only the q = 1 basis vectors are considered, with
polarization perpendicular to the photon wave vector k. Only these are of physical
(at) (ph)
relevance in the scalar products (4.93) eq e(ph) . The e0 component is dropped
since it would indicate (un-physical) linear polarization in z(ph) direction. Hence,
(at)
eq is no longer a unit vector and its magnitude depends on k . We may, however,
express it in terms of an angle dependent normalization factor fq (k , k ) and the
(ph)
general unit vector (4.15) for elliptically polarized light in the photon frame eel :
(ph)
e(at)
q = fq (k , k )eel , so that (4.98)
(at)
(ph)
r ab e(ph) = ma |rq |mb
fq (k , k )eel e(ph) . (4.99)
q
Assuming we can detect one specific transition q =
m only, and register all light
(ph)
emitted from this transition without further discrimination, i.e. e(ph) := eel , the
emission rate becomes proportional to
(at) (ph) 2
r e = ma |rq |mb
2 fq (k , k )2 . (4.100)
ab el
One easily verifies that this probability has indeed the dimension T1 . The intensity
of radiation emitted per atom seen at a detector of an area
A in a distance R is
obtained by multiplying (4.101) with the photon energy ba , dividing it by
A,
and identifying the solid angle with d =
A/R 2 :
4
(spont) ba mb |r0 |ma
2 sin2 k .
I = 2 2
(4.102)
2c R
This angular distribution corresponds exactly to the previously derived semi-
classical pattern (4.90) shown in Fig. 4.16. The angular part jb mb |C1q |ja ma
of
the matrix element mb |r0 |ma
according to (4.77) is typically on the order of 1,
while the radial matrix element b |r|a
may vary over a wide range depending on
the size of the atomic orbitals and the overlap and symmetry of the wave functions.
We now turn to the circular components of the emitted radiation pattern
(q = 1). The atomic basis vectors projected onto the photon frame are
e1 = f1 eel = f1 (k , k ) ei cos e+1 ei sin e1 .
(at) (ph)
(4.103)
(ph)
The parameters and of the unit elliptic basis vector eel (4.15) are found from
a comparison of (4.103) with (4.96) and (4.97). We see immediately that = /2
4.5 Angular Dependence of Dipole Radiation 209
(the main axis of the polarization ellipse is y (ph) ); with cos2 +sin2 = 1 and with
the trigonometric identity 2(cos4 (k /2) + sin4 (k /2)) = (1 + cos2 k ) we derive
(1 + cos2 k )1/2
f1 (k , k ) = ieik (4.104)
2
For emission at k = 0, i.e. along the z(at) axis we have cos = 1 or = 0 and
sin = 0 or = 1 for q = +1 or 1, respectively. As discussed already, this implies
emission of pure circularly polarized +(LHC) or (RHC) light. In contrast,
at k = /2 one finds cos = sin = 1/ 2, corresponding to light emission with
linear polarization vector in the x (at) y (at) plane (see Fig. 4.3).
The angular distributions for
m = q = +1 and
m = q = 1 follow naturally
by inserting f1 (k , k ) from (4.104) into (4.100) and this in turn into (4.67). Again,
(ph)
the overall intensity of the emitted light is found by setting e(ph) := eel . We finally
obtain for q = 1 transitions
3
mb |r1 |ma
2 1 + cos k d,
ba 2
(spont)
dR = (4.106)
2c2 2
and the intensity per atom at a distance R is
4
(spont) ba mb |r1 |ma
2 1 + cos k .
2
I = (4.107)
2c2 R 2 2
This confirms again the angular characteristics (4.92) of the radiation found by the
semiclassical model as illustrated Fig. 4.17.
We may integrate the radiation characteristic (4.101) and (4.106) over the full
solid angle and thus derive the overall spontaneous emission probability for one
specific transition. Both, the sin2 k distribution (4.103) for the -components as
well as the (1 + cos2 k )/2 distribution (4.107) for the transitions lead to the same
factor 8/3:
8 1
sin k d = 2
2
sin k dk =
3
= 1 + cos2 k d. (4.108)
4 0 3 4 2
Hence, from (4.101) as well as from (4.106) the integral spontaneous emission prob-
ability for a specific transition with ma + q = mb becomes:
13 The overall phase factor i exp(ik ) is here of no significance for measurable quantities, since
these are proportional to the absolute squares of the matrix elements.
210 4 Non-stationary Problems: Dipole Excitation with One Photon
3
(spont) 4 ba mb |rq |ma
2
Rq = A(ja ma ; jb mb ; q) = 2
3 c
3
4 ba b |r|a
2 jb mb |C1q |ja ma
2 .
2
(4.109)
3 c
Equations (4.95)(4.109) describe spontaneous emission processes between well
defined states |ja ma
|jb mb
. This situation is realized experimentally if a spe-
cific initial state is prepared and only one specific final state is detected. Then the
angular distributions (4.102) and (4.107) are observed. One method (by far not the
only one) to prepare and detect such specific transitions is the Z EEMAN effect as
illustrated in Fig. 4.18.
We finally mention that a more general theory of radiation emitted from excited
atoms will be presented in Sect. 9.4, Vol. 2 as an interesting example for applications
of the density matrix.
I(ba ) 2 sin2 (k )
Rba = 4 2 mb |r0 |ma
.
2
4. Finally, if the same transition can also be induced by linearly polarized light
(ph)
x (ph) , i.e. if e(ph) := ex . We find
I(ba ) 2
Rba = 4 2 ma |r0 |mb
sin2 (k ),
which obviously is a more efficient process. The pattern is the same as shown in
Fig. 4.16.
Section summary
We have derived the angular distributions of radiation emitted for q = 0 (q =
mb ma ) transitions and for q = 1 from a semiclassical picture as well as
from quantum mechanical rates for E1 transitions.
We should memorize these distributions according to Figs. 4.16 and 4.17.
The polarization depends on the emission angle k with respect to the z(at)
axis. For q = 0 transitions, only linearly polarized light ( light) is emitted,
with its polarization vector parallel to z(at) for k = /2. For q = 1 transi-
tions, at k = 0 pure circularly polarized light ( light) is emitted, while at
k = /2 one finds linearly polarized light with polarization vector perpen-
dicular to z(at) . For other emission angles the light is elliptically polarized.
The three components of radiation may be separated experimentally, e.g. by
exploiting the Z EEMAN effect as illustrated for a 1 S0 1 P1 transition in
Fig. 4.18.
For induced transitions one may derive the angular dependence of rates for
transitions induced between specific sublevels by a specific polarization. The
general case is more complex however, as we shall see in a moment, it
may be greatly simplified for an initially isotropic population of all sub-
states.
212 4 Non-stationary Problems: Dipole Excitation with One Photon
According to (4.68) the key parameter for all absorption and emission processes is
|r ba e|2 . It contains the necessary information on polarization and angular charac-
teristics for E1 transitions between specific substates. In contrast, to characterize the
overall strength of a dipole transition between levels ja and jb with several degen-
erate ma and mb substates one introduces (see also Appendix H.2) a symmetrically
defined line strength with the dimension L2 :
S(jb ja ) = |r ba |2
mb ma
= b jb mb |rq |a ja ma
2
mb ma q
2
= (2jb + 1)b |r|a
jb C1 ja
2 (4.112)
2
= (2ja + 1)b |r|a
ja C1 jb
2 S(ja jb ).
Here we have used (4.84) with (4.77), (4.82) and the 3j orthogonality relation
(B.41). The radial matrix element b |r|a
is given by (4.78). We recall that b and
a represent all quantum numbers needed to characterize the radial wave function.
We also note that ja and jb may refer to various kinds of the angular momentum
quantum numbers, representative for the states under discussion in the simplest
case these may be orbital angular momenta or L, but could also be total angular
momenta as we shall discuss in Chaps. 6 and 9.
The line strength is a useful reference quantity. However, when describing spe-
cific cases one has to distinguish between types of radiation (polarized or unpolar-
ized, unidirectional or averaged over all solid angles) as well as between types of
transitions (selected substates, sums over final states, averages over initial states).
This will be discussed below in terms of E INSTEIN coefficients. Figure 4.19 defines
these for transitions between the substates |jb mb
and |ja ma
with polarizations
characterized by q = mb ma .
jb mb
B( ja ma ; jb mb ;q)
A( ja ma ; jb mb ;q)
B( jb mb ; ja ma ;q)
ja ma
Fig. 4.19 E INSTEIN coefficients for absorption, B(jb mb ; ja ma ; q), for induced emission,
B(ja ma ; jb mb ; q), and for spontaneous emission, A(ja ma ; jb mb ; q), between substates of an up-
per and a lower level characterized by the angular momentum quantum numbers jb and ja , respec-
tively
4.6 Strength of Dipole Transitions 213
The spontaneous transition probability (4.109) for one specific transition |ja ma
|jb mb
and polarization q is now rewritten with (4.82), (4.77) and (4.112)
A(ja ma ; jb mb ; q)
3
2
4ba 2 jb 1 ja
=
b |r|a
(2jb + 1) jb C1 ja
2
3c2 mb q ma
3
4ba 3
2
= mb |rq |ma
2 = 4ba jb 1 ja
S(jb ja ). (4.113)
3c2 3c2 mb q ma
It has the dimension T1 . The 3j symbol accounts for all dependencies on angular
momenta and polarization while all specific properties of the atom (or molecule) are
included in S(jb ja ).
Total decay probability and spontaneous lifetime of the upper level are deter-
mined by summing over all final states and polarizations:
1
A ab = = A(ja ma ; jb mb ; q)
ja jb ma q
jb
2
3 3
(4.114)
4ba 1 ja 4ba S(jb ja )
= S(jb ja ) = .
3c2 mb q ma 3c2 (2jb + 1)
m q a
Here we have again used the orthogonality relation (B.41) for the 3j symbols. It is
important to note that the spontaneous lifetime ja jb is independent of the quantum
number mb of the sublevel in the excited state!
Explicitly, the prefactor in the above equations is given by
3
4ba 32 3 c 19 3 s
2 7.235 108 m
= = 1.083 10 ba 2 = . (4.115)
3c2 33ba m 3ba s
For practical use we insert the line strength (4.112) into (4.114) and introduce atomic
units a0 and Eh :
4 3 2
2
A ab = 2 3 Wba a0 b |r/a0 |a
jb C1 ja
2
3c
4 5 me c2 2
= (Wba /Eh )3 b |r/a0 |a
jb C1 ja
2 (4.116)
3
2.1420 1010 2
= (Wba /Eh )3 b |r/a0 |a
jb C1 ja
2 .
s
The reduced matrix element jb C1 ja
is typically on the order of 1. For the sim-
plest case, j = , it is given by (C.30), in the general case it may be evaluated as
explicated in Appendix C.3. The radial matrix element expresses the genuine atomic
214 4 Non-stationary Problems: Dipole Excitation with One Photon
physics: it is determined by the overlap between initial and final state wave func-
tions, weighted with the radius.
The dependence of A on the nuclear charge Z is particular interesting. For hy-
drogen like ions we have Wba Z 2 and r 1/Z, so that the spontaneous lifetime
ja jb = 1/A ab Z 4 decreases with the 4th power of the nuclear charge, a result
which is important for atomic X-ray physics.
With (4.114) one may derive the line strength directly from the experimentally
accessible natural lifetime of the excited state:
3c2 (2jb + 1) 1
S(jb ja ) = 3
. (4.117)
4 ba ja jb
On the other hand, with (4.113) the individual transition probabilities between indi-
vidual substates may be written as
2
jb 1 ja 1
A(ja ma ; jb mb ; q) = (2jb + 1) , (4.118)
mb q ma ja jb
and (4.77), the transition amplitude between substates, may be written
jb 1 ja
mb |rq |ma = (1)jb mb S(jb ja ). (4.119)
mb q ma
We also recall at this point, that even though according to (4.114) the excited state
lifetime is identical for all excited substates the angular radiation characteristic and
the polarization of the emitted light depends on mb and q as described by (4.101)
(4.109).
Experimentally one often detects the fluorescence of an atom without polariza-
tion and final state analysis. The respective transition probability is obtained by
summing over all polarizations q according to (4.101)(4.106) and over all final
states:
(spont)
dRja jb mb ( ) = d 3 S(jb ja ) (4.120)
2c2 ba
*+
2 ,
jb 1 ja
2
jb 1 ja 1 + cos2 k
+
mb 1 ma mb 1 ma 2
ma
2 -
jb 1 ja
+ 2
sin k .
mb 0 ma
Usually, this still needs to be averaged over all populated initial states mb and de-
scribes in general a non-isotropic angular distribution. Only if all initial excited
states |jb mb
are equally populated ( (2jb + 1)1 ) the 3j orthogonality relation
(B.41) may be applied and finally leads to an isotropic distribution:
(spont) 1 d
dRtot = . (4.121)
jb ja 4
4.6 Strength of Dipole Transitions 215
The general case of radiation from an anisotropically populated atom will be treated
in Sect. 9.4, Vol. 2.
We now consider induced processes, i.e. absorption and stimulated emission. For
transitions between specific substates induced by a polarization q (in respect of the
atomic frame) we rewrite (4.66) using (4.79):
4 2 c 2 2
B(jb mb ; ja ma ; q) = b |r|a
jb mb |C1q |ja ma
(4.122)
= B(ja ma ; jb mb ; q).
A(ja ma ; jb mb ; q) ba
3
4 h
= = . (4.123)
B(ja ma ; jb mb ; q) 3 2 c3 3 3ba
Note that this ratio differs by a factor 1/3 from (4.38) where isotropic radiation
and summation over all final states is assumed. In contrast, presently a transition
between specific substates is considered. The spontaneous rate has been integrated
over all emission angles to obtain A, while B has been determined for a specific
polarization and direction of the inducing light.
In full analogy to (4.113) we may write the coefficients for stimulated emission
and absorption between specific sublevels as
2
4 2 c jb 1 ja
B(ja ma ; jb mb ; q) = B(jb mb ; ja ma ; q) = S(jb ja )
mb q ma
2
3 2 c3 (2jb + 1) jb 1 ja
= (4.124)
ba3 jb ja m b q ma
2
33 jb 1 ja
= (2jb + 1) A ab ,
4h mb q ma
1
B ba = B(jb mb ; ja ma ; q)
2ja + 1 m m
b a
1 4 2 c jb 1 ja
2
= S(jb ja ) (4.125)
2ja + 1 mb q ma
mb ma
4 2 c S(jb ja ) 2 c3 1 3ab
B ab = = = A ab . (4.126)
3 (2jb + 1) ba 3
ja jb 4h
The relations (4.125) and (4.126) just derived express a very important property of
absorption and induced emission without which quantitative spectroscopy would be
much more difficult (see also Appendix H.2):
2 c3 3ab
ga B ba = gb B ab and B ab = A ab = A ab . (4.127)
ba3 4h
The agreement with (4.38) eventually justifies the hand waving introduction of nu-
merical factors for spontaneous emission in (4.67). However, we emphasize again,
that this is not a derivation of the spontaneous emission rate which we postpone to
Chap. 2 in Vol. 2.
Section summary
The line strength S(jb ja ) introduced in (4.112) characterizes the overall
strength of an atomic transition between energy levels jb and ja .
Various relations among the E INSTEIN A and B coefficients have been de-
rived which are useful for practical computations.
The spontaneous lifetime of substates in a given excited level jb is indepen-
dent of the initial projection quantum number mb .
Absorption as well as stimulated emission in an isotropically populated target
is independent of the polarization of the light used!
4.7 Superposition of States, Quantum Beats and Jumps 217
So far we have discussed only transitions between basis states in the atomic frame.
But of course, initial and final states may also be linear superpositions of these.
Even if the initial state is a basis state, absorption as well as emission of light with
polarization e(ph) may create a linear superposition of states |(b)
and |(a)
,
respectively:
(b) mb |rq |ma
eq(at) e(ph) |jb mb
(4.128)
mb q
(a) ma |rq |mb
e(at) e(ph) |ja ma
. (4.129)
q
ma q
Often one cannot choose the atomic frame arbitrarily, since the experimental
setup already defines a preferential (at) frame, e.g. if an external magnetic or electric
field is applied, or if one wants to analyze a scattering experiment where a well
defined collision plane exists. The atom has to be described then in an appropriately
chosen coordinate system, which may differ from the (ph) frame which refers e.g.
to the position of a photon detector or to a laser beam for exciting the target. We
choose atomic (at) and photon system (ph) again according to Fig. 4.3 and shall
have to transform either e(at)
q into the photon frame or e
(ph) into the atomic frame as
- ( at ) y ( at )
k k
x ( at ) k x (ph )
i
() = (b) mb |r1 |ma
eik Y11 (, ) + eik Y11 (, )
2
rba 1
= sin sin( k ). (4.137)
2
(b) 1 eik mb |r1 |ma
|jb ma 1
+ eik mb |r1 |ma
|jb ma + 1
2
1
+ mb |r0 |ma
|jb ma
. (4.138)
2
Apart from the phase factor i the first term of (4.138) and (4.139) [in square brack-
ets] is obviously identical to the wave function generated by linearly polarized light
as described by (4.136) and (4.137) and imaged in Fig. 4.20. The second term cor-
responds to a dumbbell in z(at) direction.
220 4 Non-stationary Problems: Dipole Excitation with One Photon
Fig. 4.21 Illustration of a 1 P1 1 S0 transition, induced by circularly polarized light which prop-
agates in y (at) direction
with j 0 + mj = Wj m /. The atomic frame is chosen such that z(at) B, and in all
relevant equations of the preceding section the basis states |j m
have to be replaced
by (4.141). A common trivial time dependence exp(ij 0 t) may be pulled out of
the expressions and contributes to an overall non-measurable phase. However, the
4.7 Superposition of States, Quantum Beats and Jumps 221
frequency shift mgj L = mgj 8.79 1010 s1 T1 B is different for each magnetic
substate and leads to changing amplitudes.
We first have a look at the simplest case, excitation by linearly polarized light
described by (4.136) and (4.137). The E vector of the light is assumed to be per-
pendicular to the magnetic field. Inserting the states (4.141) into (4.136) instead of
|j m
and identifying k = k (0) with the phase angle at time t = 0 we obtain the
angular part of the wave functions for a 1 P1 state:
irba
(, , t) ei10 t
6
i( (0)+g t) i(k (0)gj L t)
e k j L Y
11 (, ) + e Y11 (, ) (4.142)
rba 1 i10 t
= e sin sin k (t) (4.143)
2
with k (t) = k (0) + gj L t. (4.144)
The charge distribution |(, , t)|2 contains now a time dependent azimuthal an-
gle k (t). In order to observe this dynamics, the time origin has to be known with
sufficient accuracy in any case better than 1/(2gj L ). For a magnetic field on the
order of 1 T it should be around 0.1 ns. Assume then the atom is excited by a short
light pulse. The dumbbell shown in Fig. 4.20 will then rotate around the z(at) axis
(k (t) decreases with time) with an angular frequency gj L which may be observed
as quantum beat. In contrast, if the atom is excited by a continuous wave (CW) light
beam the L ARMOR precession will tend to generate a disc like charge cloud.
The other example discussed above, excitation by circularly polarized light prop-
agating perpendicular to the magnetic field according to (4.138), may also lead to
quantum beats. Since m = 0 states do not show an energy shift in the magnetic field
the corresponding component Y10 (, ) of the wave function remains unchanged.
In contrast, the Y11 components describe again a dumbbell rotating in the x (at) y (at)
plane as before. Thus, pulsed excitation leads to counter clockwise rotation around
the z(at) axis of the toroidal charge distribution |+1
(ph) shown in Fig. 4.21. For CW
excitation a completely isotropic distribution is expected.
No such dynamics will emerge if the exciting circularly polarized light is prop-
agating parallel to the B field (z(ph) B): the toroidal charge cloud will just rotate
around its symmetry axis without observable change of shape or alignment. The
same holds for linear polarization with e B. In summary, the dynamics in a lin-
ear superposition of non-degenerate states can only be observed if the substates are
populated non-isotropically and if a suitably geometry is chosen.
How can one actually measure such dynamics or wave-packets of the charge
cloud distribution? The so called quantum beat spectroscopy exploits these oscilla-
tions and is used today in several varieties as very efficient method to measure small
energy differences in excited states. Two (or more) neighbouring energy levels in
the excited state are needed as indicated in Fig. 4.22. The initial ground state |a
is excited coherently into the two states |b1
and |b2
by a short light pulse. For
222 4 Non-stationary Problems: Dipole Excitation with One Photon
|b 2
21
|b1
|f
|a
Fig. 4.22 Schematic of a typical four level scheme for the observation of quantum beats. Note
that the excited levels must be populated coherently. Quantum beats are generated by interference
of the emission lines from different excited states decaying into the same final state
the moment,14 the term coherent simply expresses a well defined phase relation
among the excited states as a consequence from a sufficiently short duration of the
exciting light pulse 1/21 (with 21 being the excited state level splitting).
Equivalently, this implies a F OURIER limited bandwidth, broad enough to simulta-
neously excite both levels b1 and b2 . At time t = 0 the excited state is then given
by
(0) = c1 |b1
+ c2 |b2
, (4.145)
where cj are e.g. the coefficients used in (4.142). However, the following consider-
ations are valid much more generally, for two or more states which are excited co-
herently and have slightly different energies Wj . Quantum beats are not restricted to
j mj sublevels which are split in a magnetic field, as just discussed. Other examples
are highly excited RYDBERG states in atoms, molecular rotational and vibrational
states etc.
The excited states evolve in time according to cj |bj
exp(iWj t/) with
Wj / = j . Let us now assume that these states decay by spontaneous emission
into a lower lying, final state |f
. The lifetime of the excited states j = 1/Aj
is accounted for by a damping term exp(Aj t/2) for the amplitude, i.e. by
exp(Aj t) = exp(t/j ) for the probability:
(t) = c1 |b1
e(i1 +A1 /2)t + c2 |b2
e(i2 +A2 /2)t . (4.146)
intensity
+1
0
Re (FT)
-1
intensity
0 2 4 6
t/ s
(b)
Re (FT)
0 4 8 12 0 1 2 3 4 0 10 20 30 40
t/ s / MHz / MHz
Fig. 4.23 Quantum beats in molecules according to C ARTER and H UBER (2000): (a) fluorescence
of the Z EEMAN levels for a R(0) line of the transition 17U in CS2 (magnetic field B 1.5 mT). The
laser polarization was perpendicular to B and enables coherence between the M = 1 sublevels;
(b) fluorescence after coherent excitation of the 000 101 rotational line in the vibrational band
(3,0)
231+
0 I(0,0) of the S1 S0 transition in acetone; quantum beats arise from a mixture between
singlet and triplet states; (a ) and (b ) give the real parts of the F OURIER transformed signals (a)
and (b), respectively
with ed being the polarization registered in the detector system. Obviously, the tran-
sition amplitudes for both processes |f
|b1
and |f
|b2
interfere. Inserting
(4.146) into (4.147) one obtains
Quantum beats have already been observed in the 60s of the past century. How-
ever, only after the introduction of tuneable, pulsed lasers it was possible to study
and use them in full beauty. Figure 4.23 presents two nice examples from molecu-
lar physics: (a) a rather clear spectrum is shown for a rotational transition of free
CS2 molecules in a magnetic field where exactly two Z EEMAN levels participate.
The F OURIER transform of the temporal evolution in (a ) shows essentially one dif-
ference frequency. A somewhat more complicated molecule is acetone for which
a quantum beat spectrum is reproduced in (b). It originates from contributions of
several, close lying electronic states. The temporal spectrum thus looks somewhat
intricate, and the F OURIER transform (b ) documents that more than four transitions
participate in the beats. The fact that the signals are not fully modulated from zero
to maximum is attributed to additional, incoherent contributions from other levels.
|m
itati
3
esce
on
ng
0
nce
rki
|g time / s
Fig. 4.24 Quantum jumps as observed by S AUTER et al. (1988). (a) Part of the term scheme for
Ba+ and the transitions involved. (b) One observes the fluorescence of the state |e
as a function
of time. Grey shaded are times where the fluorescence signal vanishes while the ion is parked in
state |m
Acronyms and Terminology 225
lived excited state |e
. One observes the fluorescence signal emitted from state |e
.
At the same time one also induces (with a much smaller probability) a transition
to the metastable state |m
from where the ion can decay only with a very low
probability back into the ground state (the average lifetime |m
is several seconds).
As documented in Fig. 4.24(b) the almost continuous fluorescence signal emitted
from |e
is interrupted in irregular intervals (grey shaded) a clear indication that
the ion is not available for the excitation-fluorescence cycle. Remember, there is
always one but only one ion in the trap! We thus have to conclude that during the
dark periods the ion is parked in the excited, metastable state |m
. Obviously, most
of the time one observes a fluorescence signal, i.e. the ion spends most of its time in
the ground state |g
from where it can easily be excited for 108 s into the state |e
and thus be monitored. The intermissions by excitation into |m
are rather infrequent
(intervals of up to more than 100 s) and have different duration times. A statistical
analysis of these intermissions (not shown here) simply reproduces the exponential
lifetime distribution of the state |m
.
This experiment thus documents very explicitly that the ion exists always in a
very well defined state: except for very short times (on average for some 108 s)
in the fluorescing state |e
the ion is found either in the ground state |g
and may
be detected by excitation-fluorescence cycles, or it is found in the (dark) meta sta-
ble state |m
and cannot be detected by fluorescence. There is nothing in between
and a transition from one into the other state occurs (arbitrarily fast) by a quantum
jump. It is only the probability for such a quantum jump which is described by the
exponential decay function for the state |m
.
Section summary
Absorption and induced emission with fully polarized light (linear, circular
or elliptic polarization) leads in general to a coherent superposition of basis
states. We have derived closed expressions for these states as a function of
polarization.
If at least two states b1 and b2 with slightly different energies b1 and b2
are exited coherently and with an anisotropic overall population, one may
observe quantum beats if they radiate to the same final states. Coherent exci-
tation in this context implies short pulses with a duration
t 1/
b2 b1 .
Single ion storage allows one to observe quantum jumps in an atomic system
which is excited with a certain, low probability: the atom is found either in
the excited or in the ground state. The observed change of population occurs
suddenly and in statistically determined time intervals.
CW: Continuous wave, (as opposed to pulsed) light beam, laser radiation etc.
E1: Electric dipole, transitions induced by the interaction of an electric dipole
with the electric field component of electromagnetic radiation.
226 4 Non-stationary Problems: Dipole Excitation with One Photon
References
B ETH , R. A.: 1936. Mechanical detection and measurement of the angular momentum of light.
Phys. Rev., 50, 115125.
C ARTER , R. T. and J. R. H UBER: 2000. Quantum beat spectroscopy in chemistry. Chem. Soc.
Rev., 29, 305314.
D EHMELT , H. G. and W. PAUL: 1989. The N OBEL prize in physics: for the development of the
ion trap technique, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1989/.
ISO 21348: 2007. Space environment (natural and artificial) process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
K RAMIDA , A. E., Y. R ALCHENKO, J. R EADER and NIST ASD T EAM: 2013. NIST Atomic
Spectra Database (version 5.1), NIST. http://physics.nist.gov/asd, accessed: 7 Jan 2014.
S AUTER , T., H. G ILHAUS, I. S IEMERS, R. B LATT, W. N EUHAUSER and P. E T OSCHEK: 1988.
On the photo-dynamics of single ions in a trap. Z. Phys. D, 10, 153163.
Linewidths, Photoionization, and More
5
Overview
Section 5.1 introduces realistic finite linewidths into the description of optical
transitions. It should be easy to read and is of central importance for the entire
spectroscopy. In Sect. 5.2 closely related to this topic cross sections for
excitation involving lines of finite width are discussed. The concept of optical
oscillator strength is introduced (for which Appendix H.2 provides additional
background). Section 5.3 offers a brief introduction into multi-photon pro-
cesses without which much of modern spectroscopy would not be possible.
Likewise, E2 and in particular M1 processes are important in many areas of
spectroscopy; the reader may, nevertheless, consider the somewhat mathemat-
ical Sect. 5.4 as being mainly for later reference without loosing the present
context. This holds to some extend also for Sect. 5.5 where photoionization
is treated which plays a key role in many areas of modern physics. We shall
make use of it later, e.g. in Sect. 7.6.2 and Chap. 10.
Even if we are still missing some key tools for treating spontaneous emission it
is important to develop a pragmatic way to work with it correctly in the context of
spectroscopy. One essential consequence is the finite lifetime =
t of stationary
states and the finite width =
W of atomic levels, both connected through the
uncertainty relation (1.125). In Sect. 4.7.2 we have already introduced in passing a
Wba
|a
An atom which is prepared in this excited state at time t = 0 will decay with time
t according to the probability distribution
2
wj (t) = cj (t) = eAj t = et/j .
In the spirit of perturbation theory the former 0th order approximation cj (t) = 1
now becomes cj (t) = exp(t/2j ). In the ansatz (4.41) we thus replace
j (r)eij t by
j (r)et/2j eij t = j (r)ei(j i/2j )t . (5.1)
One may read this equation as if excited states were now characterized by a complex
eigenfrequency or eigenenergy, substituting
i Aj
j j = j i or (5.2a)
2 2
j
Wj Wj i with j = Aj = . (5.2b)
2 j
dcb (t) i
= cj (t)Ubj (t)ei(bj tiAj t/2) , (5.3)
dt
j
5.1 Line Broadening 229
and instead of (4.58) the final result in 1st order perturbation theory is
i
cr = cb (t ) . (5.5)
ba iA/2
cr is a typical resonance amplitude, well known from classical physics where it de-
scribes a forced oscillator. It depends on the detuning ba and on the damping
constant A.
LORENTZ Prole
Somewhat more precisely, the decay constant A used here has to be identified with
A ab given in (4.114)(4.117) as long as only transitions between two isolated lev-
els are considered. From (5.5) follows the probability to find an atom in the excited
state if CW light of angular frequency is absorbed:
A2 /4
cb ()2 gL () = 2 . (5.6)
A ( ba )2 + A2 /4
This is the well known L ORENTZ profile (L ORENTZ distribution), describing an ab-
sorption line with a natural linewidth. The distribution is given here per unit angular
frequency, i.e. it has the dimension T. Its FWHM is
nat = A A ab . In respect of
energy the L ORENTZ profile reads
2 ( /2)2 /2
gL (W ) = = . (5.7)
(W Wba ) + ( /2)
2 2 (W Wba )2 + ( /2)2
nat is related to the frequency bandwidth
nat , the energy width of the state,
its natural lifetime nat and its spontaneous decay rate A:
1
nat = 2
nat = A = = . (5.8)
nat
Both profiles, gL () and gL (W ), are normalized to unity:
gL ()d = gL (W )dW = 1. (5.9)
I()
dRba = BgL () d. (5.10)
c
B is the E INSTEIN coefficient according to (4.124) or (4.126), depending on exper-
imental conditions.1 The rate for spontaneous emission of a photon with angular
frequency is given by
(spont)
dRab = AgL ()d,
with the spontaneous decay rate A A ab , equal to the inverse lifetime of any of the
substates of level b in respect of decay into all substates of level a.
We may now introduce a frequency dependent absorption coefficient by multi-
plying B ba according to (4.126) with the normalized line shape (5.6):
gb 3ab A 2ab /4
B ba () = B ba gL () = . (5.12)
ga 2h ( ba )2 + A 2ab /4
B ba is the averaged absorption probability from the lower levels a to the upper levels
b for a given polarization.2
The attentive reader may find the above derivation of the frequency dependent
absorption coefficient somewhat heuristic. We shall come back to this in Sect. 10.5,
Vol. 2 where a more systematic deduction will be presented which turns out to
lead to identical results as the one above.
1 Ifinstead of individual transitions between substates, averaged probabilities are studied, we have
to replace in (4.68) the squared transition matrix element
1 1
|
Dab |2 = |r ab e|2 by |
Dab |2 = |r ab e|2 , (5.11)
ga m m ga m m
a b a b
2 If one is interested in the absorption profile for one specific transition between substates
jb mb ja ma , one would have to replace in (5.12) the prefactor (gb /ga ) 3 and the linewidth
A A(jb mb ja ma ; q) according to (4.123).
3 If several excited or lower state sublevels are involved one has, in addition, to account for optical
decay constant A = Aab for this particular transition (see footnote 2). We replace
the spectral radiation density u() I /c with I0 being the intensity of the (very
narrow band) laser radiation. Then we obtain from (4.36)
B ba ()I /c
Nb /Na = ,
Aab + B ab ()I /c
from which we extract the relative population of the excited state
Nb (R /2)2
= (5.13)
Nb + Na (Aab /2)2 + R2 /2 + ( ba )2
speak of power broadening or saturation broadening: even for extremely high inten-
sity the maximum population (stationary limit) in the excited state at resonance
= ba is always Nb /(Nb + Na ) < 1/2, i.e. the transition becomes saturated. By
excitation with CW radiation one can never excite more than 50 % of the atoms!
We emphasize again, that the above derivation is to some extent guess
work: perturbation theory is strictly valid only for population densities Nb /(Nb +
Na ) 1/2. However, as we shall see in Sect. 10.5.2, Vol. 2, the non-perturbative
treatment confirms (5.13).
Some Numbers
To develop some feeling for the orders of magnitude of natural linewidths we com-
municate two typical values:
232 5 Linewidths, Photoionization, and More
For the LYMAN alpha line (1s 2p) in hydrogen like atoms we use (4.117)
with (C.56) and Table C.2. With these values4 we obtain A(1s2p) = 6.2658
108 Z 4 s1 . Specifically, for the 2p state of the H atom (H line at =
121.57 nm) the lifetime becomes nat = 1.596 ns. From this follows a linewidth
nat 99 MHz.
For other atoms one has to compute the radial wave functions and from these
the radial matrix elements numerically, as described by way of example for
alkali atoms in Chap. 3. One of the strongest atomic transitions is the Na D
line (3s 3p) at = 589 nm, with a lifetime 16.2 ns,
nat = 1/nat =
6.15 107 s1 or
nat 9.8 MHz.
5.1.2 Dispersion
For later reference we note that the complex resonance amplitude (5.5) with reso-
nance energy Wba and linewidth = / (FWHM) may be written:
absorption
-8 -4 0 8
- 0.5
-8 -4 0 4 8
(W-Wba) /
We note that the imaginary part reproduces (apart from the normalization) the
L ORENTZ distribution (5.7) and thus corresponds to the absorption probability,
while the real part describes a dispersion type line shape. Both profiles are illus-
trated in Fig. 5.2. We shall come to these again in Sect. 7.6.2 where so called FANO
resonances are treated, as well as in Sect. 8.4.3 in the context of dispersion of the
refractive index as a function of wavelength.
At higher pressures atoms and molecules in the gas phase collide frequently with
each other. One may rationalize the influence of such collisions on the linewidth by
assuming that the emission (or absorption) process is not completely disrupted, but
looses all its phase memory. One assumes that the interactions occur statistically
distributed (exponential probability distribution), with an average time interval tcol .
As will be discussed in detail in Sect. 2.1.3, Vol. 2 these considerations lead again
to a L ORENTZ distribution of the frequency spectrum with a bandwidth inversely
proportional to the collision time.
The average time tcol between two collisions is derived with (1.52) from the
particle density N of the perturbing gas, the collision cross section and the rel-
ative velocity v between radiating atoms and perturbing gas particles (being not
necessarily the same). Thus, collisional line broadening (or pressure broadening) is
characterized by a Lorentzian line shape with a FWHM
col = 1/tcol = N v
= v
p/kB T p 8/( Mk BT ) (5.16)
col p 8/( Mk B T )/2, (5.17)
where
indicates averaging over the relative velocity distribution in the gas stud-
ied. The particle density is related to the pressure p by (1.50) at temperature T . The
234 5 Linewidths, Photoionization, and More
Fig. 5.3 On the origin of the D OPPLER profile for an absorption line: A red or blue shift occurs,
depending on the direction of the atomic velocity (red and black marked atoms, respectively). No
shift occurs for atoms at rest or moving k (grey)
average (relative) velocity v
8kB T /( M) is derived from the M AXWELL -
B OLTZMANN distribution (1.57). The reduced mass M of the colliding pairs has to
be used here, since it is their relative velocity that counts. Typical gas kinetic cross
sections are 1015 cm2 (mostly elastic).
For our standard example, Na atoms, say in a cell with 100 mbar Ar buffer gas
(M = 14.6 u) at T = 554 K, the Na vapour pressure is 1 Pa = 0.01 mbar, sufficiently
high for nice spectroscopy. At these conditions the frequency bandwidth becomes
about
col = 1.84 108 Hz, about an order of magnitude larger than the natural
linewidth. It increases linearly with pressure and may become very important for
dense gases. The collisional broadening too is a homogeneous line broadening since
it is of statistical nature and concerns all atoms in a gas in the same way.
If this additional broadening is of similar magnitude as the natural linewidth
one has to convolute both profiles. Specifically, the convolution of two Lorentzian
profiles (see Appendix G.5) leads again to a Lorentzian line profile with additive
linewidths (FWHM):
1 1
1/2 =
col + A = + . (5.18)
tcol nat
vx 2
ba = ba = vx , (5.19)
c ba
where ba = 2ba is the angular transition frequency in the rest frame of the atom,
ba the corresponding wavelength, and
ba = 2
ba the shift of absorbed (or
emitted) angular frequency in respect of ba .
5.1 Line Broadening 235
1/2
-4 -2 0 2 4
- b a / 1/2
Fig. 5.4 Comparison of a L ORENTZ profile (black line), typical for the natural line profile and
collision broadening, and a Gaussian distribution (red line), arising from thermal line broaden-
ing. Compared are here profiles of the same FWHM
1/2 , renormalized to equal maxima for
comparison
According to (1.56), the B OLTZMANN distribution taken for one velocity com-
ponent vx (in the direction of light propagation k) is
1 kB T
w(vx )dvx = exp vx2 / 2v02 dvx with v02 = v 2 = (5.20)
2v02 m
where v b and v a are the atomic velocities, b and a the electronic energies in
the upper and lower state, respectively. Momentum conservation implies
mv b = mv a + n. (5.24)
c
Inserting v a from (5.24) into (5.23), abbreviating as usual ba = b a , and
neglecting terms /mc2 vx /c, one easily verifies
ba vx ba vx
= ba 1 + or
ba = = ba .
1 vx /c c 2 c
Here vx = v b n is the projection of the initial atomic velocity v b onto the direction
of propagation of the photon. Thus we have recovered the non-relativistic D OPPLER
shift according to (5.19), q.e.d.
In practice, D OPPLER broadening in the gas phase is usually much larger than
the natural linewidth or collision broadening at not too high pressures. As an
example, we refer once again to the Na D lines. At 554 K with m = 23 u the
D OPPLER width becomes
D 1.8 109 Hz at = 5.09 1014 Hz as com-
pared to
nat = 9.79 106 Hz corresponding to a natural lifetime nat = 16.2 ns.
The D OPPLER width in this case is thus two orders of magnitude larger than the
natural linewidth and one order of magnitude larger than the collision broadened
line at 100 mbar. Spectroscopy of emission or absorption lines in the gas phase
the most important source of our knowledge about the structure of atoms and
molecules has to fight massively against D OPPLER broadening. In later chapters
we get to know a variety of interesting methods which deal artfully with this prob-
lem.
In view of Fig. 5.3 one easily recognizes that D OPPLER broadening in con-
trast to natural line broadening and collisional broadening is inhomogeneous:
each atom emits its specific wavelength according to its instantaneous velocity. In
an absorption process each particular frequency excites only a specific group of
particles in a velocity interval from vx to vx +
vx related through (5.19) to the
natural linewidth by
vx = ba
nat .
In the most general case (i.e. at high temperatures and high pressures) one has to
account for both effects, D OPPLER broadening and collision broadening. One has
to convolute then a Lorentzian with a Gaussian profile. This is not completely triv-
ial, but can be done as briefly sketched in Appendix G.6. The result, a so called
VOIGT profile, is illustrated in Fig. 5.5 for the specific example of a L ORENTZ and
D OPPLER profile with equal width. A rather accurate approximate formula for the
linewidth of a VOIGT profile is (G.25).
Such delicate studies and evaluations of spectral lines were done rather exten-
sively in the second half of the past century, partially in order to better understand
5.1 Line Broadening 237
collision processes. Modern spectroscopy and collision physics have advanced over
the past decades substantially. Thus it may be said that collisional and D OPPLER
broadening have more or less been overcome by a toolbox of sophisticated spectro-
scopic methods.
Nevertheless, a detailed evaluation of line profiles still plays an important role
for remote spectroscopy of dense gases, high pressure gas discharges, and plas-
mas in general in laboratory studies as well as in astrophysics. In addition to the
effects mentioned here, several other mechanisms such as S TARK broadening in
plasmas contribute to the line shapes. In otherwise inaccessible media systematic
profile measurements of spectral lines often provide the only viable experimental
access to important physical properties such as particle densities and tempera-
tures.
Section summary
The natural line profile arises from the finite lifetime ab of the excited state. It
is described by a L ORENTZ distribution (5.12) with a FWHM
nat = A ab =
1/ab .
Collisional line broadening is also represented by a L ORENTZ profile. It is
for studies in gases at high pressure p, its FWHM being
col
relevant
p 8/( Mk B T ) at a temperature T .
D OPPLER broadening arises from transition frequencies shifted due to differ-
ent velocities of freely moving atoms or molecules in a gas. The B OLTZMANN
velocity distribution
leadsto a Gaussian line profile (5.21) with a FWHM of
D (2/ba ) 8 ln 2 kB T /m.
Typically
nat
col and
nat
D .
The VOIGT profile is a convolution of L ORENTZ and Gaussian profile and
describes a situation where both, D OPPLER broadening and collisional broad-
ening, are important, i.e. at high pressure and high temperature.
We distinguish homogeneous and inhomogeneous line broadening, if each
absorber or emitter has the same line profile or a different resonant frequency,
respectively. The natural line profile and collision broadening are typical for
the former, D OPPLER broadening for the latter.
238 5 Linewidths, Photoionization, and More
In Sect. 5.1.1 we have started to extend the concept of transition rates between dis-
crete stationary states to levels broadened by spontaneous emission. We may con-
sider this as a first step to make the transition into the continuum. The considerations
presented there need not be restricted to the natural linewidth or to homogeneous
line broadening. They can also be applied to transitions broadened by collisions or
by the D OPPLER effect as just discussed. We simply have to replace gL () in (5.10)
by the corresponding profiles. In this context one usually normalizes to energy scale.
With W = and I() = dI /d = dI /dW the transition rate (4.62) is rewritten
as
dRba
= 4 2 I(W )|
Dba |2 g(W ), (5.25)
dW
with |Dba |2 according to (4.57) in dipole approximation (see also footnote 1). As
introduced in Sect. 1.3.4, g(W ) is now an appropriate density of states which de-
scribes the probability to encounter and excite the corresponding final states. The
area under this distribution function (whatever its overall width) must of course be
properly normalized according to (5.9). In this reading dRba /dW is a transition rate
(or a rate of absorption from the photon point of view) per unit energy interval (di-
mension T1 Enrg1 ), and the density of states g(W ) may in the most general case
describe an extended continuum or quasi-continuum of very closely spaced final
states: it specifies how many states exist per unit energy.
Replacing in (5.25) the specific interaction for electromagnetic radiation with
matter by the general transition operator U as introduced in Sect. 4.3.1, this expres-
sion turns out to be just a special application of the famous
dRba 2 2
F ERMIs golden rule: = |Uba | g(W ), (5.26)
dW
which in the literature is usually written without explicit reference to its differential
nature per unit energy interval.
In Sect. 4.6.1 we have introduced the line strength S(jb ja ), which has the dimen-
sion L2 . For ease of comparing dipole transitions between different transitions and
atoms, one also defines a dimensionless quantity, called oscillator strength
2
(opt) 2Wba S(jb ja ) 2Wba
Dba
2Wba zba 2
fba = = = . (5.27)
Eh 3ga a02 E h a0 E h a0
5.2 Oscillator Strength and Cross Section 239
Briefly, the line strengths S(jb ja ) is the sum over the squares of all transition matrix
(opt)
elements between sublevels of a and b, while fba characterizes absorption (or
induced emission) from one substate |ja ma
of the initial level a to the final level b
averaged over all initial substates and summed over all final states of that level.5
The degeneracy factor ga = 2ja + 1 compensates the summation over all initial
states in the definition of S(jb ja ) according to (4.112), the factor 1/3 compensates
(opt)
summation over all polarizations. We emphasize that fba is independent of the
polarization used!
With the transition energy Wba = ba being related as usual to the term energies
by Wba = (Wb Wa ) and with the definition (5.27) we have
(opt) (opt)
Wba = Wab and ga fba = gb fab .
(opt)
Thus fba > 0 for absorption and <0 for induced emission. As shown in Ap-
pendix H.2 the important
(opt)
T HOMAS -R EICHE -K UHN sum rule fba = N , (5.28)
b
strictly holds in dipole approximation, with N being the number of active electrons.
The summation (and possibly integration) has to be carried out over a complete
basis set of states describing the system, and potentially includes positive as well as
(opt)
negative values of fba .
The term oscillator strength is best understood in the context of the classical
(opt)
picture of one oscillating electron: an oscillator strength fba = 1 for a particular
transition implies that the whole ability of an active electron to absorb electromag-
netic radiation is concentrated in that one particular transition. A nearly perfect ex-
ample provides Na. As discussed in detail in Sect. 3.2 its one valence electron is per-
fectly described by a quasi-one-electron system. The oscillator strength for its yel-
(opt)
low NaD line (which we have already mentioned several times) is f3p3s 0.98,
i.e. it contains ca. 98 % of all available oscillator strength. In contrast, for the H
atom f2p1s = 0.416, f3p1s = 0.073, f4p1s = 0.029,
(opt) (opt) (opt) (opt)
n=5 fnp1s = 0.041. In
addition one has to consider ionization for which integration over all continuum
(opt)
states gives dfba = 0.435.
When applying the T HOMAS -R EICHE -K UHN sum rule to larger systems one has,
however, to be aware that the sum (5.28) is to be taken literally and may include tran-
sitions to fully occupied levels which are forbidden transitions as a consequence
of the PAULI principle.
(opt)
A very useful numerical relation between oscillator strength fba (for excitation
b a) and spontaneous transition probability Aab (for the decay process a b) is
derived from (5.27) and (4.114):
5 In the second and third equality of (5.27) we have omitted these summations for simplicity. Precise
details and alternative expressions are summarized in Appendix H.2. The last expression refers to
linear polarization, parallel to the z-axis.
240 5 Linewidths, Photoionization, and More
(opt)
ga fba = Cgb Aab 2 (5.29)
0 me c
with C = = 1.4992 1014 nm2 s.
2e2
If one describes the absorption process so to say from the view point of the photon,
the question arises, which cross section does the photon see during its interaction
with an atom. With the photon energy W = we may rewrite (5.25) for the number
of photons absorbed per unit of time and energy interval as (see also footnote 1)
dRba I(W )
= 4 2 W |
Dba |2 g(W ). (5.30)
dW
Now, I(W )/() = is the number of photons per unit time per unit area and per
spectral energy range, i.e. photon flux per unit energy. Thus, generalizing (4.24) we
obtain the absorption cross section for a transition from |a
to |b
as a function of
the energy of the absorbed photons:
2
2W
Dba
g(W )Eh a 2 .
ba = 4 W |
2
Dba |2 g(W ) = 2 2 (5.31)
E h a0 0
On the right hand side we have rewritten this expression using atomic units a0 and
Eh , so that ba is given essentially as a product of dimensionless factors: Since
|
Dba |2 /a02 is dimensionless, Eh is an energy, and the density of states g(W ) has the
dimension Enrg1 , ba has indeed the dimension L2 of a cross section (here given
in units a02 ). If we refer here to the averaged cross section for transitions between
levels a and b (see footnote 1) the term in square brackets is the oscillator strength
as defined in (5.27) and (H.37). Thus we may write in compact form
(opt) (opt)
ba = 2 2 fba g(W )Eh a02 = 2 2 re cfba g(W ). (5.32)
ba = 4.034 1018 cm2 Eh fba g(W ) = 1.098 1016 cm2 eV fba g(W ).
(opt) (opt)
6 Note that different authors use different units. Thus, one finds different prefactors in different
publications. The value used here corresponds to FANO and C OOPER (1968), while e.g. C OOPER
(1988) uses 8.067 1018 cm2 and measures energies in RYDBERG = Eh /2.
5.2 Oscillator Strength and Cross Section 241
For excitation of a resonance level, which is determined only by the natural life-
time, the cross section (5.32) may be rewritten in an even more transparent manner.
(opt)
With (5.29) we express fba by Aab and identify g(W ) as the L ORENTZ distribution
(5.7) with a FWHM = =
nat , normalized according to (5.9). By inserting the
atomic constants the photo-absorption cross section becomes
gb 2 ( /2)2
ba (W ) = . (5.33)
ga 2 (W Wba )2 + ( /2)2
Here we have accounted for some detuning Wdet of the light source in respect of
the center of the spectral line. This convolution (see Appendix G) thus allows one
to describe a realistic spectroscopic experiment where the laser is tuned over the
spectral lines.
We assume now for simplicity that the spectral distribution of the light source
(e.g. a tuneable dye laser) is also characterized by a Lorentzian profile, its FWHM
being
light . Using (G.21) the convolution integral (5.36) can be evaluated, leading
242 5 Linewidths, Photoionization, and More
2 gb
ba () =
nat gL (;
nat +
light ) (5.37)
2 ga
with ba being the detuning of the light source from the maximum ba of
the spectral line. On resonance ( = ba ) the averaged absorption cross section
becomes
___ 2 gb
nat
A = . (5.38)
2 ga
nat +
light
Thus, in practice the huge maximum absorption cross section (5.34) is reduced to an
effectively much smaller value if the bandwidth of the light source is large compared
to the natural linewidth. The total absorbed intensity (W cm1 ) is then given by
___
I = Na A I
x.
For other line profiles similar considerations may be applied to derive averaged
cross sections. For instance, to describe a mainly collision broadened line, one
would have to replace
nat = / by
col according to (5.16) and reduce the
cross section (5.33) by
nat /
col .
Note, however, that the linear averaging described above is only applicable in
the linear absorption regime, i.e. as long as
I(W )/I(W ) 1, and the relative
absorption may be computed directly from (5.35). Otherwise one has to apply its
integrated form, the L AMBERT-B EER law (1.45) at each incident frequency, and
finally average over the line shape of the radiation.
In general, this is not a trivial task, in particular if molecular spectra with many lines
are considered. They may even arise from several different absorbing gas molecules
as it is the case in radiative-transfer of the incoming solar radiation through our
earths atmosphere. In the spectroscopic literature and in data banks (see e.g. ROTH -
MAN et al. 2009) one often rewrites (5.33) as a function of the wavenumber = 1/
gb 2 /
ba ( ) = = Sba L( , ba , ), (5.39)
ga 2 ( ba )2 + 2
where the L ORENTZ line shape L( , ba , ) is given as a function of wavenumbers,
and is the half width of the spectral line profile at half maximum (HWHM) in
wavenumbers [ ] = cm1 . Note that L( , ba , ) is measured in units [L] = cm
and normalized so that L( , ba , )d = 1.
The parameter Sba is usually also called line strength (not to be confused
with the line strength S(jb ja ) which we have used elsewhere, with the dimen-
sion L2 ). It is measured in units [S] = cm1 /(molecule cm2 ) to be identified
as wavenumber per column density. In radiative-transfer calculations one has also
5.2 Oscillator Strength and Cross Section 243
to account for changes in the density Na (x) of absorbing molecules with position
x in space and defines the column density as CD = Na (x) x, measured in units
[CD ] = molecule cm2 . Thus, (5.35) is rewritten for a given spectral line, centred
at ba , as
dI( ) = I( )Sba L( , ba , )dCD . (5.40)
If only one single absorbing line is considered, the relative intensity transmitted
over a spectral range
is determined by
ba +
/2
I 1 /
= exp CD Sba d ,
I0
ba
/2 (ba )2 + 2
where we have assumed that the incident spectral intensity is constant over
and
centred at ba .
For the general multiline problem (also accounting for D OPPLER broadening)
one has to replace Sj L( , j , ) j Sj V ( ; j ) in (5.40), with VOIGT line pro-
files (Sect. 5.1.5) centred at each relevant spectral line j . By integrating (5.40) over
CD one derives the relative transmission I( )/I0 ( ) for the spectral intensity at each
of interest. To model a genuine molecular spectrum one also has to account for the
thermal population of the initial levels, and include induced emission in addition to
absorption. Finally, this spectrum I( ) thus obtained has to be convoluted with the
resolution of the spectrometer.
For atmospheric modelling, reemission of radiation and scattering from the at-
mosphere has also to be included, as well as absorption and scattering from aerosols
and small droplets. Thus, the problem gets rather involved. References to a number
of codes available for atmospheric radiative-transfer are reported by W IKIPEDIA
CONTRIBUTORS (2013). They all rely on information from molecular spectroscopy
as documented in databanks such as HITRAN, and allow to model solar spectra on
the surface of the earth as presented in Sect. 1.4.7.
Section summary
We have related the rate for optically induced E1 transitions to the famous
F ERMI golden rule (5.26).
(opt)
The optical oscillator strength fba is a useful, dimensionless quantity which
characterizes the strength of an E1 transition. For a (quasi)single electron sys-
tem, in the classical limit of a harmonic electron oscillator it would be equal
to 1, in quantum mechanical reality it obeys (5.28), the T HOMAS -R EICHE -
(opt)
K UHN sum rule, b fba = 1 and is always <1.
The photo-absorption cross section (5.32) for an infinitely narrow bandwidth
laser beam is very large, specifically A = 2 /(2)(gb /ga ) at the maximum
of a L ORENTZ distribution if only natural line broadening (FWHM
nat =
1/ab ) is relevant. The measured cross section depends, however, strongly on
the line profile of the absorbed light (FWHM
light ). If the latter is large
___
compared to the former, the photo-absorption cross section becomes A
A
nat /
light .
244 5 Linewidths, Photoionization, and More
While in (4.46) perturbation theory was applied only in 1st order, one now has
to carry it out consequently at least up to N th order to determine the transition
probability for this particular type of process. While in 1st order the transition rate
(4.63) was proportional to the intensity I of the laser field, in N th order it becomes
proportional to the N th power of it. Thus, the N photon transition rate will be
(N ) (N )
Rba = ba N I N with = I /, (5.43)
5.3 Multi-photon Processes 245
(N )
where ba is a so called generalized cross section. The photon flux is given in
units [] = Photons m2 s1 and the unit of the generalized cross section [ (N ) ] =
m2N sN 1 is thus a little bit unused but meaningful. In the following we shall talk
about processes which are induced by an intense laser field which is, however, not
yet strong enough to completely deform the subject of investigation or even to
destroy it completely.
The question whether two photons may be absorbed simultaneously was first studied
theoretically by a student of Max B ORN, Maria G PPERT-M AYER (1931), who
later in 1963 won the N OBEL prize (together with W IGNER and J ENSEN) for her
work on the nuclear shell structure. In her PhD thesis ber Elementarakte mit
zwei Quantensprngen (On elementary acts with two quantum jumps) she laid the
foundations of this field which, however, only started to flourish with the advent
of powerful lasers, long after her pioneering work. We have already prepared in the
last chapter the tools for understanding what happens, so that now it will take only
a few steps to sketch the basic theory necessary for catching up with state-of-the-art
research in this field. In Sect. 5.5.5 we shall return to it.
We simply have to reintroduce the transition amplitude (4.52) (obtained in 1st or-
der perturbation theory) on the right hand side of the original ODE (4.45), use (4.50)
and to integrate. This leads to the probability amplitude in 2nd order perturbation
theory:
ie2 E02 t Db
D a ei(ba 2)t
D
D a eiba t
dt
b
cb (t) = + (5.44)
42 0
( a ) ( a )
+ emission terms .
The expression sums (in principle) over all basis states |
of the system.7 Charac-
teristic for all absorption terms of this sum (first line) are the resonance denomina-
tors ( a ) where a = W Wa gives the energy difference between initial
state |a
and an intermediate state |
, each with exponentials in the nominator re-
sembling those in the 1st order approximation (4.51).
We see now clearly how two photon absorption comes into play: the first ex-
ponent i(ba 2)t in (5.44) common to all terms in the sum simply replaces
i(ba )t in (4.51): the double angular frequency 2 of the electromagnetic wave
assumes the role of . These terms, characteristic to 2nd order, oscillate rapidly and
average out in the integration unless 2 = ba in analogy to terms with ba
7 We have written out here only the absorption terms; the emission terms are identical, except that
is replaced by and
D by
D (and vice versa).
246 5 Linewidths, Photoionization, and More
in 1st order. Terms with exp(iba t ) vanish in any case when the integration is car-
ried out.
Hence, for absorption only the first terms of the sum over in (5.44) contribute
to the integral, and exp[i(ba 2)t ] is pulled out of the sum. (The correspond-
ing emission terms in (5.44) contain exp[i(ba + 2)t ] and describe two-photon
emission in the case that ba < 0.) All further steps correspond to those in 1st or-
der: one carries out the integration from t = 0 to t and makes the transition from
probability amplitudes to transition probabilities in the limit t . Dividing that
by t gives a transition rate proportional to E04 (hence to I 2 ) and to the D IRAC delta
function this time of the argument ba 2.
Thus, the resonance frequency for a transition from |a
to |b
by simultaneous
absorption of two photons (both of energy) becomes 2 = ba as expected.
The necessary extension to levels of finite width or for transitions into the contin-
uum follows the scheme used in the one-photon case. Specifically, for a L ORENTZ
broadened line one replaces
(ba 2) gL (2), (5.45)
where the line profile (5.7) of the excited state |b
is now a function of twice the
photon energy. From the thus derived rate one pulls out the square of the photon
flux (I /)2 to obtain the generalized cross section for two-photon excitation in
2nd order perturbation theory:
2
b| D|
|
D|a
ba = (2)3 2 ()2 gL (2).
(2)
(5.46)
W a
As indicated in this expression, in principle one has to sum not only over infinitely
many discrete basis states |
but also to integrate over the continuum. This can
lead to rather elaborate calculations. Fortunately, however, with differing more
and more from a, the increasing denominators as well as the decreasing magnitude
of the matrix elements D a = |
D|a
usually leads to satisfactory results with only
a finite number of terms.
One verifies easily that this generalized two-photon cross section is derived in
units [ba ] = m4 s since the line profile is given in [g(2)] = eV1 and the ma-
(2)
(2)
trix elements are measured in units of length. Note that ba as defined by (5.43)
refers in principle to a strictly monochromatic electromagnetic wave (e.g. a highly
stabilized laser source) which is much narrower than the atomic linewidth given by
gL (2). For a radiation source of broader bandwidth only the fraction of the ra-
diation spectrum I() overlapping with the atomic linewidth is relevant just as
in the one-photon case. For a spectrum with finite bandwidth, in the present case of
two-photon absorption one has first to convolute I() with itself, i.e. with I( ),
and the result in turn with gL (2) to finally obtain an expression analogous to
(5.38).
Several extensions of the procedure described above are obvious. Excitation pro-
cesses with two photons of different frequency and/or polarization may be consid-
ered by adding to the interaction potential (4.55) a second electric field vector. The
5.3 Multi-photon Processes 247
resonant. Nevertheless such processes can be studied with high efficiency and have
played a key role in the ultra high precision spectroscopy of atomic hydrogen which
led to the N OBEL prize for Ted H NSCH (2005). As we shall discuss in Sect. 6.1.8,
crucial for this kind of precision spectroscopy is the possibility to study two-photon
absorption in the gas phase completely free of D OPPLER broadening.
As a somewhat exotic but interesting field of atomic physics we finally mention
spectroscopic studies relating to parity violation. They rely on such high precision
spectroscopy. In this context, e.g. so called nuclear anapole moments have been
observed in 2 photon excitation of heavy alkali atoms (see e.g. B OUCHIAT 2007).
8 The
designations 2 S, 2 P used here and 1 S0 , 3 P0 , etc. used in Fig. 5.7 will be explained in detail in
Chap. 6.
5.3 Multi-photon Processes 249
coincident
poton rate
0
1.65fs
7.80 21P 1
71ps
E1 2 3 P2
154ps 13fs 23P1
21S0 3
2.5ns 2 P0
1
10
7.75 E1E1
M2
8
2 3 S1
2E1 2.3ns
6
1
4
7.70 E1 10
/ ke
M1 8
2 6
2
V
4
58Ni26+ 2 2
/ keV
0
0 11S0 0
Fig. 5.7 Two-photon decay of the 1s2s 1 S0 state in He like 58 Ni26+ . (a) Term scheme with several
neighbouring levels and transitions. (b) Two-photon coincidences as a function of the emitted pho-
ton energies 1 and 2 . The red dotted line emphasizes the transition of interest 2 1 S0 1 1 S0 .
Figure adapted from S CHFFER et al. (1999)
Lifetimes and process types are noted in Fig. 5.7(a).9 Figure 5.7(b) shows the mea-
sured coincidence rate for detecting simultaneously two photons as a function of
their respective energies, 1 and 2 . Accidental coincidences with some of the
other processes make the results somewhat intransparent hence the red dotted
line indicates the 2E1 decay process of interest. Energy conservation requires that
1 + 2 = W2 1 S0 W1 1 S0 . One sees clearly that energy sharing among the two
photons leads to a broad distribution with a flat maximum for similar energies while
the probability that one of the photons carries all energy is zero (barely recognizable
in Fig. 5.7 due to accidental coincidences).
Precision measurements of such decay processes are of considerable interest
as an important testing ground for relativistic multi-electron quantum mechanical
computations. Interestingly, special two-photon transitions (3 P0 1 S0 otherwise
strictly forbidden for 2E1 processes) are discussed in the context of investigating
parity violation in atomic physics (B OUCHIAT and B OUCHIAT 1997).
Section summary
Multi-photon processes can be induced by intense electromagnetic radiation.
(N ) (N )
The rate for absorption of N photons is Rba = ba (I /)N (at not too
(N )
high intensities I ), the generalized cross section ba being measured in
m2N sN 1 . Resonance conditions are now N = ba .
9 We cannot enter into a discussion of the interesting details here, which require knowledge of the
underlying atomic physics still to be explored in later chapters. For the advanced reader it may
be particularly interesting that the high nuclear charge Z implies strong spin-orbit interaction and
leads to configuration mixing. Thus, the transition 2 3 P1 1 1 S0 is dominated by an E1 process,
while transitions 2 3 P2,0 1 1 S0 with
J = 2 and 0, respectively, are strictly E1 forbidden due to
angular momentum conservation.
250 5 Linewidths, Photoionization, and More
eikr ieikr
and (H.22)
D= e
p= e (5.49)
me me
contains the term exp(ik r) = 1 + ik r + , which in dipole approximation was
just replaced by 1. We shall now treat the second term of this expansion:
i
D(2) = (k r)(e
p ).
me
We first rewrite10 this by trivial addition and subtraction of an extra term:
i i
D(2) = [k r e
p e r k p] + [k r e
p +e r k
p ].
2me 2me
10 Above we have indicated once the precedence of vector multiplication by brackets. In all follow-
ing expressions we omit this notation: vector products always have precedence over scalar products
while these again have precedence over simple scalar multiplications.
5.4 Magnetic Dipole and Electric Quadrupole Transitions 251
We exploit now some rules for multiple products from vector algebra, using the fact
that k e B. The first bracket is identical to the scalar product of two vector
products (making sure not to permutate r and p):
i i
D(2) = k e r
p+ [k r e
p +e r k
p ].
2me 2me
tion of the position vector onto the polarization vector, e r, and the projection of the
momentum vector onto the wave vector, k p , represent perpendicular components
of the r and p operator, respectively. Hence, they commute and we may write
ik B 0 i
D(2) = L+ [k r e
p +k
p e r].
2me B0 2me
e ecB0 ].
ecB0
D(2) = i L B0 [H k r e r k r e r H (5.50)
2me 2
We have thus managed to expand the operator (5.47) describing the interaction of
electromagnetic waves with matter beyond the simplest dipole approximation in
a suggestive manner. Obviously, (5.50) expresses the second most important term
as two different types of interactions due to the electromagnetic field: as we shall
see in a moment, the first one is responsible for magnetic dipole (M1), the second
one for electric quadrupole transitions (E2). Both have the same time dependence
as specified in (5.47) and may be treated completely analogous to electric dipole
transitions in 1st order perturbation theory. We thus simply need to determine the
transition matrix elements between |a
and |b
of ecB0 D(2) for the two components.
The magnetic dipole term may then be written as
1
L
(ecB0 B0
D)M1 = B B 0 = M (5.51)
i
where according to (1.150) we have inserted explicitly B OHRs magneton B to
express the characteristic nature of this interaction. The magnetic dipole interac-
tion energy M B(t) induces M1 transitions in complete analogy to the electric
dipole interaction energy D E(t) which is responsible for E1 transitions. Let us
now obtain an estimate for the relative order of magnitude for these two types of
252 5 Linewidths, Photoionization, and More
|M B 0 |2 Z 2 2 (Z)2
= 1.33 105 Z 2 . (5.52)
|D E 0 |2 4m2e a02 c2 4
The probability to induce an M1 transition is thus (for light atoms) about five orders
of magnitude smaller than for an E1 transition at a given intensity of the radi-
ation. The same holds for spontaneous decay probabilities. Sure enough, for high
charge states M1 or even M2 may become appreciable, as we have already seen in
Sect. 5.3.2. The M1 transitions are crucial for all magnetic resonance spectroscopy
where transitions between different magnetic substates |j mb
|j ma
within the
same electronic level are studied (see Sects. 9.5.2 and 9.5.3) strictly forbidden for
electric dipole transitions.
For a quantitative treatment of M1 processes we also have to account for the elec-
tron spin. As mentioned already in Chap. 1, the magnetic moment associated with
the electron spin is
= 2B . This implies that the magnetic moment M L = B L/
of the orbital angular momentum has to be replaced by M = B ( L + 2S)/. Since
M1 transitions are single photon processes, selection rules and line strength for M1
transitions thus follow from the corresponding transition matrix elements. For the
determination of transition rates and cross sections one simply has to replace in all
relevant expressions of Sects. 4.35.2 the electric dipole transition matrix element
eE0 Dba by the corresponding one for the magnetic dipole interaction:
1 B B B B
(ecB0 b|L + 2
(M1)
D)ba = S|a
= b|J + S|a
. (5.53)
i ec B ec B
The components of total angular momentum and spin are thus projected onto the
magnetic field direction of the electromagnetic wave with B E k. Tools for ex-
plicit evaluation of the matrix elements are summarized in Appendix C.4.3. Details
will be discussed in several of the following chapters.
Let us come back now to the second part in (5.50), the electric quadrupole term,
from which the transition matrix element follows as
eE0
a|eE0
D|b
(E2) = (Wb Wa )b|k r e r|a
2
eE0
= b|ek r e r|a
. (5.54)
ba
With k = /c we have introduced here the unit vector ek = k/k in the direction of
the wave propagation and the transition wavelength ba . Alternatively with (5.47)
(5.49) and (4.1) the Hamiltonian for the electric quadrupole interaction may also be
written as
(r, t)(E2) = e ba E(t) r (k r).
U (5.55)
2
5.4 Magnetic Dipole and Electric Quadrupole Transitions 253
Since k and e are always perpendicular to each other, the two projections of r
onto these directions correspond to orthogonal components of r. Thus, depending
on the choice of the (at) and (ph)frame (see Fig. 4.3), the relevant matrix ele-
ments are of the type xy = Q22 / 3 (propagation in x-direction, polarization in
y-direction or vice versa), xz = Q21+ / 3 (propagation
in z-direction, polarization
in x-direction or vice versa) or yz = Q21 / 3. Light propagation under 45 in
the xy plane with a polarization angle vector also aligned within
the xy plane (az-
imuthal angle 45 ) is described by (x 2 y 2 )/2 = Q22+ / 3. The terminology
used here for the real components of the so called quadrupole tensor (an irreducible
tensor operator) is summarized in Appendix F.3. Any other geometry may be ex-
pressed as a linear combination of the components of this tensor operator, typically
by rotation of this operator with the help of the rotation matrix of rank 2 according
to Appendix E. Quantitatively the transition probabilities and cross sections for E2
transitions are obtained by replacing the electric dipole transition matrix elements
eE0Dba throughout Sect. 4.3 with
(1)q+1
b|eE0
D|a
(E2) = b|Q2q |a
. (5.56)
3ba
Here q characterizes the geometry of absorption or emission as just outlined. It
is advantageous that angular momentum algebra allows to derive the selection rules
in a relatively straight forward manner. For multi-electron atoms one has to account
for the respective angular momentum coupling schemes which will be discussed
in later chapters. A brief survey on selection rules for E2 transitions is given in
Appendix C.4.2.
We finally estimate the relative probability for such transitions. The E2 ma-
trix element (5.54) is eE0 xi xj
/ba , xi and xj being two position coordinates,
while the E1 matrix element is eE0 xi
. For a rough estimate we derive the
averaged position coordinates from (2.130) with xi xj
r 2
n n4 a02 /Z 2 and
xi
r
n n2 a0 /Z. Thus, the ratio of the transition probabilities at identical
intensities is on the order of
2
2
|a|eE0
D|b
(E2) |2 n a0
. (5.57)
|D E 0 |2 Zba
This ratio is thus determined by the square of the ratio of atomic dimension to
wavelength. For a typical transition in the VIS or UV spectral range for small nu-
clear charges Z and low principle quantum numbers n, one estimates a ratio of
(0.1 nm /300 nm)2 107 . For high Z, the probability for E2 processes even de-
creases with Z 2 in contrast to M1 transitions which increase with Z 2 as discussed
above. Thus, E2 processes have not played a major role in classical atomic, molec-
ular or solid state spectroscopy except in the X-ray region where they may be
considerable.
However, E2 transitions are of increasing interest as states with ever higher prin-
ciple quantum numbers n are studied, since orbital radii increase with n2 . A so
called RYDBERG atom with n = 100, say, has a diameter of ca. 500 nm, and thus
254 5 Linewidths, Photoionization, and More
Section summary
By expanding the spatial dependence exp(ik r) of the electromagnetic
field beyond the first, constant term one obtains magnetic dipole and elec-
tric quadrupole interaction terms, proportional to B b| J + S|a
and to
b|Q2q |a
/ba , respectively.
The rates for the resulting M1 and E2 transition can be evaluated in 1st order
perturbation theory in full analogy to those for E1 transitions. In comparison
to the latter M1 transitions are typically less probable by a factor of 105 Z 2 ,
E2 transitions by about n4 (a0 /ba )2 Z 2 .
5.5 Photoionization
11 This expression is often used like a synonym for up to now only poorly understood.
5.5 Photoionization 255
so that its dimension is L3/2 Enrg1/2 . In this normalization the density of states
(number of states per energy interval), g(W ) in (5.31), is already built into the wave
functions.12 The photoionization cross section thus becomes
a () = 4 2 |
Da |2 . (5.61)
Note that the squared matrix element has the dimension L2 Enrg1 so that a is
still a genuine cross section, with dimension L2 , measured e.g. in [a ] = b (see
Appendix A.2).
Extending now the definition (H.34) to the continuum, one introduces the optical
oscillator strength density (OOSD)
(opt)
dfa + WI
=2 2
|
Da |2 = 2 |za |2 . (5.62)
d E h a0 Eh a02
Following B ETHE and S ALPETER (1957), for high but not relativistically high pho-
ton energies
WI me c2 (5.65)
12 Alternativelyand completely equivalent one may employ continuum wave functions which are
determined as an orthogonal basis set for a large but finite volume L3 . In this case a density of
states corresponding to (5.31) has to be used in (5.61).
5.5 Photoionization 257
k ke
y
e
x
one may approximate the continuum wave function of a free electron by a plane
wave (1st order B ORN approximation, FBA):
me ke ike r
r|k e
= e . (5.66)
(2)3 2
The normalization factor under the square root is the density of states in an element
of solid angle d per unit volume (dimension L3 ) and energy according to (2.58).
The geometry for describing the photoionization process is sketched in Fig. 5.8:
the ionizing light propagates into z-direction with a wave vector k. We assume it
to be linearly polarized, its polarization vector e pointing in x-direction. After ion-
ization the electron is emitted into the direction k e . At these high energies we can
no longer use the dipole approximation. Rather the full expression (H.22) for the
transition operator has to be employed:
1 ikr 1 ikr
D= e e
p= e p E . (5.67)
me me
We denote the component of e
p in the direction of the polarization vector e as p
as introduced in (2.50). The transition matrix element for photoionization of the
state |a
is thus written
1 ikr 1 ikr
Da = k e |
D|a
= k e | e p e |a
=
pe k e | e |a
. (5.68)
me me
e e k so that p
The latter identity holds since p e and eikr commute, and we can
apply (2.35). We recall now that according to (2.51) each plane wave is an eigen-
e with the eigenvalue p cos = ke cos , where is the angle between
vector of p
polarization vector e and the direction of electron emission k e . Thus, applying (5.66)
the matrix element is written in integral form:
3/2
ke
Da = k e |
D|a
= cos (2)3/2 ei(kk e )r a (r)d3 r. (5.69)
me
With K = k e k we may abbreviate
3/2
a (K) = (2) eiKr a (r)d3 r (5.70)
258 5 Linewidths, Photoionization, and More
which is just the F OURIER transform of the initial wave function a (r), i.e. its rep-
resentation in momentum space. The matrix element Da refers to electrons which
are emitted into a specific solid angle. Hence, inserting (5.69) into (5.61) leads to
the differential cross section for photoionization in FBA:
1 ke3 2
da () = 4 2 cos2 a (k e k) d. (5.71)
me
The function a (k e k) may be derived either by (5.70) from the wave function
in space or directly from the S CHRDINGER equation transformed into momentum
space (see e.g. B ETHE and S ALPETER 1957, Chap. 8).
For the angular integration we choose z K so that eiKr = eiKr cos and
2
a (K) = drRns (r)r sin Kr (5.73)
K 2 0
which can be readily integrated. For large K = |k e k| and hydrogen like initial ns
states it may be approximated by
3/2
2 2Z 5/2 a0
ns (k e k) = (5.74)
n3/2 a04 |k e k|4
(B ETHE and S ALPETER 1957, Eq. (70.4), in a.u.). Introducing this into (5.71) leads
to a differential photoionization cross section for H like ns electrons:
1 2 ke3 8Z 5 a3
da () = 4 2 cos2 2 3 8 0 d
me n a0 |k e k|8
Z5 1 cos2
= 32 d. (5.75)
me n3 (a0 ke )5 [1 (v/c) cos ]4
The expression in brackets [ ] in the second row is obtained by exploiting (5.65)
and (5.59) = = ke /2me ; with the velocity of the ejected
so that with (5.58) 2 2
Angular distributions will be discussed in the next subsection. Here we just integrate
over the full solid angle. From Fig. 5.8 we read cos = sin cos and integration
4 2
over . . . d = 0 d 0 . . . sin d simply gives a factor 4/3. Thus, for H
like atoms in ns states the integral photoionization cross section in FBA is given
by:13
4
da 256 Z 5 a02
ns = d = 3 . (5.77)
0 d 3 n (2/Eh )7/2
Even if this result of the B ORN approximation is only valid at high (but not rel-
ativistically high) energies, it describes the general trend for H like orbitals reason-
ably well: a dramatic decrease of the cross section with photon energy ()7/2 ,
and strong dependence on nuclear charge Z 5 which leads to substantial photoion-
ization cross sections for the heavier atoms, as well as the decrease for higher or-
bitals n3 . As we shall see in Chap. 10 the inner shells of large atoms may very
well be described by H like orbitals so that the FBA leads to quite realistic first es-
timates for absorption of X- and -rays (see Sect. 10.5.3). A comprehensive data
source for such cross sections is C HANTLER et al. (2005).
Specifically for photoionization of the H 1s orbital (5.77) gives in numbers:
13 Since the integral is independent of the same expression holds for unpolarized light.
260 5 Linewidths, Photoionization, and More
4
exact calc. (NIST)
1s / cm2
10-22
10-24
2 10-26
BORN approximation
10-28
10-30
0 10-32
20 40 60 80 100 0.01 0.1 1 10 100
/ eV log( / keV )
Fig. 5.9 Cross section for photoionization of atomic hydrogen as a function of photon energy .
Exact calculations (red lines) according to C HANTLER et al. (2005) are compared with the B ORN
approximation (black lines) computed by (5.79), and with experimental data (grey squares) from
PALENIUS et al. (1976). (a) Linear plot for low , (b) logarithmic plot for energies from 13.7 eV
to 400 keV. The dashed line indicates the decrease W 7/2
first and the second ionization threshold (He+ and He++ , respectively) a rich spec-
trum of so called autoionizing lines, superposed on the general trend of the pho-
toionization cross section. For larger atoms, the photoionization cross section may
even pass at low energies through a so called C OOPER minimum as will be dis-
cussed in Chap. 10. Nevertheless, in the limit of very high photon energies one
always finds the dramatic decrease of the photoionization cross section discussed
here.
The angular distribution of the emitted photoelectrons for polarized light is charac-
terized according to (5.76) mainly by cos2 = sin2 cos2 . At high photon ener-
gies the term 4(v/c) cos leads to a shift of the angular distribution into the direction
of the ionizing light, but may in general be neglected for non-relativistic electrons.
Without derivation we note here the result for the general, low energy case. Strict
application of the dipole approximation (E1) leads to an angular dependence of the
differential photoionization cross section
da a
= 1 + P2 (cos ) where 1 2, (5.80)
d 4
which is characterized by the so called parameter (also anisotropy parameter),
with the L EGENDRE polynomial P2 (cos ) = (3 cos2 1)/2. For ionization of
pure ns states as in H(1s) case the just discussed, (5.76) is reproduced with = 2
neglecting the high energy term which is included in the B ORN approximation in
contrast to the simplified dipole approximation.
The maximum of the electron signal for ns ionization is thus found at = 0,
i.e. in the direction of the polarization vector e (see Fig. 5.8) while perpendicular
to it, i.e. parallel to the wave vector k of the ionizing light no electrons are emitted.
5.5 Photoionization 261
One may say, the angular distribution of the state generated by linear polarized light
corresponds to a |px
orbital. Its angular probability distribution is represented by
the angular distribution of the photoionization cross section. For not too high photon
energies the continuum states generated by photoionization are thus equivalent to
those obtained in dipole excitation (see Sect. 4.7).
For unpolarized or circularly polarized light one simply has to average the
differential cross sections for polarization in x- and y-direction. According to
Fig. 5.8 this implies averaging over the corresponding polarization angles with
cos2 = sin2 cos2 and = sin2 sin2 , respectively and leads to
da a
= 1 P2 (cos ) . (5.81)
d 4 2
For = 2 this characterizes a torus shaped (doughnut like) distribution around the
z-axis.
Both expressions are generally valid for not too high in photoionization of
atoms and molecules initially prepared in an isotropic population of states. One
finds, however, that usually depends on photon energy. As we shall see, is an
important quantity which describes the contributions of different partial waves in
single photon ionization processes.
We finally note that the photoelectron yield emitted at the so called magic angle
mag = 54.736 for which P2 (cos mag ) = 0 does not depend on . Systematic
studies of the integral photoionization cross section a are preferably done at this
emission angle (or at = mag , respectively).
The B ORN approximation is a high energy approximation. Its advantage is its sim-
plicity and the fact that it includes all multipole moments of the incoming plane
wave up to infinite order. We have seen, however, that it does not give reliable re-
sults at energies below about 5 to 10 WI . Yet it is this very energy range that
contains most of the continuum oscillator strengths, and provides the background for
many interesting atomic and molecular phenomena. Fortunately, one may in princi-
ple apply here the electric dipole (E1) approximation which according to (5.62) and
(5.64) just requires an accurate computation of the relevant radial matrix elements.
Thus, one only has to determine sufficiently exact wave functions for the initial
bound and the final continuum states.
As discussed in detail in Chaps. 2 and 3, the initial state |a
of the system is
described by a wave function with well defined quantum numbers n m:
un
|a
= Rn (r)Y m (, ) = Y m (, ). (5.82)
r
262 5 Linewidths, Photoionization, and More
The asymptotic behaviour of the radial functions for the bound states is according
to (2.119) and (2.120)
un (r) r +1 and un (r) exp 2|Wn |r . (5.83)
r0 r
In contrast, the continuum wave functions |
for a given energy of the free electron
may in principle contain all orbital angular momenta :
u (r)
|
= a Y m (, ). (5.84)
r
=0 m =
14 In
the literature the energy is often given in RYDBERG units so that the factor 2 under the root is
omitted.
15 Forthe ionization of H like atoms only a naked ion remains so that for the outgoing wave in
the pure C OULOMB potential = 0. For all other cases is related with the quantum defect
according to (3.26).
5.5 Photoionization 263
While the radial matrix elements |r|n
are determined by the specific prop-
erties of the atom and the photon energy, the matrix element m |C10 | m
(see
Appendix C.2.2) may be evaluated in close analogy to Sect. 4.4 where we have de-
scribed optical excitation processes. This leads to corresponding dipole selection
rules for photoionization: again
= 1 must hold, since a photon with angular
momentum is absorbed and the transition
= 0 is forbidden as a consequence of
parity conservation. We note, however, that in the continuum both final states, with
= 1 may be superposed at each photon energy above the photoionization
threshold. Also the selection rule m = m holds again for light linearly polarized in
z-direction. Evaluation of the sum and the matrix elements in (5.87) finally leads to
4 2 2
a () = r, 1 + ( + 1)r, +1
2
(5.88)
3
with the corresponding radial matrix elements
r, 1 = ru 1 (r)un (r)dr. (5.89)
0
Hence, unless the initial state is an s state with = 0 the photoionization cross sec-
tion contains contributions from two matrix elements involving continuum functions
with = + 1 and 1.
With some additional effort the continuum function (5.84) may be determined in
such a way that it describes an electron emitted into the direction k e . Without going
into details this leads to (5.80) with an energy dependent anisotropy parameter (see
Eq. (2) in C OOPER and Z ARE 1968a,b):
1
() = (5.90)
(2 + 1)[ r, 1
2 + ( + 1)r, +1
2 ]
'
( 1)r, 1
2
+ ( + 1)( + 2)r, +1
2
(
6 ( + 1)r, +1 r, 1 cos +1 () 1 () .
264 5 Linewidths, Photoionization, and More
0
- 45o 0o 45o 90o 135o 180o
polarization angle
Obviously is determined by the ratio r, 1 /r, +1 of the two radial matrix
elements involved and by the phase differences +1 () 1 () of the two partial
waves. One easily verifies that for = 0 (and thus r, 1 = 0) = 2 holds as
already found by the B ORN approximation. One also confirms without difficulty
that in the general case 1 2.
Figure 5.10 illustrates this by way of example for an experimentally determined
angular distribution in the photo-detachment from an anion
Cu [Ar]3d 10 4s 2 1 S + Cu[Ar]3d 10 4s 2 S + e
studied by C OVINGTON et al. (2007). Since the electron configuration of the copper
anion may be considered He like (two s electrons on top of a closed 3d shell), and
since one s electron is detached, one expects = 2 and hence according to (5.80)
a clean cos2 angular distribution. Figure 5.10 largely confirms this expectation.
A small deviation cannot be excluded by the experiment (the minimum at 90 may
not be exactly zero). One would attribute this to the fact that the description of
the electron wave function by a pure product wave function of all orbitals (in this
case a 4s electron has to be detached) is not completely correct for such a large
atom. Rather one has to account for some configuration mixing as to be discussed
in Sect. 10.2.3.
Figure 5.11 shows the rather typical ion beam apparatus used in this experiment.
The setup is in principle uncomplicated and self explaining. It requires, however,
high experimental precision. The determination of the kinetic energy of the electrons
is achieved by a 160 spherical sector analyzer (for a hemispherical analyzer see
Appendix B.3 in Vol. 2) which is characterized by its high resolution and good
focussing properties. The angular distribution shown in Fig. 5.10 has been recorded
as usual in such experiments: one rotates the laser polarization which is much easier
than turning the whole electron analyzer-detector setup. For = 0 the polarization
direction is parallel to the direction of electron detection.
Quantitative computations of cross sections for such multi-electron atoms require
a major effort. First, the wave functions for bound and continuum states have to be
determined accurately. Then the one electron dipole operator has to be replaced by
(H.29) summing over all electron coordinates. For a realistic description of such an
atom one also has to account for the spin and angular momentum coupling. The
ensuing matrix elements have to be evaluated using the angular momentum algebra
outlined in Appendix C.
5.5 Photoionization 265
2.4
m
4m
insertable 90o selector insertable
FARADAY magnet FARADAY
insertable cup cup
FARADAY
polarization einzel cup
adjustable
rotor lens einzel beam slits
polarizer lens
b e am l a ser
aperture anion source
ion o bea electrostatic
45 e - m steering elements
FARADAY laser power meter
cup
spherical electron energy
analyzer (160o segment)
Fig. 5.11 Typical experimental setup for studying photo-detachment from anions, here according
to C OVINGTON et al. (2007). The ion beam is focussed with several lenses and guided by deflec-
tors. A 90 magnet selects one anion species which crosses in the interaction region a CW laser
beam. The polarization of the latter may be rotated
V(r ) / Eh
0.5
e-
N Wkin
- 15 15
r /a0
WI
- 0.5
- 1.0
Fig. 5.12 Illustration of multi-photon ionization for the example of an H atom. In the continuum
the electron has a kinetic energy Wkin corresponding to (5.91) depending on the number of
absorbed photons N (the schematic is drawn to scale in a.u. for photons with a wavelength
= 800 nm). The electrons kinetic energy Wkin is indicated by downward pointing red arrows,
which indicate here that 0 to 5 additional photons beyond those necessary for ionization have been
absorbed (ATI)
2 ke2
Tg + N Tg+ + e (Wkin ) with WI + N = Wkin = . (5.91)
2me
Obviously the kinetic energy Wkin of the electron after ionization depends on the
number N of photons absorbed. The probability for a multi-photon ionization pro-
cess (MPI) depends on the photon flux = I /() just as in the case of multi-
photon excitation. In N th order perturbation theory one writes the transition rate
from a bound initial state |a
into the continuum state |
as
(N )
Rke a = (N ) N I N , (5.92)
summation over all accessible continuum states from (5.84) which will be reflected
in the differential cross section. Considering parity conservation, for a two-photon
ionization process the partial waves = and 2 contribute (as long as 0),
for a three photon process we have = 1 and 3 and so forth. When linearly
polarized light is used the selection rule
m = 0 remains valid, but for circularly
polarized light also higher values of N m N are accessible. It is obvious
that the angular distribution of the emitted electrons is correspondingly more com-
plex.16
As in multi-photon excitation, discussed in Sect. 5.3, the transition amplitudes
are now to be treated at least in N th order perturbation theory. In the most simple
case, two-photon ionization of an initial single electron n state to a final state of
energy by linearly polarized light, the generalized differential ionization cross
section according to C OOPER and Z ARE (1969) is
dn (, ) m|r0 | m
m|r0 |n m
2
(2)
, (5.93)
d
Wn W +
using the transition amplitudes in the helicity basis introduced in (4.79) and (4.77).
One has to average over all initial substates n m, to sum over all intermediate states
m (in principle over all discrete bound states as well as over continuum states),
and over all final continuum states m. One may guess that the evaluation of this
expression may become a formidable enterprize. For circular polarization, or differ-
ent polarization in each of the two steps, the expression becomes even more com-
plicated. One sees that this gets even more intricate if one of the intermediate states
may be excited resonantly or nearly resonantly ( W Wn ). Then the typ-
ical damping terms must be included, and one has to account for possible saturation
effects: in the limit one just populates (non-isotropically) an intermediate state from
which ionization occurs in the second step.
As a rule, (5.81) will not describe the angular distribution of the emitted elec-
trons after MPI. Rather, the (generalized) differential cross section for N photon
ionization will involve several parameters:
(N ) (N )
dn (, ) n 1 + 2 cos2 + 4 cos4 + + 2N cos2N
= . (5.94)
d 4 1 + 2 /3 + 4 /5 + + 2N /(2N + 1)
Owing to the continuous progress with intense, tuneable and reliable laser systems
the interest in studying such processes experimentally has increased tremendously
over the paste decades. Multi-photon ionization, in particular via resonant inter-
mediate states (so called resonantly enhanced multi-photon ionization, REMPI) is
today a universally employed tool for studying the structure and dynamics of atoms,
molecules, clusters and even biomolecules.
16 Mutatis mutandis this also holds for single-photon ionization of non-isotropically populated ini-
tial states which may e.g. be prepared by additional photons in an optical pumping process with
polarized light (see Appendix D in Vol. 2 or H ERTEL and S TOLL 1978).
268 5 Linewidths, Photoionization, and More
Fig. 5.14 Resonant multi-photon ionization of Cs atoms according to C OMPTON et al. (1984).
Measured data (points) and computed (red lines) angular distributions of the photoelectron yield
for the processes listed in (5.95)
We have to confine the discussion here on two examples. C OMPTON et al. (1984)
have studied REMPI transitions in Cs atoms. The experimental setup is similar to
that shown in Fig. 5.11, but much more simple since the target atoms are gener-
ated by a simple atomic beam oven. However, electron detection and polarization
rotation are practically the same. The light source was a pulsed dye laser, tuneable
to a number of wavelengths, with 10 ns pulse duration and an intensity of moder-
ate 108 W cm2 . Cs atoms have been excited with one or two photons being in
resonance with the intermediate state depending on selection rules. One further
photon of the same wavelength served for ionization. Observed were the following
processes:
1
1
Cs(7p 2 P3/2 ) Cs+ + e ( s and p)
1
Cs 6s 2 S1/2 1 1
Cs(8p 2 P3/2 ) Cs+ + e ( s and p)
2
(5.95)
23 1
Cs(8d 2 D5/2 ) Cs+ + e ( s and p).
Figure 5.14 shows the measured and calculated angular distributions of photoelec-
trons. Comparison with Fig. 5.10 emphasizes the rich structure of the photoelectron
signal as a function of polarization angle for the present MPI processes. This is
what we expect according to (5.94). We shall return to this in the context of intense
laser fields, e.g. in Sect. 8.5.3.
In recent years so called electron imaging spectrometers (EIS) or velocity map
imaging (VMI) devices have been developed. With these very elegant and efficient
methods one collects in one shot (e.g. for one laser pulse) a complete angular and
energy distribution of the photoelectrons such images have, of course, to be av-
eraged over many events. Several variants of these techniques are used, all of them
based on the fact thatthe photoelectrons expand around their point of origin on
spheres of radius r = t 2Wkin /me . At a well defined time t one projects this spher-
ical shell onto an extended position sensitive detector and determines the kinetic
energy Wkin and by determining x/t and y/t also the emission angles and in
respect of the polarization vector.
5.5 Photoionization 269
electron signal
screen
/ arb. un.
+1600 V MCP
6 cm
0
10
el
0V 00 0 ix
3 cm
-1 0 /p
-5 y
pp / pix0 x
0 00
50
-160 V el
10 -1
Fig. 5.15 Imaging photoelectron spectroscopy for H according to R EICHLE et al. (2001).
(a) Optical EIS system with a special field shape (red dotted the equipotential lines) and two
trajectories for equal momenta p leading to the same detector position. (b) Photoelectron signal
from H in an infrared laser field with I = 1.7 1013 W cm2 . The polarization direction of the
laser points into y-direction.
The angular distribution of the photoelectrons is projected onto the
xy plane, the distance x 2 + y 2 reflects the electron kinetic energy
Here we show in Fig. 5.15(a) a typical setup used for a nice experiment with
atomic hydrogen anions by the group of Hanspeter H ELM, one of the pioneers of
these methods. With a judiciously shaped electric field one optimizes the collec-
tion of electrons which are amplified by a multi-channel plate electron multiplier
(MCP), detected by a phosphorescent screen, and finally optically registered and
digitized by a CCD camera. Figure 5.15 shows a measured electron distribution for
the photo-detachment from H with a typical ATI structure: The distance from the
centre represents Wkin , and each concentric angular distribution corresponds to one
particular value of N according to (5.91) each with its own nontrivial angular dis-
tribution corresponding to (5.94). We shall come back to these powerful techniques
at several occasions, e.g. in Chap. 5 and Appendix B.4, both in Vol. 2.
Section summary
Photoionization processes may be described as transitions between bound and
continuum states, induced by the electromagnetic field. In dipole approxima-
tion the cross section (5.61) is simply proportional to the absolute square of
the respective transition matrix element. It may be expressed in terms of the
OOSD (5.62).
FBA may be used to obtain an estimate of the cross sections at photon energies
> 5 to 10 WI . At threshold ( = WI ) the photoionization cross section
is finite (in contrast to the prediction of FBA).
The angular distribution of the ejected electron is an important characteristic
of the photoionization process, essentially imaging the continuum wave func-
tions accessed. For ionization with linearly polarized light one observes in the
general case a distribution [1 + P2 (cos )], with being the anisotropy
parameter (1 2) and the angle between polarization and ejected
electron.
270 5 Linewidths, Photoionization, and More
References
B ETHE , H. A. and E. E. S ALPETER: 1957. Quantum Mechanis of One- and Two-Electron Atoms.
Berlin, Gttingen, Heidelberg: Springer Verlag, 369 pages.
B OUCHIAT , M. A.: 2007. Linear stark shift in dressed atoms as a signal to measure a nuclear
anapole moment with a cold-atom fountain or interferometer. Phys. Rev. Lett., 98, 043003.
B OUCHIAT , M. A. and C. B OUCHIAT: 1997. Parity violation in atoms. Rep. Prog. Phys., 60,
13511396.
B URGESS , A. and M. J. S EATON: 1960. A general formula for the calculation of atomic photo-
ionization cross sections. Mon. Not. R. Astron. Soc., 120, 121151.
C HANTLER , C. T., K. O LSEN, R. A. D RAGOSET, J. C HANG, A. R. K ISHORE, S. A. KO -
TOCHIGOVA and D. S. Z UCKER : 2005. X-ray form factor, attenuation, and scattering tables
(version 2.1), NIST. http://physics.nist.gov/ffast, accessed: 7 Jan 2014.
C OMPTON , R. N., J. A. D. S TOCKDALE, C. D. C OOPER, X. TANG and P. L AMBROPOU -
LOS : 1984. Photoelectron angular-distributions from multi-photon ionization of cesium atoms.
Phys. Rev. A, 30, 17661774.
C OOPER , J. and R. N. Z ARE: 1968a. Angular distribution of photoelectrons. J. Chem. Phys., 48,
942943.
C OOPER , J. and R. N. Z ARE: 1968b. Correction. J. Chem. Phys., 49, 4252.
C OOPER , J. and R. N. Z ARE: 1969. Photoelectron angular distributions. In: S. G ELTMAN et al.,
eds., Lectures in Theoretical Physics, vol. XI-C, 317337. New York: Gordon and Breach.
C OOPER , J. W.: 1962. Photoionization from outer atomic subshells. A model study. Phys. Rev.,
128, 681693.
C OOPER , J. W.: 1988. Near-threshold K-shell absorption cross-section of argon relaxation and
correlation-effects. Phys. Rev. A, 38, 34173424.
C OVINGTON , A. M. et al.: 2007. Measurements of partial cross sections and photoelectron angu-
lar distributions for the photodetachment of Fe and Cu at visible photon wavelengths. Phys.
Rev. A, 75, 022711.
FANO , U. and J. W. C OOPER: 1968. Spectral distribution of atomic oscillator strengths. Rev.
Mod. Phys., 40, 441507.
G PPERT-M AYER , M.: 1931. ber Elementarakte mit zwei Quantensprngen. Ann. Phys.
(Berlin), 9, 273294.
H NSCH , T. W.: 2005. N OBEL lecture: passion for precision, Stockholm. http://nobelprize.org/
nobel_prizes/physics/laureates/2005/hansch-lecture.html.
272 5 Linewidths, Photoionization, and More
H ERTEL , I. V. and W. S TOLL: 1978. Collision experiments with laser excited atoms in crossed
beams. In: Adv. Atom. Mol. Phys., vol. 13, 113228. New York: Academic Press.
ISO 21348: 2007. Space environment (natural and artificial) process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
L AMBROPOULUS , P.: 1985. Mechanisms for multiple ionization of atoms by strong pulsed
lasers. Phys. Rev. Lett., 55, 21412144.
L IPELES , M., R. N OVICK and N. T OLK: 1965. Direct detection of 2-photon emission from
metastable state of singly ionized helium. Phys. Rev. Lett., 15, 690693.
M ADDEN , R. P. and K. C ODLING: 1963. New autoionizing atomic energy levels in He, Ne, and
Ar. Phys. Rev. Lett., 10, 516518.
M ANSON , S. T. and A. F. S TARACE: 1982. Photo-electron angular-distributions energy-
dependence for s subshells. Rev. Mod. Phys., 54, 389405.
N OVICK , R.: 1972. 2-Photon decay of metastable hydrogenic atoms. Science, 177, 367.
PALENIUS , H. P., J. L. KOHL and W. H. PARKINSON: 1976. Absolute measurement of photoion-
ization cross-section of atomic-hydrogen with a shock-tube for extreme ultraviolet. Phys. Rev.
A, 13, 18051816.
P ETITE , G., P. AGOSTINI and H. G. M ULLER: 1988. Intensity dependence of non-perturbative
above-threshold ionization spectra experimental-study. J. Phys. B, At. Mol. Phys., 21, 4097
4105.
R EICHLE , R., H. H ELM and I. Y. K IYAN: 2001. Photodetachment of H in a strong infrared
laser field. Phys. Rev. Lett., 87, 243001.
ROTHMAN , L. S. et al.: 2009. The HITRAN 2008 molecular spectroscopic database. J. Quant.
Spectrosc. Radiat. Transf., 110, 533572.
S AHA , H. P.: 1989. Threshold behavior of the M-shell photoionization of argon. Phys. Rev. A,
39, 24562460.
S CHFFER , H. W., R. W. D UNFORD, E. P. K ANTER, S. C HENG, L. J. C URTIS, A. E. L IV-
INGSTON and P. H. M OKLER : 1999. Measurement of the two-photon spectral distribution
from decay of the 1s2s S-1(0) level in heliumlike nickel. Phys. Rev. A, 59, 245250.
S CHMIDT , V.: 1992. Photoionization of atoms using synchrotron radiation. Rep. Prog. Phys., 55,
14831659.
T ONG , D., S. M. FAROOQI, E. G. M. VAN K EMPEN, Z. PAVLOVIC, J. S TANOJEVIC, R. C OTE
E. E. E YLER and P. L. G OULD: 2009. Observation of electric quadrupole transitions to Ryd-
berg nd states of ultracold rubidium atoms. Phys. Rev. A, 79, 052509.
W IKIPEDIA CONTRIBUTORS: 2013. Atmospheric radiative transfer codes, Wikipedia, The Free
Encyclopedia. http://en.wikipedia.org/wiki/Atmospheric_radiative_transfer_codes, accessed:
8 Jan 2014.
Fine Structure and L AMB Shift
6
Overview
This chapter focuses on some key topics of atomic physics. It should be stud-
ied carefully. To start, Sect. 6.1 introduces some state-of-the-art experimental
methods used today in high precision spectroscopy. Section 6.2 introduces
into the phenomenology and terminology of spin-orbit interaction and an-
gular momentum coupling, while Sect. 6.3 may be seen as the quantitative
counterpart perhaps somewhat strenuous, but important. The focus is on
the H atom, described by the D IRAC theory and the only system for which
fine structure (FS) splittings can be computed exactly. Alkali atoms serve as
further examples. Section 6.4 is concerned with selection rules and intensi-
ties for FS transitions, a somewhat dry but necessary subject which the reader
may perhaps skip on first reading and come back later. The L AMB shift is
presented in Sect. 6.5 with the classical experiments as well as with some
results from modern high precision spectroscopy for the H atom and highly
charged ions. Understanding the physics should be supported by Sect. 6.5.6
with a very brief introduction to F EYNMAN diagrams and a glimpse on QED.
Section 6.6 concludes the chapter with a short excursion into the physics of
the electron magnetic moment anomaly.
exit
slit B
spherical grating
6.1 Methods of High Resolution Spectroscopy 275
sin FN
sin
Fig. 6.2 (a) Standard geometry for a reflecting grating. Schematically indicated is the optical
delay for two light rays on the way in (fat, red line) and out (fat black line). Angles of incidence
() and diffraction () in respect of the grating normal (GN) are taken as positive to the left,
negative to the right of GN. (b) Blazed grating with a blazing angle B between the facette normal
(FN) and GN. (c) Echelle grating in a near L ITTROW arrangement; shown is an R3 echelle with a
blaze angle of 71.5
exit slit B. Typically, the grating is moved along the circle to vary the wavelengths
which reaches B.
The actual diffraction and interference geometry of a reflective grating is shown
in Fig. 6.2(a). We see that the optical delay between rays 2 and 1 from neighbour-
ing grooves of the grating at a distance d is given by
s = d(sin + sin ). Note
that < 0 in the arrangement of Fig. 6.2(a). We expect constructive interference in
the outgoing wave front if the grating equation
c
p = z , or in frequency space = z (6.1)
d
s
with p = (sin + sin ),
s = pd and z = 0, 1, 2, . . .
holds, with z being the order of the interference and pd the optical delay. An im-
portant characteristic of any spectrometer is the change of the deflection angle with
wavelength. One derives the change of the interference maximum with wavelength
(at a fixed angle of incidence ) from (6.1), and finds for a grating the
d 1 z
angular dispersion = . (6.2)
d cos d
1 The concept of lateral coherence will be discussed in some detail in Chap. 2, Vol. 2. Essentially,
coherent illumination requires that s sin d holds for the width s of the spectrometer entrance
slit, where d is the full opening angle of the optics illuminating the grating.
276 6 Fine Structure and L AMB Shift
=0 =1 =2
1
(a)
0 /d /d
2
1 (b)
Fig. 6.3 (a) Normalized interference function T (p)/T0 (p) of a diffraction grating with N coher-
ently illuminated grooves in 0th, 1st and 2nd order for two wavelengths differing by
; (b) both
traces added, illustrating that the two wavelengths are (just) distinguishable when the principle
maximum of one overlaps with the first side minimum of the other (here in 2nd order)
T0 (p) represents an overall normalized intensity function for a single groove of the
grating.2 It depends on the shape and reflectivity of an individual groove and varies
smoothly.
Figure 6.3(a) shows the interference function T (p)/T0 (p) due to multi-ray in-
terference for two, slightly different wavelengths and +
, with maxima
at p = z/d for z = 0, 1 and 2 according to (6.1). According to the so called
R AYLEIGH criterium, illustrated in Fig. 6.3(b), the two wavelengths can just be
distinguished when the principle maximum of the latter coincides with the first side
minimum in the fringe pattern of the former (i.e. from N coherently interfering
rays). With (6.1) and (6.3) the principle maxima are positioned at
N pkd/2 = N pd/ = N z
while the optical delay (p +
p)d for the first side minima corresponds to
N (p +
p)d/ = (N z + 1)
p = .
Nd
On the other hand, the principle maxima for a wavelength +
correspond to
+
p +
p = z
p = z .
d d
By equating the two expressions for
p we obtain the
2 Inthe most simple case, a transmission grating consisting of slits with a width s, this would be
the single slit transmission function
T0 (p) = I0 sinc2 (ksp/2), with sinc(x) = sin(x)/x.
6.1 Methods of High Resolution Spectroscopy 277
which holds for all interference based spectrometers. For grating spectrometers it
determines the spectral resolution usually given as difference
of the wave-
length between two neighbouring spectral lines which can just be distinguished.
Modern gratings are fabricated by holographic methods or are ruled and may have
several thousands of grooves per mm and a length of up to a few 100 mm, thus
N may in principle be up to 100 000 or more, albeit difficult to illuminate coher-
ently. Typically they operate in 1st or 2nd order. Special shape and reflectivity of
the surface may help to channel all diffracted intensity into one interference order,
thus avoiding ambiguities in the observed spectra. Specifically, blazed gratings have
grooves with facets that are aligned such that the maximum diffraction angle in re-
spect of the facet normal (FN) is equal to the entrance angle in respect of FN as
indicated in Fig. 6.2(b). One reads from the figure that this blaze angle is
B = ( + )/2 (6.5)
where + is derived from (6.1) for a specific wavelength. The diffracted signal
then appears as if directly reflected from the facets. Even though such geometric
interpretation is somewhat questionable (typically the wavelengths are on the same
order of magnitude as the groove distance) detailed calculations show that one may
indeed maximize in this manner the light reflected for a specific order of interference
and wavelength B one says the grating is blazed for that wavelength.
At last we mention so called echelle gratings (from the French word chelle,
stairs or ladder) which have experienced a true renaissance in recent years. In con-
trast to standard gratings they have only a moderate number of grooves N , but very
large z 1 and large blaze angles. A typical setup is shown in Fig. 6.2(c). They
may be viewed as an intermediate between grating spectrometer and interferometer.
One fabricates them today by ruling or chemical etching. Characteristic for echelle
spectrometers is the combination of high dispersion according to (6.2) and high re-
solving power according to (6.4). A disadvantage is that due to the high value of z
the spectra may overlap in different orders. In frequency space we read from (6.1)
that frequencies and +
FSR have a maximum of order z and z +1, respectively,
at the same optical path difference
s if
c
FSR = = . (6.6)
s z
Thus, frequencies differing by more than
FSR cannot be assigned unambiguously.
Obviously, the higher the order of interference z for which the system is designed,
the more serious is the problem. One may overcome it by crossing the echelle with
a suitable second dispersive element (grating or prism) which leads to an unambigu-
ous 2D pattern, ideally suited for state-of-the-art CCD devices for detection.
Finally, one should note that in practice the degree to which the theoretical re-
solving power is attained (by any grating spectrometer or monochromator) depends
not only on the angles and (and on z and N ) but also on the optical quality of
the grating surface, the uniformity of the groove spacing, the quality of the associ-
ated optics in the system, and (most importantly) on the width of the (entrance and
278 6 Fine Structure and L AMB Shift
collimating s=s2 - s1
lens
s2
light
source beam
splitter reflecting
mirror 1
focussing lens
detector
exit) slits (or detector elements). Any departure of the diffracted wavefront greater
than /10 from a plane (for a plane grating) or from a sphere (for a spherical grat-
ing) will result in a loss of resolving power due to aberrations at the image plane.
The grating groove spacing must be kept constant to within about 1 % of the wave-
length at which theoretical performance is desired. (PALMER and L OEWEN 2005,
Chap. 2.4)
6.1.2 Interferometers
Expressions analogue to (6.4) also hold for all interference based spectrometric de-
vices, such as M ICHELSON, M ACH -Z EHNDER, FABRY-P ROT (FPI) etc. interfer-
ometers. In an genuine interferometer N may vary from 2 beams in a M ICHELSON
setup to several 100 in an FPI. On the other hand, the order of interference z is
very large in all interferometers: For an optical path difference
s of the interfering
beams it is
z =
s/ =
s/c. (6.7)
In the visible (VIS) a moderately large optical path difference of
s = 6 cm corre-
sponds already to z = 105 . One may thus in principle obtain very high resolution
with interferometers. Typically one observes fringes of equal inclination (slightly
different divergence angles lead to a modification of the optical path difference
s).
Since in various contexts M ICHELSON interferometers will play a role later on
(e.g. in F OURIER transform spectrometers, see Sect. 5.3.2 in Vol. 2), we show such
a setup in Fig. 6.4 for reference without entering here into details (N OBEL pize for
M ICHELSON 1907). The key elements are a beam splitter and two reflecting mirrors
between which the light travels an optical distance s1 and s2 , respectively. The op-
tical path difference
s = s2 s1 may be varied by moving one of the mirrors with
high precision. Unfortunately, spectra obtained with interferometers are not unam-
biguous (just as gratings used in higher order diffraction): assume that the central
6.1 Methods of High Resolution Spectroscopy 279
interference pattern
r',t' r,t
3
C . 2
B P
2
L
A 1
possibly n >1
c
FSR = , (6.9)
2nL
280 6 Fine Structure and L AMB Shift
z = 2nL/. (6.11)
which is a geometrical series that can be written in closed form. Neglecting losses,
remembering that intensities are I |E|2 , and evaluating the geometrical series
emerging from (6.13) one obtains the
It 1
A IRY function = (6.14)
I0 1 + F sin2 ( 2 )
inclined
3P 3/2
incidence
detector beam excitation emission
splitter
FABRY-PEROT CW dye laser beam 3S1/2
as marker tunable, perpendicular
Fig. 6.7 (a) Schematic of an atomic or molecular beam setup for high resolution spectroscopy.
(b) Partial term scheme of Na, with the two Na D2 HFS lines recorded
The precise observation of finer effects in atomic spectra is not only limited by the
resolving power of the spectrometers but very often also by the D OPPLER broad-
ening (5.22). By studying spectra in atomic (or molecular) beams this handicap
may be overcome. A typical experimental setup and detection scheme is sketched
in Fig. 6.7: The atomic beam emerges from an oven (preferably at high pressures
to allow for adiabatic cooling) and is collimated by a skimmer and some apertures
to a divergence angle
. The velocity distribution has its maximum at vm and
a FWHM =
v. Let the transition wavelength at resonance be ba and the angle
of light incidence . With (1.32) the resulting D OPPLER shift
ba and D OPPLER
broadening D is then
vm vm
v
ba = ba cos = cos and D = cos , (6.18)
c ba ba
respectively. When exciting (or detecting) perpendicular to the beam
marker
100 MHz
Fig. 6.8 Hyperfine splitting of the Na D2 line (32 P3/2 3 2 S1/2 transition) recorded with the
atomic beam apparatus sketched in Fig. 6.7(a). One clearly sees the two hyperfine components with
perpendicular (black) and with inclined (red) irradiation as function of the exciting frequency .
The full red line gives 4 the signal extracted from the combined signal (red dashed). The marker
shown at the bottom was simultaneously recorded with the FPI
One observes the whole fluorescence signal as a function of the incident fre-
quency and thus registers simultaneously the excitation spectrum for perpen-
dicular and inclined incidence of the light beam. Both contributions can easily be
separated as indicated in Fig. 6.8. While tuning the laser, a reference signal is simul-
taneously recorded from a small fraction of the laser beam which passes through an
FPI. This signal appears and vanishes in intervals corresponding to the free spectral
range of the FPI (here 100 MHz). It generates a convenient marker for calibrat-
ing the recorded spectrum. For perpendicular irradiation, the two HFS components
of the Na D2 line are clearly separated by ca. 1700 MHz. The observed linewidth
D 100 MHz corresponds to that predicted by (6.19). With some effort it can
be further reduced.
Inclined incidence (red line Fig. 6.8) leads to a D OPPLER shift ba ba
1800 MHz and a broadening of D 900 MHz much larger than for perpendic-
ular incidence again in full agreement with the predictions of (6.18) for the given
atomic beam conditions.
Laser spectroscopy on ions is often performed with energetic ion beams (the follow-
ing also holds for high energy neutral particle beams). In such experiments the laser
beam travels parallel or antiparallel to the fast ion or neutral beam. The method
has two advantages: On the one hand, one may tune the absorption frequency of the
ions by varying the energy of the beam and keeping the laser frequency fixed L at
some convenient laser line. On the other hand, D OPPLER broadening can be reduced
considerably (so called D OPPLER narrowing). The D OPPLER shift of the absorption
frequency ba as seen from the laser beam is according to (1.30) and (1.34)
ba = ba wk + 2wk + wk2 L 2wk (1 wk /2 + ), (6.20)
284 6 Fine Structure and L AMB Shift
for ions with a rest mass m and a kinetic energy Wkin , with wk = Wkin /(mc2 ). The
+ and signs refer to irradiation head-on or parallel to the ion beam, respectively.
This shift may be considerable for ions and allows for calibrated scanning through
the finer structures of spectra by tuning the acceleration voltage U , with Wkin = qeU
for ions of charge qe.
Typically the ions are prepared in an ion source with a finite energy width Wk at
low kinetic energies. According to (6.20) this would lead to a D OPPLER broadening
2Wk
D ba . (6.21)
mc2
During acceleration of the ion beam Wk usually stays constant, and so does wk =
Wk /(mc2 ). Hence, for higher but not too high kinetic energies (Wk Wkin
mc2 ) the D OPPLER width is derived from (6.20):
d
ba wk ba Wk
D wk ba (1 + 2wk + ) . (6.22)
dwk 2wk 2Wkin mc2
Comparison with (6.21) shows that the linewidth is reduced by a factor 4Wkin /Wk
in respect of a measurement in the ion source due to D OPPLER narrowing. Typ-
ical for ion beam spectroscopy would be a 100 keV beam and an initial width
Wk = 1 eV in the source (much above thermal energy but characteristic for ion
sources), amounting to a reduction factor of 630.
For illustration, in the case of He+ ions the thermal D OPPLER width at room
temperature according to (5.21) is D (293 K) = ba 6.1 106 . The linewidth in
the 100 keV ion beam (Wk = 1 eV) would be D = ba 3.66 108 a factor
of about 170 less than even the thermal D OPPLER width! This already allows a
convenient resolution (even if not a high precision measurement) of the L AMB shift
for the n = 2 states in He+ which is around 1.423 106 ba . The D OPPLER shift
according to (6.20) in this case is 0.007ba , which allows for sufficient tunability
(in the per mill range) by changing the acceleration voltage. The method has been
used (and is still used) extensively for the determination of HFS and FS structure as
well as of the L AMB shift for heavy ions, specifically so for their unstable isotopes.
The basis of many modern methods of D OPPLER free spectroscopy is the so called
hole burning by which a certain part of a population of the sample atoms or
molecules is depleted by monochromatic excitation. This allows a nice illustration
of the terms homogeneous and inhomogeneous linewidth.
As we have seen in Sect. 5.1.4, in gas phase the D OPPLER width is typically much
larger than the natural linewidth: D
nat . To record the absorption probability
as a function of the frequency for an ensemble of atoms or molecules we may excite
with a (quasi-monochromatic) narrow band tuneable laser in the neighbourhood of a
resonance frequency ba . As discussed in Sect. 5.1.4 and Sect. 6.1.3 the absorption
6.1 Methods of High Resolution Spectroscopy 285
(a) (b)
BOLTZMAN N BOLTZMANN
with hole
LORENTZ
0 (velocity || of light) 0
Fig. 6.9 Velocity distribution of atoms or molecules in the gas phase parallel to the wave vector
k of the light. (a) M AXWELL -B OLTZMANN distribution prior to absorption and L ORENTZ profile
for a particular velocity class vtest (dashed). (b) Velocity distribution with hole, which has been
created by the radiation absorbed at vtest
vx () = v cos = c 1 = ba ( ba ) (6.23)
ba
where vx is the projection of the atomic velocity onto the wave vector k of the
light, with k x. More precisely, atoms with velocities v slightly smaller or larger
than vx () also absorb light of the frequency with a probability given by the
Lorentzian natural line profile on the velocity scale. This is illustrated in Fig. 6.9(a)
by the red dashed line.
Hence, when irradiating with a fixed test frequency test one only excites atoms
of a particular velocity class around vtest = vx (test ) of the initial velocity distribu-
tion. If the light intensity is sufficiently high this results in a noticeable depletion
of this velocity class in the initial state D OPPLER profile as shown in Fig. 6.9(b):
the remaining velocity distribution of atoms in the initial state has then typically a
hole at vtest . The width of the absorbing velocity class corresponds to the natural
linewidth
nat = ba
nat : all atoms within the homogeneous linewidth partici-
pate in the depletion correspondingly. The same argument hold for collision broad-
ened profiles of a width (5.17) if that is much larger than the natural linewidth.
One may exploit hole burning very elegantly for spectroscopy. This has first been
realized by H NSCH et al. (1971) who coined the term D OPPLER free saturation
spectroscopy. A typical experimental setup is shown in Fig. 6.10 and the principle
286 6 Fine Structure and L AMB Shift
las beam
er splitter
phase sensitive
detection
differential
beam chopper amplifier
splitter
mirror
mirror
probe beam pumpbeam
photo
sample vapour (gas) in cell reference beam detectors
e.g. Na or H(2s,2p)
Fig. 6.10 Experimental setup for D OPPLER free saturation spectroscopy according to H NSCH
et al. (1971)
(a)
(b)
(c)
Fig. 6.11 Principle of D OPPLER free saturation spectroscopy: circles with arrows indicate atoms
and their velocity vectors, full colours indicate resonance, grey out of resonance. The incident
frequency is (a) blue shifted in respect of ba , (b) exactly ba or (c) red shifted. Consequently
(a) the probe beam excites atoms moving to the left, the pump beam only those moving to the
right; (c) the probe beam excites atoms moving to the right, the pump beam only those moving to
the left; (b) pump and probe beam excite only (the same) atoms at rest
of the method is illustrated in Fig. 6.11. One splits the laser beam into an intense
pump beam and a weaker probe beam, and sends them counter-propagating through
the cell with the sample. The pump beam burns holes into the D OPPLER profile of
the target atoms in the ground state. Ideally the transition is then saturated for that
frequency, i.e. nearly one half of the atoms in that particular velocity class are found
in the excited state. All atoms not in resonance with the pump beam frequency are
unaffected. Hence, in most situations the probe beam does not notice anything of this
preparation since it comes from the opposite side and is in resonance with a different
velocity class. Only if both, pump and probe beam match exactly a resonance transi-
tion for atoms at rest the probe beam sees the same atoms as the pump beam. This
leads to a reduction of the absorption exactly when the laser frequency corresponds
to the atomic resonance frequency for atoms at rest. One calls this reduction L AMB
dip. The signal may be recorded in comparison to a reference signal without pump-
6.1 Methods of High Resolution Spectroscopy 287
crossover
equivalent experiment:
DOPPLER profile saturation signal
probed in Na bulb
1700
MHz
100 MHz marker
Fig. 6.12 D OPPLER free saturation spectrum of the two main hyperfine lines in the
3 2 P3/2 2 S1/2 transition of sodium (Na). The spectrum has been recorded by a most unso-
phisticated setup with Na in a bulb to illustrate the robustness of the method. One recognizes
the two characteristic L AMB dips showing the HFS splitting of 1700 MHz and one (physically
insignificant) crossover signal in between
ing and typically is registered by phase sensitive detection which synchronizes the
detected signal with a chopper for the laser beam to minimize statistical fluctuations
(using a so called lock-in amplifier).
If several allowed transitions are accessible over the tuning range of the laser,
correspondingly several L AMB dips appear. This is illustrated in Fig. 6.12 for the
hyperfine structure of the Na atom which we have already encountered in Sect. 6.1.3,
Fig. 6.8. In contrast to the latter experiments, the setup for recording the saturation
spectrum Fig. 6.12 was particularly simple: just a bulb with Na vapour, no opti-
cal surfaces, and rather weak pump and probe laser beams. Consequently the two
L AMP dips in the Na D2 line (3 2 P3/2 3 2 S1/2 ) separated by ca. 1700 MHz HFS
splitting are small, but clearly visible on the D OPPLER profile, and illustrate the
robustness of the method. Unfortunately, the method is not completely unambigu-
288 6 Fine Structure and L AMB Shift
ous: so called crossover signals may arise from two closely spaced lines exactly
halfway between them.
In any case, this D OPPLER free saturation spectroscopy is an excellent method
to determine resonance lines with an accuracy on the order of the natural linewidth
irrespective of being taken in a gas cell with much larger D OPPLER width. A par-
ticular nice and important example (H NSCH et al. 1972, 1974) is illustrated in
Fig. 6.13. In this pioneering work the fine structure of the H line has been deter-
mined for the first time with high accuracy. To emphasize the resolution Fig. 6.13(a)
gives an overview for the BALMER series of atomic hydrogen. For comparison
(b) shows two D OPPLER broadened spectra in the frequency range of interest, sim-
ulated for 300 K and 50 K, respectively, which obviously do not allow to resolve
the fine structure in detail. In contrast, saturation spectroscopy (c) gives a very clear
picture of the fully resolved fine structure. We shall come back to it later.
Even if one may overcome the D OPPLER broadening and other pitfalls of spectro-
scopic observations by clever methods, one often encounters the situation that the
resolution is limited simply because not enough time is available to perform a mea-
surement since, alas, the fundamental conditions imposed by the uncertainty relation
(1.125) between time and energy provide apparently stringent boundaries:
W t .
As a consequence one obtains an overall line profile as shown in Fig. 6.14. The
practical realization for transitions in the radio frequency and microwave region
caused no problem even in 1950 since there the phase coherence is a built in prop-
erty of these waves. With modern laser techniques the required fixed phase relation
between two different laser pulses is no problem either. We shall get to know a
few practical, state-of-the-art realizations in Fig. 6.16 and Fig. 6.28 as well as in
Sect. 9.5.1.
Another very elegant way to realize D OPPLER free spectroscopy is based on two-
photon excitation (introduced in Sect. 5.3.1).
The method has been applied experimentally for the first time in 1974 nearly
simultaneously by several groups. For a detailed review the interested reader is re-
ferred to G RYNBERG and C AGNAC (1977). One uses a standing wave, e.g. a cir-
cularly polarized wave field, created by a FABRY-P ROT interferometer advan-
tageously in a confocal resonator with spherical mirrors so that focussing the laser
beam provides sufficient intensity for efficient excitation of two-photon processes.
The gas cell with the sample atoms or molecules is positioned into the focus of the
mirrors possibly the mirrors may even act as walls of the gas cell.
The basic idea of the methods is the following. If v is the velocity of an atom or
molecule in the gas cell and k the wave vector of the exciting light (in a FABRY-
P ROT resonator defined by the beam axis), then according to (1.33) the D OPPLER
shift in 1st is k v (angular frequency). For the inverse direction of light propagation
(k k) the sign of the D OPPLER shift changes for this particular atom. Let us
assume that a two-photon process is possible in the target atom between levels of
energy Wa and Wb . An atom in a resonator with the two counter-propagating waves
290 6 Fine Structure and L AMB Shift
150 MHz
5 S1/2 (a)
(b)
4S1/2 4P 4P3/2
61 1/2
6n
laser wavelength
61
602.23 nm
5n
330 n cence 3P3/2
m
3P1/2 0 300 MHz dye laser frequency
re s
m
m
m
m
589 n
fluo
590 n
330 n
(c)
1772 MHz
3S1/2
0 600 MHz atomic transition
frequency
Fig. 6.15 Two-photon transition between HFS states of Na, exemplified for the 3s 2 S1/2
5s 2 S1/2 transition, according to G RYNBERG and C AGNAC (1977). (a) Part of the term scheme
with fluorescence lines (in nm) for detection; (b) measured two-photon spectrum; (c) calibration
with a FABRY-P ROT interferometer: the marker lines are spaced by one free spectral range of the
FPI of 300 MHz, corresponding to a change in absorption frequency of 600 MHz
Section summary
High resolution spectroscopy is the key to the investigation and understanding
of finer details in atomic structure, such as FS and HFS. A brief summary of
some essential tools has been given in this section.
292 6 Fine Structure and L AMB Shift
vacuum chamber
cold finger
(a) gas discharge
H atoms 2S-
FARADAY
cage detector
dye laser SHG AOM
7500
800 Hz
2S signal / cps
50
2500
0 0
- 80 - 40 0 40 80 - 20 - 10 0 10 20
detuning / kHz @ 121 nm detuning / kHz @ 121 nm
Fig. 6.16 Two-photon excitation of the 1s 2 S1/2 2s 2 S1/2 transition in the H atom according to
G ROSS et al. (1998). (a) Scheme of the experiment. An acousto-optic modulator (AOM) generates
suitable pulse sequences at 243 nm. (b) Excitation probability as a function of detuning in respect
of the 2h resonance when irradiated by 50 s pulses at 10 kHz repetition rate; (c) excitation rate
3.2 kHz, otherwise as (b)
In Chap. 3 we have described in detail how for alkali atoms the degeneracy of energy
levels with the same principle quantum number n is removed for different orbital
angular momentum quantum numbers . However, as we have already noted in the
preceding section, on closer inspection one finds that even the n levels are again
split into a finer substructure and the same holds for the absorption and emission
lines of the H atom.
Quantitatively one finds that this so called fine structure splitting (FS) is on the
order of magnitude
WFS Eh
Z2 = (Z)2 (6.25)
Wn me c2
with the atomic unit of energy Eh , the rest energy of the electron me c2 , and =
Eh /me c2 the fine structure constant (1.10) which we have encountered already on
several occasions. This order of magnitude already suggests that FS must be of
relativistic origin.
For light atoms the splitting is rather small and non-trivial to measure in the gas
phase. For the H atom in the first excited state FS splitting is of similar magnitude
as D OPPLER broadening (5.22) at room temperature,
Figure 6.17 illustrates the term scheme. The splitting is about
17 cm1 while
the D OPPLER width amounts to D 0.06 cm1 , so that spectroscopic separa-
tion is possible without problems. One needs an experimental resolution of about
/
= /
1000. That can easily be achieved with a grating monochromator.
All this observed fine structure is connected with the electron spin. As already
discussed in Sect. 1.10 the spin (quantum number S) is accompanied by a magnetic
moment which interacts with the orbital angular momentum (quantum number L)
of the electron. This is called spin-orbit (or LS) interaction. Since two possible
294 6 Fine Structure and L AMB Shift
3s 2S1/2
orientations of the electron spin, sz = 1/2, are possible in respect of a given axis
here defined by the angular momentum, the levels are split into two components.
For the example of a 3p electron shown in Fig. 6.17 the two states are characterized
by a total angular momentum j = 3/2 and 1/2. For the 3s state the orbital angular
momentum is zero, hence j = 1/2 and no splitting is observed. We shall discuss this
in detail below.
B = M
B = M
V z B (6.29)
If only the orbital magnetic moment or only the spin magnetic moment were acting,
these expressions would represent the level splittings in an external B field, and one
would observe
The situation is further complicated by the interaction between the magnetic mo-
ment M S of the electron spin and the magnetic field B L created by the orbiting
electron. To derive this interaction quantitatively one needs to know the effective
magnetic field at the position of the electron spin. Thus, one transforms the mo-
tion into a coordinate system where the atomic nucleus rotates around an electron
considered at rest as indicated in Fig. 6.18. Without going into details of the un-
derlying electrodynamics we note: The L ORENTZ force F = qv B (1.105) onto
a charge q moving with velocity v in a magnetic field B may be viewed as the re-
sult of an effective electric field E eff = v B. Conversely, a charge moving in an
electric field E generates a magnetic field
vE
B = . (6.32)
c2
In the potential V (r) of the ionic core of an atom (atomic nucleus + core elec-
trons) the electric field E experienced by an electron is derived from
r dV
eE = F = .
r dr
Hence, the electron moving in this field generates according to (6.32) a magnetic
field
v r dV 1 1 dV 1 1 dV
BL = = (r me v) = L (6.33)
ec2 r dr eme c2 r dr eme c2 r dr
stitute this expression with a positive sign into (6.29) and obtain with (6.28) and
ge = 2:
LS = MS B L = 1 ge 1 dV L S .
2
V (6.34)
4 me c r dr 2
2 2
We have introduced here without proof the so called T HOMAS factor 1/2 which
follows directly from D IRAC equation when re-transforming V LS into the nuclear
system at rest.
Introducing Eh , a0 and the fine structure constant = Eh /(me c2 ) the spin-
orbit (or LS) interaction is
LS = a 3 ge (r) L S Eh
2 40 dV
V with (r) = a03 . (6.35)
4 0
2 r dr
Note that for an attractive, pure C OULOMB potential (r) = Zr 3 . To simplify the
writing, in the remainder of this chapter all equations will be given in a.u. (unless
otherwise indicated).
V L
LS
= a S
2
with a = ge n |(r)|n
(6.36)
4
2 1 dV 2 1 dV 2
= n | |n
= 2
Rn (r) r dr.
2 r dr 2 0 r dr
Here we have introduced the spin-orbit coupling parameter a. Only for the H
atom and H atom like ions it can be evaluated in closed form. The results are, how-
ever, characteristic and may readily be extended to quasi-one-electron systems such
as the alkalis. For V (r) = Z/r and (r) = Zr 3 the coupling constant a can
be derived from the expectation value (2.130) of r 3
for hydrogenic wave func-
tions:
2 1 2 Z4
a = ge Z 3 = . (6.37)
4 r 2 n3 ( + 1/2)( + 1)
This disappears for = 0. For > 0 we obtain a good estimate by approximating
L
S
s so that FS splitting is seen to be on the order of
2 Z4 VLS
(Z)2
LS
V and ,
2 n3 ( + 1/2)( + 1) |Wn | n( + 1/2)( + 1)
6.2 Spin-Orbit Interaction 297
where |Wn | = Z 2 /2n2 has been introduced to obtain the relative perturbation. At
small Z 11 it is very small, only about 106 . . . 103 for n 3. In general, the
FS splitting decreases 1/n3 (for any given , spectroscopically relevant being
4).
However, spin-orbit interaction becomes substantial for heavy atoms because of
its proportionality to Z 4 . Perturbation theory will break down altogether for Z 1.
The physics emerging from such a situation is subject to current research with atoms
of nuclear charges beyond the natural stability limit for 92 U. Such investigations are
carried out today at heavy ion accelerators and storage facilities where this limit can
easily be surpassed in highly energetic collisions.
r2 /2 and
2
with p L /(2r 2 ) being the radial and angular kinetic energy. To evaluate the
fine structure term 2 (r)L
S/2 we have to consider the coupling of spin and orbit
through their respective magnetic fields and derive the resulting eigenstates of the
Hamiltonian (6.38).
One formally introduces an operator of total angular momentum
L +
J = z +
S and its z component Jz = L Sz , (6.39)
for which the standard commutation rules hold in full analogy to those for orbital,
(2.84), (2.75), and spin angular momenta, (2.91):
J , Jz = 0
2
(6.40)
Jx , Jy = iJz , Jy , Jz = iJx , Jz , Jx = iJy . (6.41)
This implies that one can measure simultaneously the magnitude and one of the
components of the total angular momentum but not two components. Specifically
we shall use eigenstates |j mj
for which
by a vector model as shown in Fig. 6.19. The following constraints must hold:
|
J | = j (j + 1), |
L| = ( + 1), |
S| = s(s + 1) (6.44)
with = 0, 1, 2, . . . , n 1,
| s|, | s| + 1, . . . j | + s| (6.45)
We have already encountered the triangular relation in Sect. 4.4.1 when discussing
selection rules for dipole transitions and the photon spin sph = 1. For an electron
spin s = 1/2 and > 0 it leads to j = 1/2, while for s = 1/2 and = 0 we
have j = 1/2. The possible projections mj are illustrated in Fig. 6.20 for the case
of j = 3/2.
L+
J = (
S)2 = L
L + 2 S +
2 2 2
S (6.46)
and the scalar product of angular momentum and spin operators becomes
1 2 2 2
L
S= J L S . (6.47)
2
The whole H AMILTON operator (6.38) can then be rewritten:
2
2 2
FS (r) =
H
p 2r L 2
+ 2 + V (r) + (r) L
J
2
S . (6.48)
2 2r 4
Hence, we have to search for eigenstates |( s)j mj
created by the coupling of
L
and
S which are at the same time eigenstates of J ,L and L
S and thus of
2 2 2
S.
6.2 Spin-Orbit Interaction 299
one may see with (6.46) that L and S commute also with L S and
2 2 2
J . Hence,
L ,
S and FS according to (6.48) and j , and s are
2 2 2
J all commute with H
good quantum numbers (see Sect. 2.6.5).
L
Note that S does not commute with L z and not with
Sz as one sees from mul-
tiplying (6.50) with L x nor L
z since neither L y commute with L z , and analogical
for Sz . Thus, neither m nor ms are good quantum numbers. On the other hand,
z +
Jz = L Sz does commute with L S, since with (6.50)
1 2 2 2
L
[ S, Jz ] = J , Jz L , (Lz + Sz ) S , (Lz + Sz ) = 0.
2
Thus, mj is a good quantum number, and we note that because of Jz = L
z +
Sz
m + ms = mj (6.51)
must always hold for eigenstates of the Hamiltonian (6.38), while neither m nor ms
are good quantum numbers.
The eigenstates of the complete Hamiltonian H FS which we are looking for are
thus characterized by n, , s, j, mj :
FS |n sj mj
= Wn j |n sj mj
.
H
L
We note, that (6.42) holds for these states. They are also eigenstates of S and
according to (6.47) we have (in units of )
2
1
L
n sj mj | S|n sj mj
= j (j + 1) ( + 1) s(s + 1) . (6.52)
2
becomes diagonal. For spin and orbital angular momentum this can be achieved by
using the definitions and commutation rules for angular momenta as sketched above.
We refer to the relevant extensive literature on angular momentum algebra and to
the brief extract of the most important formulas collected in Appendix B.2.3 The
angular part of the eigenstates for (6.48) is written
( s)j mj = | m
|sms
m sms ( s)j mj or
m ,ms
mj + s s j
= 2j + 1 | m sms
(1) , (6.53)
m ms mj
m ,ms
1/2
(s1 s2 )SM = s1 m1 s2 m1 |s1 s2 SM
|s1 m1 s2 m2
m1 =1/2
meter2 =1/2
1/2
s1 s2 S
= 2S + 1 (1)M |s1 m1 s2 m2
.
m1 m2 M
m1 =1/2
meter2 =1/2
With help of the readily accessible 3j symbols (e.g. with a Java applet from S VEN
G ATO R EDSUN 2004) one verifies that the nonvanishing C LEBSCH -G ORDAN coef-
ficients are
1 1 1 1 1 1 1 1
1 1 = 1
2 2 2 2 11 = 1, 2 2 2 2
1 1 1 1 1 1 1 1
10 =
2 2 2 2 2 2 2 2 10 = 1/ 2
1 1 1 1
00 = 1 1 1 1 00 = 1/ 2.
2 2 2 2 2 2 2 2
3 Several slightly different definitions are used. We follow B RINK and S ATCHLER (1994).
6.2 Spin-Orbit Interaction 301
Table 6.1 Triplet and singlet states with C LEBSCH -G ORDAN coefficients for the coupling of
two electrons with spin 1/2. The symmetry of the spin functions is given as well as symbolic
representations of the two spin 1/2 states involved
Multiplicity Symmetry |SM
= | (1, 2)
S M
|11
= |(1)(2)
1 1
equal
|(1)(2)
+ |(1)(2)
Triplet symmetric |10
= 1 0
2 phase
|11 = |(1)(2) 1 1
opposite
|(1)(2)
|(1)(2)
Singlet antisymmetric |00
= 0 0 phase
2
The corresponding spin functions for triplet (symmetric) and singlet states (anti-
symmetric) are summarized and symbolized in Table 6.1, using the abbreviations
|
= |1/2 1/2
and |
= |1/2 1/2
as introduced in Sect. 2.5.4. The numbers in
brackets behind the spin designations and refer to electron number (1) or (2).
Note that the symmetry of a full wave function in spin and position is the product
of the symmetries in spin and position space. As we shall elaborate on in Sect. 7.3.1
and Sect. 10.2.1 the PAULI principle demands the full wave function to be antisym-
metric!
coupling scheme which is relevant for heavier atoms where spin-orbit interaction is
substantial.
Here we just communicate the general terminology of LS coupling. The number
of possible orientations of the total spin is called
multiplicity = 2S + 1 (6.54)
of the state. For a one electron system such as alkali atoms we have S = s = 1/2
and the multiplicity is always 2 1/2 + 1 = 2, one calls this a doublet. As we
have just seen in Sect. 6.2.5, atoms with two electrons have singlet and triplet states
(S = 0 and S = 1, respectively). In systems with N electrons in principle multi-
plicities up to 2N 1/2 + 1 could be constructed. Typically, quintuplets are the
highest multiplicities observed (constructed from 4 electrons, i.e. half the number
of electrons in a completely filled outer valence shell with two s and four p elec-
trons).
To distinguish the quantum number of a single electron from those of the full
atom on characterizes them with lower case and capital letters, respectively. Hence,
the total orbital angular momentum of an atomic state (quantum number L) is char-
acterized by upright capital letters S, P, D, F, G for L = 0, 1, 2, 3, 4 (in analogy to
s, p, d, f, g, . . . for the i quantum numbers of individual electrons). As superscript
on the left of the angular momentum one writes the multiplicity, as subscript on
the right the total angular momentum is given. In summary, an atomic state with
N electrons in a configuration {n1 1 . . . nN N } (see Sect. 3.1.1), total orbital an-
gular momentum L, total spin S and overall total angular momentum J is written
as
{n1 1 . . . nN N } 2S+1 LJ . (6.55)
Often the electron configuration is dropped for brevity. As additional information
for multi-electron spectra one often finds a superscript o for odd parity, especially
if the latter cannot be identified uniquely from the angular momentum quantum
number (more about parity is discussed in Appendix D).
We give a few examples:
For the H atomas well as for alkali atoms s states with = 0 and j = J =
| 1/2| = 1/2 are denoted as ns 2 S1/2 ,
and p states with = 1 and j = J = | 1/2| = 1/2, 3/2 are correspondingly
np 2 Po1/2 and np 2 Po3/2 , respectively,
while for a d state with = 2, j = | 1/2| = 3/2, 5/2 and hence nd 2 D3/2 or
nd 2 D5/2 holds.
The ground state of the He atom (two 1s electrons, total spin = 0) is characterized
by 1s 2 1 S0 ,
while the first excited states are in the singlet system 1s2s 1 S0 and 1s2p 1 Po1 ,
and the corresponding triplet states are 1s2p 3 Po2 , 1s2p 3 Po1 , 1s2p 3 Po0 and
1s2s 3 S1 . They will be discussed in detail later.
6.3 Quantitative Determination of Fine Structure 303
We mention, finally, that in the spectroscopic literature and relevant data collec-
tions the spectra of neutral atoms are marked with the roman number I, those of
singly ionized atoms with II and generally the spectra of the N fold ionized species
with roman N + 1. Thus, one speaks of the C I, O II, Fe III spectra and refers in
this way to the spectra of the neutral C atom, and of the ions O+ and Fe++ , respec-
tively.
Section summary
Spin-orbit (or LS) interaction (6.35) arises as magnetic dipole energy of the
electron spin oriented in the magnetic field due to the electron orbital angular
momentum. It is proportional to L S and to the gradient of the potential in
which the electron moves.
Quantitatively, the resulting fine structure (FS) splitting is on the order of
2 Z 4 and decreases 1/n3 with the principal quantum number (its relative
magnitude being (Z)2 /n).
With spin-orbit interaction the orbital and spin angular momenta couple to a
total angular momentum J = L+ S. With | m
and |sms
describing the
orbital and spin angular momentum, respectively, the states of the coupled
angular momenta |(ls)j mj
are a sum of products of the former (involving so
called C LEBSCH -G ORDAN coefficients or 3j symbols).
good quantum numbers are n j mj with mj = m + ms , while m and ms are
not. The L
S interaction term is diagonal in this scheme, its eigenvalue being
( /2)[j (j + 1) ( + 1) s(s + 1)].
2
electron) correctly by so called spinors (with four complex values) rather than by
one complex wave function, and it provides automatically the correct electron spin
s = 1/2 with its magnetic moment and ge = 2, as well as the concept of antimatter.
Unfortunately, an exact solution and comparison with precision spectroscopy is
only possible for the H atom and does even in this case as we shall see in Sect. 6.5
not lead to a complete agreement with experiment. Interactions with the vacuum
field, to be treated with QED, enter into the term energies with small contributions
only about an order of magnitude smaller than those introduced by the (relativisti-
cally correct) D IRAC theory. Thus, we refrain here from a detailed treatment of the
D IRAC equation which is found in all major quantum mechanics textbooks.
Clearly, computations with the D IRAC equation (even approximate ones) are sub-
stantially more complex than those with the S CHRDINGER equation. Hence, one
tries to make the latter relativistic, i.e. to include electron spin and relativis-
tic effects by two component spin functions (as outlined in Sect. 6.2.5), originally
introduced by PAULI, and by adding appropriate interactions derived from D IRAC
theory.
As summarized below, there are three relativistic terms which have to be added
to the S CHRDINGER equation. All of them are on the order (Z)2 , i.e. of similar,
rather small magnitude. If they cannot be treated exactly, perturbation theory usu-
ally leads to very satisfactory results for most problems in AMO physics and quan-
tum chemistry. In combination with QED and higher order approximations, one can
even explain state-of-the-art high precision experiments extremely well for not too
large atomic nuclei, i.e. as long as (Z) 1 can be assumed in appropriate series
expansions.
p2 1 p4
Wkin = .
2me 8 m3e c2
Thus, the relativistic, quantum mechanical correction term is to lowest order
2
2
2
4
p 1 p 1 1 Eh p2
H1 = 3 2 = = Eh ,
8me c 2 2me me c2 2 me c2 2me Eh
1 obviously depends only on r. In 1st order perturbation theory the energy shift is
H
2
1 |n sj mj
=
2
0 Z + Z |n
02 + 2H
Vrel = n sj mj |H n |H 2
2 r r
2
1 1
= Wn2 + 2ZWn + Z2 2
2 r r
with Wn = Z 2 /(2n2 ). Inserting the matrix elements for the H atom according to
(2.130) one obtains in a.u.
(Z)2 3 1
Vrel = Wn . (6.58)
n 4n + 1/2
D ARWIN Term
The so called DARWIN term emerges from D IRAC theory as in a.u.
2
3 = Z (r).
H (6.59)
2
This perturbation is non-zero only at the origin r = 0 and we only need to evaluate
it for s states ( = 0) the only states which have a finite probability at the origin.
In 1st order perturbation theory we obtain
3 |n sj mj
= Z 2
VD = n sj mj |H ns|(r)|ns
2
Z 2 2 Z 2 2 (Z)2
VD = ns (r) (r)d3 r = ns (0) = Wn , (6.60)
2 2 n
where the last identity is derived for the H atom from inserting Rn0 (0) according to
(2.122), using the explicit expressions for the L AGUERRE polynomials (2.123).
We note that the relativistic correction as well as the DARWIN term represent
( independent) shifts of the energy levels and not splittings. Only if the term posi-
tions can be computed exactly (i.e. essentially only for the H atom and H like ions)
one may hope to compare experiment and theory in respect of these shifts. Already
for the alkali atoms, where degeneracy is significantly removed, such an absolute
comparison of these very small effects is usually beyond the limits of the combined
accuracy of theory and experiment.
306 6 Fine Structure and L AMB Shift
Spin-Orbit Term
In contrast to the two terms discussed above the spin-orbit interaction V LS (6.35)
leads to a splitting which can be calculated and measured with good accuracy and
L
is relevant for all atoms. In the S coupled scheme it is diagonal in respect of the
sj mj quantum numbers:
2
LS
= n |(r)|n
sj mj |
VLS = V L
S| sj mj
. (6.61)
2
With (6.52) and with the spin-orbit interaction parameter a defined in (6.36) one
may write the energy change due to spin-orbit interaction as
a
Wn j VLS = j (j + 1) ( + 1) s(s + 1) . (6.62)
2
Obviously, the fine structure splitting depends on j and . States with = 0 and
j = s do not split, all other states do. Systems with one active electron and s = 1/2
form doublet energy levels with j = 1/2. If the potential V (r) is known, the fine
structure splitting is readily derived from the matrix element n |(r)|n
.
For the H atom and H like ions and > 0 (6.62) becomes with (6.37):
"
Z4 2 1 for j = + 1/2
VLS = (6.63)
2n3 2 ( + 1/2)( + 1) ( + 1) for j = 1/2.
Interestingly, the total energy with all FS terms included depends only on j . The
exact solution of the D IRAC equation (e.g. B JORKEN and D RELL 1964, Eq. 4.14) is
usually expanded in powers of (Z)2 and is
(Z)2 (Z)2 1 3
Wnj = 1+ + O(Z) me c2 ,
6
(6.64)
2n2 n j + 1/2 4n
Z2 Z4 2 1 3
Wnj = + + Eh . (6.65)
2n2 2n3 j + 1/2 4n
In the following we stay with a.u. unless otherwise mentioned. The terms are split
into FS levels with j = 1/2, if and only if 1, by
Z4 2 1
Wn = Eh . (6.66)
2n3 ( + 1)
6.3 Quantitative Determination of Fine Structure 307
s1/2
Wn s1/2 WNJ
BOHR -
SCHRDINGER Vrel relativistic VD DARWIN VLS spin orbit DIRAC
Fig. 6.21 Fine structure of the H atom, n = 2 levels. All energy shifts in cm1
Figure 6.21 summarizes the situation for the n = 2 levels of atomic hydrogen:
The non-relativistic S CHRDINGER equation (as well as the B OHR model) pre-
dicts identical energies Wn for the 2s and 2p levels (left). The different contri-
butions to Wnj from D IRAC theory are indicated as dotted lines. The final results
Wnj = Wn + Vrel + VD + VLS , are shown on the right. Note that D IRAC theory
predicts Wnj for s1/2 and p1/2 to be identical.
The LS splitting decreases rapidly with n and as one would expect for an in-
teraction 1/r 3 . For not too large (typical in spectroscopic studies) we memo-
rize that fine structure splitting quite generally is Z 4 /n3 . Spin-orbit coupling is
particularly important for larger atoms. As we shall see in the following chapters,
RUSSEL -S AUNDERS (or LS) coupling works very well for light atoms where spin-
orbit coupling is small. It breaks down for large Z and other coupling schemes take
over, such as the so called jj coupling.
As already mentioned, only for H and hydrogen like ions He+ , Li++ , etc. (and
a few special exceptions, see Sect. 6.5.5) absolute calculations of term energies are
possible with a precision to verify or falsify the energetic level positions shown in
Fig. 6.21. However, the FS splittings due to L S interaction for individual n levels
may be computed with sufficient accuracy for comparison with experiment. Thus,
LS according to (6.35) to the
for practical applications it is usually sufficient to add V
standard S CHRDINGER equation, which for effective one electron systems simply
amounts to (6.62). For very heavy atoms it may become necessary to also account
for the relativistic kinetic energy shift.
For effective one electron systems such as alkali atoms and > 0 the energy
levels split in doublets according to (6.62). The spin-orbit coupling parameter a
(6.36) summarizes the specific FS properties of a given atom in a state characterized
by n . Fine structure splitting for alkali atoms is significantly larger than for the
308 6 Fine Structure and L AMB Shift
Table 6.2 Fine structure splitting of the first excited p states of H and the alkali atoms
Atom H Li Na K Rb Cs
n 2p 2p 3p 4p 5p 6p
Wn j / cm1 0.365 0.335 17.196 57.71 237.595 554.039
H atom. Table 6.2 gives a survey for the first resonance lines (ns 2 S1/2 np 2 P1/2
and np 2 P3/2 ) of different alkali atoms in comparison to the H atom. Note that for
Cesium (Cs) FS splitting is already 5 % of the total transition energy not really a
small effect any more. We summarize a few general rules valid for H and the alkali
atoms but also beyond:
The higher n and the smaller the FS splitting. For example, in Na already the
5p level is split by only 2.47 cm1 and for the 3d level, where the electron hardly
ever comes close to the nucleus it is only 0.05 cm1 .
For term energies usually Wnj = +1/2 > Wnj = 1/2 holds this is called normal
ordering. For higher terms and heavier elements there are cases where the en-
ergies are inverted: the one electron picture is just an approximation which may
break down under extreme conditions.
Expression (6.62) for the FS splitting is (approximately) valid in many cases
also for more complex atoms. Then, of course, the overall total angular momen-
tum J , the total orbital angular momentum L and the total spin S > 1/2 replace
j , and s. The formula has proved to be valuable in particular for larger Z and n,
at least as long as LS coupling is still valid.
Using (6.62) one derives the distance of two neighbouring fine structure levels in
a multiplet, the so called L AND interval rule:
a
WFS = WJ WJ 1 = J (J + 1) (J 1)J = aJ. (6.67)
2
The energy differences between any two neighbouring FS levels with J and J 1
in an FS multiplet are proportional to J . For doublets, (6.67) gives the splitting.
j = n 3 Wn j (6.68)
0.4
0.02223 0.3688
0.022 np 2P
np 2P
0.3
0.020 0.00444
0.04
Na (Z=11) Rb (Z=37)
nd 2D
0.00042 0.02
nd 2D
0.00023
nf 2F
0 0.00
-3.0 -2.0 -1.0 0.0 -3.0 -2.0 -1.0 0.0
/ eV
Fig. 6.22 Effective change
j = n 3
Wn j of the quantum defect due to FS for Na and Rb.
The experimental data points have been derived from the NIST data (K RAMIDA et al. 2013). Note
that
j follows very clearly the linear least squares fits (red lines) as a function of the binding
energy Wn typical for quantum defects (compare Fig. 3.7)
each n 2 L series. As shown in Fig. 6.22, the linear least squares fits represent the
experimentally observed data extremely well. We have to recall:
j is the change
of the quantum defect due to FS splitting. It has to be compared to (n, ) as shown
in Fig. 3.8 (for Wnl 0, n 1). Obviously, for Na the FS splitting is a minor effect,
in particular for the P1/2 2 P3/2 series where (, 1) 0.8551 and
j 0.022.
In contrast, for the same series in Rb FS splitting with
j 0.369 is no longer
negligible in comparison to (, 1) 2.6535.
Section summary
An exact treatment of fine structure requires the solution of the D IRAC equa-
tion. However, very good approximations can be achieved by incorporating
electron spin and the three essential interaction terms into the S CHRDINGER
equation:
Relativistic correction to the kinetic energy, according to (6.57),
the DARWIN term, non-zero only at the origin and thus relevant only for s
electrons, and
spin-orbit interaction which can be diagonalized for one electron systems
by for |( s)j mj
states, and is given by (6.61) and (6.62).
The first two terms represent shifts and can be compared with experiment
essentially only for H and H like ions. Spin-orbit interaction leads to splittings
which are readily observed and compared to theory for all atoms.
For atomic H (and H like ions) D IRAC theory predicts a splitting between
np3/2 and np1/2 states, but equal energies for the latter and ns1/2 .
310 6 Fine Structure and L AMB Shift
For larger alkali atoms FS splitting is no longer a small effect and amounts
e.g. for the first resonance line in Cs already to 5 % of the transition en-
ergy.
For fine structure transitions all selection rules hold as derived and discussed for the
general case in Sect. 4.4. In particular, for dipole allowed (E1) transitions the trian-
gular rule (Ja Jb 1) = 1 holds for the total initial Ja and final Jb angular momenta.
Their projection quantum numbers may change according to Ma Mb = 0, 1
depending on the polarization of the emitted or absorbed light. We use here capital
letters, indicating that all selection rules also hold for systems with several electrons.
In the following we shall apply the general expressions for transition probabilities
derived in Sect. 4.6 to systems which ar well described by LS coupling. Specifically
we are interested in the relative intensities of several lines within one multiplet. This
requires some extension of the angular momentum algebra used so far. These tools
albeit they may appear somewhat uninspiring in the beginning will later on
prove very useful for describing various other, more complex situations. Details are
presented in Appendixes C and B.3.
To illustrate the present subject, Fig. 6.23 gives a schematic overview of the
dipole allowed transitions between fine structure levels in the H atom up to n = 3
(term energies are not to scale). The thickness of the arrows indicates the strengths
of the individual transitions within the different multiplets (only to be compared
within one multiplet). All transitions between the doublet levels of the H atom (as
well as of the alkalis) lead to double spectral lines or to triple lines one of which is
typically very weak (see the 3d 2 D 3p 2 P transition). We point out here that the
relative intensities within a multiplet in emission differs from those in absorption.
For example, the two components of the first resonance line in emission, 2p 2 P1/2
1s 2 S1/2 and 2p 2 P3/2 1s 2 S1/2 occur with of probability, while the ratio of the
absorption cross sections for the inverse processes is 1 : 2.
We recall that LS coupling implies that orbital angular momentum L and spin an-
gular momentum S combine to a total angular momentum J of the system. The
following considerations are independent of whether L and S refer to a single ac-
tive electron or represent several electrons which, in turn, are composed of orbital
and spin angular momenta from several active electrons. Our first task is to calculate
the line strengths S(Jb Ja ) in the (LS)J coupled scheme for transitions between in-
dividual fine structure levels with quantum numbers Ja and Jb . According to (4.112)
6.4 Selection Rules and Intensities of Transitions 311
term energies
(emission) in the H atom 3p3/2
3s1/2
not to scale
3 3d3/2
3p1/2
2p3/2
2 2p1/2
2s1/2
1
1s1/2
it is proportional to the radial transition matrix element and the corresponding re-
duced matrix element of the renormalized spherical harmonics tensor C1 :
2
S(Jb Ja ) = (2Jb + 1)b |r|a
Lb SJb C1 La SJa
2 . (6.69)
The ratios of transition probabilities within individual multiplets (as indicated by the
red arrow in Fig. 6.24) may be derived by further reduction of the reduced matrix
element in (6.69). Since the dipole operator er responsible for E1 transitions does
not act on the electron spin, S remains constant during the transition. In (6.69) this
is formally born out by C1 , a spherical tensor of rank one which only acts onto the
angular momentum part of the coupled scheme |La SJa
. Hence, using (C.46) and
the 6j symbols described in detail in Appendix B.3 we may pull out the spin:
2
S(Jb Ja ) = b |r|a
(6.70)
" #2
La Lb 1
(2Jb + 1)(2Ja + 1)(2Lb + 1) Lb C1 La
2 .
Jb Ja S
From this we may derive the spontaneous lifetime(s) of the upper level(s). First,
with (4.114) we obtain the individual spontaneous transition probability A(Ja Jb )
from an upper multiplet substate |Jb Mb
into all substates |Ja Ma
of one specific
lower Ja level. This probability is independent of Mb and Ma but still depends on
Jb and Ja . We then sum over all final Ja levels to obtain the total decay probability
of the level Jb :
43 S(Jb Ja )
A(Ja Jb ) = ba
(6.71)
3c2 2Jb + 1
Ja Ja
3
4ba
b |r|a
2 Lb C1 La
2
3c2
" #2
La L b 1
(2Ja + 1)(2Lb + 1) .
Jb Ja S
Ja
312 6 Fine Structure and L AMB Shift
Jb b L b S
A(Lb La)
A(Jb Ja)
Ja a L a S
Fig. 6.24 Transitions between multiplet levels b Lb S and a La S, explaining the terminology for
transition probabilities between individual J sublevels and overall L levels
The sign in the second line accounts for the radial transition matrix elements not
being strictly independent of Ja and Jb and the angular frequencies ba differing
slightly for different multiplet components. Significant deviations occur only for
heavy atoms.4
We now exploit the 6j orthogonality relation (B.67) and obtain:
4ba3
A(Ja Jb ) = b |r|a
2 Lb C1 La
2 = A(La Lb ) = 1 (6.72)
3c 2
Ja
3
4ba S(Lb La )
A(La Lb ) = . (6.73)
3c 2Lb + 1
2
Note that (6.72) is not only independent of Mb but also of the initial Jb level: we
thus have found the overall decay probability A(La Lb ) of the upper b Lb S level
into all substates of the lower a La S level. It only depends on La and Lb and on the
radial matrix element b |r|a
. Thus A(La Lb ) is equal to the inverse spontaneous
lifetime of the upper level (more precisely: of each substate |Jb Mb
) as far as no
other, competing decay channels exist.
With (6.73) we have introduced also the overall line strength S(Lb La ) for the
multiplet transition, analogously to (6.69). Obviously this relation is completely
equivalent to (4.114) where an uncoupled scheme was assumed. Only the initial
and final orbital angular quantum numbers Lb and La and the radial matrix element
determine the excited state lifetime: La Lb = is the lifetime of the atomic tran-
sition. Take e.g. a sodium atom in the first excited 3p 2 P1/2 or 3p 2 P3/2 resonance
state: both states decay into the 3s 2 S1/2 ground state and each of the 6 different
excited substates has the same lifetime in respect of spontaneous decay.
All above equations still contain the reduced matrix element of C1 . It may be fur-
ther evaluated by angular momentum algebra. Specifically, for quasi-one-electron
systems such as alkalis with S = s = 1/2 we put La = a and Lb = b , respectively,
4 The lifetimes of the 3 2 P1/2 and 3 2 P3/2 states in Na are reported as 16.30 and 16.25 ns, respec-
tively, while for Cs 14 % difference is observed with 34.7 ns for the 6 2 P1/2 level and 30.4 ns for
the 6 2 P3/2 level (see e.g. S TECK 2010, and references there). This warrants an improved theoret-
ical treatment with more accurate wave functions for each individual j state. These could e.g. be
derived by QDT as outlined in Sect. 3.2.6 with the quantum defect differences among FS levels as
presented in the last paragraphs of Sect. 6.3.3.
6.4 Selection Rules and Intensities of Transitions 313
the angular momentum quantum numbers of the electron. The reduced matrix ele-
ment Lb C1 La
can be evaluated explicitly with (C.30):
" #2
2 l lb 1
S(nb b sjb na a sja ) = nb b |r|na a
(2jb + 1)(2ja + 1) a
jb ja 1/2
2
b 1 a
(2 b + 1)(2 a + 1) . (6.74)
0 0 0
The 6j symbol implies the triangular relation (ja jb 1) = 1, or equivalently the se-
lection rule
j = 0, 1 (however jb = 0 ja = 0). The 3j symbol demands in
addition that a + 1 + b is even, i.e. that b = a 1 ensuring parity conservation.
Finally, with (B.53) the last line in (6.74) is simply written as
( b + a + 1)/2. (6.75)
In the more general case, where L is already a coupled orbital angular momentum
as typical for more complex atoms, one has again to apply the reduction scheme
described above onto that coupling scheme, do this step by step and exploit (C.30)
where possible.
To compute the radial matrix element nb b |r|na a
one has to know the wave
functions in detail. Useful programmes are now readily available. One may then
obtain the spontaneous lifetime = Lb La from (6.72).
Once is known (either from theory or from experiment) it is often useful to
express all quantities in terms of it. Inserting (6.72) into (6.70) one obtains for the
line strength
" #2
3c2 2Lb + 1 La Lb 1
S(Jb Ja ) = 3
(2J b + 1)(2J a + 1) (6.76)
4ba Jb Ja S
by which in turn all other relations maybe expressed: e.g. the state to state sponta-
neous transition probabilities A(Ja Ma ; Jb Mb ) according to (4.113) as well as the
corresponding coefficients B(Ja Ma ; Jb Mb ) = B(Jb Mb ; Ja Ma ) for induced transi-
tion probabilities (4.124).
5 We recall here that the direction of a transition is always read from right to left, e.g. B(Jb Ja )
implies a transition to Jb Ja from Ja .
314 6 Fine Structure and L AMB Shift
2 c3
= A(Ja Jb ), (6.79)
ba
3
2 c3 2Lb + 1 1
B(Lb La ) = B(Jb Ja ) = 3 2L + 1
(6.82)
Jb
ba a
2Lb + 1
= B(La Lb ). (6.83)
2La + 1
Obviously, the familiar E INSTEIN relations (4.127) are valid for the individual
fine structure components (6.77)(6.80) with j := J , as well as for the multiplet
averaged quantities (6.81)(6.83) where j := L.
6.4 Selection Rules and Intensities of Transitions 315
Table 6.3 Branching ratios in 2 S 2 P and 2 P 2 D transitions calculated by (6.80) and (6.78).
The relative intensities can only be compared within one multiplet. The sum over all transitions
starting from one Ja or Jb are normalized to unity
to from to from
2S 2P 2P 2P 2P 2D 2D
1/2 1/2 3/2 1/2 3/2 3/2 5/2
2S
1/2 1 1 2P
1/2 5/6
2P
1/2 1/3 2P
3/2 1/6 1
2P
3/2 2/3 2D
3/2 1 1/10
1 1 1 2D
5/2 9/10
1 1 1 1
In the older literature, branching ratios within multiplets were often presented
in extended sets of tables or derived by equating different sums from (6.81) and
(6.83). Today these relative line strengths and transition probabilities are evaluated
readily using the 6j symbols (computed e.g. with S VEN G ATO R EDSUN 2004).
Hence, we only communicate by way of example in Table 6.3 the branching ratios
for 2 S 2 P and 2 P 2 D multiplets as relevant e.g. for H and the alkali atoms. It
is interesting to note that for a 2 S 2 P transition the absorption probability for the
1/2 P3/2 is twice that for the S1/2 2P1/2 transition. In contrast, the inverse
2S 2 2
processes P1/2 S1/2 and P3/2 2 S1/2 have identical transition probabilities,
2 2 2
Section summary
Selection rules for transitions between fine structure levels follow the general
concepts derived in Sect. 4.4 from angular momentum conservation now
applied to the total angular momentum quantum number J and its projection
M onto a given axis.
E1 transitions are allowed for
J = 0, 1 (but 0 0). For linearly and cir-
cularly polarized light
M = 0 and 1 must hold, with M referring to the
directions of polarization and light propagation, respectively. In addition, par-
ity of the whole system (photon plus atom) has to be conserved which for
single electron systems requires
= 1.
Transition probabilities between individual J and J levels can be derived
from line strengths in the usual manner, involving the radial matrix element
between initial and final n and n states and the reduced matrix element
Jb C1 Ja
in the coupled system. Using angular momentum algebra the lat-
ter may be reduced to expressions which refer to the initial and final orbital
angular momenta L and L .
Spontaneous emission probability from one specific initial substate |J M
of
level nL to all substates |J M
of the final level n L is independent of J
316 6 Fine Structure and L AMB Shift
(a) (b)
crossover
Fig. 6.25 Fine structure and L AMB shift for the H atom. (a) Term scheme and transitions of the
H BALMER line between the n = 3 and n = 2 levels (energies not to scale). (b) Measurement
by D OPPLER free saturation spectroscopy and original assignment according to H NSCH et al.
(1974); positions and heights of the stick spectrum (vertical full red lines) give the most recent
theoretical data according to K RAMIDA (2010)
In Sect. 6.1 we have already familiarized ourselves with optical methods for observ-
ing E1 transitions between the n = 2 and n = 3 levels in the H atom with high pre-
cision. Figure 6.25 shows (a) the relevant term scheme and reproduces in (b) again
the well resolved spectrum from Fig. 6.13. The assignment of the spectral lines to
transitions is indicated by lower case letters. For a quantitative comparison of line
intensities one would use (6.74) with the radial dipole matrix elements from Ap-
pendix C.5. However, one important feature shown in Fig. 6.25 cannot be explained
by the FS theory discussed so far: We remember that according to (6.65), the en-
ergy terms derived from D IRAC theory depend only on j but not on . This rule
is obviously broken for the n = 2 levels states in H as revealed by the experiment,
the 2p 2 S1/2 being slightly higher in energy than the 2p 2 P1/2 level: This important
observation is called L AMB shift.
6.5 L AMB Shift 317
The main enemy for optical precision spectroscopy of the H atom, with its small
mass M and high average thermal velocity v
, is the D OPPLER effect which
according to (5.22) also depends on the transition frequency ba :
D ba v /c = ba v .
If one tries to determine the FS splitting or other small effects from the difference
of two optical transitions D OPPLER broadening enters proportionally to ba with
dramatic influence on the result. However, in some cases it is possible to access the
energetic difference between different |j mj
states within one multiplet directly,
typically by M1 transitions as described in Sect. 5.4. The necessary alternating mag-
netic fields can be generated without problems in microwave resonators.
While the corresponding optical E1 transitions are observed in the VIS or UV re-
gion (at wavenumbers between (10 000 to 100 000) cm1 ) the FS splittings amount
to some cm1 or even below. D OPPLER broadening can be neglected completely
in this microwave regime (1 cm1 = 30 GHz) and one may determine transition fre-
quencies with orders of magnitude higher precision than by typical optical spec-
troscopy and even finer effects may be measured with radio frequency transitions
(RF) at frequencies from 1 MHz to 1 GHz, depending on the external magnetic field
which one usually applies in addition.
The most prominent, and in its consequences far reaching experiment was the first
determination of the L AMB shift for which L AMB and K USCH (1955) obtained the
N OBEL prize.
In 1947, Willis E. L AMB and R. C. R ETHERFORD were looking for a transition
between the 2s 2 S1/2 and 2p 2 P1/2 level which according to D IRAC should be de-
generate. Thereby they exploited the fact that the 2p 2 P1/2 state decays directly into
the ground state ( = 1.6 ns) while the 2s 2 S1/2 state is metastable. It may only de-
cay by two-photon emission with a lifetime of 120 ms as we have learned in
Sect. 5.3.2.
The experimental scheme, sketched in Fig. 6.26, is interesting. A atomic beam
(see footnote 24, Chap. 1) of hydrogen emerges from a hot oven where H2 disso-
ciates. By electron impact a certain fraction of these H atoms is excited into the
2s 2 S1/2 and 2p 2 P levels. While the 2p 2 P1/2,3/2 atoms rapidly decay by sponta-
neous emission into the ground state, the metastable 2s 2 S1/2 atoms in the atomic
beam reach a microwave resonator. There, the microwaves if on resonance in-
duce the transition 2s 2 S1/2 2p 2 P1/2,3/2 . If one of these transitions occurs, the
2p 2 P1/2,3/2 states decay rapidly into the ground state. One detects the transition
by the loss of metastable atoms reaching the detector. The L ANGMUIR -TAYLOR
detector (see Sect. 1.9.3) is sensitive only to excited (2s 2 S1/2 ) atoms.
318 6 Fine Structure and L AMB Shift
Fig. 6.26 Scheme of the experimental setup of L AMB and R ETHERFORD for measuring mi-
crowave induced transitions from the 2s 2 S1/2 state of the H atom
transition frequency / GHz
16
QED
DIRAC theory 2s 2 S1/2 (m j = 1/2) 2 p 2 P3/2 (m j = 3/2)
12
Fig. 6.27 Results of the original L AMB shift experiment for atomic hydrogen. Transition fre-
quencies between the 2s 2 S1/2 and the 2 p 2 P1/2 and 2p 2 P3/2 levels as a function of an external
magnetic field B. Dashed lines: D IRAC theory; full lines: fits to the experiment and QED. The
remaining energy difference between the 2s 2 S1/2 and 2p 2 P1/2 levels at B = 0 (ca. 1060 MHz) is
called L AMB shift
The experiment was carried out at several fixed microwave frequencies and the
splitting of the two states was registered by tuning this magnetic field so that res-
onant transitions occurred. In Fig. 6.27 the results are compared with D IRAC the-
ory and quantum electrodynamics (QED). In the limit of vanishing magnetic field
a finite value of 1060 MHz ( = 0.0353 cm1 ) for the resonance frequency of the
2 S1/2 2 P1/2 transition remains the L AMB shift, by which the 2 2 S1/2 level
2 2
lies higher than the 2 2 P1/2 level. In contrast, D IRAC theory predicts exact agree-
ment of both energies.
6.5 L AMB Shift 319
hydrogen
cryostat
vacuum chamber cooling Cesium
atoms
FARADAY cage
laser clock
chopper
2S
243 nm time resolved
detector
2F photon counting
microwave
F
dye laser 486 nm resonator
probe laser fluorescence
detector
4/7 F 1/2 F
cooling lasers
frequency comb
9.2 GHz
70 fs Ti:Sapphire
laser from Cs clock
cooling lasers
Fig. 6.28 Schematic of an advanced experiment for determining the 1S2S transition frequency
in the H atom by D OPPLER free two-photon spectroscopy according to N IERING et al. (2000). The
setup in the vacuum chamber corresponds essentially to Fig. 6.16. The frequency comb is the heart
of the experiment, providing a high precision frequency standard which is directly synchronized
with an atomic Cesium fountain clock, sketched in the spirit of J EFFERTS and M EEKHOF (2011)
(for recent improvements see e.g. J ENTSCHURA et al. 2011; PARTHEY et al. 2011;
M ATVEEV et al. 2013, the latter being a simultaneous effort of two institutes via a
920 km fiber link). Part of it, the two-photon excitation of an atomic H beam in a
confocal resonator, has already been described in Sect. 6.1.8. A so called Cesium
fountain atomic clock is used, which is currently employed world wide as primary
time and frequency standard. It is sketched in Fig. 6.28 on the right. A compact
description is found e.g. on the NIST Web-pages (J EFFERTS and M EEKHOF 2011).
The special trick with this setup is the long effective interaction time T 1 s of
the atoms with the microwave field: one exploits the method of R AMSEY fringes as
described in Sect. 6.1.7.
To optimize this concept the Cs atoms are first cooled with 6 diode lasers. By a
brief detuning pulse applied to the bottom laser one kicks the atoms upwards,
pushes them so to say through the microwave resonator. The grey trajectory in
Fig. 6.28 shows schematically the path which the atoms take. Under earth gravity
they move freely up to a classical turning point and then fall back into the resonator
indeed some kind of an atom fountain. One induces the ground state hyperfine
transition in atomic Cs which is detected from a change of the fluorescence signal
induced by the probe laser.6 As explained in Sect. 6.1.7 the total interaction time
T in the frame work of the R AMSEY fringe method is given by the time of flight
of the atoms from the resonator to the turning point and back to the resonator. The
precision of the reference line is then 1/T , typically on the order of 1 Hz.
Based on this type of optical measurements the spectroscopy of the H and D
atom is today known with extreme precision. Specifically, the n = 2 L AMB shift
W (2s 2 S1/2 ) W (2p 2 P1/2 ) corresponds to
6 Werecall that 1 s is defined as the duration of 9 192 631 770 periods of the transition frequency
between the two hyperfine levels of the ground state in 133 Cs.
6.5 L AMB Shift 321
wavenumber / cm-1
term positions of the n = 2 BOHR DIRAC QED
states for the H atom in 0
S CHRDINGER, D IRAC and =1 =1 2P
j = 3/2 3/2
QED approximation; term
distances are shown to scale 0.365 cm-1
and may be compared directly
=0 2S
1/2
j = 1/2 2P
0.5 = 0, 1 =1 1/2
0.035 cm-1
transitions thus deriving simultaneously an ultra precise value for the RYDBERG
constant: 7
Using this value of R , numerical values for the hydrogen 1S 2S transition fre-
quencies from B OHR /S CHRDINGER (1.149) or D IRAC theory (6.65) are obtained
with appropriate accuracy:
3R c
S CHRDINGER : (1S 2S) = = 2.467381470272 1015 Hz,
4
D IRAC : (1S1/2 2S1/2 ) = 2.467411580803 1015 Hz . (6.87)
measured with a relative accuracy of 4.5 1015 ! To compare this value measured
for atomic H with the prediction from D IRAC theory (6.87) one has to correct it by
the kinematic correction factor (1.146). With the present CODATA value for me /mp
one obtains
A distinct difference remains on the one hand due to hyperfine interaction (in
(6.88) indicated by the subscripts F = 1), and to the 1S1/2 L AMB shift on the other.
A quantitative evaluation of HFS is relatively straight forward and experimentally
well under control by measurements of the corresponding level splittings (we shall
treat HFS in Chap. 9). In fact, the HFS splitting of the H 1sS1/2 ground state was for
many years one of the best determined transition frequencies until the (1S1/2
2S1/2 ) transition acquired this merit. After all this is done one obtains the ground
state L AMB shift W (1s 2 S1/2 ) W (1s 2 S1/2 D IRAC)
the presently agreed global average from several independent measurements. Note
that this is again an upward shift of the D IRAC term level towards a slightly smaller
binding energy just as for the 2s 2 S1/2 level.
We point out that the whole field of ultra high precision spectroscopy and the
interpretation of its results by QED is still rapidly evolving. Several not fully re-
solved issues and discrepancies remain. The interested reader is referred to the liter-
ature for details. Unless otherwise specified, numerical values given here reflect the
state-of-the-art according to some recent, extensive reviews by E IDES et al. (2001),
K ARSHENBOIM (2005), K RAMIDA (2010).
The critical reader may perhaps ask now, what this extreme precision might be
good for. Let us quote from the master of precision spectroscopy, Ted H NSCH et
al. (2005): Even beyond more precise measurements of fundamental constants and
searches for their possible time variations, there are many good practical reasons
for pushing the art of measuring time and frequency to the feasible limits. Advances
in time and frequency metrology will enable the synchronization of clocks over large
distances. Such synchronization is, e.g. needed in astronomy for very long baseline
interferometry. Better clocks will improve the performance of satellite navigation
systems. They are also crucial for the precise tracking of remote space probes. In
telecommunications, better atomic clocks will be needed for network synchroniza-
tion, as the need for bandwidth increases. Geologists can apply better clocks to
study the variability of earths rotation or to follow the drifts of continents with
millimeter precision. Astronomers can study irregularities in the periods of pulsars.
For physicists interested in fundamental science, better clocks will permit new strin-
gent tests of special and general relativity. Very likely, there will be new unexpected
discoveries, as the art of precision spectroscopy and optical frequency metrology
continues to advance.
And one may add in our present context, that progress in basic research has
always been nourished by quantum jumps in precision. The L AMB shift itself is an
excellent example: its discovery has paved the way for quantum electrodynamics
(QED), today one of the cornerstones of modern physics.
Atomic physics with highly charged ions (HCI) is a very active and productive
branch of modern physics. Advanced high energy ion storage rings and state-of-the-
art electron beam ion traps (EBIT), accessible as user facilities, offer challenging
perspectives for a broad range of interesting fundamental science (see also Sect. 7.5,
6.5 L AMB Shift 323
Vol. 2) as well as for many technological, biological and medical applications. Criti-
cal tests of QED and of atomic-nuclear interactions with high Z ions are one exciting
aspect of the spectroscopy of such systems (a compact review is given by B EIERS -
DORFER 2010). As we shall discuss in Sect. 6.5.6, the relevant coupling constant for
QED is also Z, and so called higher order loops (nth order perturbation theory)
enter into the term energies with (Z)n Thus QED may have a problem with highly
charged ions where Z is no longer really small, the fine structure constant being
1/137.
In the past, theory and experiment have mostly focussed on the 1S1/2 L AMB
shift of hydrogen like ions. In singly charged He+ the experimental value for the
2 2 S1/2 2 2 P1/2 transition is 14 041.13(17) MHz in excellent agreement with theory,
giving 14 041.18(13) MHz. The main contribution from QED are the one-loop ef-
fects, self-energy and vacuum polarization. Effects from the finite size of the nuclei
are included in comparisons of theory and experiment, and become equally impor-
tant for larger nuclei. We shall discuss volume effects of this kind in Sect. 9.4 in the
context of HFS.
In total, one expects for hydrogen like, (Z 1)-fold charged ions and ns levels a
L AMB shift (see e.g. J OHNSON and S OFF 1985)
(Z)4 Z42
WLamb = F (Z)me c 2
= F (Z)Eh , (6.91)
n3 n3
given in terms of the electron rest mass energy and a.u., respectively. F (Z) is a
slowly varying function of Z, but also depends slightly on n. For the neutral H
atom F (1) 10 as reflected in Fig. 6.29 for = 2. Inverting the experimental
results, (6.84) and (6.90), we obtain F (1) = 10.042037(27) and 10.39828(9) for
the 1S1/2 and 2S1/2 L AMB shift, respectively.
Note the proportionality
WLamb Z 4 2 /n3 (in a.u.) which we have found to
be characteristic also for fine structure (6.65). The additional factors / F (Z)
make the L AMB shift about an order of magnitude smaller than FS.
The overall trend of the L AMB shift is illustrated over the whole nuclear charge
range from Z = 1 . . . 91 in Fig. 6.30(a). The dominating one-loop contributions and
finite nuclear size effects are plotted together with the total theoretical value (red
line) in units of (/)[(Z)4 /n3 ]me c2 on a logarithmic scale. The vacuum polar-
ization gives a negative contribution.
Experimentally, spectroscopy in the hard X-ray regime in storage rings has pro-
vided impressive results for multiply charged, hydrogen like ions up to U91+ . Ele-
gant methods have been employed in these studies, as sketched in Sect. 6.1.4. The
1S1/2 L AMB shift is typically determined by comparing the 1S1/2 2P1/2 and
1S1/2 2P3/2 transitions with D IRAC theory (making appropriate corrections for
HFS and kinematics). Figure 6.30(b) according to S THLKER (see G UMBERIDZE
et al. 2005) compares the theory presented in (a) with these results from several
sources on a linear scale. One may describe the agreement as quite satisfactory.
However, so far truly high precision spectroscopy for HCI is not yet in sight, at
least not one competing with the extreme accuracy of the 1s 2s transition in the
neutral H atom. Thus, the best test for two-loops QED contributions would presently
324 6 Fine Structure and L AMB Shift
10
LAMB shift 10
self-energy
1 8
- vacuum
polarization
6
0.1
(a) 4
(b)
0.01 finite size
of the nucleus 2
0.001
0 20 40 60 80 100 0 20 40 60 80 100
nuclear charge Z nuclear charge Z
Fig. 6.30 1S1/2 L AMB shift for H like ions as a function of nuclear charge Z. (a) Main contri-
butions to the L AMB shift according to J OHNSON and S OFF (1985) from so called one-loop QED
effects (to be discussed in Sect. 6.5.6) and from the finite nuclear size. (b) Comparison of this
theory with experiments according to G UMBERIDZE et al. (2005)
still be the L AMB shift of the H 1S1/2 ground state (known to be ca. 4 parts in 105 )
were it is not hampered by the experimental uncertainties of the proton radius!
On the other hand, for 91 fold ionized Uranium, U91+ , the 1S1/2 L AMB shift
amounts to (460 4.5) eV by no means a tiny effect as in H atoms. And consid-
ering the difficulties of this experiment, the accuracy is impressive but with about
1 % by far not yet good enough for a test of two-loop contributions.
Recent theoretical and experimental advances appear to offer promising perspec-
tives for a direct comparison of the 2s 2 S1/2 2s 2 P1/2 L AMB shift in lithium like
highly charged ions (B EIERSDORFER 2010). A number of interesting questions also
remain to be clarified in respect of HFS in highly charged ions. Other fascinating
possibilities and experimentally demanding challenges arise in this context from
the advent of X-ray lasers, e.g. at XFEL P ROJECT (2011), for which theoretical
predictions are made e.g. for laser induced fluorescence in the strong field regime
(P OSTAVARU et al. 2011). In summary, the field remains open, highly active and
exciting.
What is the physical origin of the L AMB shift? Very generally speaking: the inter-
action of the electron with the vacuum field. The fundamental theory for describing
such processes, Quantum Electrodynamics QED and Quantum Field Theory, was
developed by T OMONAGA, S CHWINGER and F EYNMAN. In 1965 they received the
N OBEL prize for their ground breaking work, which is one of the cornerstones of
modern theoretical physics including the standard model i.e. far beyond its origin
and the applications in atomic physics which we discuss here.
However, even a brief introduction into QED would lead us far beyond the scope
of the present textbook. Thus we confine ourselves here to a heuristic presenta-
6.5 L AMB Shift 325
tion of F EYNMAN diagrams (or F EYNMAN graphs). They first appeared in Richard
F EYNMANs famous 1949 paper on Space-Time Approach to Quantum Electrody-
namics as a picturesque visualization representing the mathematical framework for
describing the interactions of electrons, photons and other subatomic particles. They
are a compact form of symbolizing perturbation integrals and will that is our hope
here help us to understand what causes the L AMB shift and other related phenom-
ena. QED expands interactions into a series of perturbation integrals accounting for
the influence of the quantized electromagnetic field. These perturbation series are
constructed for sequences of virtual or real events and warrant good book keeping
to account for all possible processes which might occur and this is done with the
help of the F EYNMAN graphs. Each node (or vertex) in such a graph characterizes
one individual interaction process, several subsequent nodes imply a higher order of
perturbation theory. One speaks about one-loop, two-loop processes and so on.
The movements of individual particles and their elementary interaction in space
and time are represented by graphical symbols which are summarized in Table 6.4.
The free electron (1) moves here from the left to the right, its antiparticle, the
positron, (2) moves from right to left. The exchange particle, for electromagnetic
interaction, the photon, (3) is characterized by a wiggly line. Graphs (49) are self
explaining, graph (10) represents the typical interaction of an electron with the elec-
tromagnetic magnetic radiation in vacuum: a free electron moving in space con-
stantly emits (virtual) photons which it quickly reabsorbs again. This is one of many
326 6 Fine Structure and L AMB Shift
processes which contribute to the self energy of the electron. It plays a key role for
the so called mass re-normalization by which QED overcomes problematic diver-
gences. Another important process is vacuum polarization (11). In 1st order (one-
loop) it generates one (virtual) e e+ pair, which is then immediately annihilated
again. Graph (11) may obviously be constructed from (8) and (9).
We have somewhat loosely used the terms quickly and immediately which
may be quantified by the energy-time uncertainty relation (1.125): the larger the
energy of the emitted virtual photon, the shorter the time between emission and
reabsorption, and similarly for pair formation and annihilation.
From these basic diagrams one may construct more complex ones to describe
various interaction processes between electrons and photons. The following rules
have to be observed:
In Table 6.5 F EYNMAN graphs are collected for several important and instructive
examples of interactions. The graphs are more or less self explaining. We refrain
here from a detailed description of the mathematical expressions represented by
these graphs. Only for C OULOMB scattering we have indicated, how one may glean
the underlying concepts: the well known RUTHERFORD scattering formula emerges
in this case without any effort.
According to QED the electromagnetic field has a finite amplitude, even in its vac-
uum ground state, determined by the zero point energy of the corresponding quan-
tized harmonic oscillator. And the electron constantly emits and reabsorbs photons
as indicated by the self energy graph in Fig. 6.31(a). Consequently, a force pulls
6.5 L AMB Shift 327
Table 6.5 Lowest order F EYNMAN graphs for several important interaction processes in atomic
physics
Example Process Graph
1 C OULOMB scatter- e- e-
ing e + Ze Z
prop 1/q 2
aga
tor
+Ze
+Ze
The scattering amplitude at the two vertices is and Z , respectively.
The
virtual photon receives from a 2recoil momentum q = 2me W sin /2.
the electron
The total amplitude is thus Z /q . Hence, the scattering cross section is:
d Z2 2 Z 2 e4
4
. (6.92)
d" q W sin4 /2
2
In addition to the C OULOMB term (as above), only two one-loop graphs are relevant:
the annihilation and re-generation of an e e+ pair by emission of a virtual photon.
3 C OMPTON scattering
+
e- e- e- e-
4 Pair production e+ e e-
e+
+Ze
The black cross symbolizes the atomic nucleus with which the electron interacts. The
following diagrams also need the interaction with an atomic nucleus to allow for en-
ergy and momentum conservation.
5 Bremsstrahlung:
e + Ze e + Ze + e-
e- + e-
+Ze e- +Ze
6 Electron bound in a C OULOMB potential. In the current literature the so called
D IRAC -C OULOMB propagator is summarized by just two parallel lines, compris-
ing several contributions
e + Ze = + + + ...
the electron back and forth into a Zitterbewegung (trembling motion), so that ef-
fectively the electron is smeared out over a (still very small) finite volume instead
of being point like. Due to this apparent charge distribution the electron sees
328 6 Fine Structure and L AMB Shift
(a) (b)
self-energy
vacuum
polarization
Fig. 6.31 F EYNMAN graphs of (a) one-loop and (b) two-loop radiative corrections for the L AMB
shift. Graphs adapted from B EIERSDORFER (2010)
with the C OMPTON wavelength C of the electron (1.76). QED provides similar es-
timates (see e.g. E IDES et al. 2001, Eq. (28))
2 2 2 3
r = ln (Z)2 2C = ln (Z)2 a02 . (6.95)
This corresponds to r 2
3.4 102 C for the characteristic effective spread
of the electron which obviously is below the C OMPTON wavelength but signifi-
cantly above the proton radius (1015 m). With (6.95) one would obtain a L AMB
shift (in a.u.)
Z4 2 4
WLamb = 3
ln (Z)2 .
n 3
If we interpret 4/3 ln[(Z)2 ] as F (Z), one may compare this directly to the gen-
eral formula (6.91) with numerical values 13 and 10, respectively which is not too
bad for such a rough estimate.
As a historical remark we point out that Hans B ETHE (1947), in his first calcula-
tion of the H 2S1/2 L AMB shift, obtained a very similar expression when calculating
the self energy of the bound electron. Instead of 1/ 2 = me c2 /Eh as argument in the
logarithm he obtained me c2 /(8.9Eh ) which led him to 1040 MHz already pretty
close to the experimental value (6.84). Today, of course, with several orders of mag-
nitude better experimental accuracy and highly developed computational methods
in QED it is obvious that in addition to the self-energy also the vacuum polarization
has to be taken into account (see Fig. 6.31(a)) as well as higher order corrections.
The curious reader may admire the diagrams for the relevant two-loop corrections
in Fig. 6.31(b). Also a range of additional effects have to be considered.
For the hydrogen atom in the n = 1 and 2 levels the present state-of-the-art is
summarized in Table 6.6. We cannot enter into a deeper discussion which is ongo-
ing among the experts. However, the agreement between theory and experiment is
already now truly impressive.
At the moment, the largest uncertainty which limits stringent tests of bound state
QED still appears to be the proton radius. Its value rp = 0.8775(51) fm in 2010 CO-
DATA (NIST 2010) already strongly depends on optical precision spectroscopy. In-
dependent measurements from nuclear physics give different values and have much
larger error bars. This is a startling situation, the proton being one of the most im-
portant building blocks of matter and its size so poorly known!
The most recent and fascinating development in this context is a measurement
of the L AMB shift in muonic hydrogen by P OHL et al. (2010). The muon (see
Sect. 1.5.2) is the bigger brother of the electron, its mass being more than 200
times larger, with otherwise more or less identical properties. Hence, in muonic
hydrogen, p, all the physics of the H atom remains in principle unchanged ex-
cept for the magnitude of the measurable quantities. In particular, the B OHR radius
is 200 times smaller in p and thus the muons probability to be close to the pro-
ton is correspondingly higher. Consequently, the L AMB shift is influenced much
more dramatically by the protons structure as in the case of atomic hydrogen. The
330 6 Fine Structure and L AMB Shift
Table 6.6 Main QED contributions to the L AMB for the 1S, 2S and 2P states of atomic hydrogen,
and comparison with the presently best experimental values
Contribution 1s 1/2 2s 1/2 2p 1/2 2p 1/2 2p 3/2
MHz MHz MHz MHz
Self energy (one-loop) 8383.339466(83)a 1072.958444f 12.84692(2)e 12.54795(2)e
Vacuum pol. (one-loop) 214.816607(15)a 26.852075(2)g 0.00035e 0.00008e
Two-loop corrections 0.7310(33)a 0.02598(7)e 0.01279(7)e
Proton sizea 1.253(50)a 0.1566(62)g 0 0
Recoil correctiona 2.401782(10)a 0.0145e 0.0177e
Radiative recoil corr.a 0.0123(7)a
Other corrections 0.002(1)a 0.0002e 0.0001e
Recent theory (total) 8172.894(51)a 1057.844(2.5)d 12.83599(8)e 12.51746(8)e
Experiment 8172.840(22)b 1057.847(9)c
a B IRABEN (2009)
b E IDES et al. (2001)
c K RAMIDA (2010)
d J ENTSCHURA et al. (2005)
e J ENTSCHURA and PACHUCKI (1996)
f J ENTSCHURA et al. (2001) (non-perturbative)
g order of magnitude: 1s1/2 scaled by 1/n3
p 2S1/2 2P1/2 L AMB shift which the P OHL et al. collaboration succeeded to
measure with high precision is 49881.88(76) GHz. On the basis of FS and HFS
calculation and QED they deduced from their measurement a value for the proton
radius rp = 0.84184(67) fm which claims an order of magnitude higher accuracy
and differs by 5 standard deviations from the above CODATA value!
Section summary
The L AMB shift manifests itself most clearly by a splitting of the 2s1/2 and
2p1/2 terms in atomic H for which D IRAC theory predicts identical energy.
We memorize (2s1/2 ) (2p1/2 ) 1060 MHz.
It was first discovered in 1947 by L AMB and R ETHERFORD using microwave
spectroscopy and triggered the development of QED. Today two-photon spec-
troscopy of the H and D atom is the method of choice. It has achieved an
amazing precision, not least due to the work of H NSCH and collaborators,
using frequency comb techniques and state-of-the-art atomic clocks.
Agreement with equally accurate QED computations is convincing. We have
encountered a couple of F EYNMAN graphs which are useful to understand the
origin of such radiative corrections for high precision atomic and molecular
spectroscopy. In these terms, the L AMB shift arises dominantly from self en-
ergy, with some correction from vacuum polarization both are single loop
terms.
6.6 Electron Magnetic Moment Anomaly 331
S = ge B s e
M with s = 1/2, ge = 2 and B =
2me
as introduced in (6.28). We have already mentioned on several occasions that the
value ge = 2 differs slightly from reality.
One defines the electron magnetic moment anomaly as
After the preceding discussion we rightly suspect the origin of this deviation to be
again the interaction of the electron with the vacuum field. But let us first have a
look at the experimental determination of ae which in itself is fascinating. Again
we encounter an experiment of unbelievable accuracy. All experiments to determine
ge 2 are essentially based on a very precise comparison of the (non-relativistic)
cyclotron frequency of the electron in an external magnetic field B according to
(1.111)
e
c = B (6.97)
me
with the L ARMOR precession frequency of its magnetic moment (1.162):
e
s = ge L = ge B. (6.98)
2me
Knowing both frequencies allows one to derive the anomaly by (s c )/c =
(ge 2)/2. An older, but still instructive experiment is shown schematically in
Fig. 6.32. A 100 ns, 100 keV electron beam is partially polarized by scattering from
a gold foil. Electrons drift on helical trajectories into a magnetic field trap, their
polarization being perpendicular to both, the field and their velocity. After a well
defined time they are pulsed out of the trap and their polarization is measured by
scattering from a second gold foil. For a hypothetical value ge = 2 the spin orienta-
tion of the electrons after each full circle would be the same as at the entrance. Due
332 6 Fine Structure and L AMB Shift
to the anomaly, spin precession time and circling time differ, thus allowing a direct
determination of ae .
Todays most precise method goes back to D EHMELT and his collaborators (VAN
DYCK et al. 1986). It too uses the difference between s and c , however, in a highly
refined and compact manner. Thereby VAN DYCK et al. (1987) succeeded to store
a single electron in a so called P ENNING trap. With this method they were, for the
first time, able to measure ge 2 with high precision. D EHMELT and PAUL (1989)
received the N OBEL prize for the development and application of such ion traps
together with Norman R AMSEY, who was honoured for his method of separated
field oscillators (which we have discussed in Sect. 6.1.7). Both inventions were in-
strumental for the development of precision spectroscopy and time measurement,
today a standard in atomic clocks as well as for high resolution quadrupole mass
spectrometers and many other applications. As indicated in Fig. 6.33(a), a magnetic
field B of several Tesla (T), parallel to the axis of the trap, leads to cyclotron mo-
tion of the electron. In addition, an electric quadrupole field (equipotential surfaces
x 2 + y 2 2z2 = const) keeps the electrons in the centre of the trap.8
The classical motion of an electron in such a trap is characterized by three eigen-
frequencies: One axial frequency,
2eU0
z = with d02 = 02 + 2z02 ,
me d02
8 The so called PAUL trap uses a similar design, however, instead of using a magnetic field the
electrons are stabilized by superposing an oscillatory field onto the static electric quadrupole field.
6.6 Electron Magnetic Moment Anomaly 333
negative potential
on the upper cap
energy
electron
positive potential trajectories
on the ring
negative potential
on the lower cap (a) (b)
Fig. 6.33 (a) P ENNING trap for the determination of g 2 according to VAN DYCK et al. (1986).
(b) Simplified scheme for the eigenfrequencies of Geonium
Quantum mechanically, the L ARMOR frequency (6.98) of the electron spin has to
be added. For sufficiently low temperatures and high magnetic field it makes sense
to look for quantum mechanical solutions to this system (using S CHRDINGER or
D IRAC equation). The electron then assumes discrete energy levels Wnm which are
a combination of electron spin states and harmonic oscillator, the so called L ANDAU
levels:
2
ge 1 1 1
Wnm = c ms + n + c n + + ms . (6.101)
2 2 2 2
The first term describes the spin states with the projection quantum number ms ,
the second term contains the modified cyclotron resonances and the third term is
the leading relativistic correction term with /c = c /me c2 . The name Geonium
atom was coined for this system. Its lowest terms are shown in Fig. 6.33(b). By
cooling the system to very low temperatures one may indeed trap one single elec-
tron and bring it into its energetically lowest states. Transitions are detected by
induced currents. The experimental challenge is a most accurate determination of
the anomaly frequency a = gc /2 c between the states n = 0, ms = 1/2 and
n = 1, ms = 1/2 as indicated in Fig. 6.33(b). From this the electron anomaly ae is
obtained, if one knows c and the axial frequency z with sufficient accuracy.
Such a high precision warrants a fundamental understanding of the physics in
these P ENNING traps which has improved during the past two decades substantially.
Many methodological improvements have been introduced. We cannot go into the
details and have to refer the interested reader to the original literature, in particular to
the work of G ABRIELSE and collaborators (P EIL and G ABRIELSE 1999; H ANNEKE
et al. 2008) who were able to overcome several imperfections of the trap by a variety
of clever experimental tricks. They improved the accuracy of the measurement by a
factor of 10.
In the course of these improvements they also succeeded to really document the
population of the quantum states shown in Fig. 6.33(b) in the P ENNING trap. This
is illustrated in Fig. 6.34 which shows the measured probability of quantum jumps
between the individual n states. The number of quantum jumps observed depends
directly on the thermal population of the excited states. It reduces with temperature
334 6 Fine Structure and L AMB Shift
4.2 K 1.0
0.0
cycltoron energy / c
3.2 K 1.0
probability, Pn
0.0
2.0 K 1.0
0.0
1.6 K 1.0
0.0
0.08 K 1.0
0.0
0 10 20 30 40 50 012
time / min n
Fig. 6.34 Quantum jumps between lowest states of the one electron cyclotron oscillator. A dra-
matic reduction with cavity temperature is seen. Based on this, on the right a population analysis
of the n states in the trap is shown. Adapted from P EIL and G ABRIELSE (1999)
as quantified by the population analysis on the right of Fig. 6.34. At the lowest
temperatures, indeed, only the ground state is still populated and no quantum jumps
are detected any more.
We recall that we have discussed quantum jumps in atoms already in Sect. 4.7.3.
However, in the present case we are discussing quantum jumps in a truly macro-
scopic quantum system. Thus, this experiment should be kept in mind in the context
of a presently very fashionable discussion about the limits (or extensions) of quan-
tum mechanics into the macroscopic world!
In the most recent high precision experiment (O DOM et al. 2006; H ANNEKE et
al. 2008) a one electron cyclotron oscillator (Geonium) was cooled down to below
100 mK and the magnetic field was 10.6 T. In this case s c 149 GHz while the
axial frequency was tuned to z 200 MHz. The magnetron frequency was m
134 kHz. The resulting presently most accurate value is given by 2010 CODATA
(NIST 2010) as9
In this context, the reader should bear in mind that the determination of fundamen-
tal physical constants is today based on an intricate system of weighted averages of
a whole number of precision measurements for different observables (ae is one of
them). They are thus not independent of each other and any change in the system
9 We note for completeness that the anomaly for the muon is also known with high accuracy
(albeit to some orders of magnitude less well): a = 1.16592091(63) 103 .
6.6 Electron Magnetic Moment Anomaly 335
1 2
g= 21 + + ... + ... ... + ....
2
Fig. 6.35 F EYNMAN diagrams for ge 2 in 1st order, 2nd order (one-loop) and some examples
for 4th order (two-loop contributions). Note, the open photon propagator indicates interaction with
the external magnetic field
of constants entering this CODATA system changes other constants as well. Specif-
ically, the correlation coefficient between ae and , the fine structure constant, is
given as r = 0.9384 by the system. This means, the recommended numerical values
for both fundamental constants depend very strongly on each other and we have
to add: on a precise theoretical interpretation.
The theory of the electron magnetic moment anomaly has first been developed by
J. S CHWINGER. According to this work the first nonvanishing term for ae is of 2nd
order in QED and amounts to ae = /2 = 1.161 103 . This is already very close
to reality. Figure 6.35 shows the graphs up to 2nd order of the perturbation series
completely, as well as two further correction terms of 4th order (the additional 5
possible graphs or 4th order are also two-loop contributions and are not difficult to
guess).
The result of such QED computations is often written as
where even the virtual generation of and Leptons are included. Each of these
individual terms is in turn expanded in a series:
2
3
(2) (4) (6)
Ai = Ai + Ai + Ai + .
Clearly, the A1 series contributes the most (in the A1 series the term of 2nd order
(2)
is the S CHWINGER term A1 = 0.5). When evaluating the perturbation integrals
behind these coefficients ab initio one major concern is good bookkeeping. To illus-
trate the efforts necessary: In 6th order there are already 72 graphs which contribute
to A(6)
1 and in 8th order all 891 F EYNMAN graphs have been included. On that basis,
according to G ABRIELSE ET al. (2006), the results for ae are inverted to determine
todays most precise value of the fine structure constant . Based on AOYAMA et
al. (2007) and H ANNEKE et al. (2008), who report 137.035999084(51),10 the 2010
10 More recently, AOYAMA et al. (2012) have performed even 10th order QED calculations (in-
volving 12672 diagrams) and report ae = 1.15965218178(77) 103 from which they derive
1 = 137.035999173(35).
336 6 Fine Structure and L AMB Shift
e2
= = 1/137.035999074(44). (6.103)
40 c
Section summary
The deviation of the ge factor from 2 is given as the so called anomaly of the
magnetic moment of the electron, ae = (ge 2)/2 1.16 103 and can be
determined today with an amazing accuracy of about 3 1010 .
Experimentally it is measured by sophisticated techniques in electron traps
(P ENNING trap, first used by D EHMELT in this context), where quasi-
macroscopic states of the electron are prepared the so called Geonium.
Essentially, one measures the difference between the L ARMOR and the cy-
clotron frequency.
The precise determination of the magnetic moment of the electron provides
another stringent test for the potential of QED calculations. While the domi-
nant contribution arises from the one second order perturbation term (single
loop), presently the complete contributions up to 8th order have been com-
puted.
HFS: Hyperfine structure, splitting of atomic and molecular energy levels due to
interactions of the active electron with the atomic nucleus (Chap. 9).
LHC: Left hand cicularly, polarized light, also + light.
M1: Magnetic dipole, transitions induced by the interaction of a magnetic dipole
with the magnetic field component of electromagnetic radiation.
NIST: National institute of standards and technology, located at Gaithersburg
(MD) and Boulder (CO), USA. http://www.nist.gov/index.html.
QDT: Quantum defect theory, interprets experimental spectra by phase shifts in
the radial wave functions and makes predictions for scattering processes (see
Sect. 3.2.6).
QED: Quantum electrodynamics, combines quantum theory with classical elec-
trodynamics and special relativity. It gives a complete description of light-matter
interaction.
REMPI: Resonantly enhanced multi-photon ionization, ionization of atoms or
molecules by several photons with one resonant intermediate state.
RF: Radio frequency, range of the electromagnetic spectrum. Technically, one
includes frequencies from 3 kHz up to 300 GHz or wavelengths from 100 km to
1 mm; ISO 21348 (2007) defines the RF wavelengths from 100 m to 0.1 mm; in
spectroscopy RF usually refers to 100 kHz up to some GHz.
RHC: Right hand cicularly, polarized light, also light.
SHG: Second harmonic generation, doubling of a fundamental frequency, for in-
frared or visible light typically by methods of nonlinear optics.
SI: Systme international dunits, international system of units (m, kg, s, A,
K, mol, cd), for details see the website of the Bureau International des Poids
et Msure http://www.bipm.org/en/si/ or NIST http://physics.nist.gov/cuu/Units/
index.html.
UV: Ultraviolet, spectral range of electromagnetic radiation. Wavelengths be-
tween 100 nm and 400 nm according to ISO 21348 (2007).
VIS: Visible, spectral range of electromagnetic radiation. Wavelengths between
380 nm and 760 nm according to ISO 21348 (2007).
VUV: Vacuum ultraviolet, spectral range of electromagentic radiation. part of the
UV spectral range. Wavelengths between 10 nm and 200 nm according to ISO
21348 (2007).
XUV: Soft X-ray (sometimes also extreme UV), spectral wavelength range be-
tween 0.1 nm and 10 nm according to ISO 21348 (2007), sometimes up to 40 nm.
References
AOYAMA , T., M. H AYAKAWA , T. K INOSHITA AND M. N IO: 2007. Revised value of the eighth-
order contribution to the electron g-2. Phys. Rev. Lett., 99, 110 406.
AOYAMA , T., M. H AYAKAWA , T. K INOSHITA AND M. N IO: 2012. Tenth-order QED contribu-
tion to the electron g-2 and an improved value of the fine structure constant. Phys. Rev. Lett.,
109, 111 807.
B EIERSDORFER , P.: 2010. Testing QED and atomic-nuclear interactions with high-Z ions.
J. Phys. B: At. Mol. Phys., 43, 074 032.
338 6 Fine Structure and L AMB Shift
B ETHE , H. A.: 1947. The electromagnetic shift of energy levels. Phys. Rev., 72, 339341.
B IRABEN , F.: 2009. Spectroscopy of atomic hydrogen how is the Rydberg constant determined?
Eur. Phys. J. ST, 172, 109119.
B JORKEN , J. D. AND S. D. D RELL: 1964. Relativistic Quantum Mechanics. New York: McGraw
Hill.
B ORN , M. AND E. W OLF: 2006. Principles of Optics. Cambridge University Press, 7th (expanded)
edn.
B RINK , D. M. AND G. R. S ATCHLER: 1994. Angular Momentum. Oxford: Oxford University
Press, 3 edn., 182 pages.
D EHMELT , H. G. AND W. PAUL: 1989. The N OBEL prize in physics: for the development of the
ion trap technique, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1989/.
VAN DYCK , R. S., P. B. S CHWINBERG AND H. G. D EHMELT : 1986. Electron magnetic-moment
from geonium spectra early experiments and background concepts. Phys. Rev. D, 34, 722
736.
VAN DYCK , R. S., P. B. S CHWINBERG AND H. G. D EHMELT : 1987. New high-precision com-
parison of electron and positron g-factors. Phys. Rev. Lett., 59, 2629.
E IDES , M. I., H. G ROTCH AND V. A. S HELYUTO: 2001. Theory of light hydrogenlike atoms.
Phys. Rep., 342, 63261.
F EYNMAN , R. P.: 1949. Space-time approach to quantum electrodynamics. Phys. Rev., 76, 769
789.
G ABRIELSE , G., D. H ANNEKE , T. K INOSHITA , M. N IO AND B. O DOM: 2006. New determi-
nation of the fine structure constant from the electron g value and QED. Phys. Rev. Lett., 97,
030 802.
G ROSS , B., A. H UBER , M. N IERING , M. W EITZ AND T. W. H NSCH: 1998. Optical Ramsey
spectroscopy of atomic hydrogen. Europhys. Lett., 44, 186191.
G RYNBERG , G. AND B. C AGNAC: 1977. Doppler-free multi-photonic spectroscopy. Rep. Prog.
Phys., 40, 791841.
G UMBERIDZE , A. et al.: 2005. Quantum electrodynamics in strong electric fields: The ground-
state lamb shift in hydrogenlike uranium. Phys. Rev. Lett., 94, 223 001 and personal commu-
nication from T. S THLKER.
H ANNEKE , D., S. F OGWELL AND G. G ABRIELSE: 2008. New measurement of the electron
magnetic moment and the fine structure constant. Phys. Rev. Lett., 100, 120 801.
H NSCH , T. W.: 2005. N OBEL lecture: Passion for precision, Stockholm. http://nobelprize.org/
nobel_prizes/physics/laureates/2005/hansch-lecture.html.
H NSCH , T. W., M. H. NAYFEH , S. A. L EE , S. M. C URRY AND I. S. S HAHIN: 1974. Precision-
measurement of Rydberg constant by laser saturation spectroscopy of Balmer alpha line in
hydrogen and deuterium. Phys. Rev. Lett., 32, 13361340.
H NSCH , T. W., I. S. S HAHIN AND A. L. S CHAWLOW: 1971. High-resolution saturation spec-
troscopy of sodium D lines with a pulsed tunable dye laser. Phys. Rev. Lett., 27, 707710.
H NSCH , T. W., I. S. S HAHIN AND A. L. S CHAWLOW: 1972. Optical resolution of Lamb shift
in atomic-hydrogen by laser saturation spectroscopy. Nat. Phys. Sci., 235, 63.
H NSCH , T. W. et al.: 2005. Precision spectroscopy of hydrogen and femtosecond laser frequency
combs. Philos. Trans. R. Soc. Lond., Ser. A, 363, 21552163.
ISO 21348: 2007. Space environment (natural and artificial) Process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
J EFFERTS , S. AND D. M EEKHOF: 2011. NIST-F1 cesium fountain atomic clock, NIST. http://
www.nist.gov/physlab/div847/grp50/primary-frequency-standards.cfm, accessed: 8 Jan 2014.
J ENTSCHURA , U. AND K. PACHUCKI: 1996. Higher-order binding corrections to the Lamb shift
of 2p states. Phys. Rev. A, 54, 18531861.
J ENTSCHURA , U. D., S. KOTOCHIGOVA , E. O. L. B IGOT , P. J. M OHR AND B. N. TAY-
LOR : 2005. The energy levels of hydrogen and deuterium (version 2.1), NIST. http://
physics.nist.gov/PhysRefData/HDEL/, accessed: 8 Jan 2014.
J ENTSCHURA , U. D., P. J. M OHR AND G. S OFF: 2001. Electron self-energy for the K and L
shells at low nuclear charge. Phys. Rev. A, 63, 042 512.
References 339
Overview
After a general introduction and a survey of the experimental observations
(Sect. 7.1) the quantum mechanical basis for treating multi-electron systems
is reviewed in Sect. 7.2. The reader should know about these tools or get
used to them by reading this section. Section 7.3 expands this theme by intro-
ducing electron exchange and the characteristic excited state configurations.
Fine structure interaction in He and He like ions are addressed in Sect. 7.4
a consolidation of our knowledge acquired in Chap. 6 which will turn out
to be essential also for later chapters. In Sect. 7.5 the most important selec-
tion rules for E1 transitions in multi-electron system are treated. This directly
leads to double excitation in Sect. 7.6. It is of importance far beyond atomic
physics: resonances of states imbedded into a continuum are found in all fields
of physics. The ensuing interference structures, known as FANO resonances
and their origin are described in a practical, easy to access approach. Finally,
in Sect. 7.7 the alkaline earth atoms and the Hg atom are treated: they are re-
lated to He in an analogous manner as alkali atoms are related to the H atom.
7.1.1 Basics
Helium, with its atomic number (and nuclear charge) Z = 2, is the most simple of
all multi-electron systems. Nevertheless, it eludes an exact calculation of its wave
functions, energies and other properties as indeed any multi-electron or manybody
system in general, starting with the classical three body system.
Helium is a very rare element. It owes its presence on earth mainly to the radioac-
tive decay of heavy elements: the He atom is an particle which has captured two
electrons. Typically it is isolated from natural gas, where at special sources, it may
be found with concentrations up to 7 %. There are two natural isotopes: 42 He (relative
probability of wrel = 99.999866(3) %, atomic mass m(42 He) = 4.00260325415(6) u,
nuclear spin I = 0), and the very rare 32 He (wrel = 0.000134(3) %, m(32 He) =
3.0160293191(26) u, I = 1/2).
In the ground state the two electrons of He are described by 1s 2 1 S0 . The first ion-
ization potential for separating one of these electrons is 24.5873876 eV. For ioniza-
tion of the second electron (in the hydrogen like ion He+ , spectroscopically called
He II) an energy Z 2 Eh /2 = 54.4177630 eV is needed. The total binding energy of
both electrons thus amounts to 79.0051506 eV.
Correspondingly, the spectrum of He extends from the infrared spectral range
(IR) through the visible (VIS), the ultraviolet (UV) far into the vacuum-ultraviolet
spectral range (VUV). Figure 7.1 shows by way of example a typical spectrum in
the UV/VIS spectral range which has been constructed with data from K RAMIDA
et al. (2013). Marked in red are a few characteristic emission lines ending on the
1s2s 1,3 S and 1s2p 1,3 P states. This is, however, only a section of the full spectrum
of He I and one can imagine clearly that, in the early phase of atomic physics, the
interpretation of such spectra was by no means trivial. In particular since all optical
emission and absorption lines from neutral He appear grouped into two seemingly
independent sets of spectra a singlet and a triplet system. Each of them demands
an independent assignment of term energies. In the early time of spectroscopy one
associated these even with two types of helium: para an ortho He. Today this puzzle
is of course completely resolved: one observes two sets of spectra with different
total spin quantum numbers: He with its two electrons is thus the prototype element
for which all basic phenomena of multi-electron systems can be studied very clearly.
Fig. 7.1 Spectrum of He I in the VIS and UV spectral range, according to K RAMIDA et al. (2013).
Only a few, particularly intense lines are designated here (red)
7.1 Introduction and Empirical Findings 343
Of particular importance (for all areas of modern physics) is the so called exchange
interaction.
In Fig. 7.2 the energy levels of the neutral He atom below the first ionization thresh-
old are summarized. This diagram represents extensive material and detective work
from generations of spectroscopists. We recognize in Fig. 7.2 the two term systems
already mentioned. According to their fine structure splitting we designate them as
singlet and triplet system. The ground state of He belongs to the singlet system.
Optical dipole transitions (E1) are observed only within each of the two systems
(indicated by red arrows in Fig. 7.2). However, one finds that collision processes
may very well induce transitions between singlet and triplet system and vice versa
we shall discuss such processes in Sect. 7.2, Vol. 2.
Obviously, degeneracy is removed in He: a consequence of a (partial) screening
of the nuclear charge which each electron experiences by the other electron. As in
the case alkali atoms one may characterize the term energies by an effective quantum
number n with a quantum defect = n n or by an effective charge Z :
Z 2 Eh Z 2 Eh
Wn = 2
= . (7.1)
2n 2n2
singlet triplet H
+) / eV
He+ + e-
W W(He
0 1S 1P 1D 1F 3S 3P 3D 3F
W
n=5
W-
n=4
-1 4 3F
2,3,4
n=3
31P1 31 D 2 3 3P 3 3D 123
1,2,3
-2 31S0 0 1,2
0,1
3 3 S1
-3 2 3 P0
-3.62318 eV
n=2
21P1
-4 2 3P
0,1,2
21S0 23P1
-3.62331 eV
23S1 2 3 P2
Fig. 7.2 Energy levels of the neutral helium atom, He I (so called G ROTRIAN diagram). Note that
the position of the ground state (1s)2 1 S0 is not drawn to scale. The spectra split into a singlet
and a triplet system. The inset shows the (inverted) triplet splitting of the 1s2p 3 P0,1,2 states. For
comparison the corresponding term energies of the H atom are shown: clearly, the quantum defects
of the He terms are small and converge rapidly with n
344 7 Helium and Other Two Electron Systems
For the ground state of neutral He the measured binding energy 24.5873876 eV
gives n 0.7439 (or = 0.256) and Z 1.344, respectively. From the view
point of one of the two electrons the C OULOMB potential of the nucleus is screened
already in the ground state quite strongly. For all singly excited states of He the
quantum defect lies significantly below 1 and converges for the 1sns 1 S0 levels
rapidly to 0.139 only. Similarly Z rapidly converges to 1 and already for n > 3
and > 0 the energies of singlet and triplet terms are already very close to those in
atomic hydrogen. If we assume the configuration of the He electrons to be 1sn this
is easy to understand: the 1s electron finds itself in a nearly H like orbit (with Z = 2,
however) and screens the charge of the nucleus very strongly. The second electron
thus sees essentially a nuclear charge Z = 1.
Nevertheless, the excited electron in He behaves different from the valence elec-
tron e.g. in alkali atoms since the interaction with the second electron is still very
direct. It is important to realize that both electrons can in principle not be distin-
guished! This will be a key issue in the quantum mechanical treatment. We shall
find that this indistinguishability is the key to the puzzle with the two term systems
of He.
Section summary
He is the prototype of two electron systems and indeed the basis for under-
standing any multi-electron system. Its first ionization potential is 24.6 eV,
the second electron requires 54.4 eV for ionization.
The spectra of He (ranging from the IR to the far VUV region) show two
characteristic types of term systems which in E1 transitions do not mix: a sin-
glet and a triplet system. Understanding this particular feature will be a key
to the understanding of electron-electron interaction and electron exchange:
both electrons are in principle indistinguishable.
deriving observables in the one electron case, specifically the Hamiltonian. We just
have to add the components: the energies for the two individual electrons 1 and 2
p 21 Ze2 p 22 Ze2
1 =
H 2 =
and H , (7.2)
2me 40 r1 2me 40 r2
respectively, and the repulsive C OULOMB energy between the two electrons
ee = e2
H . (7.3)
40 r12
The full H AMILTON operator for the two electron system is thus
p 21 Ze2
p2 Ze2 e2
=H
H 2 + H
1 + H ee = + 2 + . (7.4)
2me 40 r1 2me 40 r2 40 r12
ee depends on
Note that the interaction term H
and thus breaks the original spherical symmetry an essential which we have ex-
ploited when solving the one electron case.
In the remaining chapter we shall (unless mentioned otherwise) again write the
H AMILTON operator in a.u.:
=H
H 2 + H
1 + H ee =
Z
Z + 1 . (7.6)
2 r1 2 r2 r12
With (7.6) we have now a Hamiltonian which depends on two position vectors.
We continue to use polar coordinates and the wave functions will now depend on
the coordinates {r1 , 1 , 1 , r2 , 2 , 2 }. The most simple ansatz for a wave function
would be
(r 1 , r 2 ) = 1 (r1 , 1 , 1 )2 (r2 , 2 , 2 ). (7.7)
This product ansatz is strictly valid only for the model of independent particles, i.e.
for two electrons that move completely uncorrelated and experience from each other
only an average potential. This turns out to be a good 1st order approximation for
the 1sn singly excited states of He. However, for the ground state and for doubly
excited states where the two electrons interact closely this approximation is only of
limited value.
By consequently extending the probability definition of the wave function in the
one particle case, as introduced with (2.5), the probability to find electron 1 at posi-
tion r 1 and simultaneously electron 2 at position r 2 in a volume element d3 r 1 d3 r 2
is given by
346 7 Helium and Other Two Electron Systems
2
dw12 = (r 1 , r 2 ) d3 r 1 d3 r 2 (7.8)
2 2
= 1 (r1 , 1 , 1 ) 2 (r2 , 2 , 2 ) d3 r 1 d3 r 2 .
From this, the probability to find electron 1 at position r 1 and electron 2 anywhere
becomes
2 2
dw1 = d r 1 (r 1 , r 2 ) d3 r 2 = 1 (r1 , 1 , 1 ) d3 r 1 .
3
(7.9)
2
As usual, we assume the one electron wave functions (orbitals) 1 (r1 , 1 , 1 ) and
2 (r2 , 2 , 2 ) to be orthogonal and normalized.
The indices 1 and 2, more generally and , stand for the usual sets of quantum
numbers:
= {n m ms }, = {n m ms }. (7.10)
For completeness and later use we also include here the projection quantum number
ms = 1/2 of the electron spin. They will become important in Sect. 7.3.
In the compact bra-ket form, such two particle product states are written
(1, 2) = (r 1 ) (r 2 ) = |n m n m
, (7.11)
where the indices 1 and 2 (given in round brackets) are an abbreviation for the
coordinates of the two electrons. Assuming that the single particle orbitals (r j )
are orthonormalized, this hold also for the two particle states:
(1, 2) (1, 2) = (1) (1) (2) (2) (7.12)
= n m |n m
n m |n m
= .
It is important to note that the Hamilton operator (7.6) for the two electron system
in this form does not depend on electron spin! All influence of the relative orienta-
tion of the spins of the two electrons will enter through the anti-symmetrization rule
to be discussed in Sect. 7.3. For the moment we ignore the spin and just note that
the Hamiltonian does not change if we exchange the two electrons, i.e. their position
coordinates r 1 and r 2 .
In the spirit of perturbation theory we first define the unperturbed problem and
start with its solution. Thus, we neglect in (7.4) the repulsive screening term (7.3),
to be treated later as perturbation, and write the Hamiltonian in 0th order in a.u.:
p 2i Z
H 1 + H
0 = H 2 i =
with H (7.13)
2 ri
7.2 Some Quantum Mechanics of Two Electrons 347
Hi describes the eigenvalue problem of the one particle system, i.e. the hydrogen
like He+ ion. Its energies Wni are given by
2
i |ni i mi
= Wni |ni i mi
= Z |ni i mi
.
H (7.14)
2n2i
For the two particle system we thus have in 0th order
H0 |
= H 1 |1
|2
+ H 2 |2
|1
= H 1 |n1 1 m1
|n2 2 m2
+ H 2 |n2 2 m2
|n1 1 m1
= (Wn1 + Wn2 )|n2 2 m2
|n1 1 m1
= (Wn1 + Wn2 )|
,
Note that these energies are calibrated in respect of the fully separated He++ + e +
e system. Experimentally measured are spectral lines (i.e. energy differences).
Tabulated one typically finds term energies Wn1 1 n2 2 W1s 2 > 0 and ionization po-
tentials WI > 0 in respect of the He(1s 2 ) ground state (e.g. K RAMIDA et al. 2013).
(Alternatively, binding energies for the valence electron are reported in the literature
in respect of the first ionization threshold Wn1 1 n2 2 + WI (I).)
Figure 7.4 compares for He with Z = 2 the true, experimentally determined en-
ergies (left) with the 0th order approximation (right) for the neutral atom (He I)
as well as for the ionic system He+ + e (He II). The ionization energy of the hy-
drogen like He+ ion (Z = 2) can be computed from S CHRDINGER equation as
(0)
WI (II) = Z 2 Eh /2 = 2Eh = 54.422 . . . eV. This is more ore less exact: according
to K RAMIDA et al. (2013) at present the best measured value is1
WI (II) = WI He+ (1s) = 54.41776203 eV, (7.16)
(0) Z2
W1s 2 = Eh 2 = 4Eh = 108.8 eV (7.19)
2
^
W - W1s 2 / eV experiment H0 without e- e- interaction Wnn''' / eV
40 - 40
He+ + e-
WI (I)= 24.59 - 54.42
20 -13.60 1s - 60
1s 2s 1S0 1s2s 3S1
20.62eV 19.82eV 1s2s,p
0 He - 80
1s 2 1S 0 - 54.42
0.0eV
- 100
- 108.84
1s1s
Fig. 7.4 Experimentally determined energies of He and He+ (left, black) and ansatz for perturba-
tion theory in 0th order (right, red)
(0) (0)
and WI (I) = WI (II). And similarly, the 0th order binding energy of one electron
in an n state is Z 2 ((1 + 1/n2 ) + 1)/2Eh = 2Eh /n2 , specifically Eh /2 =
13.6 eV in the 1s2s or 1s2p states (the excitation energy in 0th order would be
Eh /2 (2Eh ) = 1.5Eh ).
As documented in Fig. 7.4 the real energies lie obviously much higher due to
screening of the attractive C OULOMB potential by the repulsive 1/r12 term. Obvi-
ously, theory faces a major challenge if one aims at an accuracy matching that of the
experiment!
Even more complex is the situation when considering two excited electrons
in excited n states. In 0th order a 2 2 configuration would lead to an excita-
tion energy 3Eh which is higher than the first ionization potential in 0th order.
Hence, it is already embedded into the continuum of the singly ionized helium and
has an excitation energy Eh in respect of the He+ 1s 2 S0 ground state as indi-
cated in Fig. 7.4. Such states really exist and we shall explore this kind of dou-
ble excitation and the ensuing phenomenon of FANO resonances in some detail in
Sect. 7.6.
Thus, the screening term (7.3) has substantial influence on the total energy of the
He ground state. In a first step one may try to approach the problem by perturbation
ee > 0 we expect indeed to come closer to the experimental value
theory. Since H
if we account for it. For the He(1s 2 ) ground state in 0th order both electrons are
7.2 Some Quantum Mechanics of Two Electrons 349
described by 1s hydrogen like wave functions (with Z = 2) and the wave function
(7.11) is written as
1s 2 (r 1 , r 2 ) = 1s (r 1 )1s (r 2 ) or 1s 2 = |100
|100
. (7.20)
In 1st order perturbation theory the energy correction term according to (3.41) is:
(1) 5
W1s 2 = Eh Z 2 + ZEh = 74.79 eV. (7.22)
8
Considering the crudeness of the ansatz this results is astonishingly close (about
5 %) to the experimental result (7.18).
It is instructive, to compare the experimental binding energies (or ionization po-
tentials) for several members of the He like isoelectronic sequence to WI (I) from
(7.18) with 0th and 1st order perturbation theory according to (7.19) and (7.22).
This is summarized in Table 7.1.
350 7 Helium and Other Two Electron Systems
Obviously, the relative accuracy of 1st order perturbation theory improves with
increasing nuclear charge Z. While for the H anion perturbation theory completely
fails, it is quite satisfactory for the four fold ionized carbon where an agreement
between experiment and perturbation theory to about 1 % is reached.
We note here in passing, that the He like 1s 2 1 S0 state of H is the only existing
and experimentally documented, stable state of the hydrogen anion.
Variational methods are very important in quantum mechanics for the determination
of energies, in particular for ground states. Helium is a particular clear example for
its efficiency, and we shall use variational procedures in many other instances as
well. The variational principle (also called the R ITZ or R AYLEIGH -R ITZ method
after the inventors) makes use of an important theorem according to which the mini-
mal value of energy W which is calculated for a given class of functions is always
the best:
|
|H H d3 r
W = min = min 3 (7.23)
|
d r
with (r) = c (r 1 , r 2 ).
Here r refers to the full configuration space of both electrons, and the coefficients
c have to be varied until W reaches its minimum.
In detail, one chooses a test wave function for the ground state which is in prin-
ciple able to describes the properties of the system under study as well as possible.
With this test wave function one calculates the energy as expectation value of the
H AMILTON operator so that W is derived as a function of the parameters ck defin-
ing the wave function and one minimizes W according to the standard procedures
of analysis. Specifically for He the so called H YLLERAAS wave functions
N
(s, t, u) = exp(ks) c ,2m,n s t 2m un (7.24)
mn
by variation of k and the coefficients c ,2m,n to obtain the minimum energy W . The
resulting wave function obviously does not describe two independent electrons
for which the wave function would be given by the product of the individual or-
bitals. Rather, the electrons in a H YLLERAAS type wave function are highly corre-
lated: they influence each other more directly as could be described by an averaged
7.3 PAULI Principle and Excited States in He 351
potential. With this ansatz and typically 5 parameters a very high accuracy can be
achieved. H YLLERAAS already obtained W1s1s = 79.001 eV, to be compared with
(7.18). Modern state-of-the-art calculations include all relevant corrections such
as the finite He mass, L AMB shift, nuclear size and the recalculation to the present
value of R . The agreement between experiment and theory is on the order of 107
(D RAKE and M ARTIN 1998). The data presently reported in
Section summary
The Hamiltonian (7.6) of a two electron system is given by adding up all
energies of the two single particle systems plus the mutual repulsion of the
two electrons.
The most simple ansatz for two particle wave functions is a product of sin-
gle particle wave functions for electron 1 and 2. Such a simple independent
particle model is often a reasonable 0th and 1st order approximation.
Neglecting the 1/r12 screening potential it leads to energies for He which are
substantially too low, but predicts interesting doubly excited states.
Including, however, the 1/r12 term in 1st order perturbation theory, a hy-
drogenic (1s)2 product wave functions already gives a total energy for the
He(1s 2 ) ground state which agrees to within 5 % with experiment.
Excellent agreement is obtained by variational theory (7.23), also called
R AYLEIGH -R ITZ method. For the He ground state, so called H YLLERAAS
wave functions (7.24) have proved very useful. They imply strong correlation
between the two electrons.
Up to now we have not yet explicitly accounted for the fact that electrons are
fermions. Before clarifying this aspect specifically for He, we take the opportunity
to discuss multiparticle wave functions in a more general context.
It is straight forward to generalize the two particle wave function introduced in
Sect. 7.2.2 to a system with N identical particles. To include the spin properties of
the particle we introduce so called spin-orbitals which are characterized by the three
quantum numbers n m referring to position space and the spin quantum number
ms = 1/2 (orientation of the spin in respect of the z-axis):
ms
n mms (q i ) = n m
s
(ri , i , i )1/2 (si ). (7.26)
Here we have introduced a coordinate q i for each particle, representing the position
vector r i and a symbolic spin variable si . The total wave function for the N particle
quantum state is then written (q 1 , q 2 . . . q j . . . q N ). In Sect. 3.1.2 we have already
mentioned the symmetry properties:
352 7 Helium and Other Two Electron Systems
for bosons
Pij (q 1 ..q i ..q j ..q N ) = (q 1 ..q j ..q i ..q N ) = (q 1 ..q i ..q j ..q N )
(7.27)
and for fermions
ij (q 1 ..q i ..q j ..q N ) = (q 1 ..q j ..q i ..q N ) = (q 1 ..q i ..q j ..q N ).
P
(7.28)
The latter relation is completely equivalent to the PAULI exclusion principle: assume
two identical particles to be in the same quantum state, say particle i and particle j .
In that case exchanging the two obviously leads to
For fermions this contradicts (7.28) unless 0. Thus, two identical fermions
cannot exist in the same spin-orbital or in the classical formulation by PAULI J R .
(1925):
There are never two or more equivalent electrons in an atom for which the
values of all quantum numbers are identical.
In contrast, for bosons no conflict arises between (7.29) and (7.27) so that any num-
ber of identical bosons may be in the same spin-orbital, i.e. have the same set of
quantum numbers.
For a two electron system such as He and He like ions it is convenient to express the
total wave function as a product2
2 If no ambiguities can arise, we abbreviate here and in the following r j and sj as well as qj by j .
7.3 PAULI Principle and Excited States in He 353
We may apply the exchange operator onto spatial and spin function separately.
With
12 s (r 1 , r 2 ) = s (r 2 , r 1 ),
P (7.31)
and since the two electrons are indistinguishable, the states s (r 1 , r 2 ) and
s (r 2 , r 1 ) belong to the same, non-degenerate eigenvalue. They may differ only
by a scalar factor:
12 s (r 1 , r 2 ) = s (r 2 , r 1 ) = s (r 1 , r 2 ).
P
Applying this operation twice must reproduce the original state, hence 2 = 1 and
= 1, i.e.
s (r 2 , r 1 ) = s (r 1 , r 2 ). (7.32)
The spatial part of the wave function, may thus be either symmetric in respect of
exchange of the two electrons or antisymmetric. We shall call these wave functions
+s (r 1 , r 2 ) and s (r 1 , r 2 ), respectively.
Since electrons are fermions (spin 1/2), for their total wave function in He
must hold, specializing (7.28). Thus, with (7.30) obviously follows that
the spatial part s (r 1 , r 2 ) of the wave function, must be symmetric if the spin
function (1, 2) is antisymmetric
vice versa, the spatial part s (r 1 , r 2 ) of the wave function, must be antisymmet-
ric if the spin function (1, 2) is symmetric.
For the spin function we have already developed the necessary tools in Sect. 6.2.5:
two electrons with spin s = 1/2 couple to a total spin
S1 +
S2 =
S (7.34)
Again, the numbers (1) and (2) indicate which of two electrons is represented by
the spin function or , with
(1)(2) = 1 1 , (1)(2) = 1 1 and so on
2 2 2 2
being compact writing for |ms1 ms2
= |s1 ms1
|s2 ms2
.
The total angular momentum of the system is also constructed from the individual
angular momenta of the two electrons:
L1 +
L2 =
L. (7.38)
For the ground state we have so far only discussed the spatial part of the wave
function (7.20), which is symmetric with respect to exchange of r 1 by r 2 . Thus,
the corresponding spin function must be antisymmetric. Due to the PAULI principle
(total anti-symmetry, at least one quantum number different) no triplet configuration
is possible for the ground state and the complete ground state wave function is
s |(1)(2)
|(1)(2)
= 1s
1s 2 (1, 2) 2 (r 1 , r 2 ) . (7.39)
2
In this case inclusion of the spin coordinates is just a formality which does not
change the energy, the H AMILTON operator (7.6) as such being independent of the
spins.3 In summary, the ground state of He with 1 = 2 = 0, a total spin S = 0 and
a total orbital angular momentum L = 0 is a singlet 1s 2 1 S0 state.
We use here the terminology introduced in Sect. 6.2.6, i.e. we characterize the
total spin and the total orbital angular momentum with capital letters S and L and
designate the states of He by n1 1 n2 2 2S+1 LJ . As already mentioned in Sect. 6.2.6
this coupling scheme is called RUSSEL -S AUNDERS coupling (or LS coupling). An
alternative would be the jj coupling, where first the spin s and the orbital angular
momentum of each individual electron couple to a single electron total angular
momentum j ; and then the resulting j i would couple to a total angular momentum
J of both electrons. In Sect. 7.3.4 we shall further analyze why He chooses to be
LS and not jj coupled.
In general, the total wave functions for He in LS coupling are
3 We mention, however, a very remarkable property of this description of singlet states: the wave
function (state) is not separable, i.e. we cannot write it as a simple direct product of states from the
two separated electrons. One calls such states entangled, see also Appendix E.3.
7.3 PAULI Principle and Excited States in He 355
We specialize this now for the excited states of He and consider electron con-
figurations {n1 1 n2 2 } for which n1 = n2 . In the following we shall approximate
the corresponding spatial part of the wave functions by products of orbitals of the
type ns1 1 (r 1 )ns (r 2 ). Since both electrons are in principle indistinguishable the
2 2
spatial orbitals have to be combined to symmetric or antisymmetric linear superpo-
sitions:
1
+s (1, 2) = ns1 1 m1 (r 1 )ns2 2 m2 (r 2 ) + ns1 1 m1 (r 2 )ns2 2 m2 (r 1 ) (7.42)
2
1
s (1, 2) = ns1 1 m1 (r 1 )ns2 2 m2 (r 2 ) ns1 1 m1 (r 2 )ns2 2 m2 (r 1 ) . (7.43)
2
We now focus on such excited states in which one of the electrons remains in its 1s
ground state configuration |100
, the other one, however, is raised into an excited
state |n m
. With the Hamiltonian (7.6) we obtain in 1st order perturbation theory
for singlet (7.40) and triplet states (7.41) a total energy
(1)
respectively. We have to compute the diagonal matrix element Hee which deter-
mines the energy shift in respect of the undisturbed system (W (0) = W1 + W2 ).
Since the H AMILTON operator does not depend explicitly on the spin and since the
spin functions (7.37) are orthonormal, they can be pulled outs of the integral:
(1) ee |S,T
= s (1, 2) 1 s (1, 2) .
Hee = S,T |H
r12
1 s
(1)
Hee =
s
100 (r 1 )n m (r 2 ) 100
s s
(r 2 )n m (r 1 ) (7.45)
2
1 s
(r 1 )n ms
(r 2 ) 100
s s
(r 2 )n m (r 1 ) .
r12 100
356 7 Helium and Other Two Electron Systems
1L
Kn
Kn
Jn 3L
1sn
Fig. 7.5 Singly excited He(1sn ) state with total angular momentum quantum number L: the
term is raised by C OULOMB screening Jn and split into one singlet 1 L and three degenerate
triplet states 3 L by exchange interaction Kn
One may thus read the C OULOMB integral as follows: {} is the repulsive interaction
energy of electron 2 in position r 2 with the charge density of the other electron in
state |100
integrated over the whole charge distribution of electron 1. The exter-
nal integration just averages this repulsion over the probability to find 2 anywhere in
space. In total, Jn is the overall electrostatic repulsion energy of the two electrons.
It leads to a raise of the terms as sketched in Fig. 7.5. In contrast, the exchange inte-
gral Kn is not open to an intuitive interpretation. Albeit also of electrostatic nature
(1/r12 term) it is generated by a typical quantum mechanical effect, the exchange
of the two electrons as a consequence of symmetrizing or anti-symmetrizing the
7.3 PAULI Principle and Excited States in He 357
spatial part of the wave function. This becomes particularly evident if one formally
rewrites the perturbation (7.46) and (7.47) as
1
ee
H (1)
= Jn (1 + 4
S 1
S 2 )Kn , (7.51)
2
so that the total Hamiltonian becomes
=H
H 2 + Jn 1 (1 + 4
1 + H S 1
S 2 )Kn . (7.52)
2
One easily verifies the identity of this expression with (7.6) using
1 2 2 2 1 2 3 3 1 2
S1S2 = S S1 S2 = S = 2 S 3
2 2 4 4 4
2
S |SMS
= S(S + 1)|SMS
.
One finds that the expectation value of (7.51) is indeed identical with (7.46) and
(7.47), respectively, depending on whether S = 0 or 1.
One speaks about exchange interaction almost like its own kind of force which
results from the anti-symmetrization rule for wave functions of fermions. It is this
exchange interaction which enforces the splitting of singlet and triplet terms. The
magnitude of this interaction energy is given by the exchange integral Kn that is
by a purely electrostatic interaction. The states which diagonalize the H AMILTON
operator obviously have to be such that also S 1
S 2 is diagonal, which implies diag-
onalization of S . Our initial ansatz to first couple the spins through S = S1 +
2
S2
thus finds its justification by the operator 2 (1 + 4S 1 S 2 )Kn which enters the Hamil-
1
tonian. We shall expand this somewhat more precisely in the next section.
We emphasize again that, according to the scheme sketched in Fig. 7.5, triplet
states generally have lower energies than singlet states (for otherwise identical quan-
tum numbers). A second look at Fig. 7.2 confirms this statement. The physical origin
is also quite clear after the preceding discussion: in triplet states the spin function
is symmetric, the spatial part of the wave function is antisymmetric. However, this
implies that the probability for both electrons to come close to each other is smaller
than for a symmetric spatial function. In a triplet state, the two electrons are never
found at the same position simultaneously. Consequently, the average repulsive en-
ergy is smaller in triplets than in singlets, and hence triplet states lie lower than
singlet states. Very generally, H UNDs rules hold:
For a given electron configuration the states with the highest multiplicity
2S + 1 have the lowest energy.
Among these, the states with highest L have the lowest energy.
We point out that the special form (7.52) of writing the H AMILTON operator ex-
plicitly with exchange interaction is valid also for multi-electron systems where one
358 7 Helium and Other Two Electron Systems
7.3.4 An Afterthought
What kind of force acts on the spins? It is definitively not magnetic, since, so far,
we have not included any magnetic moments of the spin, spin orbit interaction and
the like. Still, the spins prefer obviously to be oriented parallel or antiparallel; the
following considerations should help to understand why.
Turning this question around, one is tempted to search for a total wave function
differing from (7.30) which, nevertheless, is consistent with the anti-symmetrization
rule (7.33). One possible way to realize such states from a linear combination of
orbitals are so called S LATER determinants. They will be discussed and used in
some depth in Chap. 10. For the present case of a singly excited He atom in the
configuration {1s, n } the following expressions are possible:
s (1)(1)
1 s (1)(1) n
1 (1, 2) = 1s
2 1s (2)(2) n (2)(2)
s s
s s (1)(1)
1 1s (1)(1) n
2 (1, 2) = s s
2 1s (2)(2) n (2)(2)
s s (1)(1)
1 1s (1)(1) n
3 (1, 2) = s (7.53)
2 1s (2)(2) n (2)(2)
s
s s (1)(1)
1 1s (1)(1) n .
4 (1, 2) = s
2 1s (2)(2) n (2)(2)
s
The numbers in brackets () refer again to the coordinates of electron 1 and 2. Obvi-
ously one may pull out (1)(2) and (1)(2) from 1 (1, 2) and 4 (1, 2), respec-
tively. These linear combinations of atomic orbitals are thus identical to the triplet
states T (1, 2) = s (1, 2)11 (1, 2) with an antisymmetric spatial wave functions
7.3 PAULI Principle and Excited States in He 359
according to (7.41) just as used in the previous description of the He excited states.
However, the states 2 (1, 2) and 3 (1, 2) cannot be identified with any of the states
used so far albeit antisymmetric beyond any doubt: exchange of electron 1 and
2 implies interchanging two rows in the determinants, which by definition changes
the sign of the determinant.
To understand the significance of these state functions 1 (1, 2) to 4 (1, 2) let
us use them to express the H AMILTON operator (7.6) in matrix representation. By
simply multiplying this operator from the left with j (1, 2)| and from the right
with |i (1, 2)
and making use of the orthonormality relation (7.37) for the spin
functions we obtain the 4 4 matrix
W1s + Wn + Jn Kn 0 0 0
0 0 0 0
H=
0 0 0 0
0 0 0 W1s + Wn + Jn Kn
0 0 0 0
0 W1s + Wn + Jn Kn 0
+0
. (7.54)
Kn W1s + Wn + Jn 0
0 0 0 0
We have written it in terms of a diagonal and a non-diagonal part. One readily ver-
ifies that Jn and Kn are indeed the C OULOMB and exchange integrals, (7.48) and
(7.49), while W1s and Wn are the single particle energies in the 1s and n state, re-
spectively. Obviously, the S LATER determinants (7.54) diagonalize the Hamiltonian
only partially, they are thus not a complete set of eigenfunctions. We may, however,
try to diagonalize the non-diagonal term, i.e. the 2 2 block in the middle of the
second matrix in (7.54). Using the standard procedure we write the characteristic
equation
W) = 0
det(H
from which we obtain two solutions for W , the two missing eigenenergies W =
W1s + Wn + Jn Kn . And we happily note that this result is exactly the same 1st
order result (7.44) with (7.46) and (7.47) that we have obtained with the correct
state functions (7.40) and (7.41), respectively. The latter ones, i.e. +s (1, 2)00 (1, 2)
and s (1, 2)10 (1, 2), are of course also the result of the diagonalization procedure
which we have sketched here: they simply are linear combinations of 2 (1, 2) and
3 (1, 2) to the eigenvalues W+ and W while 1 (1, 2) and 4 (1, 2) are already part
of the triplet system.
Summarizing, it is the exchange interaction, recognized in (7.54) as the off-
diagonal term Knl , which enforces the parallel or antiparallel orientation of the
electron spins! Combination of spin-orbitals which do not diagonalize the spin and
orbital angular momentum simultaneously are no eigenvalues of the H AMILTON
operator under discussion.
Of course this statement is fully based on the fact that exchange interaction
is by far the dominant perturbation as assumed in (7.54). For high Z this will
360 7 Helium and Other Two Electron Systems
change significantly with spin-orbit (or LS) interaction increasing. As we shall see
in Sect. 10.4.1 LS interaction may indeed become of the same order of magnitude
or even larger than exchange. Then it will be reasonable to diagonalize the Hamil-
tonian first in respect of the dominant LS interaction and later on add exchange
interaction a minor perturbation. This explains without effort the transition from
RUSSEL -S AUNDERS (or LS coupling) to jj coupling for atoms with high nuclear
charge Z.
Section summary
The two electrons in the He atom are in principle indistinguishable, hence
proper antisymmetrization of the He wave functions according to the PAULI
principle is mandatory. This leads to singlet and triplet formation, with an-
tiparallel and parallel electrons spins, respectively.
Consequently, the He 1s 2 ground state can only be a singlet state as the spatial
quantum numbers of the two electrons are identical.
In contrast, for each singly excited 1sn configuration a singlet and a triplet
state exists. To a good 1st order approximation their spatial part is described
by the independent electron model as product wave functions of the respective
single electron orbitals, symmetrized and antisymmetrized according to (7.40)
for singlet and triplet states, respectively.
A reasonable estimate for the energy of these states is obtained in 1st order
as W1sn = W1s + Wn + Jn Kn with the C OULOMB integral, Jn , ac-
counting for the repulsion of the two electrons, and the exchange integral,
Kn , characterizing the electron exchange energy.
We recognize the exchange energy Kn to be responsible for the electrons to
orient antiparallel or parallel as singlet or triplet system, respectively.
Closer inspection shows of course that He levels with S, L > 0 may also have a fine
structure (FS) due to spin-orbit (LS) interaction quite analogous to that discussed
in Chap. 6. One finds that only the He triplet levels (S = 1) are split while the singlet
states (S = 0) remain un-split (as the name indicates). The order of magnitude of FS
splitting in He is similar to that in atomic hydrogen. The largest splitting is found
for the He 1s2p 3 P state (overall ca. 1.06 cm1 or 0.00013 eV, see inset in Fig. 7.2).
This has to be compared with the singlet-triplet splitting between the 21 P and 23 P
state (2048 cm1 or 0.25 eV). FS interaction is indeed a very small perturbation in
comparison to exchange interaction and the description of the He levels in terms of
singlet and triplet states is an excellent approximation.
Thus, for a description of the FS states we have to start now with the total spin
S = S1 + S 2 and the total orbital angular momentum L= L1 + L2 for the two elec-
trons which are already coupled by exchange interaction as just discussed. Under
the influence of LS coupling a total angular momentum J = L + S of the system
7.4 Fine Structure 361
is formed for each configuration {n1 1 n2 2 LS}. The rules which have been used in
Sect. 6.2.5 for the angular momenta of individual electrons apply with appropriate
modifications. The corresponding eigenstates of the two electron system which di-
agonalize J and Jz are constructed in full analogy to (6.53) and may be written
2
explicitly
(LS)J MJ = ( 1 2 )LML (s1 s2 )SMS LML SMS (LS)J MJ . (7.55)
m ,ms
This combined coupling scheme for the total L and S angular momenta is called
RUSSEL -S AUNDERS coupling (or LS coupling), as already mentioned. Again, as in
the single electron case, the multiplets split into 2S + 1 components (if S L).
However, the magnitude of the splitting can no longer be derived simply from
(r)LS as in the single electron case (at least not for low Z). The Hamiltonian
must include all magnetic interactions between individual electrons and will contain
terms like
HLS = Li
i (r i ) Si (7.56)
in analogy to the one electron systems. We illustrate this in Fig. 7.7 for the example
of alkaline earth metals. Their role in respect of He corresponds to that of the alkali
atoms in respect of the hydrogen atom: they have two valence electrons on top of a
closed shell core (see also Sect. 7.7.1). One sees that for Mg the interval rule holds
again very well.
We recall, however, that fine structure interaction grows according to (6.66) with
Z 4 /n3 . For large Z and not too large n it may become substantial. We shall come
2
back to this again in Chap. 10 and find that RUSSEL -S AUNDERS coupling breaks
down for large Z.
Section summary
For He (as well as for other light atoms) spin-orbit interaction (7.57) is much
smaller than exchange interaction (7.49). Hence, for light atom spin-orbit
coupling can be considered a small perturbation which does not change the
coupling of the spins.
A quantitative interpretation of the fine structure of two electron systems thus
starts with total spin S =S1 + S 2 and total orbital angular momenta L=
L1 + L2 which couple in turn to the overall angular momentum of the system
J = L + S.
The overall magnitude for FS splitting in the case of He is on the same order
of magnitude as for H. Unfortunately, the simple formulas used there for the
splitting as well as the L AND interval rule cannot be applied in general (for
light atoms): spin-spin interaction and spin-other-obit interactions have to be
accounted for.
At the beginning of this chapter we have already introduced selection rules for E1
transitions in He. We want to explore some basic questions about photon absorp-
tion and emission in multi-electron systems, using the two electron He system as a
(hopefully) still transparent example. We now have to expand our definition of the
dipole operator (4.54). For one electron at position r i the dipole moment was just
7.5 Electric Dipole Transitions 363
D i = er i . For N active electrons the perturbation energies (4.55) for all potentially
active electrons have to be added:
/N 0
ieE0
U (r, t) = D i E(t) = r i eeit + e eit . (7.58)
2
i=1
Here E(t) is again the electric field component of the wave and E0 represents the
amplitude. The dipole transition operator (4.56) for a multi-electron system is cor-
N (i)
respondingly D = i=1 D with D(i) = (D i e)/e = r i e. Time dependent per-
turbation theory has to be applied in full analogy to the one electron problem. Only
the dipole transition matrix element is somewhat more complicated.
The following considerations are based on the independent particle model. For
He dipole transition matrix element has the form:
Dba = b (1, 2)r 1 + r 2 a (1, 2) e.
(7.59)
Explicitly with the wave functions according to (7.40) and (7.41) we obtain
s s
(1, 2)Sb b (1, 2)r 1 + r 2 a
MS
Dba = b
M
(1, 2)Sa Sa (1, 2) e,
abbreviating again (r 1 ) = (1) and (r 2 ) = (2). We first note that neither r 1 nor r 2
act directly on the spin component of the wave function. We may thus pull the latter
out of the integral and apply the orthonormality relations for the spin functions:
s s MS M
Dba = b (1, 2)r 1 + r 2 a (1, 2) Sb b (1, 2)Sa Sa (1, 2) e (7.60)
s s
= b (1, 2)r 1 + r 2 a (1, 2) eSb Sa MSb MSa .
This is a very important result: the spin state remains unchanged by E1 tran-
sitions in He. In particular, there are no E1 transitions between singlet and triplet
system (these so called intercombination lines are forbidden). Correspondingly, the
+ and symmetry of the spatial wave functions remain also unchanged in E1 tran-
sitions (this holds even for E transitions of all orders since the spin is not involved).
This selection rule thus explains the experimental observation of the two quasi iso-
lated systems of lines historically para and ortho helium, i.e. transitions occur
only within the singlet or the triplet system.
For He this rule holds rather strictly. For He like ions with large Z one observes
more and more also (weak) intercombination lines: as Z increases, spin-orbit in-
teraction (7.57) also increases and has to be accounted for. This leads to a mixing
of the pure singlet or triplet character of states in the independent particle model
so that triplet and singlet components are added to the respective other multiplicity
(so called configuration interaction, CI). For this very reason intercombination lines
also become possible.
Beyond the forbidden intercombination lines the model of independent particles
supports another important selection rule: allowed are only pure one electron transi-
tions even within the singlet or triplet system i.e. only one of the quantum numbers
is allowed to change in a one-photon E1 transition. The proof is quite instructive:
364 7 Helium and Other Two Electron Systems
In the independent particle model the spatial part of the wave function is writ-
ten as symmetric or antisymmetric products of one electron orbitals according to
(7.42) or (7.43), respectively. Let the initial configuration be {a} = {1a, 2a} =
{n1a 1a m1a , n2a 2a m2a }, the configuration after the transition {b} = {1b, 2b} =
{n1b 1b m1b n2b 2b m2b }. The dipole transition matrix element between these is
s s
Dba = b (1, 2)r 1 + r 2 a (1, 2) e (7.61)
1 s
= 1b (1)2b s
(2) 1b
s s
(2)2b (1)r 1
2
+ r 2 s (1) s (2) s (2) s (1) e.
1a 2a 1a 2a
We note that r 1 contributes to the matrix element only through that part of the wave
function which depends on r 1 while r 2 is only relevant for the wave function de-
pending on r 2 . Hence the components of (7.61) factorize. This results in expressions
of the type
1 s s s s
Dba = 1b (1)1a (1) 2b (2)r 2 e2a (2)
2
1 s s s s
1b (1)2a (1) 2b (2)r 2 e1a (2)
2
1 s s s s
2b (1)1a (1) 1b (2)r 2 e2a (2)
2
1 s s s s
+ 2b (1)2a (1) 1b (2)r 2 e1a (2) +
2
and corresponding expressions with r 2 replaced by r 1 and vice versa, leading to
identical results (one has to integrate over the whole space). Recalling that the one
electron wave functions are orthonormal this implies
s s
Dba = 1b1a 2b (r)r e2a (r) (7.62)
s s
1b2a 2b (r)r e1a (r)
s s
2b1a 1b (r)r e2a (r)
s
+ 2b2a 1b (r)r e|1a
s
(r) ,
i.e. one of the quantum numbers has to remain the same before and after the tran-
sition. What remains of the double integrals over r 1 and r 2 are just single elec-
tron transition matrix elements: they are determined by the standard selection rules
which we have developed in Sect. 4.4 for one electron systems. Thus, the model of
independent particles allows only single electron transitions, i.e. only such transi-
tions are allowed where one electron changes its quantum number while all others
remains in its initial configuration.
Specifically, if one electron is originally in the ground state 100, only transition
{100n m} {100n m } or {100n m} {n m n m} may occur for which
the usual selection rules
= 1 and
m = 1, 0 hold, together with Sb = Sa
7.6 Double Excitation and Autoionization 365
Section summary
The dipole transition operator for an N electron system is given by D=
N (i)
i=1 D , with D = r i e.
(i)
Exploiting the structure of the wave functions one finds that E1 transitions
between singlet and triplet system (intercombination lines) are strictly forbid-
den.
The independent electron model allows only E1 transitions in which the quan-
tum numbers of only one electron change.
The usual
= 1 and
m = 1, 0 selection rules hold. In addition, for fine
structure transitions
J = 0, 1 with 0 0 must hold.
In the previous discussion of excited He states at least one of the electrons was as-
sumed to be in its 1s ground state, the other one in an excited n state. However,
one may very well envisage configurations where two electrons are excited irre-
spective of the question how such doubly excited states might be populated. The
independent particle model predicts in 0th order perturbation theory energies given
by (7.15). For the example of a series with the configuration 2 n (n 2) this
would be (in a.u.)
(0) Z2 1 1 Z2
W2 n = + 2 .
2 4 n 4
For He (Z = 2) these doubly excited states are thus expected at energies 1Eh =
27.2 eV, while the binding energy of the He+ ion is 2Eh = 54.4 eV. Hence,
the doubly excited states {2 n } lie in the ionization continuum of He+ + e and
converge to the excited He+ (2s) state at 2Eh /4 = 13.6 eV. The {3 n } se-
ries has energies 13Eh /18 = 19.6 eV and converges to the He+ (3s) state at
2Eh /9 = 6.0 eV.
C OULOMB and exchange interaction raise the lower limits of these series, but
the general feature remains: these states are embedded in the ionization continuum.
The expected limits for these series are schematically indicated in Fig. 7.4 right (red
horizontal lines, marked 2s and 3s ). Series of absorption lines are indeed
observed experimentally in an energy range indicated by the grey shaded areas in
Fig. 7.4 on the left, marked autoionization (upper limits shown as black lines).
366 7 Helium and Other Two Electron Systems
4 Note that the signals shown are not generated by differentiation as often done in spectroscopy for
better assignment of line centres the genuine line shapes are shown here.
5 Two asterisks, short for double excitation.
7.6 Double Excitation and Autoionization 367
n = 2+ He (sp, 2 n ) 1 P o WI (n = 2)
65.405
3- 3+
4+ 5+
x 15 6+
4- 5- 7+
photoionization yield
60 61 62 63 64 65
He (3,1n )
WI (n =3)
4 5 72.963
n=3 6 14
15
q =1.45 8 16
17
19
3,-13
= 0.21
Wr = 69.88
excitation energy / eV
Fig. 7.8 He autoionization spectra recorded with high resolution synchrotron radiation; top:
(sp, 2n) series (D OMKE et al. 1991), bottom left: (3, 1n ) series, inset bottom right: enlarged
detail (S CHULZ et al. 1996). Red vertical lines indicate the limits He+ (n = 2) and He+ (n = 3)
towards which the two series converge. For the state (3, 13 ) the red line shows by way of example
a FANO profile fit (7.70) with the parameters q, and Wr
for exciting the He state is partially used for ionization (creating He+ (1s) from
He(1s 2 )) while the excess energy, WI [He(1s 2 )], is carried as kinetic energy
Wkin by the ejected electron.
The peculiar shape of the autoionizing lines documented very neatly in Fig. 7.8
is explained as a special kind of interference: the final state He+ + e (Wkin ) may
in principle be reached from the ground state through two different channels:
or one excites one of these doubly excited states which subsequently decays into
the continuum:
+ He 1s 2 He He+ + e (Wkin ). (7.66)
The experimentally observed line shape results from the fact that both channels
(direct ionization and ionization via the resonance He ) can by principle not be
distinguished in the experiment a situation which is characteristic for any kind
of interference experiment, we may e.g. think of YOUNGs double slit experiment.
368 7 Helium and Other Two Electron Systems
Let us call the probability amplitudes for the two processes cd and cr , respectively,
written as
cd = Aei and cr = cr (W )ei(W ) . (7.67)
According to the general principles of quantum mechanics these amplitudes for the
two indistinguishable processes have to be added coherently. The probability for
observing an absorption signal is then given by
2
S(W ) = |cd + cr |2 = Aei + |cr |ei (7.68)
= A2 + |cr |2 + 2A|cr | cos( ).
Close to resonance the phase angle changes rapidly and one expects a character-
istic interference pattern as indeed observed in the autoionization spectra Fig. 7.8.
A quantitative treatment of this phenomenon has been described for the first time
by FANO (1961). Briefly, one considers amplitude A and phase for the direct pho-
toionization process to be essentially constant and independent of the photon en-
ergy W = over an energy range relevant for a single resonance. For cr (W ) one
assumes the typical behaviour of a (complex) resonance amplitude expected when
W passes through the resonance, as e.g. for a harmonic oscillator. We have already
discussed this in the context of ordinary optical resonance absorption in Sect. 5.1.2.
As illustrated in Fig. 5.2(a), the magnitude |cr (W )| rises and decreases rapidly close
to a resonance at energy Wr , its linewidth being , while the phase changes from
zero to . Due to the superposition of cr (W ) with cd one obtains indeed very pro-
nounced line shapes as observed in the experiment. The famous FANO line profile
(more precisely the B EUTLER -FANO profile, B EUTLER being the spectroscopist
who first observed such line shapes) is given by6
(q + )2 W Wr
with = (7.70)
1 + 2 /2
with the linewidth , the reduced energy , and the FANO line shape parameter q.
The latter includes the relative phase and coupling strength between the direct and
the resonant process. We have fitted the He(3, 13 ) line shown in the experimental
data Fig. 7.8 by way of example with this profile and obtain a nearly perfect fit! In
the Sect. 7.6.3 we investigate in a systematic manner different forms of such FANO
resonance profiles and try to understand them in some more detail.
6 For fitting a measured signal the original formula has to be slightly generalized, e.g. to
C (q + )2 C
S() = + A2 2 , (7.69)
q2 +1 1+ 2 q +1
accounting for the background signal A2 and the strength C of the resonance.
7.6 Double Excitation and Autoionization 369
=1 0.5 = -1
>>1 << - 1
Re
- 0.5 0 0.5
Resonances of the type discussed here are observed in many areas of physics. One
finds them in optical spectroscopy as discussed here, in atomic and molecular col-
lision physics with electrons, ions and atoms, but also in nuclear physics or high
energy physics, as well as in solid state physics. They always occur when a final
state may be reached on different pathways, one of which involves a resonant, quasi
stable state. To understand the behaviour described by the FANO profile (7.70) we
visualize the excitation amplitudes involved in the complex plane.
With the normalized energy according to (7.70) one writes magnitude |cr ()|
and phase () of the resonance amplitude cr or alternatively real and imaginary
part according to (5.14a)(5.15b), scaled with a resonance strength B:
1
cr () = B and tan () = or (7.71)
2 + 1
|cr ()|2 |cr ()|2
Re(cr ) = and Im(cr ) = . (7.72)
B B
We have already illustrated the resonance amplitude in Fig. 5.2. Figure 7.9 shows
an alternative visualization in the complex plane: with (7.72) one easily verifies:
2
2 B 2 B
Re(cr ) + Im(cr ) = . (7.73)
2 2
Thus, when the reduced energy is tuned through resonance ( = 0), cr rotates in
the complex plane on a circle with radius B/2 around a centre at (0, iB/2).
We have now to add this resonance amplitude cr () to the (constant) direct am-
plitude cd as suggested by (7.68). With (7.71) the profile of the resonance line (i.e.
the absorption or ionization cross section) becomes:
i i() 2 i B iB 2
(W ) Ae + cr () e
= Ae + 2 + . (7.74)
+ 1 2 + 1
Autoionization resonances are observed close to resonance 0 (W = Wr )
on a background from direct ionization. Depending on the phase and the relative
magnitude A/B of the direct amplitude vector addition of cd and cr () in the com-
370 7 Helium and Other Two Electron Systems
Fig. 7.10 Top row: absorption amplitudes (full red arrows) in the complex plane as a sum of
direct ionization amplitude cd = A exp(i) (dashed black arrow) and resonant amplitude cr ()
(dashed red circle) according to (7.71) with B = 1. Bottom row: Corresponding absorption profiles
(ionization signals) S(). Depending on the parameters very different FANO line shapes emerge as
a function of reduced energy = (W Wr )/( /2). In each case the complex amplitudes and the
profiles are marked for different energies i < ii < iii
plex plane generates different line profiles as illustrated in Fig. 7.10.7 The shapes
range from (a) a pure absorption profile for a phase angle = 90 (or whenever
the direct amplitude vanishes), via dispersion like profiles (b) and (c) to completely
destructive interference (d), i.e. to a transparent window in the absorption spectrum
cd and cr () are of equal magnitude and opposite phase at resonance.
Section summary
Doubly excited states in He, embedded in the ionization continuum of He+ +
e may be excited by E1 due to a break down of the independent electron
model.
Pronounced interference structures are observed due to coherent superposi-
tion of the amplitudes for direct ionization and autoionization. These can be
described by the FANO profile (7.70).
It may be rationalized by addition of a direct and a resonance amplitude. While
the former is constant, the magnitude of the latter changes rapidly over the
resonance according to (7.71).
7 One may derive from (7.74) with some algebra the modified FANO formula (7.69) with
2
4B B C/(AB) + B/A + 2 sin
C = AB 4 + sin + and q = for cos = 0
A A 2 cos
q = 0 for = /2 and q for = /2; C and thus q are in principle double valued.
7.7 Quasi-two-Electron Systems 371
The alkaline earth metal atoms are in respect of the He atom in the same role as the
alkali atoms in respect of atomic hydrogen: their two potentially active valence elec-
trons are screened from the nuclei by Z 2 electrons in the completed filled ionic
core levels. In full analogy to the alkali metal atoms (Sect. 3.2.3) with one electron,
the Hamiltonian for two valence electrons of alkaline earth atoms is essentially that
of He, except that in (7.6) the pure nuclear C OULOMB potential has to be replaced
by a screened one which approaches 2/r1 (and 2/r2 ) for sufficiently large distance
from the nucleus. Correspondingly, the term schemes and spectra are very similar to
those of He, but in detail even richer, including also several series of doubly excited
states. They are very well documented by K RAMIDA et al. (2013): energy levels,
wavelength tables with transition probabilities, G ROTRIAN diagrams and simulated
spectra can be obtained within seconds. Thus, here we show in Fig. 7.11 only the
G ROTRIAN diagram of Be as one typical example.
binding energy / cm -1
1Po 1 3P o 3P 3D
WI = 13.282 1 P1 0 0
13
5 5
Be+ ( 2S 4 4
1/2)
3
11 3 3
2sns 2snp 2snd 2p 2 2sns 2snp 2snd
10 1S 1P o 1D 1S 3S 3 Po 3D
WI = 0 1 2 0 0
n 0
9.32270 5
5 4 5 5
5 4 4
4 4 4 - 10000
8 3 2p 2 3
3 1D 2p2
2 3 3P
7 3 0 -20000
3
6
- 30000
2
5
- 40000
excitation energy / eV
3 Be I - 50000
singlet 2
2 - 60000
Be I
1 triplet
- 70000
2 1s 22s 2 1S0 Be ground state - 75192.63
0
Fig. 7.11 Term scheme of the beryllium atom derived from K RAMIDA et al. (2013) (G ROTRIAN
diagram for the strongest transitions, Aab > 5 106 s1 ). Singlet terms are marked by red, triplet
terms by black lines
372 7 Helium and Other Two Electron Systems
7.7.2 Mercury
As a last example for quasi-two-electron atoms we discuss mercury, Hg. Its ground
state configuration is [Xe]4f 14 5d 10 6s 2 , i.e. the K to N shells are completely filled,
and all subshells of the O shell are filled except for the 5f electrons. The building-
up of the P shell starts with the two 6s 2 electrons.
Figure 7.12 shows part of the Hg emission spectrum, constructed from the
NIST data assuming here a somewhat unrealistically high electron temperature
(Te = 15 000 K) which allows us to see also several transitions between higher lying
states. In a standard low temperature gas discharges typically the 6s 2 1 S0 6s6p 3 Po2
is by far the dominant line due to a combination of a reasonable transition probabil-
ity (Aab = 8 106 s1 ) and low enough excitation energy Wexc = 4.89 eV. Hence,
the population density of the excited 63 Po2 state exp(Wexc /kB Te ) is high enough.
Note that this line is an intercombination line which in strict RUSSEL -S AUNDERS
coupling (LS coupling) is completely forbidden. Obviously, with its nuclear charge
Z = 80 and a nucleon number from A = 196204 Hg belongs to the heavy elements
of the periodic system and LS coupling starts to break down as spin-orbit interac-
tion increases Z 4 /n3 (see Sect. 6.3.2). Nevertheless, as shown in Fig. 7.13, one
continues to classify the states as singlet or triplet system.
The term scheme shows a remarkably large FS splitting of the 6 3 PJ states. With
1767 cm1 (J = 0 J = 1) and 4631 cm1 (J = 1 J = 2) the terms are nor-
mally ordered and in fair agreement with L ANDs interval rule. Its order of magni-
tude is getting close to the splitting between singlet and triplet system, which in the
7.7 Quasi-two-Electron Systems 373
6s 6p 3P o1 - 6s7d 3D2
6s 2 1S- 6s6p 3Po
6s6p 3Po - 6s 7d 3D
6s6p 3P o0 - 6s 7s 3S1
6s 6p 3P o1 - 6s7s 3S1
Fig. 7.12 Sticks spectrum of Hg in the UV and VIS after K RAMIDA et al. (2013) (S AHA-LTE
spectrum, electron temperature 1.23 eV)
6snp 6snp 6snd 6snf 6snp 6snp 6snd 6snd 5d 9(2D5/2) cm-1
WI = 1S 1P o 1D 1F o 3S 3Po 3D 3Fo 2
10.4375 0 1 2 3 0 2,1,0 3,2,1 4,3,2 6s 6p 0
10 10 9 8 6 10 9 8 6
9 8 7 5 9 8 7 5 - 10000
9 8 7 6 8 6
7
8 7 complex - 20000
7 terms
7 - 30000
6
6
3 Po - 40000
2
5
excitation energy / eV
6 3 Po
1
3 Po - 50000
4 0
singlet triplet
3 - 60000
nm
3.7
2 25 - 70000
1
6 [Xe] 4f 14 5d 10 6s 2 1S0 Hg I ground state - 84184 - 80000
0
Fig. 7.13 Term scheme of the mercury atom derived from K RAMIDA et al. (2013) (G ROTRIAN
diagram for the strongest transitions in Hg, Aab > 1 106 s1 ). Singlet terms are marked by red,
triplet terms by black lines. Intercombination lines are dashed black
case of 6 1 Po1 and 6 3 Po2 is about 10 026 cm1 . Still, the line with the highest transi-
tion probability is 6s6p 3 Po2 6s6d 3 D3 (Aab = 1.3 108 s1 ) a clean LS allowed
intra-triplet transition.
Section summary
The two valence electrons in alkaline earth metals and Hg essentially behave
like those in He, with the nuclear charge effectively shielded by the filled inner
shells.
374 7 Helium and Other Two Electron Systems
References
D OMKE , M. et al.: 1991. Extensive double-excitation states in atomic helium. Phys. Rev. Lett.,
66, 13061309.
D RAKE , G. W. F. and W. C. M ARTIN: 1998. Ionization energies and quantum electrodynamic
effects in the lower 1sns and 1snp levels of neutral helium (4 He I). Can. J. Phys., 76, 679698.
FANO , U.: 1961. Effects of configuration interaction on intensities and phase shifts. Phys. Rev.,
124, 18661878.
H UTEM , A. and S. B OONCHUI: 2012. Evaluation of Coulomb and exchange integrals for higher
excited states of helium atom by using spherical harmonics series. J. Math. Chem., 50, 2086
2102.
ISO 21348: 2007. Space environment (natural and artificial) process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
References 375
K RAMIDA , A. E., Y. R ALCHENKO, J. R EADER and NIST ASD T EAM: 2013. NIST Atomic
Spectra Database (version 5.1), NIST. http://physics.nist.gov/asd, accessed: 7 Jan 2014.
PAULI J R ., W.: 1925. ber den Zusammenhang des Abschlusses der Elektronengruppen im Atom
mit der Komplexstruktur der Spektren. Z. Phys., 31, 765783.
S CHULZ , K., G. K AINDL, M. D OMKE, J. D. B OZEK, P. A. H EIMANN, A. S. S CHLACHTER
and J. M. ROST: 1996. Observation of new Rydberg series and resonances in doubly excited
helium at ultrahigh resolution. Phys. Rev. Lett., 77, 30863089.
Atoms in External Fields
8
Overview
Sections 8.18.4 present the essentials on these topics. The reader should be-
come well acquainted with these key tools in modern atomic physics even
though parts of it may seem somewhat strenuous at first sight. These concepts
are essential also for molecular physics, as well as for a fundamental under-
standing of many properties of condensed matter and plasmas. This will be
illustrated in Sects. 8.1.6, 8.1.7, 8.1.8, 8.2.10, 8.3 and 8.4.2 for a number of
selected examples. In these presentations the reader will also find various ref-
erences to modern developments in atomic physics which will brighten the
study of these classical themes with currently hot topics as e.g. fast and
slow light in Sect. 8.4.4 or, in Sect. 8.5, with a first approach towards the
rapidly developing modern research field of matter in intense and ultra-intense
laser fields.
Pioneering work on atoms in magnetic fields was already performed at the end of
the 19th century (the second N OBEL prize was awarded to L ORENTZ and Z EEMAN
1902). Here we recall our previous discussion on magnetic moments in magnetic
LS
= a
VLS = V J (J + 1) L(L + 1) S(S + 1) (8.2)
2
with the fine structure constant 1/137 and the spin-orbit coupling parameter
a = 2 (r)
/2. We recall that (r) = (dV /dr)/r. In the following, we shall use for
clarity SI units and rewrite (8.1):
V L
LS = (a0 )2 (r) L
S/22 = a S/2 . (8.3)
The quantum numbers for orbital angular momentum, spin and total angular mo-
mentum are L, S and J , the corresponding projection quantum numbers ML , MS
and MJ (also called magnetic quantum numbers).
B depends on the mag-
In an external magnetic field B the interaction energy V
netic moments ML and MS of orbital angular momentum L and electron spin
S.
They define the total magnetic moment MJ = ML + MS . Assuming B z we
have to add (6.30) and (6.31) with the g factors gL = 1 (orbit) and ge = 2 (spin),
and obtain (B = e/2me is the B OHR magneton)
B = M
V J B = (M L + M S ) B = B (gL Lz + ge
Sz )B (8.4)
z + 2
L Sz Jz +
Sz
= B B = B B = M J z B. (8.5)
z + 2
In the last steps we have used Jz = L Sz and defined the z-component MJ z =
B (Jz + J .
Sz )/ of M
The quantum numbers for orbital angular momentum, electron spin and total
angular momentum are again L, S and J , their projections onto the z-axis ML , MS
and MJ , respectively (magnetic quantum numbers). We consider the problem for
the completely unperturbed system as solved:
= H
H 0 LS
+V +VB
J
L Jz +
2 2 2
0 S Sz (8.6)
=H +a + B B
2 2
unperturbed spin-orbit interaction magnetic interaction.
It is important to realize at this point that neither LML SMS nor LSJ MJ is a set of
good quantum numbers for this combined Hamiltonian: as discussed in Sect. 6.2.5
the states |LML SMS
do not diagonalize spin-orbit interaction. On the other hand,
the |LSJ MJ
states do not diagonalize the magnetic interaction in (8.6): Sz and
2
J
do not commute since the latter contains also the components Sx and Sy . In contrast,
per definition Jz commutes with J as well as with L and S (since Jz = L z +
2 2 2
Sz )
, Jz ] = 0.
so that [H
Thus, in the general case where magnetic and spin-orbit interactions are equally
important, MJ = ML + MS is a good quantum number of the system, while J is not
a good quantum number.
If we want to treat the problem in perturbation theory we have to choose in 0th
order the most appropriate coupling scheme, |LML SMS
or |LSJ MJ
depending
on which approximately describes the states best. This depends on the magnitude
of the magnetic field as summarized in Table 8.1. The two limiting cases for very
small and very large external magnetic field are called anomalous Z EEMAN effect
and PASCHEN -BACK effect, respectively.
Before entering into the details it is commendable to obtain an estimate for the
order of magnitude of the magnetic interaction. According to (8.5) it will be in the
range of B B. Without much effort one may obtain in the laboratory magnetic fields
up to 1 T to 2 T, using soft iron electromagnets or permanent magnets. With state-
of-the-art super-conducting magnets one reaches 10 T to 30 T that are e.g. required
for high resolution NMR and EPR spectroscopy. With B = 5.788 105 eV T1 ,
B < 30 T and L z + 2
Sz
/ 1 one thus expects splittings up to a maximum of
VB
< 2 103 eV =
14 cm1 , (8.7)
fields typically remains below FS splitting in the first excited states. To fully treat
lower and higher Z elements and states with low as well as high principle quantum
numbers n we have to consider both cases characterized in Table 8.1.
B
= B J MJ |Jz +
VB = V Sz |J MJ
B (8.8)
B
= J MJ |Jz |J MJ
+ J MJ |
Sz |J MJ
B. (8.9)
J MJ |
Sz |J MJ
VB = 1 + B MJ B.
MJ
In the next step we use some angular momentum algebra derived in Appendix C.1.2.
By applying the projection theorem in the form (C.18) onto the matrix element of
Sz the magnetic energy becomes
J MJ |
S
J |J MJ
VB = 1 + B MJ B. (8.10)
J (J + 1)2
Using the binomial formula this can be further evaluated as 2 SJ =J +
2 2
S
J
( J +
S)2 = S
2 2 2
L . Inserting this into the above matrix element the eigen-
values J (J + 1) etc. emerge and one finds for the numerical factor in the square
2
bracket
J MJ |
S
J |J MJ
3J (J + 1) + S(S + 1) L(L + 1)
gJ = 1 + = . (8.11)
J (J + 1)2 2J (J + 1)
VB = gJ B MJ B = MJ z B, (8.12)
8.1 Atoms in a Static Magnetic Field 381
with MJ z being the expectation value for the z-component of the total magnetic
moment of the electron, defined in (8.5). Thus, in a (not to high) magnetic field each
J level in a multiplet splits into 2J + 1 equally separated components.
We point out here, that the above derivation also holds for multi-electron systems
in RUSSEL -S AUNDERS (LS) coupling in low magnetic fields. The key assumption
for deriving the gJ factor from the Hamiltonian (8.6) is that the magnetic moments
of the electron orbits and spins are strictly proportional to the respective angular
momenta, with gL = 1 and ge = 2. This implies that the total
orbital angular mo-
mentum and the total spin are exactly L = i Li and S = i S i (where i refers to
the electrons involved); only after that coupling, the total overall angular momentum
J = L + S is composed. Note that assumptions also define the RUSSEL -S AUNDERS
coupling scheme.
J = L + S.
fast
L
ML
MJz
MJ
MS
382 8 Atoms in External Fields
the averaged moments
are simply given by the respective scalar products
M J /|J |. In summary, with ML = B L/ and MS = 2B S/ one writes
(8.13)
ML J J MS J J
VB = MJ
B = + B
|J | |J | |J | |J |
B L J + 2S J (J + S) J Jz
= J B = B B = gJ B MJ B.
J 2 J 2
Obviously, the ratio of scalar products to J 2 in the squared bracket is the gJ factor.
Comparing this semiclassical relation to (8.11) and (8.12) one finds that they are
equivalent: one just has to apply again the binomial formula for the scalar product
S J , and interpret J 2 , S 2 and L2 by the respective quantum mechanical expectation
values J (J + 1)2 , etc.
Some Examples
Singlet system: We remember the so called normal Z EEMAN effect which we
have introduced already in Sect. 2.7. The standard experiment is illustrated in
Fig. 4.18. We are now able to identify the special situation in which this clas-
sically expected splitting into three components is actually observed: the total
spin has to be S = 0 so that no fine structure splitting is observed in this case. A
typical example is He, where the spins of both electrons s1 = 1/2 and s2 = 1/2
may be composed to a triplet (S = 1) or a singlet (S = 0) system as discussed
in detail in Sect. 7.3. In the latter case J = L which according to (8.11) is then
characterized by a g factor gJ = 1. Hence, independent of J (or L) all J MJ sub-
states in the singlet system have the same splitting B B between neighbouring
magnetic levels MJ and MJ + 1. In Fig. 8.2 this is illustrated for the example of
MJ = +2 "normal"
(a) +1
(b)
gJ = 1
ZEEMAN spectrum
0
equal splitting
1P : J=1 +1
gJ = 1
1
0
observed in z direction
-1 - +
MJ = +1 0 -1
- +
linearly polarized
(no emission in z direction right and left
in z direction) circularly polarized
Fig. 8.2 So called normal Z EEMAN effect exemplified by a 1 P 1 D transition (e.g. in the He
singlet system). (a) Term scheme. (b) Spectrum (schematic); since gJ is the same for the upper and
lower levels only the classical triplet of lines is observed (red: components, pink: components)
8.1 Atoms in a Static Magnetic Field 383
3S +1
1
J=1 0
gJ = 2 -1
The most important 2 S1/2 2 P3/2,1/2 transitions (e.g. the Na D line doublet)
have g factors g(2 S1/2 ) = 2, g(2 P1/2 ) = 2/3 or g(2 P3/2 ) = 4/3. This leads to
a splitting scheme in the magnetic field as shown in Fig. 8.4(a). Sketched in
2P 2D
3/2 - 5/2
2P 2D
2S
1/2 3/2 - 3/2
+1/2
gJ = 2
J = 1/2 -1/2 2P
1/2 -
2D
3/2
+ - + -
6 lines 4 lines
Fig. 8.4 Alkali doublets (a) splitting of the 2 S1/2 and 2 P1/2,3/2 states in a magnetic field with
allowed E1 transitions (e.g. for the Na D1 and D2 lines). (b) corresponding doublet spectra
(schematic) and more examples
384 8 Atoms in External Fields
Fig. 8.5 Stick spectrum (calculated) of transitions between magnetically split sublevels in a septet
7 S 7 P . This could e.g. be in Cr I between the configurations 3d 5 ( 6 S)4s 3d 5 ( 6 S)4p) at a
3 4
wavelength 425.4 nm. Pink the , red the components detected in the xy plane
Fig. 8.4(b) are the corresponding spectra. They may turn out to be rather com-
plex. For reference we summarize the whole fine structure splitting with (8.8) and
(8.2) explicitly for the example 2 P3/2,1/2 . In a low magnetic field with g3/2 = 4/3
and g1/2 = 2/3 we obtain by combining (8.8) and (6.62):
*
4
B BMJ + a2 for J = 32 , MJ = 32 , 12 , 12 , 32
VB + VLS = 32 (8.14)
3 B BMJ a for J = 12 , MJ = 12 .
Septet: For heavier atoms one also observes lines with high multiplicity. In
Fig. 8.5 we just show as one arbitrary example the characteristic splitting of a
line in the Cr I spectrum. (This calculation just demonstrates how complex such
spectra can be, without going into detail for chromium which has a rather com-
plicated electronic structure).
We finally note, that the magnetic level splitting as complicated as it may appear
from the above examples often serves as an important tool for revealing term
configurations in the spectra of complex atoms and molecules.
Line Strengths
For a quantitative evaluation of spectra one needs to calculate the transition probabil-
ities between individual magnetic sublevels J MJ J MJ splitted in the magnetic
field. These are easily obtained from (4.118) and (4.124) for spontaneous emission
and absorption, respectively. One simply has to evaluate the corresponding 3j sym-
bols for
MJ = q.
The relative strengths of different lines J J within one multiplet nLS
n L S require evaluation of (6.70), i.e. of the corresponding 6j symbols. With the
tools communicated in Appendixes B.2 and B.3 this can be done without problems.
Several Java applets for computing 3j , 6j and 9j symbols are available in the In-
ternet (e.g. S VEN G ATO R EDSUN 2004).
The case of very high magnetic fields B a/B can be of relevance for light atoms
such as H, He or Li, as well as for all highly excited states since FS interaction de-
creases with the principle quantum number as 1/n3 (see (6.66)). Also in the extreme
magnetic fields of some stellar atmospheres this case has to be considered.
8.1 Atoms in a Static Magnetic Field 385
ML
In 0th order we completely neglect now the spin-orbit coupling and start with the
Hamiltonian:
H 0 + V
P B = H B
=H0 + B (L
z + 2
Sz )B. (8.15)
Then L and S are decoupled, and this situation is called the PASCHEN -BACK effect.
In this approximation L z and
Sz as well as Jz are well determined and commute
with H . Thus LSML MS are good quantum numbers as well as MJ = ML + MS .
In the vector model one may visualize this as illustrated in Fig. 8.6: both angu-
lar momenta precess around the external magnetic field B. We may compare this
directly with Fig. 8.1, where the corresponding situation for low magnetic field is
displayed (in both figures L = 1, S = 1, MJ = 3/2 is assumed and presented to
scale). We thus choose the uncoupled states, short |ML MS
, as basis and obtain
with (8.15)
B |ML MS
= B B(L
V z + 2
Sz )|ML MS
= B B(ML + 2MS )|ML MS
,
MJ ML MS
+3/2 +1 +1/2 2
2P
+1/2 0 +1/2 1
3/2
-1/2 -1 +1/2 0
+1/2 +1 -1/2
2P
1/2 -1/2 0 -1/2 -1
- 3/2 -1 -1/2 -2
2S 0 +1/2 1
1/2 +1/2
Fig. 8.7 PASCHEN -BACK effect: splitting and E1 transitions in a high magnetic field, illustrated
for the example 2 P3/2, 1/2 2 S1/2 (cf. weak B field Fig. 8.4). One sees the classical line triplet
as in the normal Z EEMAN effect. Shown are the magnetic quantum numbers MJ = ML + MS
as well as the projection of the magnetic moment onto the B or z-axis, = ML + 2MS , that
determines the splitting. Different energies may belong to the same MJ , but one specific energy
may also be realized for different MJ
In order to identify allowed optical transitions we have to account for the fact
L and
that S are now completely decoupled. And since the E1 dipole operator acts
only on the orbital component L of the states and not onto the spin
S, the selection
rules for E1 transitions in a high magnetic field are
ML = 0, 1, MS = 0 and L = 1 (8.17)
as indicated in Fig. 8.7. Since in the upper and lower level the splitting equally
depends only on = ML + 2MS , six different transitions are possible but only
three different frequencies are observed: in high magnetic fields one thus observes
the classical normal Z EEMAN triplet.
Let the spin orientation in space at this moment be given by its polar and azimuthal
8.1 Atoms in a Static Magnetic Field 387
angles, 0 0 . This may e.g. be achieved by a passing the atom through a S TERN -
G ERLACH magnet whose axes are aligned correspondingly in respect of B. This
particular spin state prepared at t = 0, is described according to (E.16) by
0 0 1 1 0 0 1 1
|
= cos exp i + sin exp i , (8.18)
2 2 2 2 2 2 2 2
in the basis of the spin states |SMS
= | 12 12
. In the magnetic field both spin com-
ponents will change with time according to (2.16) as exp(VB t/) corresponding
to their energies VB in the magnetic field. With (8.16) VB = 2B BMS and we ob-
tain the spin state as a function of time:
(t) = cos 0 exp i 0 exp i B B t 1 1
2 2 2 2
0 0 B B 1 1
+ sin exp i exp i t .
2 2 2 2
become 30 T or even more comparable to the highest fields which can be generated
by supra conducting magnets. The precession times are reduced correspondingly.
In the theory of magnetism one occasionally uses the term exchange fields, in
particular in the context of ferromagnetism. To derive such (hypothetical) exchange
fields one simply divides the exchange interaction which is responsible for ferro-
magnetism by B , as we have done in the present context. The numerics of this
procedure may lead to fields that are orders of magnitude larger than those dis-
cussed above. Of course they are just a mathematical construct! As we have learned
in (7.49) in the context of the helium atom, exchange interaction is of purely elec-
trostatic nature, enforced by anti-symmetrization of the electronic wave function,
and definitively not generated by magnetic fields. The example of He illustrates very
clearly that exchange interaction may be on the order of magnitude of an eV, in
comparison to which spin-orbit splitting (Fig. 7.6) is minute.
and assuming that only one principle quantum number n contributes, a linear com-
bination of uncoupled states
|LSMJ
= cML MS |LML SMS
(8.21)
ML MS
is the obvious ansatz for the perturbed system. The eigenvalues of H BLS are the
sought-after splittings V,MJ (B) of the energy levels in the magnetic field:
BLS |LSMJ
= V,MJ (B)|LSMJ
.
H (8.22)
The magnetic quantum number MJ by itself does not yet fully specify an eigenstate
of the total system. Thus, we use the additional parameter which is chosen such
that in the limit of high fields = ML + 2MS = MJ + MS . Hence it corresponds
to the projection = M z /B of the magnetic moment onto the z-axis. This has
already been exemplified for a 2 P3/2 state in Fig. 8.7 (see labels on the right).
Note that there may be several solutions for each value MJ , as well as several MJ
values for one specific . In detail Fig. 8.8 illustrates schematically the connection
between quantum numbers and splittings in the transition from low to high B field:
the states that have already been identified in Fig. 8.7 for the high field case are
connected with the low field situation such that MJ = ML + MS . This is, however,
not yet an unambiguous rule for the change from low to high field. In addition, we
have applied the so called non-crossing rule according to which states with equal
MJ must not cross!
The latter rule emerges from the full treatment of the problem: we have to di-
agonalize the combined Hamiltonian (8.20) for magnetic and spin-orbit interaction
by the expansion coefficients cML MS for the states |LSMJ
according to (8.21).
Using (8.16) we write H BLS in matrix form:
LML SMS HBLS |LML SMS
(8.24)
a
= B (ML + 2MS )B ML ML MS MS + 2 LML SMS L
S|LML SMS
The matrix elements of L
S are derived in Appendix C.3.2. Specifically we use
(C.59)(C.62) and the important result that the off-diagonal matrix elements are
non-zero only for MJ = ML + MS = ML + MS = MJ :
LML SMS HBLS |LML SMS
= (M +M )(M +M ) .
L S L S
(8.25)
This H AMILTON matrix may easily be diagonalized. The ansatz (8.21) leads to
the following system of linear equations
cL,S
' ( cL+1,S
LML SMS HBLS W 1|LML SMS
= 0,
(8.26)
...
cL,S
BLS W
with {. . . } representing the matrix corresponding to H 1. The eigenenergies
W in the general case are then obtained from solving the secular equation
' (
det ML , MS HBLS W1|ML , MS
= 0. (8.27)
We want to discuss this in some more detail for the important case of a 2 P3/2,1/2
doublet, for which Table 8.2 represents (8.24). It has been obtained using Table C.1.
Obviously, the secular equation (8.27) factorizes in this case into four rather conve-
nient expressions:
2B B + a/2 W = 0
B B W a/ 2
=0
a/ 2 a/2 W
a/2 W a/ 2
=0
a/ 2 B B W
2B B + a/2 W = 0.
These have six solutions Wi = V,MJ (B) that are the sought-after energy splittings
in the B field. With x = B B/a they may be written:
1
V2,3/2 /a = 2x + (8.28)
2
2
1 1 3 4 2
V1,1/2 /a = x + 1+ x+ x (8.29)
2 4 4 9 3
8.1 Atoms in a Static Magnetic Field 391
V, M / a
J
MJ = 3/2 =ML+2MS
2
1
5 1/2
2P
3/2 5 10
0 -1/2 0
2P
1/2 1/2 0 B B / a
-5 -1/2
-1
- 3/2 -2
Fig. 8.9 Transition from a low to a high magnetic field B for the example of a 2 P3/2,1/2 system.
Plotted as full red lines are the splittings VMJ of the levels as a function of the field strength B.
The parameters MJ = ML + MS and = ML + 2MS characterize the 6 states (see text). Dashed
black (red) lines show the extrapolation in the case of a low (high) field. Two avoided crossings are
marked by pink circles. The energy splitting VMJ is given in units of the fine structure parame-
ters a, the field strength B in units of a/B
2
1 1 3 4 2
V0,1/2 /a = x 1+ x+ x (8.30)
2 4 4 9 3
2
1 1 3 4 2
V0,1/2 /a = x + 1 x + x (8.31)
2 4 4 9 3
2
1 1 3 4 2
V1,1/2 /a = x 1 x + x (8.32)
2 4 4 9 3
1
V2,3/2 /a = 2x + . (8.33)
2
The result of this diagonalization of H BLS is plotted to scale in Fig. 8.9. In a high
field the splitting is proportional to B = (ML + 2MS )B, in a low field to gJ MJ B.
An important consequence of the fact that states with equal MJ couple (as ex-
pressed by nonvanishing matrix elements in the Hamiltonian Table 8.2) is clearly
seen in Fig. 8.9:
This implies a repulsion of the corresponding terms (here MJ = 1/2). One sees
that the linear splitting V,MJ corresponding to g3/2 = 4/3 and g1/2 = 2/3 in the
case of low fields is only valid over a rather small range of the magnetic field
392 8 Atoms in External Fields
Table 8.3 Limiting values for the energy splitting V,MJ (B) of a 2 P3/2,1/2 doublet in very weak
(left) and very high (right) magnetic field B, respectively. The levels are determined by the quantum
numbers J and MJ in the weak field, and by = MJ + 2MS and MJ in the high field limit
Weak field Strong field
J MJ BB a BB a MJ
3/2 3/2 2B B + a/2 2B B + a/2 2 3/2
3/2 1/2 (2/3)BB + a/2 BB 1 1/2
1/2 1/2 (1/3)BB a a/2 0 1/2
1/2 1/2 (1/3)BB a a/2 0 1/2
3/2 1/2 (2/3)BB + a/2 BB 1 1/2
3/2 3/2 2B B + a/2 2B B + a/2 2 3/2
since it would lead to two crossings (marked with pink circles) which are avoided.
While theses states with MJ = 1/2 belonging to the 2 P1/2 and 2 P3/2 doublet
levels strongly interact, the MJ = 3/2 states of the 2 P3/2 term are not at all in-
fluenced by the 2 P1/2 level the latter simply has no MJ = 3/2 component.
Thus the splitting of these terms is at all field strengths always linear in B:
V2,3/2 = 2B B + a/2.
With the eigenvalues (8.28)(8.33) one may now solve the system of equations
(8.26) for the coefficients cML MS which according to (8.21) determine the respective
states |LSMJ
. In the case of small B fields these will approach the corresponding
C LEBSCH -G ORDAN coefficients, for large fields, however, one of the |LML SMS
states will dominate the linear superposition, in fact the state with ML = 2MJ
and MS = MJ . We refrain from entering into the details of this somewhat
tedious but trivial computation.
However, it is instructive to verify that the general solutions (8.28)(8.33) con-
verge to the limiting cases for small and large fields as previously derived in
Sect. 8.1.2 and 8.1.3, respectively. Table 8.3 summarizes the results of expand-
ing the roots for BB a and BB a into a power series of x = BB /a and
1/x = a/BB ; the states are characterized in these limits by the quantum numbers
J , MJ and , MJ , respectively.
Comparing the three left columns in Table 8.3 with (8.14) we see that indeed the
dependence of the term energies on the magnetic field is exactly reproduced. The
three right most columns agree fully with (8.16) for the PASCHEN -BACK effect if
one neglects the fine structure term for very high fields. The remaining small shifts
of the terms energies are indicated schematically in Fig. 8.8 and Fig. 8.9 most
evident for the |LSMJ
= |1 1/2 0 1/2
and |1 1/2 0 1/2
states.
Problems of the type just discussed how do degenerate or nearly degenerate en-
ergy levels of quantum systems change under the influence of external or internal
fields (magnetic or electric) is of quite general importance. It was first treated by
B REIT and R ABI (1931) for the splitting of hyperfine levels in magnetic fields (so
8.1 Atoms in a Static Magnetic Field 393
called B REIT-R ABI formula, Eq. (9.53)). Prominent examples are found in mag-
netic resonance spectroscopy, i.e. in electron spin resonance (EPR) and nuclear spin
resonance (NMR) which will be treated in some detail in Chap. 9. However, similar
problems occur also in different context e.g. in molecular bond formation or in
atomic collision processes as we shall see in Vol. 2. Whenever two or more states,
originally separated in energy, are exposed to an interaction that may become larger
than their initial separation, one may in principle expect crossings of the energies
as a function of the perturbation. Such a perturbation induced energy degeneracy
warrants detailed inspection.
We treat the problem for the general case, choosing, however, the most simple
situation of only two relevant states involved, |1
and |2
. Be H 0 the unperturbed
Hamiltonian and V (q) the perturbation as a function of a characteristic experimental
parameter q (e.g. the strength of an electric field, the distance of two atoms and so
on). Assume that in 0th order
H0 1(0) = W (0) 1(0) and H 0 2(0) = W (0) 2(0) .
1 2
(0)
W1 W + V11 W2(0) W + V22 |V12 |2 = 0.
W2 W = 2V12
W1
qx
perturbation parameter q
Fig. 8.10 Avoided crossing. As a function of a perturbation parameter q the energies W1 and W2
of two states are plotted: with (full lines) and without (dashed) interaction, respectively, i.e. with
finite or negligible off-diagonal matrix element V12 (q). The perturbation parameter q may be an
external magnetic or electric field, or e.g. the distance between two atoms
= 0, the re-
If the off-diagonal terms of the perturbation matrix vanish, V12 = V21
sults correspond to those obtained from perturbation theory in 1st order (for two
non-degenerate states):
(1) (0) (1) (0)
W1 = W1 + V11 and W2 = W2 + V22 .
Thus in this situation (V12 = 0) and depending on the behaviour of V11 (q) and
V22 (q) as a function of a perturbation parameter q, curve crossings are in principle
possible, i.e.
W (1) (q) = W1(0) W2(0) + V11 V22 = 0 may occur at a certain
value of q = qx . If, however, the two states interact and V12 (qx ) = 0, the curve
crossing is avoided as illustrated in Fig. 8.10. According to (8.36) the splitting at the
approximate crossing point qx is
W = W1 W2 = 2|V12 |. (8.37)
We summarize: The potential energy curves of two states |a
and |b
do not cross
as a function of a characteristic perturbation parameter q, if the interaction potential
at the approximate crossing point is finite, i.e. if Vab = 0. Such a situation is called
avoided crossing.
Specifically, for the change of atomic energies in a magnetic field we have seen
in Sect. 8.1.5 that the matrix elements (8.24) of the perturbation HBLS do not vanish
if MJ = ML + MS = ML + MS = MJ : states with equal ML + MS = MJ couple.
Hence, such states cannot cross as illustrated in Fig. 8.9 (non-crossing rule).
8.1.7 Paramagnetism
The magnetic properties of matter are determined by the magnetic dipole moments
of its constituents (atoms and molecules). We give here a brief outline of how to
8.1 Atoms in a Static Magnetic Field 395
B = 0 (H + M) = 0 H + M 0 H = r 0 H . (8.38)
H is the magnetic field strength in the material and M the so called magnetization,
i.e. the magnetic moment per unit volume of this material (both measured in A m1 )
with = r 0 being the magnetic permeability (0 = 4 107 N A2 ), r the
relative magnetic permeability and M = r 1 the magnetic susceptibility (both
dimensionless). Occasionally one also finds Mmol given per mol (so called molar
susceptibility).
Thus, the present task is to express the magnetization
The partition function in the denominator ensures correct normalization. At not too
low temperatures T the energy (8.12) of an individual magnet VB (MJ ) is very small
compared to the average thermal energy kB T at room temperature we typically
have VB (MJ ) 105 eV kB T 0.010 eV. Thus the partition function is 1
and the term in the nominator of (8.40) may be expanded as
VB (MJ )
MJ z exp VB (MJ )/kB T gJ B MJ 1
kB T
B
= gJ B MJ + (B gJ MJ )2 .
kB T
We insert this into (8.40) and see that positive and negative values of MJ compensate
each other in 0th order while the 1st order terms in the lead to a finite positive
contribution B/T . We finally obtain for the magnetization
C
M
= NMJ z = B,
T
396 8 Atoms in External Fields
where N is the particle density of the material (dimension L3 ). With (8.39) the well
known
C
C URIE law M = (8.41)
T
for paramagnetism follows, M being the magnetic susceptibility.
For not too dense matter the C URIE constant C may be easily evaluated for
specific cases of L, S and J . If one extends the parameter range to very low values
of the temperature T or to very high magnetic fields B the full sum (8.40) has to
be used, the so called L ANGEVIN function, with the variable B/T . It leads to a
saturation of the magnetization at large B/T . In the general case one has, however,
also to account for the fact that a more rigorous treatment gives M = M/H . As
a simple exercise the reader may evaluate in detail the case L = 0, S = J = 1/2.
Paramagnetism requires a nonvanishing total angular momentum J of the ma-
terial under investigation. In the electronic ground state in which we normally find
matter the angular momentum is often zero and the paramagnetism of the atoms or
molecules studied is determined by their electron spin. As it turns out, not many
materials in nature happen to have unsaturated electron spins. Typical examples are
the alkali metals and a few other metals (not the alkaline earth metals, however).
Among the molecules are only a very few: one important example for molecular
paramagnetism is oxygen, O2 in its 3 1 ground state, as we shall see in Sect. 3.6.9,
Vol. 2.
8.1.8 Diamagnetism
=
H
p2 LS
+ V (r) + V
2me
in an external field becomes
= p2 aL
S eA
p e2 2
H + V (r) + + + A
2me 1 23 4 2 me 2me
1 23 4 potential, 1 23 4 1 23 4 1 23 4
kin. energy, spin-orbit, paramagn., diamagn. (8.42)
etc. etc.
which correspond to >0 < 0.
As explicated in Sect. 4, Sect. 5.4 and Appendix H, the term eA p /me leads
to electric and magnetic dipole interaction, er E and gJ B B J , respectively,
which are responsible for electric polarizability (Sect. 8.2.10) and paramagnetism
8.1 Atoms in a Static Magnetic Field 397
(Sect. 8.1.7). It is also the cause for electric quadrupole interaction (Sect. 5.4). In
contrast, the quadratic term in (8.42) leads to the so called ponderomotive potential
in intense fields (to be discussed in Sect. 8.5.1), as well as to diamagnetism as we
shall show now.
In comparison to paramagnetism (as far as present in a material) diamagnetism
is a very small effect, counteracting the external magnetic field according to L ENZs
law. From a classical point of view it describes the deceleration of the electron orbits
around the nucleus as a consequence of the external B field: the magnetic moment
of one atom induced by the latter is
MJ z =
)M H =
)M B/0 (8.43)
dmag
= e2 2 2 2 e2
V B r sin = B2 r2 (8.45)
8me 12me
with r 2
being the expectation value of the squared electron distance from the
atomic centre. The diamagnetic energy (8.45) thus obtained has to be compared
to (8.44). We finally obtain with 0 = 1/(0 c2 ), the fine structure constant and
the atomic unit of length a0 :
2 2
e2 2 2 r r
)M = 0
r = a03 2 2 = 1.12 104 a03 2 . (8.46)
6me 3 a0 a0
M =
)M N and Mmol =
)M NA , (8.47)
respectively. With this and (8.47) even for solid materials in the most dense close-
packed arrangement of spheres the macroscopic diamagnetic susceptibility is only
on the order of M 0.83 104 . We mention in passing that while the value
of )M increases with r 2
the particle density decreases with 1/r 3
so that the
macroscopic M decreases with the size of the atoms.
We finally note that in standard atomic spectroscopy the diamagnetic energy is
minute and does not play a significant role. However, this is different when high
lying RYDBERG states are studied. Since the radius of these atoms increases with
n2 and r 2
a02 n4 the ratio of the diamagnetic term (8.45) to the normal magnetic
field splitting (8.8) is
dmag
V e B
Ba 2 n4 106 . . . 107 n4 . (8.48)
VB
6gJ 0 T
Section summary
The interaction energy of an electron with an external magnetic field is de-
scribed by VB = B (Lz + 2Sz )B/.
In low magnetic fields B (spin orbit interaction V L
LS = a S/2 V B ) one
uses the |(LS)J MJ
basis and considers V B as a small perturbation. The en-
ergy is then gJ B MJ B with the L AND factor gJ in (8.11).
In the limit of high magnetic fields, i.e. for very low Z atoms, extremely large
B fields, or high principle quantum numbers n, on treats V B in the uncoupled
|LML SMS
scheme, while V LS is a small perturbation. In this case, L, S, ML
and MS are good quantum numbers as well as MJ = ML + MS . In addition,
= ML + 2MS has to be introduced.
Vector models are useful to visualize the projection of magnetic moments
onto each other and onto the external field. They are based on the precession
of magnetic moments at L ARMOR frequency, which can be rationalized in
quantum mechanical terms.
By diagonalizing the full Hamiltonian (8.6), its matrix form being given in
Table 8.2, one may also treat the intermediate situation between low and
high fields exactly. The energies are given by B REIT-R ABI type formulas, see
(8.28)(8.33). Characteristic are interactions between states with equal MJ
which lead to the non-crossing rule for such states.
Avoided crossings are important also in a broader context: they occur if de-
generate or quasi-degenerate states are coupled by a non-diagonal interaction
which changes as a function of a characteristic parameter.
8.2 Atoms in an Electric Field 399
The splitting of energy terms in static electric fields, the so called S TARK effect, has
been discovered by Johannes S TARK in 1913 (N OBEL prize 1919). In traditional
spectroscopy it plays only a minor role since static electric fields available in the
laboratory (see Table 8.4) are typically very small compared to the inner atomic
fields. The latter are on the order of magnitude
Ze V
Eatom = 2
= 5.14 1011 Z . (8.49)
40 a0 m
This may be compared to the electric field strength for electric breakdown in air
which is about 10 kV cm1 = 1 106 V m1 which is thus far below the electric
field strength needed to modify the atomic energies significantly.
8.2.2 Signicance
1. Electric fields play a key role in the structure and composition of matter: be
it in molecular bonding, the forces in crystal lattices, soft matter or plasmas.
The electric fields by which neighbouring atoms and ions interact are essential
ingredients for our understanding of the properties of matter.
A static electric field which is a polar vector field in contrast to the axial magnetic
field breaks the symmetry of the H AMILTON operator. The perturbation by the
electric field may be written in 1st order approximation
VE (r) = D E = er E (8.51)
8.2 Atoms in an Electric Field 401
making use of (4.75) and of the renormalized spherical harmonics Ckq according to
(B.29). Note that the squared orbital angular momentum operator
2
L (which implies
differentiation with respect to ) does not commute with this perturbation. Hence
2
L does no longer commute with H and L is not any more a good quantum number
in contrast to the situation for an atom in a magnetic field where according to (8.5)
the relevant interaction is proportional to L z + 2Sz , which commutes with
2
L but
no longer with
2
J .
However, even with (8.52) included in the Hamiltonian it still commutes with L z
since C10 (, ) = C10 ( ) and VE (r) does not depend on , while Lz = i/ acts
only on , and Sz does not act on the spatial coordinates at all. Hence, ML and MS
remain good quantum numbers in the electric field. As we shall see, the interaction
matrix elements depend only on |ML |. Hence, it is reasonable to characterize the
eigenstates by L2z and |ML |. In the extreme case we expect a superposition of many
orbitals with different L but a constant value of |ML |: a so called hybridization
occurs a phenomenon crucial for the understanding of chemical bonding.
In the following considerations we shall concentrate for simplicity on systems
with only one active electron (the valence electron). In the general case one has to
replace in (8.51) the vector r by the sum r i over all electron coordinates. Typi-
cally, for coupled systems the reduction formulas for evaluating the matrix elements
are accordingly more complicated.
The effect is thus expected to be indeed very small, by comparison to (8.7) even
significantly smaller than the Z EEMAN effect under standard laboratory conditions.
We distinguish two limiting cases depending on the non-C OULOMB term VnC in
the perturbation hierarchy Table 8.5:
402 8 Atoms in External Fields
Table 8.5 Perturbation hierarchy for effective one electron atoms: Summary of interactions and
consequences (gqn stands here for good quantum number)
H like Alkali like FS splitting Z EEMAN effect S TARK effect
=H
H C (r) +VnC (r) +VLS +VB +VE (r)
pure C OULOMB electrostatic spin-orbit ext. B field ext. E field
Tkin + C/r not 1/r L S B (Lz + 2Sz )B er E
L degeneracy L degeneracy M degeneracy |M| degeneracy
removed removed removed
(LS)J MJ ,
[H Sz ] = 0 ,
[H
2
L ] = 0
coupling J no longer gqn L no longer gqn
VnC (r)
VE (r)
: In this case the electric field removes for the first time L
degeneracy. This has to be treated by perturbation theory for degenerate states
(see Sect. 3.3.4). One finds the so called linear S TARK effect.
VnC (r)
VE (r)
: L degeneracy is already removed when the E field is
switched on. Due to the symmetry of the perturbation potential (z has odd par-
ity) all diagonal matrix elements disappear. Hence, the energy does not change
in 1st order perturbation theory and one has to resort to 2nd order for determin-
ing the energy shift. Consequently the so called quadratic S TARK effect E 2 is
observed!
The perturbation VE by the electric field (8.51) does not act on the spin in contrast
to the B field (8.6). The spin S and its projection MS thus remain conserved. This
facilitates the evaluation of the matrix elements
J M VE (r)| J M
(8.53)
= eE J M rC10 ( ) J M = eE r|
J M C10 ( )|J M
significantly. For later use we derive these now in detail. All quantum numbers re-
lated to the radial wave functions are summarized by while J M stands here for all
angular momentum quantum numbers. Depending on the strength of the interaction
as compared to FS interaction the states are again described most suitably in the
uncoupled |LML SMS
or in the coupled scheme |(LS)J MJ
, respectively. Here L,
S and J stand again for the total orbital angular momentum, the totals spin and the
total angular momentum of the system, respectively, while the E field is assumed
to act only on one active electron with the angular momentum . For the present
derivations we shall assume for simplicity L = . However, the following may eas-
ily be extended to coupled orbital angular momenta. Obviously, the matrix elements
(8.53) are essentially the same which we have already encountered in Chap. 4 in
the context of describing E1 transitions induced by linearly polarized light. We thus
obtain the same selection rules as in Sect. 4.4.
8.2 Atoms in an Electric Field 403
and the radial wave function Rn (r). Using (C.29) and (C.30) we obtain
+M L 1 L
L M VE | LM
= eErn n (1) 2L+L
2L + 1
M 0 M
5
L 5C1 L
(8.56)
L M VE | LM
= eErn n M M L L1 (1)M 2L + 1 (2L + 1)
L 1 L L 1 L
, (8.57)
M 0 M 0 0 0
assuming for (8.57) a pure one electron system L = and L = and making
use of the 3j symbols symmetries (B.37) and (B.49): Thus, the matrix element
is only non-zero for L = L 1. Hence, as already mentioned, L is no longer
a good quantum number, while the projection quantum number M = M is con-
served. In contrast to the magnetic field where the angular momentum (an axial
vector) explicitly is included in the perturbation, in the case of an electric field the
quantity z (derived from the polar vector r) does not act on the projection of the
angular momentum. Using (B.53) for the relevant 3j symbols we finally obtain:
(L + 1)2 M 2
(L + 1)M VE | LM
= eErn n (8.58)
(2L + 1)(2L + 3)
L2 M 2
(L 1)M VE | LM
= eErn n . (8.59)
(2L 1)(2L + 1)
The matrix elements only depend on M 2 and thus on the absolute value |M| of
the projection quantum number. This is a consequence of the fact that on inversion
(thereby changing +M to M) the sign of VE changes, while at the same time one
(and only one) of the spherical harmonics of the wave functions changes since they
have different parity due to L = L 1.
According to (8.58) and (8.59) the influence of the electric field decreases with
increasing |M| L. Thus, as we shall elaborate in Sect. 8.2.7, states with the highest
projection quantum number show the weakest S TARK effect.
404 8 Atoms in External Fields
holds. Explicitly from (B.53) and (B.55) this can simply be written as
L SJ M VE | LSJ M
= eErn n M M L L1 (8.62)
(J +1)2 M 2
for J = J + 1,
(2J +2)
(1)2M1 (2J +1)
(1) for J = J, and
2J
2J (J +1)(2J +2) M
J 2 M 2
2J for J = J 1.
Here too the projection quantum number is conserved (in this case the projection
of J ), and only the matrix elements with L = L 1 are non-zero. Even the rule that
the states with highest projection quantum number |M| are least disturbed remains
valid, since a quantitative evaluation of (8.62) shows that the opposite trend for
J = J contributes only very little.
8.2 Atoms in an Electric Field 405
W n 1
According to (8.57) and (8.60) quite generally a|VE |b
1 , i.e. the diagonal
term vanishes. With (8.52) we have
|zab |2 r 2 |a|C10 |b
|2
Wa Wa(0) = |eE|2 = |eE|2 ab
, (8.63)
Wa Wb Wa Wb
b=a b=a
and the change of energy depends on the square of the electric field: if the degener-
acy is already removed, a quadratic S TARK effect is observed. For effective one elec-
tron systems, |a
and |b
represent |n m
or |n sJ M
, depending on the coupling
case. Only states with equal m or M, respectively and with
= 1 interact (mix).
Only the absolute values |m | or |M| of the projection quantum numbers matter.
(a) the S TARK effect will always lower the terms since the atomic level spacing
typically decreases with increasing principle quantum number n as sketched in
1 Note that this depends on the precision of the measurement: for very weak electric fields and very
high precision even the H atom levels of equal n are already split due to FS interaction.
406 8 Atoms in External Fields
Fig. 8.11; thus the series in (8.63) contains always many more closely spaced
levels for which Wa Wb < 0 as compared to those levels where the reverse is
true; this is particularly pronounced for the ground state;
(b) higher lying levels are stronger influenced by the S TARK effect since Wa Wb
decreases with increasing principle quantum number n;
(c) within one level states with larger |m | are lowered less as already mentioned.
In the uncoupled case the latter statement is read directly from the matrix ele-
ments (8.58) and (8.59). It is valid, however, also in respect of |M| in the coupled
case as explicit evaluation of the matrix elements (8.62) shows. One may visual-
ize this with the help of Fig. 8.12 where the components of the matrix elements
2pq |C10 |2s
are illustrated. In physical terms positive and negative charges may be
displaced by the E field more easily for the |2pz
state along the z-axis than for the
|2px
or |2py
state. In the latter case the positive charge would have to be more or
less extracted from the negative charge cloud as illustrated in Fig. 8.13.
Thus, for the quadratic S TARK effect one expects a characteristic dependence of
the energies of different |M| states as a function of the applied electric field E as
sketched in Fig. 8.14 for the example of a 2 P3/2,1/2 doublet.
Corresponding to the splitting with |M| in emission or absorption spectra one
finds in an electric field only two polarization components and : these are lin-
early polarized (perpendicular to each other) since the electric field does not enforce
an orientation onto the system again in contrast to the situation in a magnetic field
where + (LHC) and (RHC) light is emitted at two different frequencies.
8.2 Atoms in an Electric Field 407
|m|=1/2
32P1/2
|m|=1/2
without field with field E
The situation is completely different if the external E field is the dominant interac-
tion which removes degeneracy. Thus, let us assume states of different parity are
degenerate without the external E field. This is the case for the H atom and H like
ions (as long as FS can be neglected) as well as for extremely high field strengths
where the splitting is negligible (e.g. in molecular bond formation between sev-
eral atoms) but also for highly excited RYDBERG states and high as we shall see
in Sect. 8.2.9.
For the 1s 2 S ground state of the H atom all matrix elements of VE vanish ac-
cording to (8.57) or (8.62) because of L L1 . The H atom ground state has no linear
S TARK effect. However, for the first excited state (with FS neglected) there are four
degenerate states, written in the real basis,
The states |2s0
, |2pz
are characterized by M = 0, while |2px
and |2py
have a
projection quantum number |M| = 1. According to (8.57) all diagonal matrix ele-
ments disappear, as well as those to different M and M . With (8.58) or (8.59) only
two matrix elements are non-zero:
1
2pz |VE |2s0
= 2s0|VE |2pz
= eEr2s2p . (8.65)
3
The radial matrix element (8.55) between the states |2s
and |2p
may be de-
rivedfrom the radial wave functions of the H atoms (Table 2.2) and is r2s2p =
(3 3/Z)a0 .
The H AMILTON matrix is thus
may be solved algebraically. Since only two states couple, this amounts to solving a
set of the linear equations for the coefficients cn ,m
(0)
W W 3eEa0 c2s0
= 0, (8.67)
3eEa0 W (0) W c2pz
With (8.67) the coefficients c2s0 and c2pz are derived (properly normalized)
(1) c2s0 = c2pz = 1/ 2 and (2) c2s0 = c2pz = 1/ 2
Hence, the dipole moment of the H atom in the state |2
and |2+
states is 3ea0
and 3ea0 , respectively.
8.2 Atoms in an Electric Field 409
+ =
(|2pz |2s) / 2 = |2
As an experimental example for the S TARK effect we shall discuss here highly ex-
cited atomic states (so called RYDBERG atoms). This is a wide field and still subject
of present research. We present just one particularly impressive pioneering exper-
iment of Z IMMERMAN et al. (1979). It is considered a benchmark experiment for
pertinent theoretical investigations (see e.g. M ENENDEZ et al. 2005) and demon-
strates a number of important aspects of the S TARK effect. The experimental setup
is relatively simple, as illustrated schematically in Fig. 8.17. A Li atomic beam (see
footnote 24 in Chap. 1) is excited by a resonant multi-photon process with three
laser frequencies: the steps are 2s 2p (671 nm), 2p 3s (813 nm), and finally
3s 15p (626 nm). At a fixed DC S TARK field, the latter laser is tuned through
ca. 100 cm1 . Excitation of the RYDBERG states is detected by field ionization in a
pulsed electric field (HV pulser), applied shortly after the laser pulse.
The spectra, observed for a number of electric field strengths, are reproduced in
Fig. 8.18 (black, vertical traces with horizontal excitation lines). The red lines in
Fig. 8.18 give the theoretically expected energy dependence of the relevant states
as a function of the electric field strength. The observed absorption lines follow the
theoretical understanding very impressively. There are two sets of states, for M = 0
and |M| = 1, which are excited by light polarized parallel and perpendicular to the
E field ( and light), respectively. In the previously discussed case of the 2s 2p
410 8 Atoms in External Fields
delayed trigger
HV pulser
DC
atomic beam
laser beams
collecting
ion detector
electronics
Fig. 8.17 Experimental setup according to Z IMMERMAN et al. (1979) for RYDBERG spectroscopy
in an electric field. Two laser systems tuned to the transitions 2s 2p and 2p 3s, respectively,
excite the Li atoms into the 3s state. This is then further excited by a very narrow band, tuneable
laser into the n = 15 region. Shortly after the excitation, in addition to the S TARK field a pulsed,
high electric field (HV power supply) is applied which ionizes all RYDBERG atoms. Detected is
the ion signal
16s
470
n =15 15
15p
500 15p
15s
530
Fig. 8.18 RYDBERG levels n 15 in an electric field as a function of the field strength, for atomic
Li from Z IMMERMAN et al. (1979). (a) |M| = 0, (b) |M| = 1. Black spectra represent the experi-
mental excitation probabilities, red lines give the calculated term positions
states this would correspond to excitation of the |2
, |2+
states (with light) and
the |2px
and |2py
states (with light).
In the present case, for n = 15, practically all levels with 2 14 = n 1
are degenerate without electric field (a total of 13 levels). Thus, all these levels split
with increasing electric field by linear S TARK effect. For the 15p state and very
weak electric field (300 V m1 ) one recognizes that the degeneracy is already
removed and the energy decreases initially in a quadratic fashion with increasing
field. As the electric field gets larger VE supersedes the initial splitting and the
linear S TARK effect takes over also for the 15p state. The 15s state behaves analo-
8.2 Atoms in an Electric Field 411
gously, however, due to the larger initial splitting the linear region is reached only
at about 2000 V m1 . There, however, already numerous avoided crossings with the
neighbouring levels n = 14 are encountered, as evident from Fig. 8.18 left (M = 0).
A similar behaviour is seen at higher energies for the 16s levels.
It is also interesting to compare the |M| = 0 and |M| = 1 levels. Even though
the general pattern is very similar in both cases, one recognizes very clearly that the
interaction and hence the repulsion in the vicinity of avoided crossings for terms
of equal symmetry is much larger for |M| = 0 in comparison with |M| = 1: this
clearly reflects the magnitude of the corresponding matrix elements according to
(8.58) and (8.59), respectively.
8.2.10 Polarizability
We have seen that the S TARK effect is quadratic when the degeneracy is already
removed among interacting states of different parity. This is the case for almost all
atoms in their ground and lower excited states at moderate electric fields. It does not
only change the energy but also the wave functions, and thus the electronic charge
distribution around the nucleus changes. This is called polarization. Quantitatively,
the electric field induces an electric dipole
D el = E E (8.71)
|zab
|2 2 |a|C |b
|2
rab 10
= 2e2 = 2e2 . (8.74)
Wb Wa Wb Wa
b=a b=a
412 8 Atoms in External Fields
After averaging over all initial substates states with projection quantum number ma ,
inserting the transition frequencies ba = Wba / with Wba = Wb Wa , and using
(opt)
the oscillator strength fba as defined in (5.27), one may write the polarizability E
in compact form as:2
e2 fba
(opt)
E = 2
. (8.75)
me ba
b=a
We note in passing that the above discussion refers to the so called induced po-
larization. The situation is different, if the medium consists of particles with a per-
manent dipole moment (e.g. water molecules). In that case the permanent electric
dipole moments will tend to orient along the electric field. Similar to the discussion
for magnetic moments in the context of paramagnetism, one has to consider the en-
ergetics and statistics involved in this orientation processes to obtain the orientation
polarization.
It is instructive to compare this quantum mechanically exact calculation with the
classical model (J.J. T HOMSON) considering the atom as a harmonic oscillator of
eigenfrequency 0 . The driving mechanical force Fm = me 02 z must be compen-
sated by the electric force Fel = eE in the field E:
Fm = me 02 z = Fel = eE.
e2
D = ez = E,
me 02
e2 1
E = . (8.76)
me 02
Comparison with (8.75) shows that the classical formula corresponds to an atom
with only one single transition frequency ba = 0 . In contrast, the quantum me-
chanical result distributes the ability of the electrons to oscillate over all transition
2 Note that the SI unit of the polarization is [E ] = A2 s4 kg1 = C m2 V1 . In a.u. the polarization
(au)
isE = E /(40 a03 ). Often the esu system is still used in this context (see also Appendix A.3)
with
(esu)
(esu) E E 106
E = or 3
= E ,
40 cm 40 m3
(esu)
indicating that E is usually given in cm3 . This scaling allows a direct comparison of the polar-
izability with the volume of the atoms that are polarized.
8.2 Atoms in an Electric Field 413
frequencies. The T HOMAS -R EICHE -K UHN sum rule (5.28), which for a one elec-
(opt)
tron system reads fba = 1, may be seen as the mathematical manifestation of
this picture.
The classical formula (8.76) allows us to obtain a rough estimate for E in the
ground state of the H atom. Setting 0 = Eh /2 leads to
Generalizing this with the aid of (8.75) one finds the (static) polarizability to be
proportional to the third power of the extension a of the polarized object. This fully
corresponds to the findings of classical electrodynamics.
8.2.11 Susceptibility
In analogy to the magnetic properties Sect. 8.1.7 one derives the macroscopic di-
electric susceptibility from the microscopic quantity polarizability E . Electric
displacement D and electric field E are related by
D = 0 E + 0 E = r 0 E = 0 E + P, (8.78)
P = (r 1)0 E = 0 E, (8.79)
P = N E E, (8.80)
with N being again the particle density (number of atoms per volume). Thus, the
dielectric susceptibility is
N E E
= (r 1) = or per atom a = . (8.81)
0 0
To obtain some feeling for the order of magnitude of we make an estimate for
our favourite Na atoms where nearly all oscillator strength is concentrated in the
main transition 3s 3p ( 3p3s = 1.696 104 cm1 ). Setting f3p3s 1 we obtain
with E from (8.75) the static dielectric susceptibility per atom
e2
a = = 3.12 1028 m3 . (8.82)
0 me (2c 3p3s )2
3 Note that this differs slightly from the scheme in (8.39) for the magnetization M.
414 8 Atoms in External Fields
This may be compared to the atomic volume. With rNa 3a0 it is Vat 1.7
1029 m3 or roughly a /4 . This is also what classical theory predicts.
Section summary
The interaction of atoms with a static electric field E is described by the dipole
energy er E = erC10 ( )E, assuming the field to be parallel to z. This op-
erator breaks the symmetry of the Hamiltonian, so that the orbital angular
momentum L is no longer a good quantum number, in contrast to M. Inter-
action matrix elements are only finite between states with L and L 1. They
are independent of the electron spin.
Consequently, the change of the energy of LM levels in an electric field
(called S TARK effect) depends on M 2 . If L degeneracy for the energy lev-
els is already removed, the S TARK effect is quadratic in E and negative. For
L > 0 the levels split into L + 1 sublevels while no S TARK effect occurs for
isolated s states.
If L degeneracy is not yet removed, or the splitting between different L levels
is small compared to the dipole energy, the dipole operator mixes different L
states, but M is conserved. The splitting of the new states is then linear in E
(for not too high fields).
Typically, the S TARK splitting is a small effect in standard spectroscopy. How-
ever, since it depends quadratically on the radial matrix elements, it increases
with principle quantum number n4 and becomes substantial for high RYD -
BERG states.
The polarizability of atoms is following the quadratic S TARK effect also
derived from the same interaction. According to (8.75) it is obtained as sum
over all relevant oscillator strengths divided by the respective squared transi-
tion frequencies. It is related to the relative dielectric permittivity r and to
the dielectric susceptibility by (8.81).
We shall now address the question how two atoms or molecules (or their ions) inter-
act as they approach each other. At this point we are interested in distances R large
enough so that no chemical bonds can be formed yet, while on the other hand the
respective charge distributions of the particles begin to influence each other. This
range of distances is so to say the precursor for the formation of macroscopic
matter and plays a key role in plasma physics, scattering physics, kinetic gas theory
etc. A detailed derivation for the relevant interaction potentials has been given by
B UCKINGHAM (1967). We summarize here the key results and make some plausi-
bility considerations.
8.3 Long Range Interaction Potentials 415
q1 q2 e 2
V (R) = R 1 . (8.83)
40 R
For later use we also note without proof the corresponding electric field that is ob-
tained as gradient of (8.84):
1
E= 3
D 3eR (D eR ) . (8.85)
40 R
(au) (esu)
with E = E /(40 a03 ) = E /a03 (see footnote 2). Note that this so called
polarization potential is always attractive. For an estimate of the order of magni-
tude we consider an H atom in the field of a singly charged ion and use (8.77)
as an approximation for the polarizability. In this case V (R)/Eh = 2(R/a0 )4 ,
so that in a distance of one B OHR radius the polarization potential would be
just 2 a.u. Of course the formula holds strictly only for significantly larger dis-
tances.
Quadrupole Quadrupole: R 5 Important examples for this case are e.g. the
interaction of two homonuclear, diatomic molecules as well as the interaction of
one excited atom in a px state interacting with such a molecule. The geometry is
shown in Fig. 8.22. Without entering into further details, we record the dependence
on the distance:
F (1 , 2 )
V (R) = R 5 . (8.89)
R5
+ R +
Permanent Dipole Induced Dipole: R 6 An example for this case is the in-
teraction of a heterogeneous molecule with a neutral atom as indicated in Fig. 8.23.
This case too is easily derived since the induced dipole is always aligned parallel to
the field of the permanent one. In analogy to the case monopole induced dipole
one obtains the potential as
E 2
V (R) = D ind dE = E E dE = E(R)
2
8.3 Long Range Interaction Potentials 417
E 1 2
= D 3e R (D e R ) (8.90)
2 40 R 3
E D 2
= 2 6
1 + 3 cos2 E R 6 ,
2(40 ) R
where the second line follows with (8.85). In the limiting cases, D R and D R,
the angular dependence in the bracket gives a factor of 4 and 1, respectively. In any
case one finds an attractive, non-isotropic potential proportional to R 6 .
Induced Dipole Induced Dipole: R 6 The famous VAN DER WAALS potential
occurs in the interaction of all neutral atoms or molecules depending on their polar-
izability. It always ensures an attractive interaction at large distances as far as not
another of the above discussed cases dominates.
One may visualize this interaction to arise from spontaneous charge fluctuations
that lead to dipole formation in one of the atoms. This initial dipole in turn induces
by polarization a dipole in the other atom and so on, until a stable situation as de-
picted in Fig. 8.24 is reached.
The resulting dipole dipole interaction will lead to an expression similar to
(8.90) except for the fact that there is no preferred direction and one has to average
over all alignment angles . In the quantum mechanical calculation one expands the
electrostatic interaction of all charges involved (valence electrons and ion core) for
large R into a 1/R N series. Summing over all electron coordinates r A (at atom A)
and r B (at atom B) one encounters typical dipole terms H di (er A )(er B )/R 3 that
correspond to (8.86).
One has to treat these in 2nd order perturbation theory (in 1st order the dipole
terms of a neutral charge distributions disappear). The interaction energy in 2nd
order is the sought-after polarization potential:
|a|H
di |b
|2
V (R) = . (8.91)
Wa Wb
b=a
Here a specifies again the initial state (usually the ground state) and b are all inter-
mediate states. Here too, as discussed in Sect. 8.2.7, for the vast majority of relevant
terms Wa Wb < 0 holds, and the sum is negative. Comparing this with (8.74) one
418 8 Atoms in External Fields
sees that this essentially amounts to compute the polarizability of both atoms. Since
di R 3 and V (R) according to (8.91) depends quadratic on the matrix elements
H
di , the overall interaction potential is attractive and proportional to R 6 :
of H
C
VAN DER WAALS potential V (R) = . (8.92)
R6
One may evaluate (8.91) approximatively (see B UCKINGHAM 1967), and finds
3 WA WB A B
V (R) = . (8.93)
2(40 )2 WA + WB R 6
The key parameters are the polarizabilities, A and B , together with average bind-
ing energies, WA and WB , of particle A and B, respectively one typically uses the
respective ionization energies.
The force derived from VAN DER WAALS interaction is often called dispersion
force as the polarizability E () depends on the frequency of the external electric
field.
Section summary
The long range interaction potentials between ions, atoms and molecules play
an important role in spectroscopy as well as in scattering physics. They are
determined by the structure of the interacting particles and their plausibility.
For large R the interaction potential is 1/R for monopole-monopole inter-
action, 1/R 2 for monopole-permanent dipole, 1/R 3 for permanent
dipole-permanent dipole, 1/R 4 for monopole-induced dipole, 1/R 5
for quadrupole-quadrupole, and 1/R 6 for dipole-induced dipole as well
as induced dipole-induced dipole.
that for a very short time
t a virtual photon is emitted or absorbed, as we have
discussed and visualized by F EYNMAN diagrams in Sect. 6.5.6. Such transient ex-
citation is possible even for non-resonant irradiation since the energy W of states is
defined only within the limits of the uncertainty relation, so that
W > /
t holds.
On a very short time scale the levels thus get completely blurred and may be quasi
excited. One often talks in this context somewhat misguiding about virtual
intermediate levels.
Thus, we expect a shift of the atomic levels by a modified quadratic S TARK ef-
fect. A clean derivation of the dynamic polarizability warrants some quite serious ef-
forts. We reduce these here to a heuristic consideration by replacing the total energy
of the system Wa prior to absorption or emission of a photon (angular frequency )
by
Wa Wa + n
and the energies Wb of the intermediate states by
Wb Wb + (n 1).
1 |zba |2 |zba |2
E () = e 2
+ .
ga Wb Wa Wb Wa +
b=a,ma
The first term stands for absorption, the second for emission. As a last step we
replace again Wb Wa = ba :
1 |zba |2 / |zba |2 /
E () = e2 + (8.94)
ga ba ba +
b=a,ma
(opt)
We have again introduced the optical oscillator strength fba according to (5.27).
In the static limit 0, expression (8.75) for the static polarizability is recovered.
Conversely, in the limit of very high frequencies we make use of the T HOMAS -
(opt)
R EICHE -K UHN sum rule (5.28) for fba (considering only one active electron) and
obtain
e2
E () for ba , (8.96)
me 2
where the minus sign indicates that the induced dipoles have opposite direction as
the polarizing electromagnetic field.
420 8 Atoms in External Fields
Finally, the shift of the atomic energy levels is derived from (8.73), now using
the dynamic polarizability (8.94):
E () 2 E () 2 E () I
Wa Wa(0) = E = E0 =
2 4 2 0 c
e2 I fba
(opt)
Wa Wa(0) = . (8.97)
20 cme 2 2
b=a ba
We have averaged here over the square E(t)
2 = E0 cos(t)
2 of the electric field
strength and introduced the intensity I of the electromagnetic radiation, using (4.2).
From elementary optics we know that the index of refraction n and the relative
dielectric permittivity relate as
n = r (8.98)
(assuming for the relative magnetic permeability = 1 in very good approxima-
tion). Hence, with the dynamic polarizability (8.95) and (8.81), the index of refrac-
tion n is obtained from
N e2 fba
(opt)
N E ()
n2 1 = r 1 = = = 2 2
. (8.99)
0 0 me
b=a ba
The resonance frequencies for typical optical materials are usually far away from
the visible region and dispersion dn/d < 0 (so called normal dispersion). One may
express (8.99) in terms of the wavelength, = 2c/ and write it as so called
S ELLMEIER equation, which is often used in optics:
Bi 2
n2 = 1 + . (8.100)
2 Ci
i
Usually one treats this as an empirical relation with typically three pairs of S ELL -
MEIER coefficients Bi , Ci obtained from experiment (a comprehensive set of data
for optical materials is given by P OLYANSKIY 2012).
For thin media (gases) n 1, so that n2 1 = (n 1)(n + 1) 2(n 1) and we
obtain as a good approximation
N e2 fba
(opt)
n1+ . (8.101)
20 me 2 2
b=a ba
an important modification: in dense media the atoms experience an electric field, the
so called L ORENTZ field, which is already modified as compared to the field in the
surrounding vacuum. According to C LAUSIUS -M OSSOTI this leads to a modified
formula:
N e2 fba
(opt)
n2 1
= . (8.102)
n2 + 2 30 me 2 2
b=a ba
So far we have tacitly assumed that the frequency of the radiation is far away from
resonances, ba = . In order to include also frequency regions close to resonance
we have to consider damping, i.e. we must account for the finite lifetimes b =
1/b of excited states. As in Sect. 5.1.1, we simply introduce a complex energy (or
transition frequency) ba ba ib /2 so that
1 1
.
ba ba ib /2
Inserting this into (8.94) and (8.99) leads to a complex index of refraction
nc = n + i. (8.103)
For thin media, (8.101) is now replaced by real and imaginary parts:
(opt)
N e2 fba ba ba +
n=1+ + (8.104)
40 me ba ( )2 + b2 ( + )2 + b
2
b=a ba 4 ba 4
N e2 fba
(opt)
b b
= + . (8.105)
40 me 2ba ( )2 + b2 ( + ) 2 + b
2
b=a ba 4 ba 4
Far from all resonances, i.e. for negligible b , one recovers from (8.104) the original
relation (8.101). In the general case the wave vector is now also complex and one
has to replace k (2/0 )nc ek = (2/0 )(n + i)ek , with ek being the unit vector
in the direction of propagation. Thus, the electric field (4.1) in a medium becomes
i
E(r, t) = E0 eei(2nz/0 t) + e ei(2nz/0 t) e2z/0 (8.106)
2
where we have assumed the light to propagate along the z-axis. The vacuum wave-
length 0 is shortened to = 0 /n, while implies exponential damping of the
wave (in physics texts is often called extinction coefficient). The intensity of the
radiation I |E(r, t)|2 decreases with distance as
I = I0 exp(4z/0 ). (8.107)
422 8 Atoms in External Fields
Fig. 8.25 Schematic of the complex index of refraction nc as a function of the incident radiation
frequency ; (a) real part, n, and (b) imaginary part, . The absorption lines shown in the
extinction coefficient clearly correspond to regions of anomalous dispersion in the index of
refraction n
Comparing this to the L AMBERT-B EER law (4.21) we see that we have found a
microscopic interpretation of the absorption coefficient introduced there.
In the vicinity of an isolated resonance, ba , only one term in the sums
(8.104) and (8.105) dominates and the real index of refraction is
(opt)
N e2 fba ( ba )/b
n=1+ , (8.108)
40 me ba b [( ba )/b ]2 + 1/4
(opt)
4 N e2 fba
= = . (8.109)
ba 40 cme [( ba )/b ]2 + 1/4
As illustrated in Fig. 8.25, the real part n (8.104) of the index of refraction
may have several regions of normal and anomalous dispersion, while the imagi-
nary part (8.105) represents absorption, with several L ORENTZ profiles of dif-
ferent strength. Clearly, these absorption lines correspond to regions of anomalous
dispersion in the real part n().
with 0 and = 0 /n being again the wavelengths in vacuo and in the medium,
respectively, while k() is the real (propagation) and () the imaginary part (ab-
sorption) of the wave vector in the medium, with n and according to (8.104)
and (8.105). The positions of constant phase = t kz are given by z =
(/k)t /k and propagate with the
dz c
phase velocity vp () = = = , (8.111)
dt =const k() n()
where
)
E() = E0 (t)eit dt, (8.114)
When describing also the spatial dependence, we have to account for dispersion:
as we have seen in the previous section, in a medium neither n nor vp are constant
4 Weuse here the notation of Appendix I.2 where the relations to the full description are given
which in the present case would just be space consuming without leading to further insight.
424 8 Atoms in External Fields
with , and the wave vector k() is not simply proportional to . However, as only
a small interval of frequencies around c contributes, we may expand k around
c dk c
kc = : k() = kc +
= + , (8.116)
vp (c ) d c vp (c ) c vg
1 ) i(t vzg )
= ei(c tkc z)z E(
)e d(
) .
2
In the last step we have exploited the fact that d(
) = d. Comparing the above
expression in square brackets with (8.115) we realize that it is just the field ampli-
tude shifted in time by z/vg . Hence, we may write the electric field of the wave at
t kz/:
z i(kc zc t)z
E0+ (t kz/) = E0 t e . (8.117)
vg
We clearly recognize that the carrier wave with carrier frequency c propagates with
the wave vector kc just as (8.110), while the E0 (t z/vg ) describes the form of the
wave-packet. If E0 (t) has its maximum at t = 0, as determined by an appropriate
)
choice of E(), at a different position z in space one finds this maximum E0 (0)
obviously for z = vg t. Thus, characteristic structures of the wave-packet travel with
the so called
d c c
group velocity vg = = = . (8.118)
dk n + d
dn ng
Here we have used d/dk = 1/(dk/d) and differentiated k = n/c from (8.111)
in respect of . In analogy to n = c/vp (which may be called phase index), one
introduces a
c dn
group index ng = =n+ . (8.119)
vg d
In the range of visible light for transparent materials as a general rule one finds
n > 1 and dn/d > 0 (normal dispersion), so that typically vg < vp < c: light pulses
propagate at velocities smaller than in vacuum. So far the canonical discussion on
phase and group velocity.
8.4 Atoms in an Oscillating Electromagnetic Field 425
The discussion about slow and fast light starts with the fact that under cer-
tain conditions dispersion may be negative as already illustrated in Fig. 8.25 for
the standard dispersion behaviour of conventional materials close to a resonance.5
A look at the formula for the group index (8.119) shows, that one may then indeed
find n to be smaller than 1 or even negative which in principle implies the possi-
bility of group velocities larger that the speed of light in vacuum vg > c. One speaks
about superluminal propagation phenomena. And if n < 0, and hence vg < 0, this
would even lead to light travelling backward before it has arrived. In contrast, in
other regions of the spectrum vg may be extremely small so that light crawls so to
say, or is being stopped. And indeed, such phenomena have been observed during
the past years in fancy experimental setups (see e.g. B OYD and G AUTHIER 2002).
To caution the reader quite in the beginning: E INSTEINs paradigm according to
which information can never be communicated faster that the speed of light in vac-
uum is by no means violated as a consequence of group velocities vg > c.
We want to exemplify these phenomena quantitatively on the basis of a simple
case study. Let us consider Na vapour as medium through which light is supposed
to propagate. For this consideration Na represents to a good approximation a two
level system if we allow only light frequencies in the vicinity of the 3 2 S 3 2 P
transition (Na D-line). Let us assume a target density of N = 2 1013 cm3 which
may be achieved in a gas cell without problems. The other relevant parameters are
the oscillator strength for the transition, fNaD = 0.98, and the decay probability for
the excited state A = = 6.15 107 s1 (corresponding to a lifetime = 16.2 ns),
0 589 nm (in vacuo) and 0 3.2 1015 s1 . Hence, the prefactor in (8.108) is
Ne2 fNaD /(40 me 0 ) = 0.08 and 0 / = 5.2 107 .
In Fig. 8.26 the situation close to the resonance line is depicted to scale as a
function of 0 in units of the linewidth. Figure 8.26(a) shows the real index of
refraction (phase index), and Fig. 8.26(b) the absorption coefficient. One recognizes
that n undergoes the typical change between normal and anomalous dispersion, and
finally back to normal with values close to 1 for which (8.108) is valid. Absorption,
on the other hand, is quite substantial contrasting the very moderate changes of the
index of refraction. Figure 8.26(c) reports the group index ng according to (8.119). It
shows dramatic changes in the resonance region. Correspondingly, the group veloc-
ity (8.108) also varies very fast across the resonance as shown in Fig. 8.26(d). When
approaching a resonance vg becomes very small, increases again, passes through
zero for 0 = /2 accompanied by two singularities in each case, and as-
sumes a very small negative value exactly on resonance, 0 = 0.
Generally speaking, experiments trying to observe the expected phenomena are
extremely difficult, the main problem being that the most interesting effects, relying
on |vg | c or |vg | > c, or even vg < 0, are expected where absorption is particularly
high. Thus, typically instead of passive two level systems one investigates three and
more levels schemes which also act as amplifiers when pumped with a suitable laser.
Beyond atomic gases, today specially designed solid state materials are employed,
5 Today, in addition a variety of artificial, specially designed smart, meta and nano materials
exist, as well as photonic fibres, with extended regions of quite unusual optical properties.
426 8 Atoms in External Fields
1.04 -4 -2 2 4
( 0) / - 106
1.0 ng - 2106
-4 -2 2 4
- 3106
0.96
(d) - 4106
0.92 vg /c = 1/ng
210- 5 1.0
(b) / 1000 cm-1
15
110- 5
10 0.5
5 ( 0) / ( 0) /
0
-4 -2
- 410- 6
-4 -2 0 2 4 0 2000 4000
Fig. 8.26 (a) Absorption and (b) index of refraction (showing changing dispersion) in the
vicinity of a resonance line. The data is shown to scale for the example of the Na D line at
N = 2 1013 cm3 . For comparison (c) the group index and (d) the group velocity are shown;
note the different scales in the left and right half of (d), the overall profile being symmetric around
0 = 0
such as doped light transmitting fibres, photonic crystals, and assemblies of quantum
dots.
Propagation of light pulses faster than light as well as slowing down of such
pulses has been observed. A detailed analysis allows to understand all observations
reported in the literature from the basic principles of optics even if they may ap-
pear as mysterious superluminal effects: information can be transported by such
schemes, but not faster than the speed of light in vacuum. Genuine discontinuities
within a wave always propagate with velocities less than c, while precursors of the
wave may reach the observer much earlier. Without proof we mention that energy
transport in an electromagnetic wave is always characterized by a propagation ve-
locity
2n
c,cf = (8.120)
n2 + 1
which is definitively less than c, the speed of light in vacuum.
While such superluminal effects may just be a fascinating theme of fundamental
research, the inverse case, light propagating at extremely low velocities, vg c, al-
beit perhaps less spectacular, could possibly be of much more practical relevance in
data transmission technology: One may envisage special delay lines, data switches
or optical storage devices that profit from these phenomena. One could e.g. slow
down a data stream temporarily by a control laser if the traffic somewhere down
the transmission lines is jammed and later on release it again on demand. Fig-
ure 8.26(d) shows in principle, how low the group velocity can be for close to a
resonance line.
8.4 Atoms in an Oscillating Electromagnetic Field 427
With the tools developed in the previous sections we are now also able to treat the
important subject of elastic scattering of light. Light cannot only be absorbed, excit-
ing atoms or molecules in the absorbing medium as presented in detail in Chap. 4.
It can also be scattered elastically.
We first note that the very intense light scattering one observes from dust and
smoke particles or from small water droplets in vapour or fog (e.g. in the head-
lights of cars, in discos and laser shows) will not be subject to this section. This
kind of elastic light scattering is called M IE scattering. It dominates if the scattering
particles have dimensions on the order of the wavelength or larger. The angular dis-
tribution of M IE scattering depends on the size and form of the scattering particles
and has to be calculated directly from M AXWELLs equations with corresponding
boundary conditions. Such computations may become quite complicated. For the
identification and analysis of nano and micro particles M IE scattering is an often
used important tool. Its theoretical treatment leads, however, beyond the scope of
this textbook (a very detailed treatment is e.g. found in B ORN and W OLF 2006).
Here we shall discuss light scattering from atoms (and molecules) which es-
sentially arises from atoms being polarized by the electromagnetic radiation. More
precisely, the oscillating electric field E(t) of the wave induces a time dependent
dipole moment D(t) = E E(t) in the atom, with E being the polarizability of the
428 8 Atoms in External Fields
atoms discussed above. These dipoles in turn radiate at the frequency of the irradi-
ating field. One may calculate the emission classically from (4.33). The power P
emitted per solid angle is
dP
|D(t)| 2 E2 |E|2 4
= sin2
= sin2 (8.121)
d (4)2 0 c3 (4)2 0 c3
and depends on the angle between the polarization of the incident radiation and
the direction of the detected radiation. The latter is often replaced by the angle
between the polarizations of incident and scattered radiation, with sin2 = cos2 .
The differential cross section for this so called R AYLEIGH scattering of polarized
light is obtained by dividing dP /d by the incident light intensity I = c0 |E|2 =
c0 E02 /2:
2 4
E
dR 3
= 2
cos2 = R cos2 . (8.122)
d 2
16 0 c 4 8
Integration over all azimuthal and polar angles gives a factor 8/3, so that the inte-
gral R AYLEIGH cross section is given by
2 4
E 2
8 3 E
R = = . (8.123)
602 c4 302 4
Equation (8.122) represents again the typical doughnut like radiation characteristic
which we have already encountered with resonant fluorescence in Sect. 4.5. One
may observe R AYLEIGH scattering quite conveniently from a laser beam propagat-
ing in air. Depending on the laser intensity it can be seen in a darkened laboratory
even in completely dry and dust free air, and one easily verifies that the emission
parallel to the direction of linear polarization of the scattered light vanishes com-
pletely (i.e. perpendicular to the laser beam in one particular azimuthal direction).
For unpolarized light one has to average (8.122) over the two directions of po-
larization and obtains
dR 3
= R 1 + cos2 , (8.124)
d 16
where is now the angle between incident and scattered light direction.
For low frequencies , in the IR and often even in the VIS spectral region, E
is essentially independent of incident frequency and corresponds to the static value
(8.75). Thus, if we define a mean oscillator frequency 0 , characteristic for each
atom, by
1 f (opt)
2
= ba
2
, (8.125)
0 b=a ba
8.4 Atoms in an Oscillating Electromagnetic Field 429
4
R = e . (8.126)
04
This is formally identical to the classical relation already found by Lord R AYLEIGH,
who considered 0 to be the eigenfrequency of the atomic electron.
The proportionality R 4 in (8.123) gives a very clear answer to the often
posed question why is the sky blue?: blue light is scattered much more efficiently
than red light essentially by molecular oxygen and nitrogen being the main con-
stituents of air. Their absorption bands are in the VUV region and do not influence
the visible spectrum. The sky, which we register as elastically scattered light from
the sun, thus appears blue. Conversely, this also explains the red colour of the sun
rising and setting.
Note, however, that for higher photon energies, in contrast to the classical low
energy limit (8.126), the elastic scattering cross section (8.123) shows pronounced
structures over a wide range of wavelengths as expected from the polarizability
(8.94). In the vicinity of resonance frequencies particularly intensive scattering phe-
nomena are observed even where absorption is still negligible.
For high enough photon energies, W = 0 , the polarizability simply be-
comes E = e2 /(me 2 ) according to (8.96), with 0 being representative for the
binding energy of any relevant electron in the atom. Then the integral cross section
in (8.122)(8.124) approaches the so called T HOMSON cross section (measured in
units of barn, see Appendix A.2)
e4 8 4 2 8 2
R e = = a0 = r = 0.665 b, (8.127)
ab 2 2
60 me c 4 3 3 e
de
so that for polarized light = re2 sin2 . (8.128)
d
The electrons thus behave quasi like free electrons for which this cross section was
first derived by J.J. T HOMSON on a fully classical basis. Here is the fine structure
constant, a0 the atomic units of length, and re the classical electron radius (see
Appendix A). Note that (8.127) is a rather suggestive relation, but we recall that re
has nothing to do with a radius of the electron the latter being as point-like as we
can measure today.
So far, in all above discussion we have neither considered the particle properties
of the photon nor used special relativity for treating its motion. Both will become im-
portant at high photon energies W , i.e. when = W/me c2 is no longer very small.
For really free electrons energy and momentum conservation must be satisfied si-
multaneously, so that the electron will accept and the photon will loose some energy:
C OMPTON scattering occurs, which we have introduced briefly in Sect. 1.4.2. We
rewrite (1.75) as
W 1
= (8.129)
W 1 + (1 cos )
430 8 Atoms in External Fields
where W = and W = are the incident and scattered photon energy, re-
spectively, while is the photon scattering angle in respect of the incoming light
beam. Without derivation we communicate here the (relativistically correct) K LEIN -
N ISHINA formula for the differential C OMPTON scattering cross section:
dC re2 W 2 W W
= + 2 + 4 cos .
2
(8.130)
d 4 W W W
Note, that in the low energy, non-relativistic limit = W/me c2 1 the C OMPTON
shift vanishes according (8.129), i.e. W /W 1, and the K LEIN -N ISHINA formula
(8.130) reproduces the differential T HOMSON cross section (8.128) for elastic pho-
ton scattering from a free electron.
Still, the treatment of elastic light scattering from atoms (or molecules), i.e. from
bound electrons is not yet complete. First, up to now we have treated only one single
active electron. In the long wavelengths range and even in the region of atomic
resonances this can (in principle) be cured by simply summing (8.125) or (8.95)
over a sufficient number of absorption frequencies and thereby accounting for all
Ne electrons of a system. Nevertheless, the scattering process will be dominated
by only a few resonance transitions that are closest to the frequency of the incident
radiation.
At higher energies, W = ba , specifically for the scattering of X- and
gamma-rays where the wavelength becomes comparable or even smaller that atomic
dimensions, and energy and momentum conservation has to be considered, the sit-
uation is much more complicated (a detailed discussion is found e.g. in K ANE et al.
1986). Most commonly, the form factor approximation is used: one sums explicitly
over all electrons, thereby accounting for their probability density, and superposes
the scattered radiation appropriately. Two cases are distinguished (see e.g. H ANSON
1986, who also describes explicitely how the polarization angle is converted into
scattering angle):
dR r2
= e F 2 (q, Z) 1 + cos2 , (8.131)
d 2
where F(q, Z) is the atomic form factor as defined in (1.97), here emphasizing
its dependence on atomic number Z.
2. C OMPTON scattering from atoms, called incoherent scattering, is described in a
analogue manner, applying the K LEIN -N ISHINA formula (8.130) but now adding
cross sections for photons of different final energy incoherently. As detailed e.g.
8.4 Atoms in an Oscillating Electromagnetic Field 431
by H UBBELL et al. (1975) one generalizes the definition of atomic form factors
to include excited states, with |a
describing the ground and |b
any excited state:
Z
F (b) (q, Z) = b| exp(iq r i )|a
so that F (a) (q, Z) = F(q, Z). (8.132)
i=1
daC r2 W W W
= e S(q, Z) + sin 2
. (8.134)
d 2 W W W
State-of-the-art numerical values of form factors for all atoms are found e.g. at
C HANTLER et al. (2005) and B ERGER et al. (2010), including scattering angle in-
tegrated coherent and incoherent scattering cross sections, photoionization data etc.
One finds that the integrated cross sections decrease with increasing photon energy
for both, coherent and incoherent light scattering.
We finally mention T HOMSON scattering from a relativistic electron beam as an
interesting topic with potential applications for the generation of short pulsed X-ray
radiation. We shall briefly discuss this in Sect. 10.6.2.
Section summary
In an oscillating electromagnetic field (angular frequency ) the energy of
atomic energy levels also changes (so called dynamic S TARK effect). Due to
the quadratic nature of the interaction the effect is obtained by simply replac-
2 with 2 2 ( denotes the relevant excitation frequencies). The
ing ba ba ba
atomic polarizability E () in the field is obtained by the same substitution
from its static limit E .
The index of refraction n = r , in a medium of density N , is obtained from
the atomic polarizability by n2 = 1 + N E ()/0 .
In spectral regions with dn/d > 0 we speak about normal dispersion, while
dn/d < 0 corresponds to anomalous dispersion. The latter situation is as-
sociated with absorption. Introduction of the respective damping constants
(linewidths of the transitions) leads to a complex index of refraction nc =
n + i.
We distinguish phase and group velocity, vp = /k = c/n and vg = d/dk =
c/(n + dn/d), respectively. Typically vg < vp < c. Near a resonance vg is
very small. However, in the region of anomalous dispersion it rapidly changes
and we can find vg > c or even vg < 0. Still, information transport can only
occur with velocities c.
432 8 Atoms in External Fields
Modern developments in ultrafast laser science open new dimensions of light matter
interaction and are the basis for an active, cutting edge field of research (a glimpse
of the kind of processes studied today may be obtained e.g. from H ICKSTEIN et al.
2012). The radiation intensity I = W/(A
t) in a focussed laser pulse scales with
the total energy W of the pulse, its focal area A, and pulse duration
t. Thus, the
intensity of a 10 fs laser pulse is e.g. 106 times higher than that of a 10 ns pulse
which contains the same energy W . As discussed in Sect. 8.2.2 and exemplified in
Table 8.4, gigantic electric field strengths can be generated in this manner, presently
surpassing the field that an electron experiences on the first B OHR orbit in the H
atom by two orders of magnitude. And the experimental limits are being pushed
further.
Atoms and molecules exposed to such extreme conditions, react with a wealth of
astonishing new phenomena warranting also theoretical approaches quite different
from that outlined in Sect. 8.2.
dv
me = eE0 cos t,
dt
and obtain velocity and kinetic energy of the electron in the stationary case:
eE0 1 e2 E02
v(t) = sin t me v 2 = sin2 t. (8.135)
me 2 2me 2
The deflection of the electron from its average position is given by
eE0
x= cos(t) = x0 cos(t). (8.136)
2 me
For the oscillation amplitude one calculates with (4.2)
eE0 e 2I e2 2I
x0 = 2 = 2 = 2 2
, (8.137)
me me 0 c 4 c me 0 c
8.5 Atoms in a High Laser Field 433
UP / eV x 0 / nm
IH 1800 nm 800 nm
400 nm 1800 nm 800 nm
mec 2 200 nm 400 nm
106
50nm (b) 200 nm
10 50nm
104 (a) 13nm
1 13nm
102
1 0.1
10-2 0.01
1012 1014 1016 1018 1020 1012 1014 1016 1018 1020
I / Wcm-2 I / Wcm-2
Fig. 8.28 Ponderomotive potential (a) and maximum amplitude of an electron (b) in the field of
a short laser pulse of intensity I and wavelength ; the full red line corresponds to the wavelength
= 800 nm of the Ti:Sapph laser
e2 I e 2 I 2
Up = 2
= I 2 , or (8.139)
20 cme 8 2 0 c3 me
Up / eV = 9.3375 108 [/ nm]2 I / 1012 W cm2 . (8.140)
We point out that this expression is completely identical to (H.20), formally derived
from the term quadratic in the vector potential A in the exact (semiclassical) Hamil-
tonian (H.1) for an atom in an electromagnetic field.
The order of magnitude of Up and x0 is illustrated by Fig. 8.28 for a number
of wavelengths as a function of laser intensity I . The full red lines refer to the
fundamental of the Titanium-sapphire laser (short Ti:Sapph) at = 800 nm the
work horse of ultrafast laser science. As an example, an electron in the focus of
a laser beam at an intensity of 1014 W cm2 experiences a ponderomotive potential
of Up = 5.976 eV, and according to (8.138) the corresponding excitation amplitude
is x0 = 0.87 nm a huge electron motion as compared to typical atomic radii of
0.1 nm to 0.25 nm (see Sect. 3.1.5).
Clearly, electrons bound to an atom or molecule exposed to such field strengths
will experience dramatic changes of their wave functions and term energies. We thus
have to compare the ponderomotive potential (8.139) to the binding energies of the
electrons in atoms. In the low intensity limit, we expect energy shifts as we have
434 8 Atoms in External Fields
derived them for the dynamic S TARK effect in (8.97). And indeed, a comparison
(opt)
with (8.139) shows that in the limit of high frequencies ba , with fba = 1
both expressions become identical.
However, for really intense laser fields, in particular at longer wavelengths as
characterized in Fig. 8.28, a complete break down of the bound state description
developed so far in atomic physics is expected. Two specific limits are indicated:
on the one hand, the system becomes highly relativistic if Up > me c2 . The intensity
necessary to reach this limit decreases according to (8.139) with the square of the
wavelength. On the other hand, for intensities above IH the electric field in the laser
focus is larger than the atomic field EH that an electron experiences in the H atom
at a distance a0 from the nucleus. This intensity is independent of wavelength:
2
0 c 2 0 c e
IH = E = = 3.51 1016 W cm2 . (8.141)
2 H 2 40 a02
There are other aspects for considering a laser field to be high. One of these is
derived from the ratio of ionization potential WI to ponderomotive potential Up .
For reasons to be discussed in Sect. 8.5.4 one defines the so called
WI 0 cme 2 WI
K ELDYSH parameter = = (8.142)
2Up e2 I
WI / eV
or in numerical terms = 2.31 10 3
I /1012 W cm2 / nm2
As also seen in Fig. 8.29 the rise of the ion yield above ca. 1014 W cm2 con-
tinues to decrease dramatically. One may conclude that at these intensities almost
all atoms are already ionized in the centre of the laser focus one speaks of satu-
ration. The continuing rise with intensity is essentially due to a geometrical effect:
enhanced ionization now also occurs of the peripheral zone of the laser beam (hav-
ing a Gaussian profile). Thus, the volume in which saturation intensity is reached
increases, and with it the total ion yield. The upper scale in Fig. 8.29 gives the
K ELDYSH parameter (8.142) for comparison. Saturation intensity obviously corre-
sponds to 1, that is to say saturation sets in where the moderate field becomes
a very high one.
At very high intensities the internal atomic field will be modified substantially by the
external (oscillating) electric field. Lets assume the atomic potential to be Coulom-
bic with charge Ze. When adding a linearly polarized laser field, an atomic electron
sees a time dependent, overall potential
Ze2
V (r, t) = eE(t)z with z = r cos , (8.143)
40 r
2 c03
Icr = (WI )4 (8.144)
2Z 2 e6
Icr 4.0 109 WI 4
.
W cm2 Z2 eV
tu = WI /(eE0 ) (8.145)
8.5 Atoms in a High Laser Field 437
Fig. 8.30 Model to understand atomic ionization in a high electric field, in particular in an intense
laser field: (a) tunnelling (b) electron emission above-barrier. Sketched are cuts through the
potential parallel to the direction of the E field at the time of maximum field in z-direction
the electron leaves the atom its kinetic energy is Wkin6 = WI , its velocity v =
As
2WI /me , and consequently the tunnelling time becomes
tu me WI 0 cme WI
ttu = = = . (8.146)
v 2eE0 2 e2 I
In order to allow the electrons to leave that atom for good, the tunnelling time must
be distinctively smaller than one half of the oscillation period, say ttu < 1/(2). One
then defines the K ELDYSH parameter as
0 cme 2 WI
= 2ttu = , (8.147)
e2 I
in agreement with (8.142). Saturation of the ion signal, observed in Fig. 8.29 for
high laser intensities ( 1), is thus found to happen at intensities and frequencies
for which the electron has sufficient time to escape when the oscillation field reaches
its maximum value E0 . With this visualization of the ionization process it is obvious
that ionization is more probable when the field oscillates less frequently, i.e. at larger
wavelengths.
The essentially classical ADK theory (A MMOSOV et al. 1986), often used with
astonishingly good results for atoms, neglects the frequency dependence of the pro-
cess completely. It considers saturation to be reached when the field is high enough
for direct above-barrier ionization as sketched in Fig. 8.30(b). To give a typical
value: in the H atom at = 800 nm, = 0.9 when this critical field intensity (8.144)
is reached. A detailed understanding of the relevant processes is subject to current
research. Some important concepts and consequences will be discussed in the fol-
lowing.
6 One should take this with a grain of salt: Tunnelling is a quantum mechanical process, while in a
classical picture the electron can only leave the atom above-barrier.
438 8 Atoms in External Fields
seq
dashed red lines (seq)
represent a theory with one
eq
active electron only, the full
10-1
ns
seq
black line (nseq) is a model
calculation for non-sequential
ionization
10-3
1014 1015 1016
intensity / Wcm-2
8.5.5 Recollision
If the timing between laser field and electron ejection is favourable, the electron may
even return to the atom. This so called rescattering of electrons was first discussed
in a pioneering paper by C ORKUM (1993). In a high, oscillating electric field the
trajectory of an electron depends of course on the exact point in time when it starts.
A simple classical calculation shows that the electron returns indeed to its starting
point, if is has not yet travelled too far when the sign of the electric field is reversed.
C ORKUM found, that (at the origin) the rescattered electron can acquire a kinetic
energy of up to
(el)
3.17 Up Wkin . (8.148)
This happens for an electron ejected at time t = 0, if the phase angle of the field
E(t) cos(t + ) is 17 . The physics of these backscattered electrons is very
interesting and continues to be a hot topic in current research.
One phenomenon associated with back scattered electrons is the ejection of a
second electron, leading to the so called non-sequential double ionization. It can be
recognized by a very special behaviour of the MPI cross section as a function of
laser intensity, illustrated for He as an example in Fig. 8.31.
Generally speaking, one expects processes of the following type for a multi-
electron system A:
A + N1 A+ + e
A+ + N2 A2+ + e
...
Aq+ + Nq+1 A(q+1)+ + e .
8.5 Atoms in a High Laser Field 439
Rescattered electrons cannot only eject a second electron. They may also be recap-
tured by the ion and emit electromagnetic radiation during this process: this leads
to the generation of electromagnetic waves with frequencies that are multiples of
the original laser frequency (fundamental). This process is called high harmonic
generation (HHG) and has attracted worldwide considerable interest during the past
years.
The HHG mechanism is illustrated schematically in Fig. 8.32. The recolliding
electron has a potentially high excess energy that may be emitted during the capture
process as radiation. According to (8.148), the energy of the recolliding electron can
be as high as 3.17Up . Thus, photon energies up to HHG 3.17Up + WI may be
emitted upon capture of the electron.
This HHG process is used in current research very successfully to generate short
pulses in the soft X-ray region (XUV). An intense femtosecond laser pulse is fo-
cussed into a dense gas target (e.g. a gas jet, a gas filled cell or capillary). One
440 8 Atoms in External Fields
(a) (b)
/ 107 nm-1
intensity
H31 H43
2 H37
4 H49
H55
0
26 24 22 20 18 16 14
HHG frequency wavelength / nm
Fig. 8.33 (a) Schematic HHG spectrum with plateau and cutoff at 3.17Up + WI . The frequency
distances are 2. (b) Example of an experimentally observed HHG spectrum from BALCOU et al.
(2002). 30 fs pulses at ca. 800 nm were focussed into a Ne gas jet. Different focussing conditions
(full and dotted lines) lead to quite different efficiencies
obtains the XUV radiation in forward direction. Typically it contains a broad spec-
trum of harmonics HHG = (2N + 1) of the fundamental as shown schemati-
cally in Fig. 8.33(a). For symmetry reasons, usually only odd harmonics are emit-
ted.
The scheme indicates the particularly high efficiency for low harmonics, fol-
lowed by a long plateau with frequencies at a distance of 2 up to the so
called cutoff at 3.17Up + WI , which is easily understood in view of Fig. 8.32. In
Fig. 8.33(b) gives as a typical experimental example the spectrum of Ne. As shown
in the figure one may modify the emitted output by judicious choice of focussing
conditions. This is a consequence of the highly nonlinear process. Optimization of
HHG generation for practical application is currently a hot topic in AMO research.
Special temporal and spatial pulse shaping may be used to improve the conversion
efficiency substantially. HHG is currently being used as a convenient, table top, time
resolved short pulse radiation source in the near X-ray region. It has considerable
application potential for X-ray spectroscopy. The shortest wavelengths achievable
depend on the target, on the pump laser intensity, as well as on its frequency (since
UP 2 ).
During the past decade the generation of attosecond laser pulses (1 as = 1018 s)
by superposition of several harmonics has developed very successfully (see e.g. the
reviews by K RAUSZ and I VANOV 2009; S ANSONE et al. 2011). As it turns out, the
harmonics generated are coherent; superposing them artfully (see e.g. T ZALLAS et
al. 2003) and filtering the resulting radiation suitably corresponds to interference in
a F OURIER series, leading to a sequence of pulses with an individual pulse duration
below 1 fs. As always, a new method that improves earlier techniques by one or
two orders of magnitude opens new perspectives with an unforeseeable potential in
basic research and applications. Certainly we shall witness exciting developments
of attosecond science in the years to come (as illustrated e.g. by S ANSONE et al.
2010; B OGUSLAVSKIY et al. 2012; V RAKKING and E LSAESSER 2012).
8.5 Atoms in a High Laser Field 441
Before ending this chapter we return briefly to ATI processes which we have intro-
duced already in Sect. 5.5.5. The question here is, how these processes change as
the laser intensity is increasing from MPI through the tunnelling regime and to
above-barrier ionization?
As a particularly suggestive example we show in Fig. 8.34 the spectra for Ar that
were studied by PAULUS et al. (1994) with beautifully resolved ATI peaks. Argon
has an ionization potential of 15.4 eV, according to (8.147) the laser intensities
used here thus correspond to K ELDYSH parameters of (a) 1.88, (b) 1.33, (c) 0.94
and (d) 0.7 a range covering the critical transition from moderate to above-barrier
behaviour. This is reflected quite evidently in the electron spectra: while at the low-
est intensity (a) an unspectacular ATI spectrum is observed, quite comparable to that
shown in Fig. 5.13 for Xe, the higher intensities promote very pronounced structures
as a function of electron energy, that remind us of the plateau seen in HHG which we
have discussed in the last section. There, the cutoff was identified as corresponding
to the maximum energy of rescattered electrons.
Thus it appears self-evident to attribute the plateaus or beats in the ATI spectra
also to recollision: obviously recolliding electrons too may absorb further photons.
Without going into the finer details of these observations we indicate by arrows
in Fig. 8.34 the maximum kinetic energy 3.17 Up of the rescattered electrons.
Obviously, there is even more structure and one should not over-stress the simple
rescattering model for such a highly complex process. We just mention that seri-
ous quantum mechanical model calculations achieve very good agreement with the
experimental data.
It is interesting to note that ATI may also be observed when ionizing quite large
molecules by (moderately) high intensity lasers. This is exemplified in Fig. 8.35 for
C60 according to C AMPBELL et al. (2000). The ionization potential is in this case ca.
7.6 eV, much lower than for argon. The intensities are thus equivalent to those used
in Fig. 8.34 as documented by the corresponding K ELDYSH parameters . Here too
one may recognize, albeit weakly, something like a prolonged plateau for the higher
442 8 Atoms in External Fields
101
0 5 10 15 20 25 30
electron energy Wkin / eV
intensities. Clearly the decrease of the electron signal beyond the 3.17 Up limit
is significantly slower for higher intensities. It is also evident that in this case the
laser intensity must not get too high as the clear ATI peaks in the electron spectrum
smear out: this large, finite system has many active electrons (there are 240 valence
electrons in C60 ). Interaction among them thermalizes the electron motion at the
highest laser intensities. Similar trends may be recognized for Ar: in that sense, C60
with its high symmetry may be seen as a kind of super atom.
Section summary
Todays short pulse lasers allow one to generate extremely high intensities I
of electromagnetic radiation. The corresponding electric field strengths can
easily surpass the inner atomic electric fields, even by orders of magnitude.
Correspondingly, in the Hamiltonian the term quadratic to the vector potential
of the field must be considered. It gives rise to the ponderomotive potential
(8.140), Up I 2 .
The K ELDYSH parameter = WI /2Up characterizes the field strengths:
it is considered to be high for 1.
Recollision of the electron ejected by the field and forced to return to the
atom by the field provides a useful concept for understanding non-sequential
ionization, HHG and ATI in high intensity fields. A recolliding electron may
acquire a kinetic energy up to 3.17 Up .
HHG by intense femtosecond laser pulses offers excellent perspectives for
time resolved spectroscopy with short X-ray pulses. It is also the basis at-
tosecond pulses.
ADK: A MMOSOV, D ELONE, and K RAINOV, (1986) theory for strong field ion-
ization (see e.g. Sect. 8.30).
Acronyms and Terminology 443
XUV: Soft X-ray (sometimes also extreme UV), spectral wavelength range be-
tween 0.1 nm and 10 nm according to ISO 21348 (2007), sometimes up to
40 nm.
References
A MMOSOV , M. V., N. B. D ELONE and V. P. K RAINOV: 1986. Tunnel ionization of complex
atoms and of atomic ions in an alternating electromagnetic field. Sov. Phys. JETP, 64, 1191
1194.
BALCOU , P. et al.: 2002. High-order-harmonic generation: towards laser-induced phase-matching
control and relativistic effects. Appl. Phys. B, 74, 509515.
B ERGER , M. J., J. H. H UBBELL, S. M. S ELTZER, J. C HANG, J. S. C OURSEY, R. S UKUMAR,
D. S. Z UCKER and K. O LSEN: 2010. XCOM: Photon cross sections database (version 1.5),
NIST. http://physics.nist.gov/xcom, accessed: 8 Jan 2014.
B OGUSLAVSKIY , A. E., A. E. B OGUSLAVSKIY, J. M IKOSCH, A. G IJSBERTSEN, M. S PANNER,
S. PATCHKOVSKII, N. G ADOR, M. J. J. V RAKKING and A. S TOLOW: 2012. The multi-
electron ionization dynamics underlying attosecond strong-field spectroscopies. Science, 335,
13361340.
B ORN , M. and E. W OLF: 2006. Principles of Optics. Cambridge University Press, 7th (expanded)
edn.
B OYD , R., O. H ESS, C. D ENZ and E. PASPALKALIS: 2010. Slow light. J. Opt., 12, 100301.
B OYD , R. W. and D. J. G AUTHIER: 2002. Slow and fast light. In: Progress in Optics,
vol. 43, 497530. Amsterdam: Elsevier.
B REIT , G. and I. I. R ABI: 1931. Measurement of nuclear spin. Phys. Rev., 38, 20822083.
B UCKINGHAM , A. D.: 1967. Permanent and induced molecular moments and long-range inter-
molecular forces. Adv. Chem. Phys., 12, 107.
C AMPBELL , E. E. B., K. H ANSEN, K. H OFFMANN, G. KORN, M. T CHAPLYGUINE
M. W ITTMANN and I. V. H ERTEL: 2000. From above threshold ionization to statistical elec-
tron emission: the laser pulse-duration dependence of C60 photoelectron spectra. Phys. Rev.
Lett., 84, 21282131.
C HANTLER , C. T., K. O LSEN, R. A. D RAGOSET, J. C HANG, A. R. K ISHORE, S. A. KO -
TOCHIGOVA and D. S. Z UCKER : 2005. X-ray form factor, attenuation, and scattering tables
(version 2.1), NIST. http://physics.nist.gov/ffast, accessed: 7 Jan 2014.
C ORKUM , P. B.: 1993. Plasma perspective on strong-field multi-photon ionization. Phys. Rev.
Lett., 71, 19941997.
H ANSON , A. L.: 1986. The calculation of scattering cross-sections for polarized X-rays. Nucl.
Instrum. Methods A, 243, 583598.
H AU , L. V., S. E. H ARRIS, Z. D UTTON and C. H. B EHROOZI: 1999. Light speed reduction to
17 meters per second in an ultracold atomic gas. Nature, 397, 594598.
H ICKSTEIN , D. D. et al.: 2012. Direct visualization of laser-driven electron multiple scattering
and tunneling distance in strong-field ionization. Phys. Rev. Lett., 109, 073004.
H UBBELL , J. H., W. J. V EIGELE, E. A. B RIGGS, R. T. B ROWN, D. T. C ROMER and R. J. H OW-
ERTON : 1975. Atomic form factors, incoherent scattering functions, and photon scattering cross
sections. J. Phys. Chem. Ref. Data, 4, 471538.
ISO 21348: 2007. Space environment (natural and artificial) process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
K ANE , P. P., L. K ISSEL, R. H. P RATT and S. C. ROY: 1986. Elastic-scattering of gamma-rays
and X-rays by atoms. Phys. Rep., 140, 75159.
K ELDYSH , L. V.: 1965. Ionization in the field of a strong electromagnetic wave. Sov. Phys. JETP,
20, 1307.
K RAUSZ , F. and M. I VANOV: 2009. Attosecond physics. Rev. Mod. Phys., 81, 163234.
References 445
Overview
This is not a particularly easy chapter to study. Nevertheless, the reader
will earlier or later have to deal with this important and methodolog-
ically fundamental topic. After an introduction into the underlying interac-
tions in Sects. 9.1 and 9.2 we shall discuss in Sect. 9.3 the Z EEMAN effect,
again for low, high and arbitrary magnetic fields, following the discussion in
Sects. 8.1.28.1.6 there applied to the electron spin, here to the nuclear spin.
Electric interactions (Sect. 9.4) and isotope shift (Sect. 9.4.2) are specialties in
the interaction of atomic nuclei with atomic electron and require some math-
ematical effort making use of formulas derived in the appendices. Finally,
Sect. 9.5 introduces some state-of-the-art experimental techniques: molecular
beam spectroscopy, EPR and finally NMR spectroscopy. All sections build
on the previous ones, but the text may be read without having mastered each
preceding detail.
9.1 Introduction
Table 9.1 Hierarchy and relevance of interactions in atomic physics (ordered according to relative
strength)
Interaction Characteristic Remarks
1. C OULOMB degeneracy only for H atom and H like
Ions
2. deviation from Z/r removal of degeneracy for alkali atoms and all others
potential
3. spin-orbit fine structure splitting (2S + 1) FS levels (S electron
J =L + S for low Z spin), each (2J + 1)-fold
J = J i for high Z degenerate
4a. external field magnetic (2J + 1)-fold degeneracy
removed
4b. external field electric (2J + 1)-fold degeneracy
removed but states with
M = |M| remain degenerate
Depending on field strength 3. and 4. may change sequence
5. radiative corrections L AMB shift sensitive test for QED
6. atomic nucleus electron hyperfine structure 2I + 1 states (I nuclear spin),
cloud (in analogy to 3. but F =J +I each (2F + 1)-fold degenerate
much weaker)
a volume effects form and mass of atomic isotope shift of spectral lines
nucleus
b magnetic dipole B j field hyperfine splitting
of electron cloud
c electric quadrupole moment additional shift
E field of electron cloud
7a. external field magnetic (2F + 1)-fold degeneracy
removed
7b. external field electric analogue to 4b
6. with 7.: Important spectroscopic tool for nuclear physics, also for chemistry (atomic nuclei
probe their chemical environment, NMR)
and molecular energy levels, subject of the present chapter. The magnetic moment
of an atomic nucleus is are related to its spin I and the nuclear g factor as
I = gI N
M I / z = gI N Iz /
and has the projection M (9.1)
Note the positive sign of the nuclear magnetic moment,1 in contrast to the magnetic
moment of the electron:
S = gB
M S/ (9.3)
e mp
B = = N (9.4)
2me me
= 927.4 1026 J T1 = 5.788 105 eV T1 =
14.00 GHz T1 .
Precise values N and B are found in Appendix A. We emphasize that the nuclear
magneton (9.1) is very small, N B /1836, and hence HFS is small too.
As a quantitative basis for the following discussion, Table 9.2 presents a range
of typical examples for nuclear moments. We remember that protons, neutrons and
atomic nuclei are not elementary particles in contrast to the electron. Hence, the
nuclear g factors, gI = MI /I N are not even approximately integer numbers, as
documented in Table 9.2.
The usual eigenvalue equations for angular momenta hold also for the nuclear
spin. For the magnitude we have
2
I |I MI
= 2 I (I + 1)|I MI
, (9.5)
=
F I +
J. (9.7)
In full analogy to FS coupling (Chap. 6) we just need to replace:
L
S
J
(9.8)
J
IF
1 Some authors use a different definition for gI : gI N (see e.g. the review of A RIMONDO
et al. 1977).
450 9 Hyperfine Structure
Table 9.2 Properties of some hadrons and atomic nuclei (S TONE 2005). The notation A Z X refers
to an atomic nucleus X with Z protons (atomic number) and a total of A nucleons (atomic mass
number). The unit of area, 1 b (see Appendix A.2) corresponds to the area of an average size atomic
nucleus
Nucleon or Spin Land factor Magnetic moment Quadrupole NMRb
atomic nucleus I gI = MI /(I N ) MI /N momenta
Q / eb
Proton p 1/2 5.58569471(5) 2.79284736(2) 0 +
Neutron n 1/2 3.8260854(10) 1.9130427(5) 0
Deuteron 21 D 1 0.8574382284 0.8574382284 0.0286(2)
3 He
2 1/2 4.25499544(6) 2.12749772(3) 0
4 He
2 0 0 0
6 Li
3 1 0.8220473(6) 0.8220473(6) 0.00083(8)
7 Li
3 3/2 2.1709513(13) 3.256427(2) 0.0406
12 C
6 0 0 0
13 C 1/2 +1.4048236(28) +0.7024118(14) 0 +
6
14 N
7 1 0.40376100(6) +0.40376100(6) +0.02001(10)
15 N
7 1/2 0.56637768(10) 0.28318884(5) 0 +
16 O
8 0 0 0
19 F
9 1/2 +5.257736(16) +2.628868(8) 0 +
23 Na
11 3/2 1.478348(2) +2.217522(2) +0.109(3)
29 Si
14 1/2 1.11058(6) 0.55529(3) 0 +
31 P
15 1/2 +2.2632(6) +1.13160(3) 0 +
39 K
19 3/2 0.26098(2) +0.39147(3) +0.049(4)
67 Zn
30 5/2 0.3501916(4) +0.875479(9) +0.150(15)
85 Rb
37 5/2 0.541192(4) +1.35298(10) +0.23(4)
129 Xe
54 1/2 1.555952(16) 0.777976(8) 0
133 Cs
55 7/2 0.7377214(9) +2.582025(3) 0.00371(14)
199 Hg
80 1/2 1.0117710(18) +0.5058855(9) 0
201 Hg
80 3/2 0.3734838(9) 0.5602257(14) +0.38(4)
235
92 U 7/2 0.108(10) 0.38(3) 4.936(6)
Comparison ge = |MS /(SB )| MS /N
Electron e 1/2 2.002319 . . . 1838.2819709(8)
Muon 1/2 2.002331 . . . 8.8905971(2)
Note: for I = 0 or I = 1/2 the quadrupole moment is always Q 0
aA precise definition for Q gives (9.71); see also Appendix A.2
b Isotopes marked with + are particularly useful for NMR
|J I F MF
= 2 F (F + 1)|J I F MF
F with (9.9)
F = J I, J I + 1, . . . , J + I for I < J and
F = I J, I J + 1, . . . , J + I for I > J.
9.1 Introduction 451
(a) z (b) z
Fz MF
Fz 3 3
F I F
2
1 1
0
- 1 -1
J S -2
- 3 -3
L
Fig. 9.1 (a) Vector model for coupling L and S to J , and of J and I to F , (b) F has 2F + 1
possibilities of orientation in space
The vector model in Fig. 9.1(a) visualizes these relations for the example of a 2 P3/2
level: first the orbital angular momentum L (here L = 1) couples with the spin S
(here S = 1/2) forming a total angular momentum J (here J = 3/2) of the electron
charge cloud. Orbital angular momentum and spin precess around J , which finally
couples with the nuclear spin I (here I = 3/2) forming a total angular momentum
F (here F = 3) of the atom. J and I precess around F , whose absolute value is
|F | = F (F + 1) (here = 3.46). In turn, F has 2F + 1 possible orientations in
space, MF = F, F + 1, . . . F , as illustrated in Fig. 9.1(b).
Including spin-orbit (LS) and hyperfine interaction terms (J I ), as well as (if
appropriate) magnetic interactions, we write the H AMILTON operator:
=H
H LS + V
0 + V MD + V
B + V
vol + V
Q . (9.10)
In the following we consider, one by one, the different contributions to HFS: the
magnetic dipole interaction (V MD ), the HFS-Z EEMAN effect (V
B ), volume shifts
(Vvol ) and finally the electric quadrupole interaction (VQ ).
Section summary
Hyperfine structure is caused by the interaction between electron cloud and
atomic nucleus. The nuclear magnetic properties are summarized by the nu-
clear magneton N and the gI factor. While N is about three orders of mag-
nitude smaller than B for the electron, the g factors are of similar magnitude.
Angular momentum coupling between the nuclear spin I and the total angular
momentum J of the electron charge cloud to an overall angular momentum
F of the atom occurs in complete analogy the coupling of electron spin S and
orbital angular momentum L to
J.
In addition to magnetic dipole interaction and interaction with external fields,
HFS is also determined by volume shifts and electric quadrupole interac-
tion.
452 9 Hyperfine Structure
The main contribution to HFS is due to the nuclear analogue of spin-orbit coupling
in FS. It is treated in a similar spirit as the latter in Sect. 6.2.3. We write this magnetic
dipole hyperfine interaction between the magnetic field of the atomic electron cloud
B J and the magnetic dipole moment of the atomic nucleus MI as
I
J
V J = gI N
I B
MD = M I BJ = A 2 with (9.11)
the magnetic dipole HFS coupling constant A = gI N J . (9.12)
J is the operator for the average magnetic field of the electron cloud, projected
B
onto J . It is caused by the electron orbital motion and their spin magnetic moments.
Its relation with J is written as
J = J
B J /. (9.13)
1 2 2 2
I
J= F I J . (9.14)
2
The change of the eigenenergy is then derived from (9.11), and we obtain
A
WMD = F (F + 1) I (I + 1) J (J + 1) , (9.15)
2
corresponding to (6.62). This expression gives already a good phenomenological de-
scription of the empirically observed hyperfine structure splitting. The computation
of J and thus A will be subject to the next sections.
As for fine structure splitting, from (9.15) follows the L AND interval rule:
n =2
2701 2p 2P3/2 F =2
23.65
2p 3/2 9911.201 F =1 F =1
2s 2S1/2 excited
13 670
1057.847(9) 177.56 F =1 states
2s1/2 , 2p1/2 F =0 59.17
2p 2P1/2
243 nm F =0
n =1
2 466 061 413.187 074(34) 243 nm
43 770
1s 2S1/2 F =1 ground
1420.41 state
1s1/2 8173
F =0
BOHR DIRAC LAMB shift hyperfine structure
Fig. 9.2 Fine and hyperfine structure of the H atom (compare to Fig. 6.29). All energy splittings
are given in MHz as taken from the compilation of K RAMIDA (2010). The nuclear spin of the
proton (H+ ) is I = 1/2. The ground to excited state distance is not to scale. The scale for the
splittings in more and more magnified towards the right.
The deuteron, d = pn, has a nuclear spin I = 1. As shown in Fig. 9.3, the HFS
structure of the hydrogenic isotope deuterium (D) is thus quite different from atomic
H. The 1s 2 S1/2 ground state and the excited states with J = 1/2 (e.g. 2s 2 S1/2 and
2p 2 P1/2 ) split also into doublets, however, with F = 1/2 and 3/2. In contrast, the
2p 2 P3/2 state forms a triplet with F = {J I, J I + 1, J + I } = {1/2, 3/2, 5/2}.
Somewhat more complicated is the situation e.g. for sodium (Na), as shown in
Fig. 9.4. With a nuclear spin I = 3/2 the maximum number of levels is (2I + 1) = 4.
The 3s 2 S1/2 ground state is again an HFS doublet (since J = 1/2) with F = I
J = {1, 2}. The same holds for the lower level of the excited FS doublet 3p 2 P1/2 .
However, the 3p 2 P3/2 , J = 3/2 level forms a quartet with the components F =
{J I, J I + 1, J I + 2, J + I } = {0, 1, 2, 3}.
Atomic hyperfine structure has been studied in the past in great detail, and the rel-
evant parameters are known for most atomic systems rather well. Today HFS is
used as a very sensitive probe in molecular and condensed matter spectroscopy. For
a quantitative comparison with theory, i.e. for the computation of A, one has to
evaluate (9.13), i.e. to determine the magnetic field B J of the electron cloud at the
position of the nucleus.
454 9 Hyperfine Structure
2s 2S F = 3/2
1/2
40.92 excited
1059.28
F = 1/2 F = 3/2 states
2p 2P1/2
2s1/2 , 2p1/2 13.63
2 466 732 407.521 71(15)
F = 1/2
1s 2S1/2 F = 3/2
n =1
327.38
F = 1/2
8184 ground
1s1/2 state
Fig. 9.3 Fine and hyperfine structure of the deuterium atom. The nuclear spin is I = 1. Otherwise
as in Fig. 9.2
In addition to the magnetic field from the electron orbits (similar to the situation
for spin-orbit interaction treated in Sect. 6.2.3) we have to account for the magnetic
field from the dipole moments of the electron spins. That makes it somewhat more
complicated. We restrict the discussion to a single active electron. Fortunately, the
magnetic fields of orbit and spin are additive and we treat these contribution, B L and
B S , now one after the other.
9.2 Magnetic Dipole Interaction 455
For B L the situation is even somewhat less complicated than in the FS case:
According to classical electrodynamics the magnetic field of a rotating shell with a
magnetic moment M and a radius r is inside the shell
0 M
B= 2 .
4 r 3
With (1.150), M =M L = B L/, the field of the orbiting electron at the nu-
cleus is thus represented by the operator
L = 0 2B L =
B
e 1
3L, (9.17)
4 r 3 40 me c 2 r
with the B OHR magneton B = e/2me and 0 = 1/0 c2 . The respective contribu-
tion to the hyperfine dipole energy is
a3
N = 0 2B gI N 1 L I = gI me 0 2L I Eh .
2
LI = B
V L M (9.18)
4 r 3 2 4 mp r 3 2
We may compare this to (6.35) for the spin-orbit interaction in the FS case.2
We point out that the above derivation of (9.18) has been obtained as energy of the
nuclear dipole M N in the magnetic field B L of the electron orbit. It is interesting to
realized that the same result is obtained if one computes the energy of the electronic
dipole moment M L in the field B I of the localized nuclear dipole moment.
Let us now turn to the electron spin. Its magnetic moment M s of is associated
with a magnetic dipole field (e.g. JACKSON 1999, Eq. (5.56))
s = 0 1 (3er M
B s er M
s ), (9.19)
4 r 3
with er = r/r, being the unit vector in r direction. The resulting interaction does
not only depend on the angle between S and I and on the distance r of the electron
from the nucleus. It also depends on the angle between r and S. By inserting the
magnetic moment M s = ge B S/ of the electron spin S with ge = 2.0023, and
adding the orbital field B L (9.17), one obtains the total field B J of the electron
cloud for r > 0. The singularities at r = 0 (vanishing distance between nucleus and
electron) require special attention. We suppress here the details of the analysis and
just communicate the result for the total magnetic field caused by the electron at
arbitrary distance r:
0 B L ge 4ge
BJ = + (3er S er S) + S (r) . (9.20)
2 r 3 2r 3 3
2 The main difference is the ratio of electron to proton mass, me /mp 1/1836, determining the
order of magnitude for the HFS splittings. The electron spin S has been replaced by
I and corre-
spondingly the ge factor by gI . We also note the factor 2, since no T HOMAS factor has to be applied
presently (compensation of the coordinate transformation is not needed). And finally, a03 /r 3 has
now replaced the more complicated (r) a consequence of the localized nuclear dipole.
456 9 Hyperfine Structure
8ge
+ (r)I S . (9.21)
3
The (r) term, only relevant at the nucleus, is called F ERMI contact term. Note that
this expression is strictly valid only for effective one electron, one nucleus systems.
In the multi-electron case one has to add the fields from all individual electron or-
bits and spins, projected onto L z and Sz , respectively. In addition, for molecules
and solid state materials one has to sum over all relevant nuclei at their respective
positions.
We note in passing, that for L = 0 and apart from the F ERMI contact term, (9.21)
is completely analogue to the electric dipole-dipole interaction, where (8.86) de-
scribes the interaction of two electric dipoles at a distance r.
For vanishing angular momentum, L = 0 (a situation typical for many atoms and
most, non-radical, organic molecules), one may write (9.21)
MD =
V S A
I, (9.22)
0 ge B gI N 1 8
A= (3e r e r 1) + (r) . (9.23)
4 2 r3 3
This tensor plays a key role in the theory of all magnetic resonance methods
(EPR, NMR, etc.). It contains only operators derived from the components of the
position vector r. In order to determine the hyperfine splitting in 1st order, one just
has to determine the diagonal matrix elements of (9.22), i.e. one has to average over
all angular momenta and all position coordinates. Averaging over the latter needs
only to be done for the A tensor. As we shall see in a moment, for atoms which are
described in spherical coordinates this implies averaging 1/r 3
, quite similar to the
fine structure problem. In the case of complex molecules or solid state materials this
averaging becomes, however, much more elaborate. Apart from the F ERMI contact
term containing (r) to be treated below, one obtains an average hyperfine tensor
2 2
3x rr
3xy
3xz
0 ge B gI N
5 r 5 r 5
3yx 2 r 2
A
=
3y
3yz
. (9.24)
4 2 r 5 r 5 r 5
3z rr
2 2
3zx
r5
3zy
r5
5
It is this local environment that determines the HFS. The nuclear spin is thus a sen-
sitive probe for such charge distributions and hence an important tool for structural
analysis of complex systems.
Let us return to atoms and focus our attention on r > 0. This is reasonable if the or-
bital angular momentum is larger than zero. Electrons with > 0 almost never come
close to the nucleus, and the terms containing (r) in (9.20) and (9.21) are obsolete
when determining the magnetic field and the HFS dipole interaction, respectively.
To compute the matrix elements in the coupled (effectively one electron) scheme
|n(( S)J I )F MF
some serious angular momentum algebra has to be applied, with
6j or even 9j symbols (as sketched in Appendix B.4, the latter allow the recoupling
of schemes with four angular momenta, SJ I to F , in a similar manner as the 6j
symbols for the coupling of three angular momenta discussed in the context of fine
structure). Here we just outline the crucial steps of such an analysis.
First we derive an estimate for the average magnetic field of the electron cloud
B J
by recalling the projection theorem (C.17) used already successfully in the
context of spin-orbit interaction. We apply it to the tensor operator B J ,
J M|BJ
J | J M
J M BJ q | J M
= J M Jq | J M
,
J (J + 1)2
and write this symbolically for all components q in vector form
BJ J J
BJ = . (9.25)
J (J + 1)
Comparing this expression with the original ansatz (9.13), we obviously have found
an expression for
BJ J
J = . (9.26)
J (J + 1)
Inserting (9.20) for r > 0 and
J = L +S, we have to evaluate:
0 B
J = 2L + g e (3S e r e r
S) (
L +
S) . (9.27)
4 r 3 J (J + 1)2
where the averaging over the radial dependence and angular components can be
carried out independently. For s = 1/2 we find with (2.105), that
+3(
S er )2
2
S = 0, (9.29)
458 9 Hyperfine Structure
0 2 ( + 1)
J = B n |r 3 |n
, (9.30)
4 J (J + 1)
and the magnetic dipole HFS coupling constant (9.12) for effective one electron
systems with > 0 becomes
0 2 ( + 1)
A= B N gI n |r 3 |n
, or (9.31)
4 J (J + 1)
2 me 2 ( + 1) 1
= gI n | 3 |n
in a.u.
4 mp J (J + 1) r
In hindsight, this remarkable result justifies our somewhat relaxed dealings with
averaging processes: since is a good quantum number in any coupling scheme, we
have to consider only diagonal matrix elements of the HFS dipole interaction (as
long as no electric field is active). For comparison with FS, and specifically with the
spin-orbit coupling constant a according to (6.36), see footnote 2. The additional
factor ( + 1)/J (J + 1) originates from the projection of the magnetic field B L
onto J .
Specifically, for H and H like ions we find with the expectation value 1/r 3
according to (2.130)
0 1 Z3
A= gI N B 3 , or (9.32)
J (J + 1)(2 + 1) a0 n3
me 1 Z3
= 2 gI in a.u.
mp J (J + 1)(2 + 1) n3
We return once more to the Hamiltonian (9.21), and discuss now the term with the
delta function (r), the F ERMI contact term. It plays a role only for = 0, since
according to (2.120) the radial wave function at small r is Rn (r) r . Only s
states have a finite probability |n 0 (0)|2 at the origin. As just outlined, for the one
electron case all other terms in (9.23) disappear for = 0. Hence, the contact term
remains the only contribution to HFS, and (9.21) becomes
MD = 0 ge B gI N 2 (r) S I ,
V for = 0. (9.33)
3 2
The diagonal matrix element is easily evaluated. In the |(LS)J I F MF
coupling
scheme we have as usual
9.2 Magnetic Dipole Interaction 459
1
S
I
/2 = F (F + 1) S(S + 1) I (I + 1) and
2
2 2
(r) = n00 (r) (r)d3 r = n00 (0) in general, (9.34)
Z3
and = specifically (9.35)
a03 n3
for H and H like ions according to (2.127). Thus, we finally obtain for effective one
electron systems the HFS splitting for = 0 in 1st order perturbation theory:
20
WMD = ge B gI N (r) I /2
S (9.36)
3
20 2
= ge B gI N n00 (0) F (F + 1) S(S + 1) I (I + 1) .
3
With ge 2 and (9.15) the magnetic dipole HFS coupling constant becomes
16 0 2 4 2 me 2
A= B gI N n00 (0) = gI n00 (0) a03 Eh . (9.37)
3 4 3 mp
Specifically for hydrogen and hydrogen like ions, |n00 (0)|2 is given by (2.127).
Interestingly, one obtains for the present case with = 0 (i.e. for J = S = 1/2)
exactly the same expression (9.32) for A already found for > 0. For precision
measurements one has, of course, to replace again a0 a0 me /m e.
For the H atom in its ground state (F = 0 or 1 and J (J + 1) = 3/4) we obtain with
gI 5.586 the hyperfine coupling constant A/ h 1421 MHz. With the interval
rule (9.16) this value is identical to the splitting of the HFS doublet. The HFS split-
ting in the ground state of the H atom is thus nearly one order of magnitude smaller
than the L AMB shift (6.90). Its experimental value is today one of the best known
spectroscopic quantities.3 It can be determined directly as microwave M1 transition,
and the first high precision measurement was carried out by R AMSEY and collabora-
tors (C RAMPTON et al. 1963), using a hydrogen maser. The presently recommended
best value (K ARSHENBOIM 2005) for the ground state HFS splitting of H is
3 Presently best known is the frequency of the 1s 2 S1/2 2s 2 S1/2 transition, see (6.88).
460 9 Hyperfine Structure
and thus it is also about an order of magnitude smaller than the respective L AMB
shift. For the 2p 2 P1/2 and 2p 2 P3/2 states the splitting is again smaller, by a factor
3 and 7.5, respectively.
The HFS splitting of the deuterium atom in its ground state (F = 1 and 2) is sig-
nificantly smaller than for hydrogen, due to the smaller gI = 0.8574382284 factor.
Its value is also known with high precision:
In alkali atoms the HFS is comparatively large, in spite of the larger principle
quantum number n. This is attributed to the much higher effective nuclear charge
Zeff , which according to (9.32) enters as A (Zeff /n)3 . Also, the higher values of F
lead to larger splitting. According to the interval rule (9.16) by a factor of 2 for 23
11 Na
in the ground state (I = 3/2 F = 2) in comparison to 1 H (I = 1/2 F = 1).
1
already mentioned above, is plausible: from (9.32) we derive with this value an
effective charge of Zeff 4.6 to be compared with Z = 11 for Na.
Selection rules and probabilities for optical (E1) dipole transition between two HFS
states |a
and |b
are derived in full analogy to those for FS transitions, as detailed
in Sect. 6.4. One just has to replace the quantum numbers according to the scheme
(9.8). In particular, the triangular relation (Fa Fb 1) = 1 holds as well as the selec-
tion rule for projection quantum numbers
MF = 0, 1 with the corresponding po-
larization of emitted or absorbed light. As an example, Fig. 9.5 shows the optically
allowed HFS transitions for the 2 P1/2 , 2 P3/2 2 S1/2 doublets in Na(I = 3/2). Part
of an experimental spectrum has already been presented in Sect. 6.1.3.
Line intensities are expressed again by reduced matrix elements and 6j symbols
from which intensity ratios are derived in the same manner as for FS multiplets.
If one wants to determine the matrix elements quantitatively, one has to apply the
9.3 ZEEMAN Effect of Hyperfine Structure 461
F=2
2S F=1
1/2
reduction formalism twice in order to resolve the dependence on hyperfine and fine
structure coupling. The final expressions can be evaluated with the aid of 6j and
3j symbols except for the radial matrix elements that are characteristic for each
individual atom and transition. They are responsible for the overall strength of a line.
Of course, parity conservation also holds. In the case of one single active electron
this implies b = a 1.
Section summary
Magnetic hyperfine dipole interaction V MD = M I BJ is treated in the
same spirit as spin-orbit interaction for FS, with B J = J J / being the
effective magnetic field of the electron cloud. Consequently, the level splitting
is given by (9.15), in complete analogy to (6.62) for FS.
The main difference to FS interaction is the contribution of the electron spin
J . This leads to somewhat more complex expressions for the interaction
to B
energy (9.21)(9.23).
We distinguish the case of nonvanishing orbital angular momentum, in which
expectation values 1/r 3
determine the splitting, while for zero orbital angu-
lar momentum the F ERMI contact term is relevant, i.e. the electron density at
the origin. Interestingly, the final result (9.32) for HFS splitting in the case of
H and H like ions is identical.
HFS splitting is nearly an order of magnitude less than the L AMB shift.
For optical transitions between HFS multiplets standard selection rules hold,
i.e. (Fa Fb 1) = 1 and
MF = 0, 1.
The problem is solved very similar as for the Z EEMAN effect of fine structure. How-
ever, some specialties and the fact that it forms the basis for NMR spectroscopy,
make a detailed discussion advisable.
462 9 Hyperfine Structure
We have to add the interaction (6.29) between the external magnetic field B and
the magnetic moments of electron cloud M J and nucleus M I to the H AMILTON
operator (9.11) in the field free case:
MDB = V
H MD + V J B M
B = MI B J M I B
I
J Jz Iz
= A 2 + gJ B gI N B. (9.39)
In the second line, we have assumed B z as usual. The first term has just been
treated in detail. And for simplicity sufficient in many cases one often ne-
glects the third term, since the gyromagnetic ratios for electrons and nuclei differ
by three orders of magnitude, gJ B gI N , and also A gI N B for not too
high B.
Note, however, that this is not acceptable for high field NMR and EPR, using
super-conducting magnets (B some tesla) today standard methods in analytical
organic chemistry. Also, the objects of NMR are often characterized by J = 0, so
that the Z EEMAN splitting is determined exclusively by the interaction of the nu-
clear moment with the external field, which in turn is modified by the local field
anisotropy due to a molecular environment. Thus, we shall slightly different from
the usual treatment in atomic physics text books distinguish between weak, high,
and very high external field,
gJ B B A, gI N B A gJ B B, and A gI N B,
respectively. In the latter case we have to account fully for the term Iz B.
For all cases, we note that Jz and Iz do not commute with
I J . Strictly speaking,
neither F nor MF nor MS or MI are good quantum numbers.
2 2 Jz
MDB A I J + gJ B J B = A F
H
J
2
I + gJ B B . (9.40)
2 2 2
The squared angular momenta have the well known eigenvalues 2 F (F + 1), etc.,
and gJ B B Jz / is treated as a perturbation.
9.3 ZEEMAN Effect of Hyperfine Structure 463
In spite of the formal similarity with fine structure splitting in a magnetic field,
there is a significant quantitative difference. To see this, we rewrite quite formally
MD term in (9.39) in two ways:
the V
I J
I
A = g I N B J or as (9.41a)
2
I J
J
A 2 = gJ B B I . (9.41b)
In our above derivation we used view point (a): the nuclear spin I orients itself in the
magnetic field B J of the electron cloud. The alternative view point (b) considers the
splitting of the substates |J MJ
of the electron cloud in the field B I of the nuclear
moment.
(a) From (9.41a) we get an estimate for the order of magnitude of the field BJ of the
electron cloud at the nucleus. Specifically, for the H atom in its 1s 2 S1/2 ground
state with A = 5.87 106 eV and gJ 2 we obtain:
BJ A/(gI N ) 33.3 T. (9.42)
Thus, the transition from the low (B BI ) to high field case (B BI ) in HFS
Zeeman splitting occurs at magnetic fields that can be achieved conveniently in
the laboratory in contrast to the PASCHEN -BACK effect for FS.
Jz F MF | |F MF
J F
gJ J I F MF | |J I F MF
= gJ MF = g F MF (9.45)
F (F + 1)2
464 9 Hyperfine Structure
(a) z B (b) z B
Fz F Jz
I BJz J
J
S
L Iz
I
Fig. 9.6 (a) Vector model for the HFS Z EEMAN effect in a low magnetic B field, (b) Vector model
for the HFS Z EEMAN effect in a high B field
F (F + 1) + J (J + 1) I (I + 1)
gF = gJ . (9.46)
2F (F + 1)
As shown in Fig. 9.6(a), one may visualize this again by a vector diagram. The
magnetic moment M J = gJ B J / of the electron cloud is averaged due to the
precession of J and
I around F . The contribution of M I may be neglected here
in contrast to MS when treating the fine structure splitting: hence the small but
important difference between the g factors for FS according to (8.11) and for HFS
according to (9.46). The hyperfine splitting (9.44) in a low, external magnetic field
B finally becomes:
A
WMDB = F (F + 1) J (J + 1) I (I + 1) + gF B BMF . (9.47)
2
As a typical example Fig. 9.7 shows the HFS splitting of the Na D lines.
Let us now consider the opposite case: high external fields B BI , for which the
coupling of I and is broken. However, the field is assumed to be not yet high
J to F
enough for breaking the spin-orbit coupling (LS)J . This implies B a/B , a be-
ing the spin-orbit coupling parameter (6.36). We thus use the H AMILTON operator
in its original form (9.39), and have to evaluate it somewhat differently: strongest is
now the interaction of the electron clouds magnetic moment MJ with the external
field B. As visualized by the vector model in Fig. 9.6(b), this leads to a fast preces-
sion of J around B, which is assumed parallel to the z-axis, as usual. This involves
the formation of 2J + 1 (anomalous) Z EEMAN sublevels. At not too high external
magnetic fields B < BJ the next smaller interaction is that between the magnetic
9.3 ZEEMAN Effect of Hyperfine Structure 465
WHFS MF = 3
F= 2
1
3
0 0.94 MHz / 10 - 4 T
-1 MF
-2 F= 2
3 2P
3/2
62.5 MHz 2 -3 1
2 0
2 1
0 3 2P1/2 -1
-1 192 MHz -2
36.1 MHz -2
1 0.23 MHz / 10 - 4 T
1
0 1
16.4 MHz
nm
-1 1
589
0
0
9.6
0 -1
.0 n
58
m
F= MF = 2
2 1
0 0.7 MHz / 10-4 T
3 2S1/2 -1
-2
1772 MHz
1 1
0
-1
Fig. 9.7 Hyperfine structure of the Na D lines with splitting in a low magnetic field for the 3 2 S1/2
ground and the 3 2 P1/2 and 3 2 P3/2 excited states (schematic, not to scale)
MDB = A Jz Iz J+ I J I+ + gJ B Jz gI N Iz B.
H (9.48)
2
The appropriate basis states are now of course |J MJ I MI
for uncoupled J and I .
In 1st order perturbation theory the HFS energies are given by the diagonal matrix
elements of HMDB which in this basis are
since the terms J+ I and J I+ have, according to (B.17), no diagonal compo-
nents. As we have seen in the last section, they cannot be neglected for small fields.
However, for high B fields their contribution averages out (the vanishing diago-
nal matrix elements are the mathematical equivalent for the fast rotation in the xy
plane); hence, the term AMJ MI describes the remaining interaction of the electron
466 9 Hyperfine Structure
3/2
WHFS
MJ + 3/2 1/2 MI
- 1/2
- 3/2
2B B 3/2
F MF + 1/2 1/2
- 1/2
3
2P 3 -3 2B B/3 - 3/2
3/2 3 A
2
2 -2 2B B/3
I = 3/2 2 A
1 1 -3/2
A 0 -1 -1/2
0
-1/2 1/2
2B B 3/2
- 3/2
- 3/2 - 1/2
1/2
zero field low field high field 3/2
Fig. 9.8 Hyperfine splitting in low and high magnetic field, schematic for a 2 P3/2 state with
I = 3/2 (e.g. Na)
cloud with the nuclear spin. We recognize (8.12) in the term gJ B BMJ , which re-
flect spin-orbit with the external magnetic field B, while the nuclear spin interaction
with B is given by gI N MI B relevant only for very high external fields.
Figure 9.8 shows schematically the transition from low to (medium) high
field gI N B A gJ B B for the already rather complex example of the Na
32 P3/2 state. According to (8.11) we have gJ = 4/3, from Table 9.2 I = 3/2 and
gI = 1.478, and the experimental value of the HFS constant is A 20 MHz. For
B 1.8 T this leads to gI N B = A, while for B 103 T one finds gJ B B = A.
The right part in Fig. 9.8, where we have neglected gI N MI B completely, is thus
valid for B = 102 to 101 T. Each FS term is spit here into (2I + 1) = 4 sub-
levels. The distances between the terms are 3A/2 and A/2, for |MJ | = 3/2 and 1/2,
respectively. For the transition from low to high field the non-crossing rule holds
just as for FS splitting in an external magnetic field: States with equal MJ + MI
do not cross (corresponding to ML + MS in Sect. 8.1.5). This leads to the transitions
from MF MJ + MI as illustrated schematically in Fig. 9.8.
For yet higher fields (above a few tesla), today accessible with superconduction
magnets, one has to account also for the term gI N MI B. However, even the total
electronic angular momentum J is then no longer a good quantum number, since
at B 10 T the spin-orbit coupling constant becomes a = gJ B B. This makes
the whole evaluation again more complicated. One obtains a picture similar to the
PASCHEN -BACK effect in Fig. 8.8 on the right. However, each of the 6 sublevels
shown there is now split again into 4 sub-sublevels. In each case the highest belongs
to MI = 3/2.
9.3 ZEEMAN Effect of Hyperfine Structure 467
As the splitting patterns in the case of high fields tends to become rather complex,
we shall now focus on two significantly simpler model cases, and follow their term
energies as a function of the magnetic field in detail. We have already noted that the
magnetic fields needed to decouple I and J are not particularly high. The general
case, where the field may neither be considered low nor high, is thus most often
encountered when studying or applying HFS in contrast to FS. One has to diago-
nalize the full Hamiltonian (9.39) in analogy to the procedure for FS described in
Sect. 8.1.5. And one may simultaneously also solve the case of very high fields. As
a basis, either |(J I )F MF
or |J MJ I MI
states may be used.
Using the |(J I )F MF
basis, I J is diagonal, and one has to evaluate the
matrix elements of Jz and Iz . This is done in a manner quite similar to that in
Sect. 8.1.5. We sketch this here for the most simple case, nuclear spin I = 1/2 and
charge cloud angular momentum J = 1/2 with gJ 2. This is e.g. the case for the
ground state of atomic H with = 0 and J = S = 1/2.
As no complications from coupling electron spin and orbit arise, the splitting
pattern becomes rather clear. The matrix elements Jz / obtained from (C.66) and
(C.67) are summarized as follows:
F MF 11 1 0 1 1 00
F MF
J I F MF |Jz |J I F MF
12 0
= 1 1 0 0 .
1 0 0 0 0 1
2
1 1 0 0 1
0
2
1
0 0 0 2 0 0
Since in this case J and I have the same value, the corresponding matrix for Iz
is almost identical: only the off-diagonal matrix elements have inverse signs cor-
responding to (C.46) and (C.47). The field free term in (9.40) is diagonal in the
(J I )F MF coupling scheme:
MD |J I F MF
J I F MF |V A
= MF MF F (F + 1) I (I + 1) J (J + 1) .
2
For J = I = 1/2 we have F = 1 and 0, and this becomes A/4 and 3A/4, respec-
tively. With the abbreviation
gJ gI
= B N , (9.50)
2 gJ
4 Note that the in front of the root has to be applied judiciously: only (2J + 1)(2I + 1) different
WHFS / A MJ MI A
+1/2 +1/2 MJ MI
2
} A /4
} - A /4
J = 1/2 MF = 1 1/2 -1/2
1
I = 1/2 2 BB
F=1 0
A /4 {
- 3A /4
F=0
{ -1
1 2 B / AB-1
-1
-1/2 -1/2
0 } A /4
} - A /4
-2
-1/2 +1/2
Fig. 9.9 Z EEMAN effect for HFS: transition from low to high magnetic field for I = J = 1/2. Left:
energies according to the B REIT-R ABI formula (9.53). Red and black, full lines refer to MI = +1/2
and 1/2, respectively. Dashed red lines show the energies one would expect without HFS. Right
(not to scale): extrapolation to (moderately) high B fields according to (9.49) however, without
the small gI N MI B term
WHFS / A MJ MI
+1/2 -1/2
NMR } gI N B - A / 2
+1/2 +1/2
164
B / AB-1 EPR
164
-1/2 -1/2
NMR } gI B + A / 2
N
-1/2 +1/2
Fig. 9.10 Z EEMAN effect for HFS at very high B fields. Left: extension of Fig. 9.9 to the (very)
high field situation for the example of the H atom according to (9.51)(9.52). For sufficiently high
field strength B (9.49) holds, now including the gI N MI B term (not to scale). Right: limiting
case A gI N B. Shown are also the potential M1 transitions in the microwave (EPR) and radio
frequency range (NMR)
with equal total projection quantum number, here MF = 0, in the transition to high
fields.
While the term gI N B is indeed negligible for low and moderately strong fields,
it becomes important for extremely high fields today exploited e.g. in EPR spec-
troscopy. This transition from moderately high to very high field is illustrated in
Fig. 9.10. For the limit of very high fields the splitting is indicated on the right
(not to scale): the sign of the nuclear spin orientation for the level with highest
energy has changed as compared to lower magnetic fields! According to (9.49),
470 9 Hyperfine Structure
- 1/2 - 1
-4 - 1/2 0
- 1/2 1
Section summary
The hyperfine Hamiltonian (9.39) in an external magnetic field B consists of
three terms:
I J interaction ( A), interaction of B with the electron spin
( Jz B) and interaction of B with the nuclear spin ( Iz B).
In weak magnetic fields, B BI , the latter may be neglected, the coupled
(I J )F MF HFS-levels split linearly with B (with BI A/(gJ B ) being the
magnetic field the electron feels from the nucleus).
9.4 Isotope Shift and Electrostatic Nuclear Interactions 471
Table 9.3 H HFS for 6 Li in the electronic ground state 2s 2 S1/2 , written in the uncoupled
|J MJ I MI
basis. The nuclear spin is I = 1, the HFS coupling constant A = h 221.864(64) MHz
(WALLS et al. 2003), and with (9.50) and Table 9.2 we have + = 1.0016B , and = 1.0007B
MJ MI 1
2 1
1
2 0 12 1 1
2 1 12 0 12 1
2 + B
1 A
1 0 0 0 0 0
2
1 A
0 0 B B 2 0 0 0
2
2
12 1 0 A
2 2 A2 + B 0 0 0
2 1 A2 + + B
1 A
0 0 0 2 0
2
12 0 0 0 0 A
2 2 B B 0
12 1 0 0 0 0 0 A
2 B
Aside from magnetic dipole interactions V MD and external fields, atomic mass A
(for equal atomic number Z) and shape of the nuclei may also influence the energy
levels of atoms. The so called isotope shifts (IS) and quadrupole interaction do
not change the number of hyperfine levels, but modify their overall position and
splittings. Most of the following considerations are presented again for effective one
electron systems. They may, however, be generalized for multi-electron systems by
summing over all electron coordinates.
The interaction energy between an atomic electron at position r and the charge
distribution n (R) of the nucleus is read from Fig. 9.12:
Ven (r) = V (R, r)n (R)d3 R, (9.54)
e
with V (R, r) = (9.55)
40 |r R|
and the nuclear charge Ze = n (R)d3 R. (9.56)
472 9 Hyperfine Structure
x y
nucleus
Let us discuss the potential created by the electron close to the nucleus. Since the
extension of the nuclear charge cloud is extremely small compared to the electron
cloud, we can safely expand the potential around the origin,
e
3
V 1 2V
3
V (R, r) = + X + X X + , (9.57)
x 0
40 r 2 x x 0
=1
With the first term in (9.58) we have recovered the C OULOMB energy of a point
charge Ze, which is already included in the unperturbed atomic Hamiltonian (and
close to the nucleus is correct also for multi-electron systems). The second term
would be proportional to the electric dipole moment of the atomic nucleus which
is zero for symmetry reasons. The first nonvanishing terms, relevant in the present
context, are the second order terms (9.59) obviously tensor components. We can
always find a coordinate system in which the axes are aligned such that this tensor
is diagonal:
1
V n (R)X X d3 R
2
1
= V n (R)X2 d3 R (9.60)
2
1 2
= V n (R)R d R + n (R) 3X R d R
2 3 2 3
(9.61)
6
A = 234 236
233 235 238
0 1 2 W / cm-1
Fig. 9.13 Example for isotope shift (IS) of spectral lines in the VIS range, here for uranium where
HFS IS. The measured isotope shift of a spectral line at 424.44 nm according to S MITH et al.
(1951) is shown for different isotopes with nucleon number A indicated by vertical black lines.
For comparison a scaling A2/3 is shown by red vertical lines below (see explanation on p. 476)
The judiciously rewritten second line consists of two terms. The first is the so called
volume term, the second is a sum over the components of a traceless tensor, the so
called quadrupole tensor. The corresponding energy shifts are obtained in 1st order
perturbation theory by multiplying Vvol (r) and VQ (r) with the electron probability
distribution |el (r)|2 and integrating them over all r space. This will be explicated
in the next two subsections.
Isotope effects arise for atoms with equal nuclear charge Z and different atomic
mass numbers A. We distinguish two effects of spectroscopic relevance: the so
called mass effect and the volume shift, which in turn are composed of different
contributions. While the former is of simple kinematic nature, the volume shift iden-
tified by Vvol (r) in (9.61)(9.62) is due to the finite size of the nucleus and reflects
small deviations from a pure Z/r potential. It is responsible for the major part of IS
for atoms with large Z.
Valuable information about the atomic nuclei can be gleaned from the IS. A typ-
ical example for an isotope specific optical spectrum is shown in Fig. 9.13. Isotope
shifts for atomic nuclei of different mass and shape have been determined since
the 1950ies quite intensively, and have led to a wealth of spectroscopic data which
contributed substantially to a systematic investigation of nuclear structure (see e.g.
A NGELI 2004). Recent investigations focus on a comprehensive study of artificial
isotopes over a large range of atomic numbers and mass numbers, allowing e.g. to
determine the radii of such species. On the other hand, one performs high precision
measurements for light atoms, in particular for the isotopes of He+ and Li+ . As
already mentioned in Chap. 6, these quantities allow very sensitive tests of theory,
including relativistic and radiative contributions (QED) to atomic energies. Below
we present only a few details.
Mass Effect
As explained already in Sect. 1.8.5 and Sect. 2.6.1, for a precise calculation of
eigenenergies of states in one electron systems one has to replace the electron mass
474 9 Hyperfine Structure
Thus, to compare the optical transition frequencies of different isotopes one does
not need to know the absolute values of the term energies. This fact is of particular
importance for the precise interpretation of the spectra of small atomic numbers.
The shift is largest of course for atomic hydrogen between 1 H and 1 D with
W 1 1 1
= 2.72 104 .
W 1836 1 2
For the LYMAN line this amounts to 22.4 cm1 , in excellent agreement with ex-
periment. A precise comparison requires of course a detailed knowledge of HFS and
L AMB shift for both nuclei, as detailed in Sect. 9.2.5 and Sect. 6.5.4 (for a review
on recent precision measurments see J ENTSCHURA et al. 2011).
Quite generally, a good 1st approximation for the mass effect is obtained by
rescaling a 0 = a0 me /m e and E h = Eh m e /me . It is largest for small atoms. How-
ever, strictly speaking, the above kinematic correction is only exact for one elec-
tron atoms. For multi-electron systems the separation of the nuclear motion is no
longer trivial and one has to consider the motion of the nucleus in respect of all
electrons (see e.g. D RAKE et al. 2005). A detailed analysis shows e.g. for the
He atom that separation of the centre of mass motion leads to an additional term
(m e /M) r 1 r 2 (in a.u.), the so called mass polarization. Its expectation value is
not trivial to compute and depends on the electronic state of the system. Fortunately,
it leads to significant contributions only for small A. Beyond the middle of the pe-
riodic system of elements one may apply (9.63) to very good approximation. (We
note in passing that the mass effect for the uranium case, displayed schematically
in Fig. 9.13, does amount to just 0.0012 cm1 between 233 U and 238 U as com-
pared to the experimental value of 2.20 cm1 , i.e. it is without any spectroscopic
relevance).
On the other hand, only for the small atoms one may hope at all to achieve a
quantitative comparison of two isotopes with high precision. Recent work concen-
trates on the isotopes of He and Li where a precision below 1 MHz can be reached
today. However, as it turns out, even for these seemingly simple atoms a theoret-
ical calculation on such level of precision becomes a formidable task. In addition
to a precise computation of mass polarization it requires inclusion of higher order
radiative corrections (QED), a complete analysis of all HFS contributions as we
have discussed them for the H atom, and finally of the volume terms to be treated
next.
9.4 Isotope Shift and Electrostatic Nuclear Interactions 475
Volume Shift
We now discuss the volume term in (9.61) which describes the interaction of the
nuclear charge distribution with the field of the electron close to the origin. It is only
relevant if the electron has a finite probability at the origin, i.e. for s states.5 We
write the expectation value of R 2 for the atomic nucleus
n (R)R 2 d3 R = Ze R 2 ,
and use the P OISSON equation for the electron charge cloud at r = 0,
el (r) e 2
V |0
V |0 = = el (0) .
0 r=0 0
The volume term complements the 0th order pure C OULOMB term of the inter-
action in (9.58), and contributes to the isotope shift: Due the small but finite ex-
tension of the atomic nuclei, s electrons experience close to the origin a potential
which deviates from Z/r. For larger nuclei this becomes significant. The expecta-
tion value R 2
is directly related to the charge distribution, and thus to the number
of nucleons in the atomic nucleus. It should be noted that (except for the simplest
atoms) experimental measurements can only be compared with theory as difference
Wvol between different isotopes with the same nuclear charge. However, in pre-
cision measurements even higher terms Wvol = C1 R 2
+ C2 R 4
+ can be
recorded (F RICKE et al. 1995).
Figure 9.14 illustrates very schematically the potentials for two different
isotopes A = i and A = k at equal Z, in and near to the nucleus resulting from this
finite extension of the nuclei. One may estimate Wvol from the difference R 2
5A relativistic treatment shows that, in addition to s1/2 electrons, a small contribution also arises
for p1/2 electrons.
476 9 Hyperfine Structure
Table 9.4 Isotope shift for the example of 3 He and 4 He according to M ORTON et al. (2006).
Measured and calculated (Wvol ) values and determination of the average nuclear radius Rc for
3 He, using the known value R (4 He) = 1.673(1) fm
c
of the expectation values for the squared nuclear radii. Summing (9.64) over all s
electrons in the electron charge cloud (j ) leads to
Ze2 2 2
Wvol = R j (0) .
60
The volume shift according to (9.64) should e.g. explain the isotope shift for
uranium, shown in Fig. 9.13. The nuclear radius Rc may be estimated to a rough
approximation by the liquid drop model for atomic nuclei with
which according to A NGELI (2004) reproduces the whole periodic system rela-
tively well6 with RLD = 0.9542 fm (for comparison, the proton radius is Rp =
0.8775(51) fm). Hence, we expect the volume shift to be
Wvol R 2 RLD2
A2/3 . (9.66)
As documented in Fig. 9.13, the isotope shifts for the 233 U to 238 U follow this
scaling rather well.
If one wants to derive reliable estimates for the squares of the nuclear radii R 2
from a measurement of the isotope shift, the electron density at the nucleus has to
be known very precisely and one has to know of course all other parameters which
determine the measured value for a particular transition. Todays experimental ac-
curacy warrants correspondingly elaborate, non-perturbative computational efforts
for some low Z atoms. Table 9.4 illustrates the state-of-the-art. Precision measure-
ments for several 2 3 S1 2, 3 3 P transitions in 3 He and 4 He are shown as derived
6 However,including terms proportional to A1/3 and to A1 in (9.65) leads to better fits for the
experimental data.
9.4 Isotope Shift and Electrostatic Nuclear Interactions 477
by M ORTON et al. (2006). To extract from these data the nuclear radius Rc for the
helium isotope 3 He, one has to account for the hyperfine structure of 3 He (transi-
tions F = 3/2 F = 1/2 and F = 1/2 F = 1/2), and to use the known
value Rc for 4 He.
While for electronic s states and more generally for s1/2 and p1/2 states the
electrostatic HFS interaction between the electron and the nucleus is completely
characterized by the volume shift (9.64), quadrupole interactions originate from a
not completely spherical shape of the nucleus which may be prolate or oblate. If
the potential of the electron cloud has also a non-spherical component at the position
of the nucleus (which requires j > 1/2), a quadrupole shift of the energies arises.
(The quadrupole term disappears for electronic states with j = 0 or j = 1/2 which
have no quadruple moment).
The reformulation of (9.60) into (9.61) corresponds to writing it in irreducible
form as outlined in Appendix F. It is thus advantageous to replace (9.57) by ex-
panding the potential (9.55) in terms of reduced spherical harmonics according to
(F.2). Outside the nucleus (i.e. for r > R), the quadrupole term VQ (r) in (9.61)
may be written as a scalar product of two irreducible tensor operators of rank 2,
as defined by (C.20). Explicitly, the energy shift due to quadrupole interaction is
extracted from (F.5) and becomes in 1st order perturbation theory
(e) 1 (r) 2 (R) Ze2
WQ = 3
C 2 ZeR C 2 = U2 Q2
, (9.67)
40 r 40
where the brackets . . .
indicate averaging, i.e. multiplication with the probability
densities n (R)/(Ze) |el (r)|2 and integration over electronic and nuclear coor-
(r) (R)
dinates, r and R, respectively. The rank 2 tensor operators, C2 and C2 , are the
renormalized spherical harmonics (B.29) in respect of the electronic and nuclear
coordinates. We also have defined two rank 2 tensor operators, Q2 and U2 , which
allow a concise description of the nuclear and electronic charge clouds and their
interaction. Their components are7,8
and the nuclear quadrupole moment Q whose definition (9.71) differs by a factor 2eZ from the
expectation value Q20
.
8 While r k C (, ) = r k 4/(2k + 1)Y (, ) = C (r) are the rescaled (regular) solid har-
kq kq kq
monics, r k1 Ckq (, ) = r k1 4/(2k + 1)Ykq (, ) = Ikq (r) are called (rescaled) irregular
solid harmonics.
478 9 Hyperfine Structure
x y x y
The former definition is identical to (F.14) for = 2, while the latter, albeit similar to
the quadrupole tensor of the electron charge cloud, differs from it by the weighting
factor 1/r 3 instead of r 2 . Since electronic and nuclear coordinates, r and R, are
independent of each other, the averaging . . .
over the respective wave functions is
done independently.
Let us first consider the nucleus: Adopting the direction of the nuclear spin as
axis of reference, and assuming rotational symmetry, the nuclear quadrupole tensor
has only one nonvanishing component:
3X32 R 2
Q20 = R 2 C20 (cos ) = (9.70)
2
Its expectation value in a well defined state |I M
is Q20
= I M|Q20 |I M
. In
principle, this has to be averaged over the population probabilities w(M) unless a
pure |I M
state is prepared. One defines the so called
nuclear quadrupole moment Q = n (R) 3X32 R 2 d3 R (9.71)
= 2eZQ20 ,
with the normalization (9.56). Numerical values are found e.g. in Table 9.2.9
Q may be computed if one knows the structure of the nucleus, i.e. the nuclear
charge distribution n (R). However, in practice one rather uses the HFS to deter-
mine Q and to compare it with suitable nuclear structure models.
We distinguish oblate nuclear shape (pancake like), shown in Fig. 9.15(a),
which corresponds to Q < 0, and prolate (cigar like) shape with Q > 0, as illus-
trated in Fig. 9.15(b).10 The nuclear spin I may be considered responsible for such
deviations of the shape from a pure sphere (see Table 9.2). With (F.25) the matrix
9 Usually,e is not mentioned in published data. As the expectation values Q2q
of the quadrupole
tensor are correctly measured in units of b, to avoid misunderstandings we use for the quadrupole
moment Q the (non-SI) unit eb (see Appendix A.2).
10 It
is important to note that this definition of prolate and oblate only holds for the quadrupole
moment (9.70), defined by the shape of the charge cloud. The sign changes when general tensor
operators are used, constructed from angular momenta (Appendix F.3.2).
9.4 Isotope Shift and Electrostatic Nuclear Interactions 479
3M 2 I (I + 1)
I M|Q20 |I M
= I Q2 I
, (9.72)
(2I + 3)(I + 1)I (2I 1)
with M being the projection quantum number of the nuclear spin I onto a given
laboratory axis (defined e.g. by the z-axis of an oriented atom). The maximum value
of I M|Q20 |I M
is obtained for M = I . Multiplied with 2eZ it is called the spec-
troscopic (nuclear) quadrupole moment:
I (2I 1)
Qs = 2eZI I |Q20 |I I
= 2Ze I Q2 I
. (9.73)
(2I + 3)(I + 1)
With this we may rewrite (9.72) for the nuclear quadrupole moment in a state |I M :
3M 2 I (I + 1)
Q(M) = 2ZeI M|Q20 |I M
= Qs .
I (2I 1)
For further details we refer to Appendix F.3.1 and classical textbooks (e.g. B LATT
and W EISSKOPF 1952).
Summarizing, the spectroscopic nuclear quadrupole moment Qs is the expecta-
tion value of 2ZeQ20 in a nuclear state |I M = I
with respect to the laboratory
frame. It has to be distinguished from the intrinsic nuclear quadrupole moment Qi
with respect to the nuclear symmetry axis which is not identical to the nuclear spin.
Typically, the nuclear spin I does not point into the direction of the and the nuclear
symmetry axis. If the nucleus is assumed to be rotationally symmetric (which is of-
ten the case but not necessarily so) spectroscopic and intrinsic nuclear quadrupole
moments are related by (see e.g. N EYENS 2003)
3MK2 I (I + 1)
Qs = Qi , (9.75)
(2I + 3)(I + 1)
where MK is the projection quantum number of I onto the nuclear symmetry axis.
Even for the maximum value of MK = I we note that Qs < Qi . This is a conse-
quenceof the nuclear axis being statistically oriented on a cone around the nuclear
spin ( I (I + 1) > I ). Only in the limit of large I = MK both quantities become
equal.
11 Note that in the nuclear physics literature several types of normalization of the reduced matrix
We emphasize again that the atomic nucleus (as well as the electron cloud) can
have a quadrupole moment only if its spin is I > 1/2 (or its total angular momentum
J > 1/2, respectively): the reduced matrix element obeys the triangular relation
(I 2I ) = 1, i.e. I 1 must hold.
(r)
Let us now turn to the atomic electron charge cloud: The tensor U2 defined
by (9.69), characterizes its electric field. Its zero component may be written
3z2 r 2 1
U20 = 5
= 3 C20 ( ), (9.76)
2r r
and its matrix elements are
2[3MJ2 J (J + 1)]
J MJ |U20 |J MJ
= J U2 J
. (9.77)
(2J + 3)(2J + 2)2J (2J 1)
may be computed for one electron systems in a straight forward manner. Again,
the expectation value of 1/r 3 has to be evaluated. In the s J coupled scheme
(r)
s J C2 s J
is given by (C.51) and (C.54).
We can now evaluate the overall HFS level splitting due to magnetic dipole and elec-
tric quadrupole interaction. For the quadrupole interaction energy (9.67) we have to
evaluate the matrix elements
I J F MF |Q2 U2 I J F MF
Ze2
WQ = (2I + 1) (2J + 1)
40
" #
I I 2
(1)F +J +I J U2 J
I Q2 I
. (9.79)
J J F
This is evaluated by inserting (9.78), (9.74) and the explicit expressions for the 6j
symbol (B.72). WQ is additive to the magnetic dipole interaction (9.15) as long
as WQ WMD . Except for really heavy atomic nuclei this is usually a reasonable
assumption. Summation of the two terms gives
9.4 Isotope Shift and Electrostatic Nuclear Interactions 481
1
WHFS = WMD + WQ = AK
2
3
2 K(K + 1) 2I (I + 1)J (J + 1)
+B , (9.80)
2I (2I 1)2J (2J 1)
with K = F (F + 1) I (I + 1) J (J + 1).
A is the magnetic dipole hyperfine constant, given by (9.31) and (9.37) for > 0
and = 0, respectively, while for the so called quadrupole coupling constant one
obtains
e 2J 1 1 2J 1 a03 Qs
B= Qs = Eh . (9.81)
40 2J + 2 r 3 2J + 2 r 3 ea02
Sometimes (9.80) is cast into the more general form
6( J )2 + 3
I I
J 2
I
2 2
J
WHFS = AI J + B , (9.82)
2I (2I 1)2J (2J 1)
where the averaging process is again interpreted in the spirit of the vector model,
and (9.80) is recovered by inserting
I
= I (I + 1),
J
= J (J + 1) and I
2 2
J
=
K/2.
A typical example for the influence of quadrupole interaction on HFS is sketched
in Fig. 9.16 as a function of B/A. The coupling corresponds to the first excited states
in the main isotopes of the smaller alkali atoms: for the 3P3/2 state in 23 Na one finds
B/A = 0.155, for the 4P3/2 state in 39 K, B/A = 0.472 (A RIMONDO et al. 1977).
We emphasize again that (9.80) is only valid for I > 1/2, as the quadrupole mo-
ment disappears for I = 0 and 1/2. Correspondingly, (9.78) describing the electron
cloud disappears if J = 0 or = 1/2, as well as for = 0.
Section summary
Mass, size and shape of the atomic nucleus contribute to shifts of hyperfine
energy levels.
Only for small atoms, the simple kinematic mass effect due to the finite nu-
clear mass is of spectroscopic relevance.
To identify the influence of size and shape of the nucleus one expands the
interaction potential between nucleus and electron in a multipole series and
finds that quadratic terms give the first nonvanishing contributions.
482 9 Hyperfine Structure
The finite size of the nucleus, decreases the overall binding energy of s elec-
trons by the so called volume shift, Wvol R 2
which is on the order of
Wvol /Wvol 104 for the largest atoms. For high Z it is the largest contri-
bution to the isotope shift observed for atoms with equal Z and different A.
The quadrupole shift reflects the non-spherical shape of atomic nuclei with
finite spin I > 1/2, represented by the nuclear quadrupole moment (9.71).
It changes the energy of individual hyperfine levels F if the electron charge
cloud has also a quadrupole moment (J > 1/2 and > 0).
The overall HFS splitting is given by (9.80), with the quadrupole and magnetic
dipole HFS coupling constants, A (9.31) or (9.37) and B (9.81), respectively.
R AMSEY (1989) and his collaborators have performed since 1946 ground breaking
work in microwave and radio frequency resonance spectroscopy on magnetically or
electrically prepared and guided atomic and molecular beams. Typical experimental
equipment and methods which are used still today even for quite complicated
molecules are still based essentially on the original design from the 1970ies (see
e.g. G ALLAGHER et al. 1972). Of course, detection schemes, data acquisition and
storage procedures have been improved by orders of magnitude in the mean time.
Figure 9.17 illustrates the principle of an electrically focussing molecular resonance
apparatus for studying magnetic dipole transitions (M1) in the RF and MW domain.
First, one has to create a molecular beam of the atoms or molecules to be stud-
ied (see footnote 24 in Chap. 1). They are then state selected in electric quadrupole
fields and focused onto the detector. Without going into details, we note that this
9.5 Magnetic Resonance Spectroscopy 483
A z C B
quadrupole DC and HF field quadrupole
beam detector
source
beam axis
DC mixer
power supply detection
and data
HF acquisiton
ramp for tuning synthesizer
Fig. 9.17 Scheme of a R AMSEY molecular beam resonance apparatus for the observation of M1
transitions in molecules according to G ALLAGHER et al. (1972); for details see text
signal
0
41705 41707 41709
transition frequency / kHz
Fig. 9.18 Experimentally observed and fitted radio frequency spectrum (red dots and lines) of nine
S TARK split HFS transitions (F1 = 3/2 7/2) in LiI according to C EDERBERG et al. (2005). The
LiI molecules are in the ground vibrational state (v = 0), slightly rotationally excited (J = 3). The
black lines represent a computed stick spectrum
S B/ h = ge B MS B/ h 2 14 GHz 1 B/ T,
W S / h = M (9.83)
2
where holds, depending on the spin orientation being parallel or antiparallel to
the direction of the field. Transitions may be induced between both states by a
microwave field. In this case, the electromagnetic field acts with its magnetic field
component (see Appendix H.1.1) onto the magnetic moment of the spin and induces
a magnetic dipole transition (M1) as described in Sect. 5.4. The selection rules do
9.5 Magnetic Resonance Spectroscopy 485
MS = - 1/2 - 300
- 500
MI = +1/2
- 750 - 500
Fig. 9.19 Typical (M1) transition frequencies in NMR (left scale, black) and EPR (right scale,
red) as a function of the external magnetic field B. In EPR the electron spin is flipped, in NMR the
nuclear spin (specifically, this graph is for 1 H)
not require a change of , so that they typically occur within one n level (also pos-
sible for = 0). Of course, again the total angular momentum, including the photon
angular momentum Jph = 1 is conserved: The triangular relation (Ja 1Jb ) = 1 must
hold. The most simple case is Ja = Jb = S = 1/2, where for the spin projection we
have
MS = 1. (9.84)
WS (b)
MW source 1 circulator 3 detector fixed frequency
B
2
irradiation reflected reflected microwave
from source signal power
(c)
MW cavity B
(a) with probe differentiated signal (- 1)
Fig. 9.20 (a) Schematic of the microwave bridge in a CW-EPR spectrometer. The microwave
enters at point 1 the so called circulator which guides it a point 2 completely onto the probe. In
contrast, the reflected signal, entering at point 2, is routed towards point 3. (b) Spectra are recorded
at fixed frequency, the magnetic field B is tuned so that the resonance energy WS changes (straight
red lines). (c) The resulting absorption signal (full black line) as a function of B; the signal is
differentiated so that a dispersion like line shape (red, dashed) emerges
circulator
main magnet
Fig. 9.21 Scheme of the whole CW-EPR spectrometer with the microwave bridge Fig. 9.20. The
magnetic field is provided by the main magnet and the modulation coils. The probe is positioned
in the resonator cavity and the signal is detected by a phase sensitive detector (PSD)
plex molecule the unpaired electrons interact e.g. with the nuclear spins of different
atoms, leading to characteristic spectroscopic patterns, from which one may glean
very detailed information about the molecular structure. In the standard, continu-
ous EPR method (CW-EPR) a spectrum is obtained at fixed frequency of the mi-
crowave radiation by tuning the magnetic field. A characteristic setup is shown in
Figs. 9.209.21. At each resonance, losses occur in the microwave resonator which
can be detected with high sensitivity. In contrast to optical spectroscopy, one typ-
ically observes here differentiated signals (dispersion like) which often allow for
better resolution of the structures. Figure 9.22 shows an example for an experimen-
tally obtained EPR spectrum. We cannot go into the details here.
As the resolution increases with B (requiring correspondingly higher resonance
frequencies for the same transitions), modern EPR spectroscopy uses frequencies
9.5 Magnetic Resonance Spectroscopy 487
(differentiated)
anion as an example obtained
absorption
in the X band, after SDBS
(2013). The different HFS
coupling constants a for the
various atoms lead to a
pronounced structure of the
spectra 0 20 40 60 80
B / mT
The frequencies for nuclear magnetic resonance (NMR) transitions with protons,
1 H, are also indicated in Fig. 9.19. For the energies of the nuclear spin states |I M
I
in an external magnetic field and the transitions between them, rules hold in com-
plete analogy to those for the electron spin states. Specifically, for an uncoupled
proton spin (e.g. in an atomic 1 S0 state), the energy splits into two levels in an ex-
ternal B field with
1
Wp / h = gp N MI B/ h = 42.56 MHz B/ T,
2
488 9 Hyperfine Structure
RF oscillator RF receiver
receiver coil
tuning
Fig. 9.24 Simplified schematic of an NMR spectrometer. A (magnetic) RF field at fixed frequency
is applied perpendicularly to the main magnetic field B. The latter is tuned through the resonances
which induce a current in the receiver coil due to the spin flip upon RF absorption. This signal is
amplified, filtered and recorded in a PC
with gp = 5.58569471(5) being the g factor of the proton and MI = 1/2 its spin
projection quantum number (magnetic quantum number). Again, M1 transitions
may be induced by interaction of the nuclear magnetic moment with magnetic field
component of an electromagnetic wave or with a local, oscillating magnetic field.
The transition frequencies for these nuclear magnetic resonances lie in the RF range
up to several 100 MHz:
The experimental methods of NMR are in principle similar to those for EPR
except that microwave radiation and techniques are replaced by RF. Schematically,
a typical NMR spectrometer is shown in Fig. 9.24.
NMR spectroscopy has been developed to one of the most powerful tools for
structural analysis of complex molecules, up to proteins, since the resonance fre-
quency of each observable nucleus is highly sensitive to its local environment. Char-
acteristic is a diamagnetic screening, the so called chemical shift , typically given
in ppm in respect of the standard reference material tetramethylsilane (TMS), so
that the modified NMR transition frequencies are given by:12
Wp / h = gp N (1 )B/ h. (9.87)
Characteristic features in the spectra reflect the magnetic fields of neighbouring nu-
clear moments.
Figure 9.25 shows a typical experimental 1 H-NMR spectrum for ethanol as ex-
ample. It is characterized by the influence of neighbouring protons: the N = 3
protons of the methyl radical form a quartet (Itot = N 1/2 = 3/2, degeneracy
2 Itot + 1 = N + 1 = 4; so called N + 1 rule). They change the NMR signal
12 If the electron charge cloud involves unsaturated p orbitals, the chemical shift may also be para-
magnetic.
9.5 Magnetic Resonance Spectroscopy 489
absorption
at 89.56 MHz in CDCl3 , CH2 sees
according to SDBS (2013) CH3
OH
4 3 2 1 0
NMR shift / ppm
A numerical value is ca. 6.9 106 for protons in a magnetic field of 1 T at room
temperature. This means, the relative difference in population density is less than
105 . The expected absorption signal is then
with the E INSTEIN Bba absorption coefficient and the spectral radiation density
() (see Chap. 4). The number of absorption and emission processes is propor-
tional to the occupation probability of the states. Hence, both processes occur nearly
equally often. The net absorption rate which may be detected is thus very small. The
fact that NMR is nevertheless very sensitive is due to the large number of absorbing
molecules in the solid or liquid probe (order of magnitude 1022 1023 ). In contrast,
in atomic beam experiments one always deals with just a few particles, which may,
however, be detected with a 100 % efficiency, due to the high population differ-
ence. With (9.88) and (9.89) one immediately sees that for a good NMR signal high
magnetic fields and low temperatures and crucial.
A significant enhancement of the detection probability for NMR (as well as for
EPR) is achieved by pulsed excitation with an RF pulse of significant bandwidth.
One detects the time dependent response of the system to this excitation and av-
erages over many such measurements. The F OURIER transform of this signal gen-
erates the true NMR spectrum, providing simultaneously information on the relax-
ation properties of the system. Significant further improvements have been intro-
duced by so called 2D or even multi dimensional procedures which employ several
RF pulses of different frequencies in judiciously chosen time sequences. In this way,
one may follow the systems response to an excitation of a specific nuclear spin, and
identify correlations among different, active nuclei. Spatial anisotropy in molecu-
lar crystals may significantly blur the resolution, particularly so in solid state NMR
490 9 Hyperfine Structure
Section summary
Magnetic resonance spectroscopy exploits M1 transitions between different
magnetic sublevels of one n level in the RF and MW region (thus avoiding
D OPPLER broadening). Today, magnetic fields up to 30 T are used. The
transition frequencies for electron spin and proton spin flip are summarized in
Fig. 9.19.
Classical molecular beam spectroscopy, still used today, employs electric or
magnetic deflection and focussing to detect atoms and molecules sensitively
in well defined magnetic substates. These are depleted in a magnetic field by
the transitions studied.
Acronyms and Terminology 491
In EPR spectroscopy one studies electron spin flip transitions in the MW spec-
tral region. It is applicable to molecules with unpaired electrons in the liquid
or solid state. The g tensor of the electron acts as a highly selective probe for
structural analysis of its chemical environment, e.g. in large molecular radi-
cals.
NMR spectroscopy is today one of the most powerful methods for chemical
analysis. Mostly 1 H spin flips are used, but other nuclei such as 13 C are also
serve as probes in large molecules. The observed transition frequencies are
very sensitive to the local environment of the nuclei.
References
A NGELI , I.: 2004. A consistent set of nuclear RMS charge radii: properties of the radius surface
r(N, Z). At. Data Nucl. Data Tables, 87, 185206.
A RIMONDO , E., M. I NGUSCIO and P. V IOLINO: 1977. Experimental determination of hyperfine-
structure in alkali atoms. Rev. Mod. Phys., 49, 3175.
B LATT , J. M. and V. F. W EISSKOPF: 1952. Theoretical Nuclear Physics. New York: John Wiley
& Sons, Inc., 1979, Springer-Verlag New York edn.
B REIT , G. and I. I. R ABI: 1931. Measurement of nuclear spin. Phys. Rev., 38, 20822083.
C EDERBERG , J. et al.: 2005. An anomaly in the isotopomer shift of the hyperfine spectrum of
LiI. J. Chem. Phys., 123, 134321.
C RAMPTON , S. B., N. F. R AMSEY and D. K LEPPNER: 1963. Hyperfine separation of ground-
state atomic hydrogen. Phys. Rev. Lett., 11, 338.
D RAKE , O. W. F., W. N RTERSHUSER and Z.-C. YAM: 2005. Isotope shifts and nuclear radius
measurements for helium and lithium. Can. J. Phys., 83, 311325.
E RNST , R. R.: 1991. The N OBEL prize in chemistry: for his contributions to the development of
the methodology of high resolution nuclear magnetic resonance (NMR) spectroscopy, Stock-
holm. http://nobelprize.org/nobel_prizes/chemistry/laureates/1991/.
F RICKE , G., C. B ERNHARDT, K. H EILIG, L. A. S CHALLER, L. S CHELLENBERG, E. B. S HERA
and C. W. D EJAGER: 1995. Nuclear ground state charge radii from electromagnetic interac-
tions. At. Data Nucl. Data Tables, 60, 177285.
G ALLAGHER , T. F., R. C. H ILBORN and N. F. R AMSEY: 1972. Hyperfine spectra of 7 Li35 Cl
and 7 Li37 Cl. J. Chem. Phys., 56, 59725979.
ISO 21348: 2007. Space environment (natural and artificial) process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
JACKSON , J. D.: 1999. Classical Electrodynamics. New York: John Wiley & sons, 3rd. edn., 808
pages.
J ENTSCHURA , U. D., A. M ATVEEV, C. G. PARTHEY, J. A LNIS, R. P OHL, T. U DEM, N. KO -
LACHEVSKY and T. W. H NSCH : 2011. Hydrogen-deuterium isotope shift: from the 1s2s-
transition frequency to the proton-deuteron charge-radius difference. Phys. Rev. A, 83, 042505.
K ARSHENBOIM , S. G.: 2005. Precision physics of simple atoms: QED tests, nuclear structure
and fundamental constants. Phys. Rep., 422, 163.
K RAMIDA , A. E.: 2010. A critical compilation of experimental data on spectral lines and energy
levels of hydrogen, deuterium, and tritium. At. Data Nucl. Data Tables, 96, 586644.
K RAMIDA , A. E., Y. R ALCHENKO, J. R EADER and NIST ASD T EAM: 2013. NIST Atomic
Spectra Database (version 5.1), NIST. http://physics.nist.gov/asd, accessed: 7 Jan 2014.
L AUTERBUR , P. C. and S. P. M ANSFIELD: 2003. The N OBEL prize in physiology or medicine:
for their discoveries concerning magnetic resonance imaging, Stockholm. http://nobelprize.
org/nobel_prizes/medicine/laureates/2003/.
M ORTON , D. C., Q. W U and G. W. F. D RAKE: 2006. Nuclear charge radius for 3 He. Phys.
Rev. A, 73, 034502.
N EYENS , G.: 2003. Nuclear magnetic and quadrupole moments for nuclear structure research on
exotic nuclei. Rep. Prog. Phys., 66, 633689.
References 493
Overview
In this final chapter of Vol. 1 much comes together what has been introduced
in the preceding text. In Sects. 10.110.2 we give an overview on the classi-
cal methods for computing multi-electron wave functions, applying what we
have learned in Chap. 3 and Chap. 7. In Sect. 10.3 a short excursion into den-
sity functional theory is added. With increasing atomic number Z not only
the number of electrons grows and hence the complexity of the problem.
Also the significance of different types of interaction changes: for light atoms
exchange interaction was dominant, and LS coupling gave a good descrip-
tion as described in Chap. 6. With increasing spin-orbit interaction, LS cou-
pling is no longer adequate, as we shall illustrate in Sect. 10.4. The energy
scale as a whole changes roughly Z 2 . Transition energies for quantum
jumps within the outermost electron shells are still in the VIS or UV spectral
range; the reader may familiarize him- or herself by way of a few characteris-
tic examples with the zoo of energy levels and coupling schemes in complex
atoms. Changes in the inner shells are associated with emission or absorption
of X-ray radiation, which treated in Sect. 10.5 complementing Chap. 4. In
addition, our understanding of photoionization (Chap. 5) is deepened by some
examples of multi-electron systems. Finally, Sect. 10.6 introduces the reader
to modern sources for the generation of X-rays.
The most generally used approach for solving the multi-electron problem remains
the independent particle model as in essentially all the preceding chapters: Each
of the N electrons moves in the C OULOMB potential of the nucleus and in the av-
eraged potential of all other (N 1) electrons (also called mean field). Any direct
correlation between the electrons due to their momentary, individual coordinates is
neglected this is the key assumption. The motion of each electron thus depends
on those of the other electrons only in as far as they contribute to the average re-
pulsion potential. The character of radial wave functions (electron orbitals) in such
a screened C OULOMB potential will be very similar to those in the hydrogen atom.
This will be manifested e.g. in the asymptotic behaviour at small and large r, and in
the number of zero-crossings. Nonetheless, pure hydrogen like orbitals are typically
no longer a good 0th order approximation for larger N and Z. One thus has to solve
the S CHRDINGER equation numerically, and then to refine the independent parti-
cle model step by step. The most prominent observations may be described already
in this manner without further effort.
We consider an atom with nuclear charge Z and N electrons and start again by
just including electrostatic forces. For the moment, we neglect are all magnetic in-
teractions as well as relativistic and QED effects. Following the approach taken in
Chap. 7 for the two electron system He, the total energy is obtained as the sum over
all energies of each electron: kinetic energy, C OULOMB attraction by the nucleus,
and C OULOMB repulsion from all other electrons. Extending (7.4), the H AMILTON
operator for the multi-electron system becomes:
N 2
N N
=
p j Ze2 1 e2
H + (10.1)
2me 40 rj 2 40 rj k
j =1 j k=j
with rj k = |r j r k |. (10.2)
The factor 1/2 in front of the double sum compensates the double appearance of the
repulsive terms 1/rkj and 1/rj k , thus allowing to write the equation symmetric in j
and k. In atomic units (10.1) reads
N
N N
= 1 1
H hej + (10.3)
2 |r j r k |
j =1 j k=j
1 Z
with
hej =
j , (10.4)
2 rj
10.1 Central Field Approximation 497
the latter representing the unscreened single electron H AMILTON operator in the
field of the nucleus. The full S CHRDINGER equation to be solved is
(q 1 , q 2 , . . . , q N ) = W (q 1 , q 2 , . . . , q N ),
H (10.5)
q j referring to position (rj , j , j ) and spin (sj ) coordinates of electron j .
In contrast to the simple models for He in singly excited states, for multi-electron
systems in general, the repulsion terms in (10.3) can no longer be regarded as a
(small) perturbation especially when summed over all electrons. Strictly speaking,
the terms |r j r k |1 make the potential non-isotropic and different for each
electron.
In addition, we have to recall that the N electrons are indistinguishable and oc-
cupy N different sets of quantum numbers, each of which is characterized by a
specific set of quantum numbers, different from all others (PAULI exclusion princi-
ple). The total wave function must be antisymmetrized with respect to exchange of
any two electrons.
This all leads to substantial complications, and any attempt to solve the
S CHRDINGER equation (10.5) in a direct way even in numerical form will be
prohibitively elaborate. One thus has to resort to intelligent approximation strate-
gies, the two most important of which will be sketched in Sect. 10.2 and Sect. 10.3.
1 Hereand in the following the index stands for a set of quantum numbers {n m } (later also
for {n m ms }. If no ambiguities can arise, we abbreviate q j , as well as r j and sj by j as
we also have done in Chap. 7.
498 10 Multi-electron Atoms
where | |
stands for integration over all other electrons k = j . For very small
as well as for very large rj this potential will behave as discussed for the active
electron in alkali atoms (see (3.10) and Fig. 3.10).
With this potential the full Hamiltonian (10.3) may be rewritten as composed of
a centrally symmetric part
N
c = 1
H hj + Vj (rj )
e
(10.9)
2
j =1
and a (hopefully) small perturbation term which is the difference between (10.3)
and (10.9):
+N
N
,
1 1
nc = H
H H c = Vj (rj ) (10.10)
2 rj k
j k=j
Hnc contains all non-spherical parts of the interaction. Its diagonal term disappears
in 1st order perturbation theory and is neglected in the central field approximation.
However, one may considered H nc as a perturbation in higher order. Typically it
is small if one has already used reasonable orbitals for determining Vj (rj ). The
so called M LLER -P LESSET perturbation theory accounts for just this remaining
difference, by computing 2nd (MP2), 3rd (MP3) and 4th order (MP4) corrections.
Obviously one has to find the one-particle wave functions (orbitals) 1 (r), 2 (r)
to N (r) for all N electrons ignoring for the moment the spin dependence. In
order to derive the corresponding one-particle S CHRDINGER equations, one fol-
lows a variational approach, similar to that introduced in Sect. 7.2.5. Starting from
the product ansatz (10.6) and the Hamiltonian (10.9) one minimizes the total energy
|H|
of the system. We do not go into details, but the general idea outlined here
leads to plausible results.
The orbitals (r) are found as eigenfunction of N S CHRDINGER equations
e
h + V (r) (r) = Wn (r), (10.11)
Of course, the radial functions Rn (r) are here not simply hydrogenic wave func-
tions. Rather, one has to compute
Rn (r) = un (r)/r
numerically as already practiced for quasi one electron systems in Sect. 3.2.4 from
the radial S CHRDINGER equation
d2 un Z ( + 1)
+ 2 Wn + V (r) un (r) = 0. (10.13)
dr 2 r r2
So far, this all looks straight forward. However, the determination of the averaged
centrally symmetric repulsive potentials according to (10.8) is by no means trivial:
in principle one already needs to know the orbitals of all electrons for the averaging
procedure!
Assume for a moment we know the orbitals. Then, averaging over the repulsion
from all other electrons k = j implies to insert (10.6) into (10.8) from left and
right and to integrate over all other coordinates. As the orbitals are orthonormal, the
relatively simple expressions
N
N
1 2
3 | (r )|
V (r) = r
r = d r (10.14)
|r r | |r r |
= =
emerge. While refers to the quantum numbers of the orbital for electron j , the
summation has to be carried out over all other N 1 occupied orbitals, from = 1
to N .
Now, one may formally insert this expression back into (10.11) and obtains for
every occupied orbital = 1 to N :
+ N ,
)|2
3 | (r
he (r) + d r (r) = E h (r). (10.15)
|r r |
=
These coupled integro-differential equations for (r) are called H ARTREE equa-
tions. Each of them contains the solutions (r ) of all others! Its a major task to
solve them. In practice this is done iteratively, as we shall discuss in the next sec-
tion. We point out again that the potentials V (r) may be different for each electron
orbital.
After all this is done, in 1st order (H ARTREE approximation) the total wave func-
tion defined by (10.6) is obtained from the N orbitals . The corresponding
expectation value of the total energy WHartree {1, 2, . . . , N } of the system is derived
from (10.9), using (10.14) for the mean repulsive potentials:
N
N N
c |
= 1
WHartree = |H + J (10.16)
2
=1 =1 =
500 10 Multi-electron Atoms
Fig. 10.1 Scheme for finding the self-consistent field potential for a multi-electron system (SCF
method) illustrated for the H ARTREE approximation
e
with = (r)
h (r) = d3 r (r)
he (r) (10.17)
1
and J = (r) r r r (r) r (10.18)
| (r)|2 | (r )|2
= d3 rd3 r .
|r r |
Here is the energy a single electron would have in the unscreened C OULOMB
potential of the nucleus, and J the averaged repulsions between two electrons
which we have already introduced as C OULOMB integral in Sect. 7.3.3.
Note that WHartree is not equal to the sum over all orbital energies E h from
(10.15) due to the factor 1/2. Without this factor the repulsive energies would
be counted twice.
The search for the best set of one-particle wave functions for all N occupied or-
bitals rests upon an iteration procedure first introduced and explored by H ARTREE.
It was later on justified by S LATER, based on the variational method. It is sketched
schematically in Fig. 10.1. One first guesses a plausible 0th order approximation
for the central field potentials V (r) for all electrons of the configuration under in-
vestigation.
10.1 Central Field Approximation 501
A possible ansatz for this potential is the T HOMAS -F ERMI model, to be dis-
cussed in the next section. With this potential one computes all electron orbitals in
the configuration, and uses these in turn to derive improved potentials V according
to (10.14). With these potentials one repeats the procedure. At the end of each in-
tegration one checks how well the wave functions obtained in the calculation agree
with those at the input. As long as the agreement is poor, one repeats the computa-
tion until it converges, i.e. until the potential reproduces itself within a preset limit.
The whole procedure is called self-consistent field method (SCF).
The classical textbook approach for the interaction potential in 0th order is the
so called T HOMAS -F ERMI model. One assumes that the (many) electrons inside
the atom follow essentially the rules for a F ERMI electron gas, as introduced in
Sect. 2.4.3.
However, in the present case the electrons are not contained in a box. Rather they
are distributed in a potential V (r) < 0 which is determined by the C OULOMB poten-
tial of the nucleus and the screening from all electrons, just as outlined in Sect. 3.2.3.
Thus, the maximum kinetic energy, F , electrons may have inside the atom is as-
sumed to be position dependent, and given by F (r) = V (r). The electron density
(2.59) becomes also position dependent:
1 2me 3/2 3/2
.
Ne (r) = V (r) (10.19)
3 2 2
On the other hand, one assumes the electrostatic potential V (r)/e to be related to
the charge density eNe (r) by the P OISSON equation:
e2
V (r) = Ne (r). (10.20)
0
We may eliminate Ne (r) and obtain the differential equation
e2 8 2m3e 3/2
V (r) = V (r) or (10.21)
40 3 3
8 2 3/2
V (r) = V (r) in atomic units, (10.22)
3
from which V (r) can be derived. Since the potential is considered spherically sym-
metric, it depends only on r and
= (1/r)d2 /dr 2 . The boundary conditions at
very small and very large r are those for the screened C OULOMB potential given by
(3.10). Thus, one scales this potential as
Z
V (r) = (r/b), (10.23)
r
502 10 Multi-electron Atoms
V(r) / Eh
valence electron sees. For -40
comparison also shown is the
C OULOMB potential of the -60
neon nucleus 10/r (limit
for r 0) and the -80 -Z/r
completely screened potential
1/r (limit for r )
-100
d2 3/2
T HOMAS -F ERMI equation = . (10.25)
dx 2 x 1/2
The limiting values of (x) are 1 and 0 for x 0 and x , respectively. The
T HOMAS -F ERMI equation is treated in the literature on differential equations (see
e.g. Z WILLINGER 1997) and describes with (10.23) a universal potential which
scales with Z. Its relation to the electron density (in a.u.) is
3/2
Z
Ne (x) = for x 0. (10.26)
4b3 x
In the literature, one finds numerical data for the T HOMAS -F ERMI function (x)
as well as approximate analytical formulas, e.g. in L ATTER (1955):
(x) 1 + 0.02747x 1/2 + 1.243x 0.1486x 3/2 (10.27)
1
+ 0.2302x 2 + 0.007298x 5/2 + 0.006944x 3 .
Note that this (x) with (10.23) gives the full atomic model potential with
the nuclear charge Z, screened by Z electrons as experienced by an infinitely
small probe charge. For one of the atomic electrons one might e.g. use the poten-
tial [(Z 1)(r/b) 1]/r as a initial input into the SCF calculation of atomic
orbitals. Figure 10.2 shows this model potential for the example of Ne (Z = 10), in
comparison to the pure C OULOMB limits for very small and very large distances.
Section summary
The Hamiltonian (10.1) for a multi-electron system with nuclear charge Z and
N electrons contains in addition to a sum over all single electron Hamiltoni-
10.2 HARTREE-FOCK Method 503
ans the repulsive potentials among the N electrons which make it the overall
interaction potential non-isotropic.
A surprisingly good key approximation is the independent particle model, ac-
cording to which the total wave function may be written as product (10.6) of
all single electron orbitals. It is based on the central field approximation which
represents the electron repulsion by a centrally symmetric screening potential.
Mathematically this may be cast into a set of N coupled integro-differential
equations. Without antisymmetrization (i.e. neglecting electron exchange)
these are the H ARTREE equations (10.15).
Typically, they are solved by an iterative procedure in which at each iteration
step the N orbitals are obtained from numerical integration of the radial equa-
tion; the screening potentials are obtained from averaging over the orbitals of
the previous iteration step. This is repeated until the orbitals are reproduced
in each step, i.e. until they are self consistent (SCF approximation).
An initial estimate for the repulsive screening potential may be obtained from
the T HOMAS -F ERMI model, which is based on a statistical electron distribu-
tion in the C OULOMB potential Z fold charged nucleus. A closed expression
may be obtained from (10.23) with (10.27).
The above discussion of the H ARTREE SCF method was meant as a first introduc-
tion on how to calculate wave functions and energies for multi-electron systems.
So far, we have completely neglected the PAULI principle. It demands for some
major additions to the procedure, first introduced by F OCK and independently by
S LATER the so called H ARTREE -F OCK (HF) method (also HF approximation or
HF theory). It provides the basis for essentially all advanced methods of atomic and
molecular structure computation and modern quantum chemistry. In detail it is a
rather sophisticated theory and we cannot give here a tutorial for those who actu-
ally want to use it (we refer to the numerous textbooks on quantum chemistry, e.g.
S ZABO and O STLUND 1996). In the following we just sketch some of its essential
ingredients.
To account for the fact that the N electrons in an atom are indistinguishable
fermions, we have to anti-symmetrize the total wave function in the S CHRDINGER
equation (10.5). The H AMILTON operator is still represented by (10.3) it is not
explicitly spin dependent and we start with product wave functions similar to
(10.6).
However, HF theory introduces so called spin-orbitals (q ) = s (r j ) (j )
(as already applied for He in Sect. 7.3.1) and uses products of these:
504 10 Multi-electron Atoms
(q 1 , . . . , q N )
= 1 (q 1 )2 (q 2 ) . . . (q 2 ) . . . N (q N )
= 1s (r 1 )1 (1)2s (r 2 )2 (2) . . . s (r j ) (j ) . . . N
s
(r N )N (N ).
(q 1 , q 2 , . . . , q N )
1 (q 1 ) 2 (q 1 ) ... (q 1 ) . . . N (q 1 )
1 (q 2 ) 2 (q 2 ) ... (q 2 ) . . . N (q 2 )
1 . . . ... ... ... ... . . .
= . (10.28)
N ! 1 (q j ) 2 (q j ) ... (q j ) . . . N (q j )
... ... ... ... ... . . .
1 (q ) 2 (q N ) ... (q N ) . . . N (q N )
N
1
|
= det 1s (1)1 (1) . . . s (j )j (j ) . . . N
s
(N )N (N ) . (10.29)
N!
In general, the spatial orbitals for different electrons will be different. If one fully
allows for this possibility, one speaks of unrestricted H ARTREE -F OCK (UHF).
For closed shells, one may assume that each spatial orbital is occupied twice,
once by an electron with spin and once by an electron with spin. This approxi-
mation is called (closed shell) restricted H ARTREE -F OCK (RHF). It is obviously a
10.2 HARTREE-FOCK Method 505
reasonable approximation for the He ground state, since the two electrons are abso-
lutely identical, both being in a 1s orbital:
1 s
1s 2 (1, 2) = det 1s (1)(1)1s s
(2)(2)
2!
(1)(2) (1)(2)
= 1s
s s
(r 1 )1s (r 2 ) .
2
This is exactly the result (7.39) for the He ground state: a singlet state with total spin
S = 0, MS = 0. Things are more complicated in the excited state apart from the
trivial cases (1) = (2) and (1) = (2). The S LATER determinant in UHF reads
now
1 s
1sn (1, 2) = det 1s s
(1)(1)n (2)(2)
2!
s (r ) s (r )(1)(2) s (r ) s (r )(1)(2)
1s 1 n 2 1 1s 2
= n .
2
Note that this is neither a singlet nor a triplet state. Thus, as we know from Sect. 7.3,
these are not eigenstates of the Hamiltonian (10.9) for the excited He system. We
have already discussed in Sect. 7.3.4 how to amend this fact by constructing singlet
and triplet states from linear combinations of S LATER determinants.
Now let us have a brief look at Li. Its ground state configuration being 1s 2 2s, the
appropriate S LATER determinant would read
1 s
1s 2 2s (1, 2, 3) = det 1s s
(1)(1)1s s
(2)(2)2s (3)(3)
3!
1 s
or alternatively = det 1s (1)(1)1s s s
(2)(2)2s (3)(3) .
3!
In the first case electron 1 and 3 have identical spin, in the second case electron 1
and 3 have opposite, 2 and 3 identical spin. Obviously the two electrons in the 1s
orbital are no longer fully equivalent (as indicated by 1s ). Thus it makes sense to
apply UHF. We shall see in a moment how this works out when actually determining
the spin-orbitals in a H ARTREE -F OCK calculation. Clearly, the UHF procedure will
be more expensive than RHF, and hence one often restricts calculations for large
system with many electrons to RHF.
We finally mention, that again orthonormality is required for all sets of orbitals.
However, orthogonality between orbitals with opposite spin is automatically en-
sured since
=
= 0. This eventually leads to one set of orthogonal spatial
orbitals for spin and one orthogonal set for spin which are not mutually or-
thogonal.
506 10 Multi-electron Atoms
One thus may generate fully antisymmetric states | LSML MS
, which are eigen-
states of the H AMILTON operator and simultaneously eigenstates of the total an-
gular momentum and of the total spin.
We have already constructed such states for atomic He. For the ground state of
He the above S LATER determinant provides such a state. For the excited states,
however, one needs for the diagonalization of H c linear combinations of several
S LATER determinants in order to obtain RUSSEL -S AUNDERS (LS) coupled eigen-
states | LSML MS
. In Sect. 7.3.4 we have discussed this already in detail for the
example of He atoms in their first excited states.
In the H ARTREE -F OCK approximation (HF) one determines the S LATER determi-
nant (10.29) in such a way that it minimizes the expectation value of the H AMILTON
operator (10.3) which (to 1st order) is then the total energy of the system, replacing
(10.16):2
N
N N
|
= 1
WHF = |H + [J K ]. (10.31)
2
=1 =1 =1
2 Onemay wonder why the sum does not exclude = . Note, however, that in this case
C OULOMB and exchange integral just cancel each other.
10.2 HARTREE-FOCK Method 507
We have met these already in Sect. 7.3.3 for the He case. They are so to say
representatives of exchange interaction. The scalar product in front of the integral is
=1 if the orbitals have the same spin, it is =0 if they have opposite spin.
We cannot go into details of the variational procedure. The main outcome is again
a set of N coupled integro-differential equations, this time for the spin-orbitals.
These H ARTREE -F OCK equations may be written as
+ N , + N ,
he (r) + Vd (r) (r) Vex (r) (r) = E h (r). (10.33)
=1 =1
The summations are over all occupied spin-orbitals (see footnote 2).
The functions Vd (q j ) and Vex (q j ) are called direct potentials (or C OULOMB
terms) and exchange potentials, respectively. They are, however, not potentials in the
standard sense. Rather, they are defined through the spin-orbitals (q j ), with q =
r, s, which in turn are to be determined from the HF equations. The direct potential
corresponds essentially to the screening term in the central potential (10.14) of the
H ARTREE method:
(r ) (r )
1
Vd r = q r r q = d3 r . (10.34)
|r r |
In principle, this matrix element of 1/|r r | implies integration over the spatial
coordinates r and the scalar product of spin functions. However, since bra and ket
vector in Vd are the same, the spin product always gives 1. In contrast, the exchange
potential is defined as a so called non-local potential
1
Vex (r) = q r r q (10.35)
(q ) (q ) s s
(r ) (r )
3
= dq = |
2
d r ,
|r r | |r r |
with the characteristic exchange of the sets of quantum numbers vs. . It involves
the orbital to be determined by (10.33) (and not just the average over all other
orbitals as the direct potential).
Thus, (10.33) is not even an eigenvalue equation for the spin-orbitals. We men-
tion in passing that, very formally, one defines a so called F OCK operator which
allows one to rewrite (10.33) so that it looks like an eigenvalue equation. Since from
our perspective no deeper insights into atomic and molecular structure are gained
by this concept, we suppress further details.
The solution of the H ARTREE -F OCK equations for the spin-orbitals may in prin-
ciple be done by iteration following closely the scheme of the H ARTREE method
described in Sect. 10.1.4. Alternatively, one introduces predefined basis sets (e.g.
special Gaussian type functions) from which the orbitals are constructed as linear
508 10 Multi-electron Atoms
combinations. The variational procedure is then applied to the coefficients, and thus
an algebraization of the problem is achieved.
In any case, one finally obtains, in addition to the binding energies W of all
electrons, the symmetries of the states, their orbitals, and the effective self consistent
potentials. We recall again, that the total energy is then obtained from (10.31) and
not (!) by simple summation over all orbital energies E h (which would account for
the repulsive terms twice).
The S LATER determinants (HF states) are, however, not yet necessarily a correct
description of the atom. Rather, they provide, a good starting point. As mentioned
above, in the next step one has to diagonalize the total Hamiltonian. As long as spin-
orbit interaction does not play a major role (i.e. for light atoms where the L
S terms
can be neglected or may be treated as small perturbations), one may diagonalize
L and
2 2
S by linear combinations of several HF states within one configuration,
invoking some angular momentum algebra with C LEBSCH -G ORDAN coefficients.
We have explained this in detail in Sect. 7.3.4 for the He atom.
The states thus obtained describe the system as good as possible in the framework
of the independent particle model and Russel-Saunders coupling (LS coupling).
For He in its ground state it amounts to Wcor = 0.114 eV (1.4 %), for neon already
10.3 eV (already 3 % of the total energy Wexact 3507 eV). Clearly, for atoms
of the size of Ne and larger, efforts beyond the HF method are required. For even
larger atoms the additional problem of spin-orbit coupling and relativistic effects
10.2 HARTREE-FOCK Method 509
Helpful in this context is the so called KOOPMANs theorem: It states that the ener-
gies E h calculated for each orbital with the HF equations (10.33) for each electron
(more precisely for each set of quantum numbers) is approximately equal to the
ionization energy of that electron. Consequently
This is, however, just a 1st order approximation, since changes in the atomic
structure due the excitation or ionization process are not accounted for.
Today, the H ARTREE -F OCK method is almost exclusively used as an initial basis
for often very elaborate CI procedures. These are by no means limited to atoms.
Similar and much extended methods are used in molecular physics and quantum
chemistry, as well as in solid state physics. A variety of powerful programmes are
even commercially available (see also Sect. 3.6.8, Vol. 2). A detailed description
would, however, reach far beyond the scope of the present book.
Section summary
The H ARTREE -F OCK (HF) method employs electron wave functions cor-
rectly antisymmetrized as S LATER determinants (10.28) constructed from
spin orbitals (q ) = s (r j ) (j ).
The latter are derived by solving the H ARTREE -F OCK equations (10.33), a set
of coupled integro-differential equations which include a non-local exchange
potential.
The total HF energy is obtained from (10.31) as a sum of single (unscreened)
electron energies, C OULOMB screening energies and exchange integrals.
However, in the general case one has to diagonalize the Hamiltonian (in LS
coupling together with L and
2 2
S ) by a linear superposition of S LATER deter-
minants for a given electron configuration.
These HF energies are still limited by the model of independent electrons
moving in a central field. To account for electron correlations one has to go
further, in the simplest case to M LLER -P LESSET perturbation theory. A rig-
orous treatment requires CI, i.e. a linear superposition of several electron con-
figurations.
510 10 Multi-electron Atoms
Walter KOHN (1998) received the N OBEL prize for the development of density func-
tional theory (DFT). This method is today the method of choice for a range of ap-
plications in atomic, molecular and solid state physics (in their ground states) for
energy and structural determination of complex systems. Different from the proce-
dures discussed so far, DFT does not focus on the wave functions of a system for
physical interpretations these are anyhow often not of primary relevance. Rather,
DFT as its name already implies, is a method to determine electron densities.
The key goal is the characterization of the ground state properties for (interact-
ing) many-particle systems, based on their single electron densities. The starting
point of DFT is the fact (as proven by H OHENBERG and KOHN), that knowing the
electron density (r) of the ground state for an electronic system (with or without
interaction) characterizes the system completely.
In the model of independent particles, which interact only via an averaged po-
tential with each other and with an external field or the field of many atoms (in a
molecule or solid state material), the electron density of a system with N electrons
is given by
N
(r) = j (r)2 . (10.37)
j =1
The so called KOHN -S HAM orbitals j are obtained from the eigenvalue equations
(in a.u.)
1
+ vKS (r) j (r) = j j (r). (10.38)
2
The density dependent (effective) KOHN -S HAM potential
(r )d3 r
vKS (r) = v0 (r) +
+ vxc (r) (10.39)
|r r |
is composed by an external potential v0 (for a single atom of nuclear charge Z
this is simply Z/r), the repulsion of the electrons among each other (the second
term) and a non-classical exchange potential vxc ((r)). The latter is the functional
derivative of the exchange energy function Wxc ((r)) in respect of (r):
Wxc ((r))
vxc (r) = . (10.40)
(r)
The nature of exchange interaction has been discussed already in Chap. 7.
10.3 Density Functional Theory 511
Equations (10.37)(10.39) are called the KOHN -S HAM equations. Without ex-
change they are equivalent to the H ARTREE method. The key problem, and the main
efforts of DFT thus is, to find an adequate treatment of the exchange potential. In the
most simple variety, the so called local density approximation (LDA), one chooses
a local potential derived from the local density functional which similar as in the
T HOMAS -F ERMI model (Sect. 10.1.5) is directly related to the electron density:
Wxc (r) = xc (r) (r)d3 r. (10.41)
Here xc () is the exchange correlation energy per electron at constant electron
density .
In any case, the total state of the system is described by the corresponding
S LATER determinant (r) constructed with KOHN -S HAM orbitals according to
(10.28). The procedure then follows a variational approach: the electron density
(r), which ultimately defines the KOHN -S HAM orbitals, is optimized until the ex-
pectation value of the H AMILTON operator (10.3)
|(r)
W = (r) |H (10.42)
Then (10.42) gives the best ground state energy within the framework of DFT, and
the corresponding (r) describes the best electron distribution for the ground state.
Unfortunately we cannot go into more detail of this method which becomes
increasingly important in modern structure theory of atoms, molecules and solid
states. Figure 10.3 shows some typical results for atomic electron densities ob-
tained with a simple PC programme (S CHUMACHER 2011) found on the Inter-
net. According to the authors, the programme uses the Xalpha functional and
The model is not spin polarized and non-relativistic. Overall, it has an accu-
racy between LDA and LSDA. Plotted in Fig. 10.3 is the radial electron density
4r 2 Ne (r) = 4 nj j r 2 Rn2j j (r), summed over all occupied orbitals nj j so that
4 0 r 2 Ne (r)dr = Z. The maxima of this distribution show the different, occu-
pied shells. For comparison the completely smooth T HOMAS -F ERMI electron den-
sity (10.26) is also plotted clearly it may only be used as a 0th order approxima-
tion.
A standard textbook on DFT is PARR and YANG (1989). In the mean time,
this powerful method has developed in a variety of brands. Particularly important
progress has been made in recent years with time dependent density functional the-
ory (TDDFT) which has become a key tool for physical and quantum chemical
structure calculations and simulations of dynamic processes in complex systems.
With certain tricks it is now possible to study excited states and, very important, op-
tical transitions. The review of M ARQUES and G ROSS (2004) provides an excellent
introduction, while more recent advances are summarized by C ASIDA (2009).
512 10 Multi-electron Atoms
2 1s 2 2s 2 2p 6
1s 2 6C
12 1s 11Na 20 18Ar
6 10 2s 2 2p 6 10 Ne
8 10 15
1s 2
4r 2 Ne(r) / a0-1
2He
4
6 10 2s 2 2p 6
2s 2 2p 2
2 4 3s 2 3p 6
10 5
2 3s only 2
1s
0 0 0
0 2 4 0 2 4 0 2 4
r/ a0
Fig. 10.3 Computed radial electron density for several atoms (full lines) in atomic units. For
comparison the densities obtained from the T HOMAS -F ERMI potential (dashed lines). For Na,
with its isolated valence electron in a 3s orbital, its radial density is multiplied by a factor of 10
(black, dash dotted line)
Section summary
The density functional theory (DFT) provides a very efficient, self consistent,
widely used method for electronic structure calculations (mainly for electronic
ground states) in atoms, molecules and condensed matter.
It concentrates on electron densities, rather than on wave functions, and ap-
proximates exchange by different types of judiciously chosen functionals, in
the simplest case by a local potential (LDA).
N
N
S=
Si and
L=
Li .
i=1 i=1
We recall that this kind of coupling by exchange interaction dominates for small
nuclear charge Z and is called RUSSEL -S AUNDERS coupling or (somewhat am-
biguously) LS coupling.
In contrast, spin-orbit interaction is of magnetic origin. As long as it is small (e.g.
in the case of He) it may be added as a perturbation, typically being proportional
L
to S and to Z 4 (see Sect. 6.2.4). Under the influence of this perturbation, total
10.4 Complex Spectra 513
spin and total orbital angular momentum finally couple to an overall total angular
momentum of the active electrons:
L +
J = S.
Except for very small Z one typically considers only the interaction of spins with
their own orbit and neglects spin-spin as well as spin-other-orbit interactions. This
leads to fine structure splittings which may be described very similar to those of
quasi-one-electron systems. This also holds for the Z EEMAN splitting observed in a
magnetic field.
However, for large Z spin-orbit interaction terms become comparable or even
larger than exchange interaction so that RUSSEL -S AUNDERS (LS) coupling does
no longer describe the physical reality appropriately. Strictly speaking, L and S are
then no longer good quantum numbers even if they are often still used to charac-
terize the atomic states. We have already reported deviations from the standard (LS)
selection rules for optical transitions and we shall see more of it below, e.g. in the
case of the Hg atom, where strong intercombination lines are observed even though
in the term scheme one still distinguishes a singlet and a triplet system (Fig. 7.13).
In the extreme case for very high Z one typically finds spin-own-orbit interac-
tion to dominate and the coupling scheme changes completely. Most appropriately
it is described by jj coupling of the N atomic electrons: First all individual orbital
angular momenta Li (quantum number j ) couple with their respective spin Sj
(sj = 1/2) to form an individual total angular momentum J i (quantum number ji ).
Only then the individual J i couple together (under the influence of exchange inter-
action which may now be considered to be a perturbation) and form the total angular
momentum J of the atom:
N
Ji =
Li +
Si
J=
J i. (10.44)
i=1
spin-orbit interaction
exchange interaction
for the example of the terms
np(n + 1)s in the 4th main Ge 3P
group of the periodic system. 2
Plotted is the difference of the 2000 C
term energies in respect of the Si
3 P state 1P
0 1
3P 3P
2 1
0 3P
0
0 10 20 30 40 50
Z
periodic system (Table 3.3). While FS is essentially negligible for the element boron
(B), in the case of thallium (Tl) it reaches an order of magnitude of eV.
A direct comparison of spin-orbit and exchange interaction becomes possible in
the 4th main group: in Fig. 10.5 the excited ns 2 np(n + 1)s configuration for sin-
glet and triplet states with their respective fine structure splitting are shown. For the
carbon atom (small Z) exchange interaction clearly dominates, separating the 1 P1
and 3 PJ states by nearly 0.2 eV, while the fine structure splitting of the 3 PJ states is
nearly negligible. In contrast, for large Z fine structure splitting (i.e. spin-orbit in-
teraction) becomes comparable to the singlet-triplet splitting (i.e. to exchange inter-
action) or even larger in particular so for tin (Sn). Thus, RUSSEL -S AUNDERS cou-
pling hardly describes reality any longer correctly. The next element in this group,
lead (Pb), is not shown here. Event though spectroscopic data are still denominated
as singlets and triplets (K RAMIDA et al. 2013), missing terms in this scheme clearly
document the transition from LS to jj coupling.
The following Figs. 10.610.10 represent some characteristic examples for com-
plex atomic spectra. We shall discuss these briefly. The presentation used here to
show the term positions together with observed transitions are called G ROTRIAN
diagrams. However, to avoid confusion we communicate only a few of the most im-
portant transitions and by no means all known energy terms. The diagrams contain
in any case a lot of information as a result of many years of spectroscopic work
of many research teams. The full data sets are collected by K RAMIDA et al. (2013),
who have analyzed and evaluated the literature sources critically and presented them
in a user friendly manner.
We start with a typical and particularly important example, carbon (C), an el-
ement of the 4th group, for which a G ROTRIAN diagram is shown in Fig. 10.6.
According to H UNDs rules, the ground state is a triplet with the configuration
1s 2 2s 2 2p 2 3 P0 , where the two spins of the 2p electrons are parallel (symmetric),
the orbital wave function is, however, antisymmetric (PAULI principle). Since car-
bon is a light element (small Z = 6), spin-orbit interaction is weak, and hence the
10.4 Complex Spectra 515
2P
110
100
2S
90
80
2D 70
60
2 2 3
4P
50
22 2P 40
2 1/2,3/2 C+ ionic ground state 1Po
120 2 2 2 30
2 22 2 22 2 22 3So 20
110 5P
3
100 1 o
P 3P
o 1
S 1 P 1 D 3 S 3P 3D 1 o o
P 1D 1Fo 3P 3D 3F
o o o
10
I 5 0
5 4
4 3Po
4 3
-1
1189
I / 1000 cm
6
70 3Do
11 .5
143
3 940 5 94 3
14 131
60 17 13
6
156
14
50
6
24
40 displaced
8
30 terms
5So
19
16
3
20 2
10 2
0 2 [He] 2 2 2 2 3P C ground state
o
Fig. 10.6 G ROTRIAN diagram for carbon, 6 C. WI ( 2 P1/2 ) = 90820.42 cm1 . Energies are
given in 1000 cm1 : left excitation energies of the neutral atom (C I), right the excitation ener-
gies of the ion (C II); wavelengths marked on the transitions are given in nm (rounded to integers)
2s 2 p 3
50
160 5 So
[He] 2s 2 2p 2 3P 0,1,2 N+ ionic ground state 40
150
30
2s 22p 2(3P) 2s 22p 2(1D) 2s2p 4
ns np nd ns np 20
4 o 4 o 2 o
2 P 4P 2So D 4Po S 2Do P 2P 4F 2F 4P 4D 2D 2D 2Fo 2 10
2Do 2Po D
120
WI 0
5 5
110 5 4
4 3 displaced
4 410
W - WI / 1000 cm-1
100 1558
terms
3 528 4P
90 859
123
132
7
4
6
3 1487
12
95 10
13
80
14
9
70
11
W - W(2 4S) / 1000 cm-1
60
13
3
11
50 4
11
8
17
40 5
2
12
30
0
20 2
10
2 [He] 2s 2 2 p 3 4S0 N ground state
0
one of the three 2p electrons is excited may be divided into three groups, related to
the N+ ionic state towards which they converge: 3 P, 1 D or 1 S. On the left of the dia-
gram the configurations 2s 2 2p 2 ns, 2s 2 2p 2 np and 2s 2 2p 2 nd with their respective
multiplets are given, converging to the 2s 2 2p 2 3 P configuration of the N+ ion. On
the right of these the less well known configurations 2s 2 2p 2 ns and 2s 2 2p 2 np are
displayed. They converge towards the N+ (2s 2 2p 2 1 D) state. Terms converging to
the 2s 2 2p 2 1 S state of the N+ ion are not known for nitrogen. However, such terms
are observed for the heavier elements of the 5th group (e.g. for As). The 4 P and 2 D
states on the right side of the diagram belong to the configuration 2s 2p 4 , where
one of the stronger bound 2s electrons is excited. As for carbon these are called
displaced states.
The G ROTRIAN diagram of oxygen (O), typical for the 6th group in the periodic
system, is shown in Fig. 10.8. The ground state configuration 1s 2 2s 2 2p 4 leads to
the same three states 3 P, 1 D and 1 S as in the case of the 4th group. The two extra
electrons have opposite spin and fill two orthogonal p orbitals. The ordering of
the 3 P multiplet is now inverted: for oxygen the 3 P2 term is energetically lowest,
while for carbon the terms are normally ordered. Since the O+ ion is isoelectronic
with neutral N, three terms (4 S, 2 D and 2 P) are possible in the O+ ionic ground
state configuration 2s 2 2p 3 . Thus, there are three ionization limits towards which
the excited state configurations of the neutral atom 2s 2 2p 3 nl may converge.
10.4 Complex Spectra 517
2 22 3(2Po)
O+ ion ground state 60
o 2 22 3(2Do)
2Po) [He] 2 2 2 3 4S
I ( 1Po o 1S 1,3D 1,3 o o
3P 1 P 1,3D
= 150.3 P
1,3D 1,3Po 1,3Fo 40
2Do) 1Do 3Do
I ( 1P 1,3F 1,3 o 1,3
S D
o 1,3Go
2 22 3(4S)
5
= 136.6 5 4 2 2 20
4 3 3Po
120 o o
3P 5 P
3 S 5S 3D 5Do
o
I 0
5 3
100 395
-1
4
I / 1000 cm
88
77
80
94
80
80
11 5
3 8 79
77
( 2 3P) / 1000 cm-1
99
95
103
60 displaced terms
98
1 30
40
2s2 2p4 1S
92
20
2s2 2p4 1D
o
Fig. 10.8 G ROTRIAN diagram for oxygen, 8 O. WI ( 4 S3/2 ) = 109 837.02 cm1 . Otherwise as
Fig. 10.6
Next we discuss the term scheme of neon (Ne) as a typical representative of the
rare gases. All rare gases are characterized by a closed rare gas outer shell (ns 2 np 6 ).
The neutral ground state is thus always a 1 S0 state. The G ROTRIAN diagram of neon
is shown in Fig. 10.9. Its neutral ground state configuration is 1s 2 2s 2 2p 6 . Excita-
tion of one electron from the 2p shell leads to terms which converge to the Ne+
ion ground state 2s 2 2p 5 2 Pj with j = 3/2 and 1/2. The coupling scheme cannot be
expressed in the form of the typical RUSSEL -S AUNDERS coupling, rather one uses
a j scheme: In this case the ionic ground state configuration 2s 2 2p 5 2 Pj cou-
ples with the excited n electron to form an angular momentum K. Each [K] state
splits under the influence of spin-orbit coupling (K couples with the electron spin
of the excited electron) into a doublet with J = K 12 . One uses the terminology
(2S+1 Lj )nl 2S+1 [K]J , e.g. the first excited state of neon is (2 P3/2 )3s 2 [ 32 ]2 . In all
rare gases some of the first excited states, those with J = 0, 2 or 3, are metastable:
they cannot decay to the ground 1 S0 state due to the selection rule
J = 0 1, but
0 0.
The famous red line of the helium-neon laser is a transition between two higher
lying excited states in neon: (2 P1/2 )3p 2 [ 32 ]2 (2 P1/2 )5s 2 [ 12 ]1 , marked red in
Fig. 10.9. The population of the upper level is achieved by collisions with metastable
He in the 1s2s 1 S0 excited state. Depopulation of the lower laser level is supported
by collisions with the glass wall of the laser cell.
518 10 Multi-electron Atoms
370 200
3
360 190
230
[He] 2s 2 2p 5 2Po1/2, 3/2 Ne+ ionic ground states 50
220
2s 2 2p 5 (2Po 3/2) 2s 2 2p 5 (2Po 1/2) 40
210 ns np nd nf ns np nd nf
30
200 2[7/2]o 2
[9/2]o 27
2[5/2]o 2[5/2]o 2[7/2]o 20
190
2[3/2]o 2[3/2]o 2[5/2]o 2[3/2]o 2[5/2]o 2[7/2]o
2[3/2]o 2
10
180 [1/2]o 2[1/2]o 2[3/2]o 2[1/2]o 2[1/2]o 2[3/2]o 2[5/2]o
WI 0
170 6 6
5 5 4 4
4 3 1838
W - WI / 1000 cm-1
160 4
W - W(2 1 S0) / 1000 cm-1
150 3 627
140
3 displaced
130 terms
63 62
120
110 74
[He] 2s 2 2p 6 1S0
Ne ground state
0
o o
Fig. 10.9 G ROTRIAN diagram of neon, 10 Ne. WI ( 2 P3/2 ) = 173 929.75 cm1 , WI ( 2 P1/2 ) =
174 710.17 cm1 . The so called j - coupling scheme holds. Otherwise as Fig. 10.6
In neon one also observes some weak absorption lines in the XUV range. These
transitions lead from the ground state to highly excited states with the configuration
2s 2p 6 np. These are again displaced terms where one of the 2s electrons is excited
to an np level. Such terms are also found for heavier rare gases.
As a last example we consider aluminum (Al). Its G ROTRIAN diagram is shown
in Fig. 10.10. The ground state configuration [Ne]3s 2 3p contains only a single elec-
tron in the 3p shell. The excited states where this 3p electron is excited thus form
a simple, alkali like term scheme as indicated on the left in the diagram Fig. 10.10.
All these excited terms converge towards the Al+ ionic ground state 3s 2 1 S0 . In ad-
dition, there are series of complex terms, all except one embedded in the ionization
continuum, as shown on the right hand side of the diagram. A 3s electron is ex-
cited in this case. There are three terms with the electron configuration 3s 3p 2 , the
4 P multiplet having the lowest energy. Terms with higher excited configurations
J
3s 3p n finally converge all towards the 3s 3p 3 P excited state of the Al+ ion.
10.5 X-Ray Spectroscopy and Photoionization 519
o
100 Al+ ionic ground state: 3s 3p(3P ) 50
[Ne] 3s 2 1S 3p ns np nd
0
WI ( 3P )
o
Al+ ionic ground state 2 S 2P 4
P 4 Po 2S 2 2
P D 2D o4
2F o4
P D 40
= 85.7 5 30
3s 2(1S0) 5 4
5 4
70 ns np nd nf 20
3
2S 2 2 3 4
1/2 P3/2,1/2 D5/2,1/2 F7/2,5/2
W - W(3 2 P1/2) / 1000 cm-1
60 10
WI 0
W - WI / 1000 cm-1
6 4 5
40 5 5 4 08
5 3
30
4
5
3 112 3
4 131 displaced terms
20 4
34
39
4
10
[Ne] 3s 2 3p 2P1/2 Al ground state
3
0
Fig. 10.10 G ROTRIAN diagram for aluminium, 13 Al. WI ( 1 S0 ) = 48 278.37 cm1 . Otherwise
as Fig. 10.6
Section summary
With increasing atomic number spin-orbit interaction increases Z 4 and
RUSSEL -S AUNDERS (LS) coupling gives a less and less appropriate descrip-
tion.
In the limit if very high Z the orbital angular momentum and spin of each
electron couple with each other to j under the influence of spin-orbit inter-
action, and only at the end the different j couple by the weaker exchange
interaction in a jj coupling scheme.
We have discussed several examples of term schemes (as G ROTRIAN dia-
grams) for more complex atoms with rich structure and increasing complexity.
Excited states of noble gases are described by a j coupling scheme.
We memorize that C, N and O each have 3 lowest multiplet levels with the
same electron configuration 2s 2 2p 2 , 2s 2 2p 3 , and 2s 2 2p 4 , respectively. These
levels are spaced by about 1000 to 2000 cm1 and follow H UNDs rules. In
the case of O the lowest level is 3 P followed by 1 D and 1 S.
important asset for todays manifold applications of X-ray spectroscopy is the abun-
dant accessibility to state-of-the-art electron storage rings which have been built
exclusively for the generation and applications of synchrotron radiation, in partic-
ular in the VUV, XUV and X-ray spectral range. Since the methods for generating
and detecting this kind of electromagnetic radiation may typically be considered
part of optical physics (in its broadest sense) and since atoms with high Z play an
important role in this context, this subject has to be discussed here. We can, however,
only touch some aspects of this important area of research and collect some of the
most important basics. In the present section we focus on terminology, methods, and
key observations of spectroscopy. In Sect. 10.6 we shall then elaborate on some of
the pertinent sources for X-ray radiation as prerequisite for spectroscopy. For more
details the reader is referred to specialized literature (e.g. the rather comprehensive
book of ATTWOOD 2007).
Up to now we have been concerned almost exclusively with the spectra of valence
electrons, i.e. from electrons in the outermost atomic shell. We now turn to the inner
shells, in particular to those of larger atoms. Due to the characteristic dependence of
atomic energies on the square of the nuclear charge Z, we expect the corresponding
spectra at very short wavelengths. In uranium, e.g., an electron in the K shell has a
binding energy of more than 110 keV. Interestingly, in certain respects these spectra
are much easier to understand than those of the complex outer shells of intermediate
size atoms which we have discussed in the last section with the help of G ROTRIAN
diagrams: since all inner shells are filled, transitions between these levels can only
occur when a hole is created in one of the shells typically by collisions or by
photoionization.
We discuss these processes on the basis of a very schematic term scheme for
an element Tg (target) with high atomic number Z, as sketched in Fig. 10.11.
Orbitals and total angular momentum for each electron are again characterized by
the quantum numbers n j , the designation of the shells as K, L, M, N, . . . and
their respective subshells, following the usual schematic which we have introduced
in Sect. 3.1.3. Let us assume that all these inner shells are filled. Three types of
processes may be observed:
Absorption of a photon from one of the inner shell levels can (essentially) oc-
cur only into the continuum, since all inner shells are occupied. This leads to
photoionization
Tg(n ) + Tg+ (n )1 + e
(10.45)
with the energy balance WI = ,
where WI = WK , WL , WM etc. is the ionization potential from the respective
shell. The kinetic energy of the electron ejected is . The initial electron configu-
ration is described by (n ), and the symbol (n )1 indicates that a corresponding
10.5 X-Ray Spectroscopy and Photoionization 521
n j shells
OI - O...
OI ...
3 3 75/2 NVII
3 3 5/2 NVI
3 2 3/2 NV
3 2 3/2 NIV
3 1 1/2 NIII
3 1 1/2 NII
3 0 /2 NI
NI ...
WM 2 2 5/2 MV
2 2 3/2 MIV
2 1 3/2 MIII
2 1 1/2 MII
2 0 1/2 MI
MI MII ...
WL edges
2 1 3/2 LIII
2 1 1/2 LII
2 0 1/2 LI
LI LII LIII 2 LI LII LIII
1
WK edges 2 L series
K1
1 0 1/2
K edge K series
Fig. 10.11 Absorption and emission of X-ray radiation: (a) absorption is only possible into the
ionization continuum, since the intermediate levels are fully occupied; (b) emission can only occur
after generation of holes in the K, L, M, . . . shells. Lines with red arrows correspond to photon
energies, black lines characterize the energetics of the system
hole has been created in the atomic shell. Since the energy balance has to be
maintained by the ejected electron, one observes as a function of the photon en-
ergy so called absorption edges: for energies below WI there is no absorption.
As soon as > WI , the corresponding shell can be ionized and the cross section
for photoionization jumps to a finite value. For still larger it decays again, as
we have already discussed in Sect. 5.5. We shall come back to this in a moment.
But first, we discuss the other two processes.
If such a whole exists in an inner shell it may be refilled by spontaneous de-
cay with an electron from a higher shell n . This process generates so called
characteristic emission lines
' ( ' 1 (
Tg+ (n )1 . . . n Tg+ n n +
(10.46)
with the energy balance Wn Wn = .
In contrast to the absorption edges just described, one now observes a spectrum
of discrete emission lines. Due to the well known 3 factor (4.38) this process
is particular efficient in the X-ray region and many orders of magnitude more
522 10 Multi-electron Atoms
probable than the absorption of a second photon. The corresponding line spectra
may be read in principle from Fig. 10.11. They are essentially one electron spec-
tra which may be understood more or less as directly as the spectra of hydrogen
like ions or better, alkali like ions if we want to account for the screening of
the nuclear Z/r potential. One may also discuss these spectra in terms of holes
being excited instead of electron transitions. We come back to this concept
quantitatively in a moment.
Finally a third process may occur. The hole is refilled by an electron from a
higher shell, but instead as in reaction (10.46) where a photon is emitted the
excess energy is used to eject a second electron. This so called AUGER electron
emission process may be visualized as
' ( ' 1 (
Tg+ (n )1 . . . n Tg+ n . . . n + ()
' 1 1 ( (10.47)
Tg++ n . . . n n + e ,
where () indicates a virtual photon which is exchanged between the two
electrons, giving the latter sufficient energy to leave that atom. As one reads
from Fig. 10.12 the energy balance is given by
In comparison to the rich material obtained from spectroscopy of outer shells, far
less detailed data exist for the absorption and emission of X-ray radiation from inner
shells for obvious reasons: only during the past decades dedicated synchrotron
radiation sources have provided intense, tuneable X-ray sources of excellent quality.
Fortunately, the available material is well documented and excellently backed up by
theoretical computations. We use here the NIST X-ray data-bank (C HANTLER et
al. 2005), which offers characteristic emission lines in tabulated form, as well as
10.5 X-Ray Spectroscopy and Photoionization 523
10 K
5 data )
.56
(0.924 , 7 LI
fit parameters LII
(n*, qs) K
4 LIII
)
5.21
,1
9)
3 1
1.4
.95
keV) 1/2
2, (1
.06
MI
(1
) )
2 .82 .98 MV
, 21 99 , 24
(Wedge
6 0
5) (3.5 (4.1
, 4.9 ) NI
1 5 98 4 , 29.26 NIV
(2. 8) (6.4 5
11.7 NVII
13 ,
(5.1 ) 6 5 .89)
40.0 3 9 ,
0 (7.305 , (5.01
0 10 20 30 40 50 60 70 80 90
Z
Fig. 10.13 M OSLEY diagram for the lowest X-ray absorption edges for all elements (compiled
from the tables in C HANTLER et al. 2005). Grey and black lines give the results of fits with (10.50).
The numbers in brackets are the parameters (n , qs ). One sees that the inner shells cannot be fitted
optimally over the whole periodic system with one set of parameters
absorption edges and photoionization cross sections over a wide range of energies,
for all elements. The data are based mostly on state-of-the-art ab initio computations
which have been tested extensively with the available experimental material.
Figure 10.13 shows the X-ray absorption edges of the K, L, and some higher
shells of all natural elements of the periodic system. This so called M OSLEY dia-
gram is based on the assumption that the transition energies are obtained in analogy
to the RYDBERG -R ITZ formula (1.149) for an effective one electron system with an
effective nuclear charge Z = Z qs :
(Z qs )2 Eh
W = . (10.49)
n2 2
Note that this also implies KOOPMANs theorem (see Sect. 10.2.4) assuming that
holes behave just complementary to electrons. The idea is thus to interpret the em-
pirically determined ionization energies for the different shells (i.e. the position of
the absorption edges) grosso modo by a combination of a screening parameter qs
and the quantum defect n n . One plots the square root of the energies as a func-
tion of Z, as suggested by (10.49):
E h Z Z qs E h
|Wn | = = . (10.50)
2 n n 2
524 10 Multi-electron Atoms
We have already used this so called M OSLEY formula (3.34) for the alkali metal
atoms. Ideally one would expect straight lines, as roughly confirmed by Fig. 10.13.
However, the K and L edges cannot be fitted optimally with one set of parame-
ters over the whole periodic system. For example, at low Z the effective quantum
number of the K shell is nicely n = 1.062 and the screening parameter qs = 1.49
is moderate. In contrast, for larger Z the best fit is obtained with n = 0.924 and
substantial screening qs = 7.56.
For Z > 13 one finds a somewhat crude, but rather useful rule of thumb for
obtaining the K energy:
For aluminum (Al) with Z = 13 this gives 1400 eV (true value 1560 eV, soft X-
ray) and for tungsten (W) with Z = 74 the formula leads to 70 574 eV (true value
71 676 eV, hard X-ray).
BORN L
L Al
106 BORN K
K
photo ionization
104
100 elastic
(coherent)
sum
a / b
photo
1 COMPTON ionization
pair
production
0.01 1 100 104
photon energy / keV
Fig. 10.14 Photo-absorption cross section (in barn) for aluminum as a function of photon energy.
The main contribution is due to photoionization. At higher energies pair production dominates
(according to C HANTLER et al. 2005). Black dashed lines give the photoionization cross section in
B ORN approximation
one should not loose time with such approximation when attacking practical prob-
lems. We use he NIST X-ray data bank of C HANTLER et al. (2005) and B ERGER
et al. (2010) and show in Fig. 10.14 the total photo-absorption cross section a for
aluminum (Al) over a broad range of photon energies. Al is an example for a light
atom (Z = 13). In contrast, Fig. 10.15 presents a for lead (Pb), a heavy element
(Z = 82) well known as a good X-ray absorber and frequently used for protection
from X- and -ray radiation.
For the L edge of lead at 16 keV one reads e.g. in Fig. 10.15 a 52 000 b
(roughly 200 that for Al at the same photon energy). With ma = 207 u this corre-
sponds to a mass absorption coefficient of ca. / = 15.0 m2 / kg. This implies e.g.
that a protection west with just 67 g lead per m2 the irradiation with a 16 keV X-ray
source is already reduced to 1/e 37 %.
As indicated Fig. 10.14 and Fig. 10.15 by dashed lines, the trend and the order
of magnitude of the B ORN approximation. The real cross section differs quantita-
tively somewhat and is of course much more structured. With increasing energy the
cross section drops typically ()5/2 2.5 (and not as predicted by B ORN ap-
proximation ()7/2 ). It jumps, however, at each absorption edge significantly:
i.e. the cross section assumes a finite value whenever the energy of the photon is
high enough to photoionize the respective shell as evident from Fig. 10.11. We re-
member: considering the energetics of individual electrons independently from the
reaction of the whole atom is the very content of KOOPMANs theorem according to
Sect. 10.2.4. The edge energy is then simply the HF orbital energy Wn according
526 10 Multi-electron Atoms
M N
106 BORN L Pb
photo ionization
L
BORN K
104
elastic K
(coherent)
100
sum
pair
production
a / b
1 COMPTON photo
ionization
Fig. 10.15 Photo-absorption cross section (in barn) for lead as a function of photon energy. Oth-
erwise as Fig. 10.14
to (10.33). One recognizes the K and L edges very clearly for Al, for Pb in addition
also the N, M and O edges.
Figure 10.14 and Fig. 10.15 illustrate that energetic photons are not only attenu-
ated by photoionization. Without going into details we summarize the most impor-
tant mechanisms for absorption and scattering of X-ray photons:
2. C OMPTON scattering
3. Pair production
e + e+ for 2me c2
4. T HOMSON scattering (elastic)
must be fulfilled simultaneously, pair production can only occur in the presence of
a third particle, preferentially an atomic nucleus with high Z. Correspondingly the
absorption cross section increases again for hard radiation, as seen in particular
for Pb. For energies around 1 MeV incoherent photon scattering (C OMPTON effect)
plays a central role, as described by the K LEIN -N ISHINA formula (8.130).
In addition to these three processes elastic (also coherent or T HOMSON) scat-
tering of photons plays a (minor) role. As discussed in Sect. 8.4.5, at energies sig-
nificantly below 2me c2 = 1.022 MeV, but far above typical atomic resonances one
expects cross sections el Z 2 e . With the T HOMSON cross section e = 0.665 b
one verifies this for Fig. 10.14 (Z = 13, Al) and Fig. 10.15 (Z = 82, Pb) at the lowest
energies. As energy increases into the relativistic regime (L ORENTZ factor 1)
the elastic cross section drops dramatically (as expected: oscillation amplitudes of
relativistic electrons are much smaller than at classical energies). One roughly finds
el ()2 1/ 2 .
In Sect. 5.5 we have discussed in some detail how (in principle) to compute pho-
toionization cross section specifically for the particularly simple example of the
H atom using the B ORN approximation. In the previous section we have presented
the photo-absorption cross sections for two selected metal atoms and focussed on an
overview from VUV to hard X-ray photons. We now present in some more detail,
by way of example, a non-trivial atom of intermediate nuclear charge. We take the
rare gas atom argon (Ar) with Z = 18 in the interesting range of photon energies
from threshold up to some keV.
Figure 10.16 shows the dependence of the photoionization cross section for Ar
in various ranges of energy. Experimental results and theoretical calculations from
a number of different groups are compared. They supplement each other obviously
quite well, even though a perfect agreement cannot be expected for such a rather
complex element. We do not enter into a discussion of the many details, both in ex-
perimental data as well as in theory. In recent years, high quality data are obtained
almost exclusively at synchrotron radiation electron storage rings (see Sect. 10.6.2),
and in Fig. 10.16 these data are recognized by their small error bars and the con-
sistent trends reported. Notwithstanding this progress, many data have still been
obtained with selected lines from spectral lamps or X-ray tubes. Figure 10.16(a)
gives a nice example. Shown there is the photoionization process
Ar [Ne]3s 2 3p 6 1 S + Ar+ [Ne]3s(3s)1 3p 6 + e (p) (10.53)
from a single subshell, the MI shell (3s 2 ). One distinguishes this process from oth-
ers by measuring the kinetic energy of the outgoing, free electron. The final state
is here a 1 P0 . The term scheme of Ar is quite similar to that of Ne (see Fig. 10.9).
In addition to the full K and L shells, [Ne] = 1s 2 2s 2 2p 6 , in Ar the M shell is filled
with 3s 2 and 3p 6 electrons. The first ionization potential, related to the configura-
tion [Ne]3s 2 3p 5 , is WI = 15.76 eV. To ionize a 3s electron, additional 13.48 eV are
528 10 Multi-electron Atoms
1.5
1.0 (a) dipole-length appr. (b)
0.8 dipole-velocity
Ar 3s
0.6 Houlgate et al. 1.0
0.4 Samson and
Gardner
0.2
0.5
0.0
30 50 70 90 50 100 150 200
6 6
(d)
photionizaton cross section / Mb
2p (c')
3 1.0
4 2p
0
245 250 0.5
2
2s
(c) 0.0
0
300 500 600 800 1000
J.H. McCrary
(e') G.V.Marr
0.20 1s np 0.10 A.J.Bearden
(f) CXRO
0.08
0.15 ZCZ06
3205 3210 (e) 0.06 R.H.Millar
0.10 FFAST
0.04 XCOM
0.05 K edge K+M
double excitation 0.02
0.00 0.00
3210 3230 3250 2400 3600 4800 6000
100
(g)
1.00
0.01
1E-4
from FFAST (NIST)
1E-6 photon energy
10 100 1000 104 105 / eV
Fig. 10.16 Photoionization cross section of argon (Ar) in different energy regimes compari-
son of experiment and theory adapted from several authors. S AHA (1989): (a) partial cross sec-
tion for the 3s ionization above-threshold, showing a typical C OOPER minimum with MCHF
theory (full and dashed lines). S UZUKI and S AITO (2005): (b) total cross section up to 200 eV
(experiment only), (c) experiments from L edge up to 500 eV, (c ) blow up close to L edge,
(d) 500 eV to 1150 eV. Z HENG et al. (2006): (e) near K edge with high and (e ) very high res-
olution, (f) K edge over a broader energy range. Derived from the FFAST data bank (C HANTLER
et al. 2005): (g) overview for whole energy range
necessary. The data shown in Fig. 10.16(a) show the energy dependence of the pro-
cess (10.53) directly above its energetic threshold. Note that this particular process
contributes only very little to the total photoionization cross section.
10.5 X-Ray Spectroscopy and Photoionization 529
However, this particular partial cross section shows in an impressive manner how
the photoionization cross section starts with the usual jump to a finite value a thresh-
old, drops down rapidly to nearly zero just above 40 eV photon energy (ca. 12 eV
electron energy), and finally rises again. This so called C OOPER minimum is not un-
usual in the photoionization of complex atoms above ionization threshold. It can be
understood easily from the theoretical background: The photoionization cross sec-
tion (5.88) contains two matrix elements (5.89) that essentially describe the overlap
between bound state and continuum wave functions (weighted by the distance r).
Depending on the positions of the nodes in the wave functions they may be positive
ore negative what can change with the energy of the continuum electron. Thus,
it may happen, that the two matrix elements in (5.88) just compensate for one par-
ticular energy. Alternatively one dominant matrix element may change from a pos-
itive to a negative value, passing necessarily zero at some energy which then leads
to zero contribution from this matrix element. In the present case things are even
a little bit more complex since these wave functions are, strictly speaking, multi-
electron wave functions comprising more than the ideal configuration according to
(10.53). Here the dominant matrix element is . . . 3s . . . 1 S|r|p 1 P0
where the con-
tribution from the 3s 2 3p 6 configuration is partially compensated by strong CI with
the 3s 2 3p 5 3d configuration. Correspondingly expensive is the multi configuration
H ARTREE -F OCK (MCHF) calculation shown in Fig. 10.16(a), which interprets the
experimental results rather well not differing much in its dipole lengths and veloc-
ity form which is a characteristic for any high quality calculation (see footnote 10
in Chap. 4).
Figure 10.16(b) shows the total photoionization cross section for the energy
range between 40 and 100 eV, which is dominated by ionization of the six 3p elec-
trons. Here too one recognizes a C OOPER minimum, this time at 50 eV, followed
by a maximum at ca. 80 eV so to say its counter part. In Fig. 10.16(c) and (c ) the
details in the vicinity of the L edge are illustrated: the ionization thresholds (termi-
nology see Fig. 10.11) for LIII , LII and LI at 248.4, 250.6 and 326.2 eV, respectively.
The contribution of the 2s electrons (LI ) is very small similar as in the M shell.
Below the LIII and LII thresholds one sees in (c ) indications of autoionizing reso-
nances, which we got to know in Sect. 7.6.2.
Figure 10.16(d) shows the rather boring decrease above the L edge, but still sig-
nificantly below the K edge. The latter is finally reached at 3205.9 eV as shown in
detail in Fig. 10.16(e), (e ) and (f). In (e) one recognizes the double excitation of K
and M shell 1s3p 4p 2 (again an autoionizing state). The most pronounced struc-
ture in this energy range is doubtless the excitation of RYDBERG states 1s np for
(n 4), shortly below the K edge, shown on an expanded scale in (e ).
Figure 10.16(g), finally, gives a full survey over the whole energy range from
threshold up to 100 keV in a loglog plot, extracted from the B ERGER et al. (2010)
data bank. These data do, of course, only reproduce the rough tendencies.
We cannot end this section without mentioning that in addition many measure-
ments of the anisotropy parameter have been reported (see Sect. 5.5.3), containing
additional information we refrain, however, from reproducing such data here.
530 10 Multi-electron Atoms
Section summary
Absorption and emission of X-ray radiation from (and to) inner atomic shells
is governed by the PAULI principle which forbids transitions into completely
filled electron shells.
Absorption (at not too high photon energy, ) is mostly due to photoioniza-
tion. The absorption cross section as a function of shows a typical edge
structure displaying the ionization potentials of different shells. Beyond these,
the cross section decreases rapidly (Figs. 10.14 and 10.15).
At intermediate photon energies, photoionization cross sections of complex
atoms show a rich structure. E.g. so called C OOPER minima may occur,
shortly above an ionization threshold. They are attributed to cancellation of
two radial matrix elements for bound-free transitions, or to zero passage of
one dominant matrix element.
The absorption
edges WI are reasonably well described by M OSLEYs law,
predicting WI Z qs .
Based on this, a very rough estimate for the energy of K radiation is obtained
from (10.51).
At > 100 keV, the cross section starts to be dominated by C OMPTON scat-
tering, while for > 1 MeV pair production sets in.
Transitions of electrons from higher to lower levels can only occur into elec-
tron holes. This leads to characteristic X-ray emission spectra, similar to those
from quasi-one-electron spectra (alkali atoms).
Alternatively, the filling of inner shell holes may be accompanied by emission
of AUGER electrons which carry the excess energy.
The classical device for generating X-rays is the X-ray tube (originally RNTGEN-
Rhre, named after its inventor RNTGEN 1901, who received the N OBEL prize for
the discovery of X-rays). Even today this very simple method is still widely used.
A typical example of a state-of-the-art, small X-ray tube for analytical purposes is
illustrated by a photo of the device as well as by a schematic in Fig. 10.17. An elec-
tron beam generated by a hot cathode wire an a negatively biased W EHNELT cylin-
der hits a metal anode. Inside the metal it is strongly decelerated in the electric field
of atomic nuclei (preferably with high Z). This leads to X-ray bremsstrahlung
(the German verb for decelerating is bremsen). At the same time, a fraction of
the atoms in the anode is ionized by electron impact, which leads to hole formation
in the inner shells of the atoms. As a consequence emission of characteristic X-ray
radiation becomes possible.
Figure 10.18 shows characteristic spectra emitted from a Rhenium anode (Rh) at
different electron energies. The measured X-ray spectrum presented here has been
10.6 Sources for X-Rays 531
e-
filament
e- HV
WEHNELT cylinder
Be window screening
X-ray radiation
recorded at 90 detection angle to reduce the signal for the spectral analysis with a
crystal monochromator. One detects elastically as well as inelastically (C OMPTON-
Effekt) scattered X-ray radiation. Indicated in Fig. 10.18 are the literature values
for the K , K and L, emission lines for Rh as well as the energies expected
according to (1.75) for inelastically scattered photons. The weak signal at 8.02 keV
(CuK line) originates from a small impurity in the anode.
The spectrum clearly documents that the X-ray bremsspectrum cannot have a
higher energy than the generating electron beam at the anode, with Wkin = eU cor-
responding to the voltage U (HV in Fig. 10.18) applied to the anode. This implies
that the shortest wavelength emitted in the X-ray bremsspectrum follows the so
called rule of D UANE -H UNT:
hc
min = . (10.54)
eU
Today, for the spectroscopy with VUV, XUV and X-ray radiation electron storage
rings are available and intensively used at many places around the world, dedicated
to generate synchrotron radiation. In so called 3rd generation storage rings one uses
extremely well focussed, highly relativistic electrons. Typically, the electrons are
stored in a large, overall circular or elliptic structure in which they are guided by
deflection magnets (magnetic dipole fields). They are arranged with long straight
sections in between, where the electrons are focussed with different magnetic struc-
tures, mainly with quadrupole magnets, but other structures such as sextupole or
octupole magnets are also needed to control the chromaticity, i.e. the variation of
532 10 Multi-electron Atoms
Rh K (inel.)
40 keV
K (inelast.)
20 keV
L L K (elast.)
10 keV
0 5 10 15 20 25 30
/ keV
Fig. 10.18 X-ray bremsstrahlung and characteristic radiation (top) from the tube shown in
Fig. 10.17 with a rhenium (Rh) anode at three different electron energies (the spectra have been
kindly provided by H ASCHKE and L ANGHOFF 2007). The bremsstrahlungs continuum as well as
characteristic radiation (K , K and L, ) can be clearly recognized
3 The quality of the focussing is characterized by the so called emittance. That is the spatial exten-
sion of the electron beam multiplied by its divergence angle. As an example, BESSY II is charac-
terized by an emittance of (3 to 6) nm rad in horizontal and <0.1 nm rad in vertical direction.
10.6 Sources for X-Rays 533
deflection
synchrotron magnets
circumference
ce
96m
undulatorr
deflection
magnets storage
g ringg
circumference 240 m
be 32 deflection magnets
am
line 16 straight sections
s
beamlines
nes
s
ne
mli
a mli bea
target hall be target hall
Fig. 10.19 Schematic of the topography in BESSY II. The electrons are pre-accelerated in a
microtron, further accelerated in a small synchrotron ring to the nominal energy, and then injected
into the storage ring. The radiation energy emitted is compensated by high frequency generators.
In between the deflection magnets, marked in red, are straight sections are arranged for insertion
devices and multipole focussing magnets. The beam lines exit tubes of radiation for the users
are shown here for clarity only in the lower half of the figure. Only one undulator is indicated
as an example for an insertion device (We thank Prof. JAESCHKE et al. 2007, for providing details
and sketches)
d3 N d
R= = , (10.55)
dtdAd d
with = W = hc/ = hc being the photon energy. The spectral radiance
)W = d ,
2
R (10.56)
ddW
534 10 Multi-electron Atoms
i.e. it refers to radiance per unit energy interval (or per frequency, wavenumber or
wavelength). The synchrotron radiation community uses a special quantity called
spectral brilliance
3
B= , (10.57)
W/W
which refers to the relative bandwidth dW/W = d/ = d/. Usually, brilliance
is measured in units
1photon/ s
= 1 Sch. (10.58)
mm2 mrad2 0.1 % bandwidth
4 This unit may be somewhat confusing. What it implies is simply that B gives the number of
photons emitted per second into a solid angle
= 1 mrad2 from a source area
S = 1 mm2
into an energy interval
W = 103 W . Thus, into an arbitrary solid angle and energy interval the
source emits
photons
S
W
=B 3 .
s mm2 mrad2 10 W
Expressed in terms of standard spectral radiance (10.56) one would write
photons R )W
=
S
W,
s
so that with (10.56), (10.57) and W =
d2 )W .
B = 103 W mrad2 mm2 = 103 R
ddW
10.6 Sources for X-Rays 535
wavelength / nm
100 1 0.01 1000 100 10 1 0.1
FLASH DESY
Y XFEL
fs-Ti:Sapph DESY 1012
1033 (seeded) XFEL
LCSL
1031 LCSL
109
peak brillance / Sch (see text)
DESY
1029 DESY FLASH
FLASH
peak power / W
1027 6
plasma lasers 10
PETRA III HHG
1025 20 m ID SPring-8
X-ray Und.
exzimer
103
BESSY laser
Rubin
U49
U-49 U125
1021 U-125
ALS BESSY
1019 U5.0 100
1013
101 102 103 104 105 106 1E-3 0.01 0.1 1 10
photon enenergy / eV
Fig. 10.20 (a) Brilliance for various synchrotron radiation sources of the third generation (in-
cluding insertion devices ID and specifically undulators U) and for Free Electron Lasers (FEL) as
a function of photon energy. (b) comparison of the peak power in these sources with a variety of
laser systems
the present case (and assuming that no photons are lost), B is thus a constant
too. Synchrotron radiation from third generation sources typically emerges from
an area of only some 102 mm2 and is extremely well collimated. Compared to
conventional light sources, SR sources are superior by many orders of magni-
tude.
Note, however, that the angular divergence of laser sources is usually even better
than SR. Also, from an experimental point of view the peak power of a source is
often more relevant than its brilliance for obtaining an experimental signal. Fig-
ure 10.20(b) thus compares typical laser systems with representative SR sources
from this point of view. Peak power determines the total number of photons avail-
able for an experiment per unit time. Also the total number of photons, i.e. the
available measuring time may be crucial, so that even X-ray tubes are still in use
and allow long time experiments, in particular for such experiments where SR
peak power cannot be exploited e.g. due to damaging side effects on delicate
targets.
536 10 Multi-electron Atoms
A
(e)
2
} time window for observation
2 t = 2sin( )/c
p
2 2 te = 2 /v
Fig. 10.21 For understanding the generation of synchrotron radiation (SR): polar plots for the
angular distribution of (a) the classical radiation characteristic from an electron on a circular orbit,
observed in its rest frame and (b) its L ORENTZ transform; (c) components of the wave vector
in a frame moving with the electron and (d) the same in the laboratory system; (e) observation
window for the radiation from this electron
Starting point for the strictly relativistic derivation of the intensity and angular dis-
tribution of SR is (4.32) which gives the radiation characteristic in the moving rest
frame of the electron. This has to be L ORENTZ transformed into the laboratory sys-
tem. Consequently a highly collimated radiation characteristic tangential to the orbit
of the electrons is emitted. The not completely trivial theory has been developed for
the first time in a famous paper by S CHWINGER (1949). The formula derived there
for the emission characteristic is so precise that synchrotron radiation is used to-
day as a calibration standard in radiation metrology. Without going into details we
just present the key ingredients by some hand waving arguments, summarized in
Fig. 10.21.
10.6 Sources for X-Rays 537
for the radiation emitted, as one reads from Fig. 10.21(c) and (d). To illustrate this
by numbers: at BESSY II with its electron energy of 1.7 GeV one has 3300
and the full divergence angle of the radiation at a deflection magnet is 2 1 (in
words: one arc minute!).5
From (10.59) one may even obtain a simple estimate of the bandwidth. One has
to realize that an observer at large distance sees the radiation from a single electron
only for a very short time. As sketched in Fig. 10.21(e) this time represents the time
the electron needs to pass an arc segment 2. The electron needs a time te =
2/v, the light the time 2t = sin(2 )/c 2/c. An observer, looking
at the electron tangentially from a far distance, sees the light emitted from that
electron smeared out over a time
1 1 1 1 1
t 2 = = 3 ,
v c v c c 2 c
where we have used (1.15) for 1.6 Applying the uncertainty relation (1.125)
for time and energy one estimates from this the characteristic bandwidth of SR to
5 Of course, the emission angle effective for the experiment also depends crucially on the experi-
mental setup, which includes a certain length of the electron orbit. The angle given here thus
only holds for the vertical divergence perpendicular to the plane of the ring. In horizontal direction
the experimental apertures limit the SR beam.
6 Clearly,this time refers to the light emitted from a single electron and has nothing to do with
the actually measurable duration of the SR light pulses from bunches of electrons. Typical pulse
duration of modern synchrotrons are several ps.
538 10 Multi-electron Atoms
3c 3 3 2 eB
Wc = c = = , (10.60)
2 2me
enters as key parameter into the S CHWINGER formula for the intensity distribution.
The corresponding critical wavelength is
4 4me c
c = = . (10.61)
3 3 3 2 eB
To characterize the angular distribution of the radiation, instead of referring to
polar ( ) and azimuthal angle () one uses for practical purpose two orthogonal
angles (in the ring plane) and (normal to it). In the literature the original for-
mula (Eq. (II.32) in the paper of S CHWINGER 1949) for the emitted photon flux is
rewritten in a variety of different ways.
A rather concise form is the following:7
d3 3 I /e 2 2 2
= 2
1 + x2 (10.62)
ddd/ 4
S c
x2
K2/32 [ ] + K 2
[ ] .
1 + x 2 1/3
Here I is the electron current in the ring and
S the emitting area. The unit Sch
(S CHWINGER), introduced above, is obviously compatible with this formula writ-
ten essentially dimensionless (apart from cm2 s1 ). Kn is the generalized second-
order B ESSEL function, implemented in standard mathematical programs, and the
variables x and are defined by
3/2
x = and = 1 + x2 . (10.63)
2Wc
_ _2 _ _2
As it turns out, the two components in (10.62), K2/3 (/2) and x 2 K1/3 (/2)
(1 + x ), respectively, represent distributions of radiation intensity for two different
2
polarizations: the first one describes light linearly polarized parallel to the plane of
the ring, while the second term corresponds to a polarization normal to the ring
which is nonvanishing only outside the ring plane and has a phase shift of /4
in respect of that in the plane. Superposition of both components thus makes SR
7 For readers who want to follow these reformulations: S CHWINGER uses the esu system, and
his formula (II.32) gives the energy which an electron emits along the orbit over = 2 . For
the determination of the brilliance it has to be divided by 2 and by the photon energy. Now,
2/c = te is the period time for an electron on a circular orbit and, on the other hand, Ne =
2I /(ec) gives the number of electrons in the ring at a current I . With an emitting area
S and
the fine structure constant = e2 /(40 c) one obtains (10.62).
10.6 Sources for X-Rays 539
light emitted above or below the ring plane elliptically polarized: it is a unique
potential of SR to provide short wavelength, tuneable radiation with elliptic or even
circular polarization. Typically such radiation is offered at dedicated beam lines. For
compact description of the radiation characteristic one usually presents the radiation
with polarization parallel to the ring plane as a function of the photon energy ,
i.e. one sets = 0 and obtains as a convenient numerical relation:
d3 I/A
= 1.326 1013 (We / GeV)2 Shor ( ) Sch. (10.64)
ddd/
S/ mm2
d2 I/A
= 2.4577 1013 (We / GeV)Sint ( ) mrad Sch. (10.65)
dd/
S/ mm2
While from (10.62) one obtains Shor ( ) directly for x = = 0 and = W/2Wc ,
one finds Sint ( ) after some reformulation by integration:
Shor ( ) = 4 2 K2/3
2 ( )
(10.66)
Sint ( ) = 2 K5/3 (z)dz.
Both functions are illustrated in Fig. 10.22(a). One sees that the maximum bril-
liance of the SR is reached just below the critical energy, at = W/2Wc 1. For
BESSY II e.g., Wc = 2.5 keV, and the useful spectral range at a dipole deflection
magnet extends up to about 10 keV.
Finally, by integrating (10.62) over all photon energies , one obtains the over-
all angular distribution in respect of (angle normal to the ring plane) which is
plotted in Fig. 10.22(b):
d
2 5/2 5 x2
1+x 1+ . (10.67)
d 7 1 + x2
The FWHM angle is 1.3/ , in fair agreement with the rough estimate 2 = 1/
according to (10.59).
The total radiation emitted may after appropriate rewriting be obtained from
(4.32) or by integration of (10.65) over all frequencies:
e2 2 dp 2 e2 c 4 e4 We2 B 2
P= = = .
60 m2e c3 dt 60 2 60 m4e c3
This expression is obviously the relativistic analogue to (4.32), where in the last
steps (1.110) and = We /(me c2 ) has been inserted. Note the strong, inverse pro-
portionality on the 4th power of the radiating particles mass, here the electron.
Thus, the electron is the particle of choice, if one wants to generate synchrotron
radiation.
540 10 Multi-electron Atoms
(a) (b)
S int ( ) 2.146 d
1/3
2.910 2/3
1.0
S hor ( ) d
1 0.8
0.1 0.6
1.339 1/2 e- F W H M ~_ 1.3/
0.4
0.01
0.2
0.001
0.0001 0.01 1.00 -2 0 2
= /2 Wc
Fig. 10.22 Characteristic energy and angular distribution of SR generated at a deflection magnet.
(a) Brilliance as a function of photon energy in the plane of the ring, Shor corresponding to (10.64)
and integrated over angle, Sint corresponding to (10.65). (b) Energy integrated intensity (10.67)
as a function of the angle above and below the plane of the ring, showing a characteristic angular
FWHM 1.3/
(a) u 1
c =
N S N S N S u
e-
N S N S N S
1/ u (c)
intensity
(b)
Fig. 10.23 (a) Scheme of an undulator with Nu (here =3) periods of aligned, alternately poled
magnetic fields. They force the electrononto an oscillatory trajectory. The opening angle of the
central cone is narrowed by a factor 1/ Nu as compared to a deflection magnet. (b) Photograph
of a U49 undulator for X-ray spectroscopy at BESSY II. (c) Typical spectrum from an undulator
in the oscillation plane with odd harmonics. (We thank Eberhard JAESCHKE et al. 2007, for the
photograph)
eBu B u
K= = 0.93373 (10.69)
2me c T cm
with B being the magnetic field in the undulator magnets. The resonance wavelength
r thus depends on the undulator wavelength u , the beam energy , the undulator
parameter K, and, remarkably, it also depends on the emission angle . A detailed
analysis shows that one has to distinguish two quite different situations: undulators
with K 1 and wigglers for which K 1.
Essential for an undulator (K 1) is, that the oscillation amplitudes of the elec-
trons remain small, so that the amplitudes of the radiation emitted at each of the Nu
magnet pairs can interfere constructively. This induces partially coherent radiation
with a
lators are very popular for intense short wavelength applications due to their superior
performance (see Fig. 10.20).
In contrast, for K 1 the emission cones from the individual undulations be-
come wider and do no longer superpose coherently. Such a situation is then called a
wiggler. A key feature is the generation of very high harmonics with broader band-
width. With increasing K they dominate the overall wavelengths spectrum which
then becomes a broad continuum. In principle, this spectrum does not differ much
from that at a bending magnet except for the fact that it is much more intense and
is shifted to much higher photon energies than achievable at bending magnets with
a given electron energy. Wigglers thus provide convenient sources for soft and hard
X-rays.
In a long undulator with many periods of magnets and suitable design the coherently
superposed radiation from all the magnets may become so intense that it acts back
onto the electron beam which has generated it: a bunch structure emerges, and the
electrons ride so to say on the electromagnetic wave. This implies the feedback and
amplification of the radiation which is needed to achieve complete coherence and
which is characteristic for a laser. Such devices are called free electron laser (FEL).
The setup is very similar to that for an undulator shown in Fig. 10.23, however, with
many magnets and typically a highly relativistic electron beam designed and ded-
icated to this purpose. Such a facility is, in principle, a single user facility, even if
as rule precautions are taken that several users may work quasi in parallel at such
a system (i.e. intermittently). The first systems which have been working very suc-
cessfully for many years by now had been designed for the infrared spectral range.
Presently, worldwide a number of FELs are under construction or already in opera-
tion. They generate intense soft or even hard X-ray short pulses (some fs). A variety
of techniques are applied, one of these being the SASE principle (Self-Amplified
Spontaneous Emission), which was used e.g. at DESY for the VUV laser FLASH
(Free Electron LASer in Hamburg) and will also be implemented at the XFEL (X-
ray Free Electron Laser) there. More efficient, but also more delicate, are concepts
where the FEL action is seeded initially by an already coherent, albeit weak, laser
pulse which may be generated from high harmonics (HHG) of a Titanium-sapphire
laser, as discussed in Sect. 8.5.6. Such concepts are planned, under construction or
even already in operation at all FEL facilities worldwide.
A look at Fig. 10.20 reveals that with these new sources at methodological quan-
tum jump for the interaction of light with matter of various kinds is under way. Not
only is the brilliance at these facilities many orders of magnitude higher than any-
thing one may have dreamed of some years ago. Perhaps even more important is the
temporal structure of such sources, which are designed for coherent femtosecond
pulses in the soft and hard X-ray radiation. They will allow for a new generation
of experiments which we may look forward to with great expectations. To mention
just one out of a host of dream experiments: it should become possible to generate a
10.6 Sources for X-Rays 543
complete X-ray diffraction image from a single, large biomolecule during one sin-
gle laser pulse and thus one would circumvent many efforts and imponderabilities
which arise from the present necessity to crystallize such objects.
Without entering into great details we just mention this interesting and timely sub-
ject, following the introduction to coherent, elastic light scattering in Sect. 8.4.5. The
basic idea is quite simple: if one collides a laser beam (photon energy L ) with a
highly relativistic electron beam ( = We /me c2 1) the radiation is exposed to the
relativistic D OPPLER shift twice on the way in as well as upon reemission. Thus,
for a head on interaction of laser and electron beam and observation in the direction
of the electron beam, the emitted radiation has, according to (1.31), an energy
X 4L 2 . (10.70)
One easily verifies that even at moderate electron energies multi keV radiation may
be obtained in such a setup, e.g. from a 50 MeV electron beam, Titanium-sapphire
laser radiation (800 nm) is scattered as 60 keV X-ray photons. The problem with
this concept as a short pulsed X-ray source is of course the low magnitude of the
T HOMSON cross section. However, if intense laser fields are used, nonlinear effects
come into play (nonlinear, relativistic T HOMSON scattering), essentially related to
the magnetic field component of the radiation. The topic is subject to state-of-the-art
research (see e.g. C HUNG et al. 2011, and references there).
As already indicated these big, new facilities will provide only limited amount of
measuring time, and many experiments stand already in line waiting for their turn.
Thus, it is worthwhile to search for alternatives, which do not necessarily offer the
ultimate specifications of the new XFELs, but are more readily accessible. One class
of such sources are laser based X-ray sources. We specifically mention such setups
where a strongly focussed, very short and intense Titanium-sapphire laser pulse
generates a hot plasma of elements with high Z. The electrons thus generated may
induce bremsstrahlung or characteristic X-ray radiation just as in an X-ray tube
with the essential difference that in this cases an X-ray pulse is generated with a
duration on the order of some 100 fs. This allows one to study dynamical processes.
One example of such a source is shown in Fig. 10.24(a). Its essential ingredients are
a Cu target which is locally ionized by an intense laser pulse (5 mJ, 50 fs, 800 nm),
focussed onto a few m2 . Since the target is vaporized during the laser pulse, one has
to use a moving band-target. A second, this time transparent, rapidly moving tape
protects the laser focussing optics from all debris particles which emerge from the
Cu target. As illustrated by the measured X-ray spectrum, shown in Fig. 10.24(b),
544 10 Multi-electron Atoms
tranparent
protection Cu band / keV
against debris target 2 4 6 8 10 12
Fig. 10.24 Laser based X-ray source for short pulses according to Z HAVORONKOV et al. (2005).
(a) Scheme of an experimental setup with a moving Cu tape source, and moving transparent tape
to protect the optics for focussing the Titanium-sapphire laser pump beam. (b) Measured charac-
teristic X-ray radiation spectrum
References
ATTWOOD , D.: 2007. Soft X-rays and Extreme Ultraviolet Radiation, Principles and Applications.
Cambridge, UK: Cambridge University Press.
B ERGER , M. J., J. H. H UBBELL, S. M. S ELTZER, J.C HANG, J. S. C OURSEY, R. S UKUMAR,
D. S. Z UCKER and K. O LSEN: 2010. XCOM: photon cross sections database (version 1.5),
NIST. http://physics.nist.gov/xcom, accessed: 8 Jan 2014.
C ASIDA , M. E.: 2009. Time-dependent density-functional theory for molecules and molecular
solids. J. Mol. Struct., Theochem, 914, 318.
C HANTLER , C. T., K. O LSEN, R. A. D RAGOSET, J. C HANG, A. R. K ISHORE, S. A. KO -
TOCHIGOVA and D. S. Z UCKER : 2005. X-ray form factor, attenuation, and scattering tables
(version 2.1), NIST. http://physics.nist.gov/ffast, accessed: 7 Jan 2014.
C HUNG , S. Y., H. J. L EE, K. L EE and D. E. K IM: 2011. Generation of a few femtosecond keV
X-ray pulse via interaction of a tightly focused laser copropagating with a relativistic electron
bunch. Phys. Rev. Spec. Top., Accel. Beams, 14, 060705.
H ASCHKE , M. and N. L ANGHOFF: 2007. Mini power X-ray tube and spectrum private
communication, see also: IfG Institute for Scientific Instruments GmbH. http://www.ifg-
adlershof.de/, accessed: 8 Jan 2014.
ISO 21348: 2007. Space environment (natural and artificial) process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
JAESCHKE , E., W. E BERHARD and M. S AUERBORN: 2007. Technical data, sketches, and pho-
tographs about BESSY II. We acknowledge grateful the material provided.
KOHN , W.: 1998. The N OBEL prize in chemistry: for his development of the density-functional
theory, Stockholm. http://nobelprize.org/nobel_prizes/chemistry/laureates/1998/.
K RAMIDA , A. E., Y. R ALCHENKO, J. R EADER and NIST ASD T EAM: 2013. NIST Atomic
Spectra Database (version 5.1), NIST. http://physics.nist.gov/asd, accessed: 7 Jan 2014.
L ATTER , R.: 1955. Atomic energy levels for the Thomas-Fermi and Thomas-Fermi-Dirac poten-
tial. Phys. Rev., 99, 510519.
M ARQUES , M. A. L. and E. K. U. G ROSS: 2004. Time-dependent density functional theory.
Annu. Rev. Phys. Chem., 55, 427455.
PARR , R. G. and W. YANG: 1989. Density Functional Theory of Atoms and Molecules. Interna-
tional Series of Monographs on Chemistry. New York, Oxford: Oxford University Press, 333
pages.
RNTGEN , W. C.: 1901. The N OBEL prize in physics: in recognition of the extraordinary ser-
vices He has rendered by the discovery of the remarkable rays subsequently named after him,
Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1901/.
S AHA , H. P.: 1989. Threshold behavior of the M-shell photoionization of argon. Phys. Rev. A,
39, 24562460.
References 547
S CHUMACHER , E.: 2011. FDAlin programme, computation of atomic orbitals (Windows and
Linux), Chemsoft, Bern. http://www.chemsoft.ch/qc/fda.htm, accessed: 5 Jan 2014.
S CHWINGER , J.: 1949. On the classical radiation of accelerated electrons. Phys. Rev., 75, 1912
1925.
S UZUKI , I. H. and N. S AITO: 2005. Total photoabsorption cross-section of Ar in the sub-keV
energy region. Radiat. Phys. Chem., 73, 16.
S ZABO , A. and N. S. O STLUND: 1996. Modern Quantum Chemistry. Dover, first revised edn.
Z HAVORONKOV , N., Y. G RITSAI, M. BARGHEER, M. W RNER, T. E LSSSER, F. Z AMPONI,
I. U SCHMANN and E. F RSTER: 2005. Microfocus Cu K source for femtosecond X-ray
science. Opt. Lett., 30, 17371739.
Z HENG , L., M. Q. C UI, Y. D. Z HAO, J. Z HAO and K. C HEN: 2006. Total photoionization cross-
sections of Ar and Xe in the energy range of 2.16.0 keV. J. Electron Spectrosc., 152, 143147.
Z WILLINGER , D.: 1997. Handbook of Differential Equations. Boston, MA: Academic Press, 3rd.
edn.
Appendices
Overview
These appendices offer a collection of concepts and relations which belong
to the standard mathematical tools in atomic, molecular and optical physics.
On many occasions we refer to these in the main text. We do not attempt here
to derive them in a closed, axiomatic manner rather a compact summary of
useful and important formulas and equations is presented. Usually they are
accompanied by some hints about their origins and connections to make them
plausible for the reader without requesting the substantial efforts needed to
go into the depth of derivations or proofs. Literature references are given for
further information.
Appendix A offers a tabulated overview of current values for the most important
physical constants used in atomic, molecular and optical physics and in these
textbooks. Some remarks about systems of units and dimensional analysis are also
included.
Appendix B defines the angular momentum operators and introduces some ba-
sics about angular momentum algebra with 3j symbols (respectively C LEBSCH -
G ORDAN coefficients), 6j symbols (respectively R ACAH functions), and 9j sym-
bols.
Appendix C is concerned with the rules of evaluating and transforming matrix
elements of irreducible tensor operators. This includes the very important W IGNER -
E CKART theorem, matrix elements for electromagnetically induced transitions and
the reduction of coupling schemes as well as some useful geometrical relations.
Finally we give some directions for the evaluation of radial matrix elements.
Appendix D formulates the concept of parity and introduces as an alternative
to the usual complex spherical harmonics real basis states for angular momenta,
typically used in quantum chemistry. The concept is also extended to half integer
angular momenta.
Appendix E offers a recipe type of access to the important theme of coordinate
rotation for angular momenta and irreducible tensor operators.
550 Appendices
Appendix F introduces the concept of multipole expansions and gives the most
important relations derived from it. Two alternatives to define multipole tensor op-
erators are discussed: based on spatial coordinates and on combinations of angular
momentum operators.
Appendix G describes the concept of convolution and illustrates it for the most
important distribution functions.
Appendix I presents some essentials about F OURIER transformations with spe-
cial emphasis on electromagnetic fields.
Appendix H provides the formal basis for treating electromagnetically induced
transitions are presented in the main text somewhat heuristically. It also derives
the important T HOMAS -R EICHE -K UHN sum rule.
In Appendix J, finally, we discuss the behaviour and the proper normalization of
continuum wave functions. Both are important for a correct treatment of ionization
and scattering processes.
Constants, Units and Conversions
A
Table A.1 2010 CODATA recommended values of physical constants extracted from NIST
(2010), published by M OHR et al. (2012)
Constant Value Equation
Universal constants
speed of light in vacuum c 2.99792458 108 m s1 defined
magnetic constant 0 4 107 N A2 defined
107
electric constant 0 8.854187817 . . . 1012 F m1 1
0 c2
= 4 c2
characteristic vacuum Z0 376.730313461 . . . " c0 = 1
impedance 0 c
= 0 /0
Newtonian constant of G 6.67384(80) 1011 m3 kg1 s2
gravitation G/c 6.70837(80) 1039 (GeV /c2 )2
P LANCK constant h 6.62606957(29) 1034 J s
4.135667516(91) 1015 eV s
angular momentum quant 1.054571726(47) 1034 J s h/2
c 197.3269631(49) MeV fm
P LANCK mass mP 2.17651(13) 108 kg c/G
P LANCK length P 1.616199(97) 1035 m /(mP c)
P LANCK temperature TP 1.416833(85) 1032 K mP c2 /kB
P LANCK time tP 5.39106(32) 1044 s P /c
Electromagnetic constants
elementary charge e 1.602176565(35) 1019 C
In general, throughout these text books we consequently use the SI units, meter
(m), kilogram (kg), second (s), ampere (A), kelvin (K), mole (mol) and candela
(cd) the international standard of units.
In addition to these 7 base units (NIST, 2000a, Table 1), we employ
We point out here that all values of constants and physical quantities given in the
table above and elsewhere in these text books refer to the currently valid definitions
of the SI. These definitions may possibly change in 2014 and be replaced by a new
set of definitions, completely based on fundamental constants (see Table A.1) as has
been decided by the CGPM (2011) (see also W IKIPEDIA CONTRIBUTORS, 2013).
Alternatively, we often use atomic units (a.u.) which are commonly used in
atomic and molecular physics. We consider these, however, to be just simplifying
554 A Constants, Units and Conversions
abbreviations: one reformulates equations such, that for lengths, times and energies
only the combinations
occur in the final expressions which then become otherwise free of units and
include only some dimensionless parameters, such as the fine structure constant .
It is often helpful to leave the equations in this form which simultaneously includes
a dimensional analysis. However, for simplicity of writing we occasionally omit the
respective units a0 , t0 , and Eh (e.g. in a S CHRDINGER equation), if no ambiguities
can arise. The respective quantities are then measured in a0 , t0 , and Eh . For more
details on a.u. see Sect. 1.8.3.
We emphasize, however, that it is rather misleading, to talk about setting =
e = me = 1 as occasionally formulated in (mostly theoretical) literature. The
same holds for relativistic expressions where the confusing notion that one sets
c = 1 is often used even in general texts, specifically on particle physics.
The attentive reader will find a number of useful references to standard (mostly
older) and sometimes important literature, where G AUSS, electrostatic (esu) or elec-
tromagnetic (emu) units are still used (G AUSS units are equivalent to esu and emu
for electrostatic and electromagnetic units, respectively). Even in current literature,
specifically in theoretical papers, these systems are still used. Obviously, the seem-
ing advantage of writing a few equations in a simpler way than SI demands, ap-
pears so attractive to some authors that they renounce general compatibility. Thus,
Table A.2 communicates some conversion rules between the SI and the G AUSS
systems, following essentially the Appendix on units and dimensions in JACKSON
(1999).
Radian (rad) and steradian (sr) are two somewhat conspicuous units which are fre-
quently used in AMO physics. They are, however, genuine SI derived units (NIST,
2000a). A brief clarification may be useful.
Radian and steradian may be visualized as shown in Fig. A.1. The radian (rad) is
defined in two dimensions for the plane angle by the length of an arc on a circle
of radius r, divided by this radius r:
= . (A.2)
r
With the full circumference of the circle being 2r, the full angle thus becomes
= 2 rad = 360 , so that 1 rad = 360 /(2) = 57.296 .
A.4 Radian and Steradian 555
Table A.2 Relations and conversions between the SI and the Gaussian system
Quantity/relation SI G AUSS
velocity of light c= 1/0 0 c
dielectric permittivity r 0 r
magnetic permeability r 0 r
electric field E, displacement D, D = 0 E +P = r 0 E D = E + 4P = r E
polarization P unit E esu = E SI 40 E SI = E esu / 40
magnetic H and B field, B = 0 (H + M) B = H + 4M = r H
magnetization M unit = r 0 H B SI = B emu 0 /4
B emu = B SI 4/0
M AXWELL equations D = D = 4
E = B
t E = 1c B
t
B = 0 B = 0
H =J + D
t H = 1c (J + D
t )
L ORENTZ force F = q(E + v B) F = q(E + B) v
c
charge, current, etc. q esu = q SI / 40 q SI =
40 q esu
unit 1C = 1As 1esu =1 dyn cm
=
2.9979 109 esu 40 109 N m
=
= 3.3356 1010 C
q1 q2 q1 q2
C OULOMB potential V (r) = 4 0 r V (r) = r
unit electric dipole moment Cm esu cm
D EBYE 1D = 3.33564 1030 C m 1D = 1018 esu cm
Polarizability esu = SI /(4 )
E SI = (4 ) esu
E
E 0 0 E
B OHR magneton B = e/(2me ) e/(2me c)
(a) (b)
rad =
m
In three dimensions the steradian (sr) is defined as the solid angle which char-
acterizes any surface area of the size S on a sphere of radius r. In other words, the
solid angle is given as the surface S on a sphere of radius r divided by the square
556 A Constants, Units and Conversions
of this radius:
S
= . (A.3)
r2
For a cone with a full opening angle 2 as shown in Fig. A.1 we have
= d = 2 sin d = 2(1 cos ). (A.4)
0
With the surface of a sphere being 4r 2 the full solid angle is 4 sr, and 1 sr cor-
responds to an angle = arccos(1 1/(2)) = 0.57195 rad = 32.77 or equiva-
lently to a cone with a full angle = 2 = 65.54 .
Note that expressed in SI base units, both rad = m m1 and sr = m2 m2 have
the unit 1, i.e. are dimensionless. The symbols rad and sr may thus be used where
advantageous to distinguish the respective observable quantities from others . . . or
be omitted wherever the expressions are unambiguous.
For small angles (A.4) may be expanded as
lim = 2 + O 4 .
0
The divergence of well collimated beams (light beams, electron or ion beams
etc.) is often given in mrad2 , i.e. in terms of the full divergence angle = 2 mea-
sured in mrad. In the case of axially symmetry, i.e. for paraxial rays in a conical
beam, the solid angle of divergence is thus
2
= 106 sr.
4 mrad
Thus, a (full) divergence angle of 1 mrad corresponds to a solid angle of =
(/4)106 sr. Note, however, that for a rectangular beam with a small diver-
gence 2 , measured in mrad2 , the solid angle is just = 106 (/ mrad)2 sr.
2. In dimensional analysis we try to avoid specific units for mass, length, and so on,
since physical laws must be independent of the specific choice of units. Rather
we use a basic set of dimensions: mass (M), length (L), time (T), charge (Q),
and temperature (') all written in block letters without serifs. For most of our
purposes all other dimensions of physical quantities can be expressed as a prod-
uct of powers of these basic dimensions, e.g. the dimension of force is MLT2 ,
of energy ML2 T2 , of electric current QT1 , or with the thermal energy being
Wtherm = kB T , the B OLTZMANN constant kB has the dimension ML2 T2 '1
and so on.
3. However, for convenience or for clarity we shall sometimes use more letters or
abbreviations for the dimensions of physical quantities, e.g. for volume V = L3 ,
for energy Enrg = ML2 T2 or R = L3/2 for the dimension of a radial wave func-
tion.
4. The units of a physical quantity A or of a dimension A are represented by square
brackets, [A] and [A], respectively. For instance the unit of a potential V (r)
would be [V (r)] = eV and more generally we have [Enrg] = eV or [Enrg] = J,
both are acceptable.
References
B OUCHENDIRA, R., P. C LADE, S. G UELLATI -K HELIFA, F. N EZ and F. B IRABEN: 2011. New
determination of the fine structure constant and test of the quantum electrodynamics. Phys.
Rev. Lett., 106, 080801.
CGPM: 2011. Resolution 1 of the General Conference on Weights and Measures (2011) CGPM:
On the possible future revision of the International System of Units, the SI, Bureau Interna-
tional des Poid et Messure (BIPM). http://www.bipm.org/en/si/new_si/, accessed: 8 Jan 2014.
558 A Constants, Units and Conversions
JACKSON, J. D.: 1999. Classical Electrodynamics. New York: John Wiley & sons, 3rd edn., 808
pages.
M OHR, P. J., B. N. TAYLOR and D. B. N EWELL: 2012. CODATA recommended values of the fun-
damental physical constants: 2010. Rev. Mod. Phys., 2013, 15271605. http://physics.nist.gov/
constants, accessed: 8 Jan 2014.
NIST: 2000a. Reference on constants, units, and uncertainties: SI base and derived units, NIST.
http://physics.nist.gov/cuu/Units/units.html, accessed: 8 Jan 2014.
NIST: 2000b. Reference on constants, units, and uncertainties: SI prefixes, NIST. http://physics.
nist.gov/cuu/Units/prefixes.html, accessed: 8 Jan 2014.
NIST: 2000c. Reference on constants, units, and uncertainties: Units outside the SI, NIST.
http://physics.nist.gov/cuu/Units/outside.html, accessed: 8 Jan 2014.
NIST: 2010. Reference on constants, units, and uncertainties, NIST. http://physics.nist.gov/
cuu/Constants/, accessed: 7 Jan 2014.
W IKIPEDIA CONTRIBUTORS: 2013. Proposed redefinition of SI base units, Wikipedia, The Free
Encyclopedia. http://en.wikipedia.org/wiki/Proposed_redefinition_of_SI_base_units, accessed:
8 Jan 2014.
Angular Momenta, 3j and 6j Symbols
B
We note here the most important relations for angular momenta in compact form
for reference and use throughout these books. Closely related are the 3j , 6j and 9j
symbols relevant for the coupling of angular momenta. They play an important role
in atomic and molecular physics. We follow here with one important exception
concerning the normalization of the spherical components of angular momenta
strictly the definitions of B RINK and S ATCHLER (1994) and E DMONDS (1996). Ex-
plicit expressions for the 3j , 6j and 9j symbols are somewhat complicated but well
documented in both references as well as in many textbooks on quantum mechan-
ics. For practical evaluation purposes one may use today readily accessible, compact
3j , 6j and 9j calculators (see e.g. S VEN G ATO R EDSUN, 2004). Also, they are of-
ten integrated in standard mathematical programmes. Thus, we refrain here from
extensive tabulations and restrict ourselves to definitions, symmetry properties, or-
thogonality relations, a few matrix elements and some important special cases (as
collected from several sources: E DMONDS, 1996; B RINK and S ATCHLER, 1994;
W EISSTEIN, 2004a,b).
or in compact form:
2
J
J = i
J and J ,
J = 0. (B.2)
Alternatively one may construct J from three spherical components J and J0 (so
called helicity basis) which are an irreducible representation of the rotation group
of rank 1. They are derived from the real components:1
1 1
J+ = (Jx + iJy ), J = + (Jx iJy ) and J0 Jz . (B.3)
2 2
The inverse relations are:
1 i
Jx = (J+ J ), Jy = (J+ + J ). (B.4)
2 2
We note that these operators are constructed in complete analogy to the spherical
components (4.75) of the position operator r in the helicity basis which we have
introduced in the context of dipole transitions in Sect. 4.4.
The spherical components J are not Hermitian, with
The operators J+ and J play a key role in the construction of angular mo-
mentum states. Using their commutation relations (B.6)(B.7) one may construct
angular momentum states without using spherical harmonics (2.79)(2.83) which
1 Very often one finds in textbooks (even in B RINK and S ATCHLER, 1994) these ladder operators
defined as J = Jx iJy . These are, however, not spherical components of the
J vector operator
and lead to asymmetric commutation relations. With the present definition which is also used
e.g. by B LUM (2012), W EISSBLUTH (1978) J fulfill both functions: as ladder (or stepping)
operators and spherical components of J.
B.1 Angular Momenta 561
exclusively represent orbital angular momenta with integer quantum numbers . All
relevant algebra for angular momentum states may be derived in that way. Specifi-
cally, if |J M
are eigenstates of Jz = J0 with eigenvalues M, it follows from (B.6)
that J+ |J M
and J |J M
are also eigenstates of J0 with eigenvalues (M + 1) and
(M 1), respectively:
With some algebra one may show that the only nonvanishing matrix elements of the
spherical angular momentum components J0 , J+ and J are
One special kind of angular momenta are the orbital angular momenta introduced in
Sect. 2.5. As described there, their respective operators may be represented in terms
of spatial coordinates more precisely, the operator of the squared orbital angular
momentum L and the components of
2
L can be expressed as differential operators
with respect to the azimuthal and polar angles in polar coordinates. Of particular
importance are
2 1 1 2
L = Lx + Ly + Lz =
2 2 2 2
sin + 2 (B.18)
sin sin 2
z = i
and L (B.19)
with the spherical harmonics Y m (, ) as joint eigenfunctions. These operators
have the eigenvalues 2 ( + 1) and m, respectively, corresponding to (B.12). Here
= 0, 1, 2, . . . and m is any integer value between m .
Without proof we communicate here the spherical harmonics,
m 2 + 1 ( m)! m
Y m (, ) = (1) P (cos ) exp(im) (B.20)
4 ( + m)!
2
with d Y m (, )Y m (, ) sin d = m m . (B.21)
0 0
1 d 2
P (x) =
x 1 . (B.23)
2 ! dx
As a function of cos the first L EGENDRE polynomials are:
P0 (cos ) = 1
P1 (cos ) = cos
3 1
P2 (cos ) = cos2 (B.24)
2 2
5 3
P3 (cos ) = cos3 cos
2 2
1
P4 (cos ) = 35 cos4 30 cos2 + 3 .
8
B.1 Angular Momenta 563
Y m (, ) = (1)m Y m (, ), (B.26)
Y m ( , + ) = (1) Y m (, ). (B.27)
The spherical harmonics are said to have positive or negative parity depending on
whether is even or odd, respectively (see detailed discussion in Appendix D). In
compact form one obtains well suited for computational use
(1) +m (2 + 1)( m)! d +m (sin )2
Y m (, ) = (sin )m exp(im)
2 ! 4( + m)! d(cos ) +m
(1) (2 + 1)( + m)! d m (sin )2
= (sin )m exp(im).
2 ! 4( m)! d(cos ) m
(B.28)
For simplicity of writing one often also uses renormalized spherical harmonics
4
C m (, ) = Y m (, ). (B.29)
2 + 1
B.2.1 Denitions
They are called C LEBSCH -G ORDAN coefficients. They are non-zero if and only
two conditions are fulfilled:
1. for the Ji quantum numbers (J1 J2 J ) = 1 must be fulfilled, with the triangular
relation defined by:
"
1 if |J2 J1 |, |J2 J1 | + 1, . . . J |J2 + J1 |
(J1 J2 J ) = (B.34)
0 else
Ji , Ji + 1, . . . Mi Ji and (B.35)
M = M1 + M 2
must hold.
The C LEBSCH -G ORDAN coefficients disappear if the triangular relation is not sat-
isfied, i.e. if (J1 J2 J ) = 0, or if M = M1 + M2 .
Alternative to the C LEBSCH -G ORDAN coefficients the W IGNER 3j symbols
J1 J2 J
M1 M2 M
are used. They are defined completely symmetric in respect of the three angular
momenta, and are related to the C LEBSCH -G ORDAN coefficients by
J1 J2 +M J1 J2 J
J1 M1 J2 M2 |J M
= (1) (2J + 1)1/2
. (B.36)
M1 M2 M
Note the minus sign in front of M in the last column. The 3j symbols disappear
unless
(J1 J2 J ) = 1 and M1 + M2 = M. (B.37)
J1 J2 J
2
(J1 J2 J )
and thus also = . (B.42)
M1 M2 M (2J + 1)
M1 M2
The 3j symbols can be computed either by suitable recursion formulae or from the
R ACAH formula:
J1 J2 J
(B.46)
M1 M2 M
= (1)J1 J2 M (M1 +M2 )M
(J1 J2 J )
(J1 + M1 )!(J1 M1 )!(J2 + M2 )!(J2 M2 )!(J + M)!(J M)!
'
(1)t t! (J J2 + t + M1 )!(J J1 + t M2 )!
t
(1
(J1 + J2 J t)!(J1 t M1 )!(J2 t + M2 )!
One has to sum over all values of t which lead to non-negative terms for the faculty
expressions in (B.47). The 3j symbols can be obtained very fast from simple com-
puting programmes, as already mentioned (e.g. S VEN G ATO R EDSUN, 2004, Java
script, also for 6j and 9j symbols).
Closed expressions for some frequently used cases are often useful. The 3j symbol
becomes specially simple if one of the angular momenta is zero:
J1 J2 0 (1)J1 M1 J1 J2 M1 M2
= . (B.48)
M1 M2 0 2J1 + 1
Often one has to couple three angular momenta. A typical case is the hyperfine
structure (HFS) of atoms in one electron systems, where orbital angular momentum,
electron spin and nuclear spin have to be considered. In the case of two electrons,
one may have no orbital angular momentum (ns orbital), the other electron may
be in a n configuration, e.g. Hg(6sn ). Here the spins of the two electrons cou-
ple to a total spin and the latter couples with the orbital angular momentum of the
n electron. In molecular physics too, a host of problems involves similar coupling
requirements, e.g. the molecular rotation may couple with the electrons orbital an-
gular momentum and its spin. And even in optical transitions between spin-orbit
coupled atomic states the photon adds an extra angular momentum from which the
optical selection rules are derived.
Generally, in order to determine the states of three angular momenta one starts
with the product states
and M. This may be done in two alternative schemes, based on the coupling of
two angular momenta described in Appendix B.2. One may first couple J 1 and
J2
to
J 12 and then combine J 12 with
J 3 to
J . According to (B.31) from the product
states |j1 m1
|j2 m2
one obtains
B.3 RACAH Function and 6j Symbols 569
(j1 j2 )J12 M12 = |j1 m1
|j2 m2
j1 m1 j2 m2 |J12 M12
,
m1 m2
j1 (j2 j3 )J23 ; J M = |j1 m1
(j2 j3 )J23 M23 j1 m1 J23 M23 |J M
. (B.60)
m1 M23
For the 6j symbols not to disappear, the 6 angular momenta must be coupled accord-
ing to the above definitions. This requires triangular relations between the angular
momenta. Schematically the following triangular relations must be satisfied:
" # " # " # " #
* * * * * *
, , , . (B.64)
* * * * * *
570 B Angular Momenta, 3j and 6j Symbols
The 6j symbols are invariant against exchanging any two columns or any two rows:
" # " # " # " #
j1 j2 J12 j1 J12 j2 j2 J12 j1 J J23 j1
= = = , etc.
j3 J J23 j3 J23 J J J23 j3 j2 J12 j3
(B.65)
Finally, a number of sum rules hold for the 6j symbols. We note
" #
j1 j2 k
(1) (2k + 1)
2k
= 1, (B.66)
j2 j 1 J
k
One may develop explicit expressions for the 6j symbols from (B.59)(B.61). One
finds according to R ACAH:
" #
j1 j2 J12
(B.68)
j3 J J23
(1)t (t + 1)!
=
(j1 j2 J12 )
(j1 J J23 )
(j3 j2 J23 )
(j3 J J12 )
t
f (t)
and the triangular function
(abc) according to (B.47). The summation is again
over all nonvanishing terms. Again it is commendable to use one of the available
Java-applets for practical evaluations.
The 6j symbols may also be expanded as sums over products of 3j symbols. An
important relation is the contraction formula:
j1 +j2 +j3 +1 +2 +3 1 j2 j3
(1)
m1 2 3
1 2 3
j1 2 j3 j1 j2 3
1 m2 3 1 2 m3
" #
1 2 3 1 2 3
= . (B.69)
m1 m2 m3 j1 j2 j3
B.4 Four Angular Momenta and 9j Symbols 571
For the special case that 1 + 2 + 3 is even and j3 = 1/2, a particularly simple
relation emerges:
j1 j2 3
= (1) 1 2 j2 j1 (2 1 + 1)(2 2 + 1) (B.70)
1/2 1/2 0
" #
1 2 3 1 2 3
.
j2 j1 1/2 0 0 0
Again, some closed expressions for special cases are occasionally useful. With s =
a + b + c and X = b(b + 1) + c(c + 1) a(a + 1):
" #
a b c (1)s (abc)be cd
= (B.71)
0 d e (2b + 1)(2c + 1)
" #
a b c 2(1)s+1 X
= (B.72)
1 c b 2b(2b + 1)(2b + 2)2c(2c + 1)(2c + 2)
" #
a b c 2(1)s
= (B.73)
2 c b (2b 1)2b(2b + 1)(2b + 2)(2b + 3)
[3X(X 1) 4b(b + 1)c(c + 1)]
(2c 1)2c(2c + 1)(2c + 2)(2c + 3)
* -
a+ 1
a 1
(1)s+1 (s 2c + 1)(2 + s)
2 2 = (B.74)
b+ 1
2 b c + 12 2 (2a + 1)(a + 1)(2b + 1)(b + 1)
* -
a+ 1
2 a 1
2 = (1) s+1 (s 2c + 1)(2 + s)
. (B.75)
b b+ 2 c
1 2 (2a + 1)(a + 1)(2b + 1)(b + 1)
For later use, we also communicate here one useful formula. In the case that one of
the j s is zero, the 9j symbol reduces to a 6j symbol:
a b c " #
cf gh (1)b+d+c+g a b c
d e f = . (B.78)
[(2c + 1)(2g + 1)]1/2 e d g
g h 0
References
B LUM, K.: 2012. Density Matrix Theory and Applications. Atomic, Optical, and Plasma Physics,
vol. 64. Berlin, Heidelberg: Springer Verlag, 3rd edn., 343 pages.
B RINK, D. M. and G. R. S ATCHLER: 1994. Angular Momentum. Oxford: Oxford University Press,
3rd edn., 182 pages.
C ONDON, E. U. and G. S HORTLEY: 1951. The Theory of Atomic Spectra. Cambridge, England:
Cambridge University Press, 441 pages.
E DMONDS, A. R.: 1996. Angular Momentum in Quantum Mechanics. Princeton, NJ, USA: Prince-
ton University Press, 154 pages.
R ACAH, G.: 1942. Theory of complex spectra II. Phys. Rev., 62, 438462.
S VEN G ATO R EDSUN: 2004. 3j 6j 9j -symbol java calculator, Sven Gato Redsun. http://www.
svengato.com/, accessed: 8 Jan 2014.
References 573
W EISSBLUTH, M.: 1978. Atoms and Molecules. Student Edition. New York, London, Toronto,
Sydney, San Francisco: Academic Press, 713 pages.
W EISSTEIN, E. W.: 2004a. Wigner 3j -symbol, Wolfram Research, Inc., Champaign, IL, USA.
http://mathworld.wolfram.com/Wigner3j-Symbol.html, accessed: 8 Jan 2014.
W EISSTEIN, E. W.: 2004b. Wigner 6j -symbol, Wolfram Research, Inc., Champaign, IL, USA.
http://mathworld.wolfram.com/Wigner6j-Symbol.html, accessed: 8 Jan 2014.
Matrix Elements
C
In this appendix some tools are collected which are useful for evaluating matrix ele-
ments of irreducible tensor operators. These matrix elements are needed to compute
transition probabilities for optical excitation of atomic and molecular states as well
as for the analysis of collision induced processes.
C.1.1 Denition
1 We refer to J = 1 (Jx iJy ) while R ACAH gives the relation for (Jx iJy ). See also
2
footnote 1 in Appendix B.
Here Tkq is the component of the tensor in the new coordinate system which arises
by rotating the old, un-dashed system through the E ULER angles . It is rep-
resented in term of the components Tks of the tensor in the old system. Important
examples of tensor operators are the spherical harmonics Y m ( ) (rank ) them-
selves, or somewhat more general the eigenstates |j m
of the angular momentum
operator (rank j ).
The adjoint (or Hermitian conjugate) operator Tkq
of a tensor operator Tkq is
defined by (see Sect. 2.3.1)
J M Tkq |J M
= Tkq J M J M = J M|Tkq JM , (C.4)
Tkq
= (1)q Tkq . (C.5)
1 = 1 [V
V x iV
y ]; 0 = V
V z . (C.6)
2
When evaluating matrix elements for a tensor operator, the W IGNER -E CKART theo-
rem is very useful. It says, that for any irreducible tensor operators of a given rank k,
the matrix elements between states |J M
and |J M
depend in identical form on
the projection quantum numbers M, M , and q:
J M Tkq | J M
5
= J Mkq J M J 5Tk J
(C.7)
J J k 5
= (1)J k+M 2J + 1 J 5Tk J
(C.8)
M M q
J M
J k J 5
= (1)
2J + 1 J 5Tk J
. (C.9)
M q M
C.1 Tensor Operators 577
The W IGNER -E CKART theorem allows one to relate two matrix elements of dif-
ferent tensor operators, Sk and Tk of the same rank k with each other in a simple
manner. The ratio R of the reduced matrix elements of two tensor operators of the
same rank k may be computed from any pair of (nonvanishing) matrix elements of
their components:
J Tk J
J M |Tkq | J M
R J , J, T, S = = . (C.12)
J Sk J
J M | Skq | J M
Our definition of reduced matrix elements follows B RINK and S ATCHLER (1994), as-
suming, however, integer rank only, so that (1)2k = 1. Note that these reduced
matrix elements are not symmetric in J and J , rather (2J + 1) J T J
=
k
(2J + 1) J Tk J
. R ACAH (1942), E DMONDS (1996) and others use a symmetric
definition: J Tk J
Racah = 2J + 1 J Tk J
.
578 C Matrix Elements
J M|V J | J M
= R J J M J M
J M| J | J M
(C.15)
M
= R J M|
2
J | J M
= RJ (J + 1)2 .
J M|VJ | J M
R= . (C.16)
J (J + 1)2
Inserting this back into (C.13) gives the very useful projection theorem
J M|VJ | J M
J M Vq | J M
= J M Jq | J M
. (C.17)
J (J + 1) 2
J M|VJ | J M
z | J M
=
J M|V M. (C.18)
J (J + 1)
This is the generalization of the standard scalar product of two vectors R and S (see
e.g. (B.8)). The result is a single scalar quantity.
(1) (2)
C C = (1)m C m (1 , 1 )C m (2 , 2 ). (C.21)
m=
P (cos ) = C m (1 , 1 )C m (2 , 2 ) (C.22)
m=
= C mp (1 , 1 )C mp (2 , 2 ). (C.23)
m=0,p=1
The second line is easily verified with the definition of real tensor operators (D.5)
applied to the (renormalized) real spherical harmonics C mp (, ) (D.7). One also
derives from (C.19) a useful expansion formula for the product of two spherical
harmonics at the same angle
kq m m C m (, )C m (, ) = k0 00 Ckq (, ), (C.24)
m m
where the normalization constant k0| 00
has been derived from the definition of
spherical harmonics at = = 0. Using the orthogonality relations (B.38) for the
C LEBSCH -G ORDAN coefficients one also obtains the inversion of this relation:
C m (, )C m (, ) = Ckq (, ) kq m m k0 00 . (C.25)
kq
580 C Matrix Elements
The often used matrix elements of the spherical harmonics in an | m
basis, nor-
malized as
(2 + 1)(2 + 1)
m Ckq | m
= C m (, )Ckq (, )C m (, )d,
4
(C.26)
may be evaluated by replacing the latter two functions with the expansion (C.25):
(2 + 1)(2 + 1)
m Ckq | m
=
4
C m (, )Ck q (, )d k q k qm k 0k 00
k q
2 + 1
=
k q m k q k qm k 0k 00 .
2 + 1
kq
The last equation is simply the W IGNER -E CKART theorem for the matrix element.
Comparing the latter two equalities gives the reduced matrix element of Ck :
5 k
5
Ck
= (1) (2 + 1) . (C.30)
0 0 0
Note that this expression only holds for even k. The reduced matrix element vanishes
for odd k. For the special case C2 one obtains:
( + 1)
C2
= . (C.34)
(2 1)(2 + 3)
Note the minus sign!
Of special importance in respect of the selection rules for electric dipole transi-
tions is according to (4.79) the case k = 1. Because of (B.34) and (B.49)
(see also Appendix D). With (B.50) the only nonvanishing matrix elements are
5 +1
for = + 1
5C1
=
2 +3
(C.36)
(2 1) for = 1,
Hence, in dipole transitions the selection rule for the projection quantum number
becomes
m = q. The matrix elements vanish unless m = m 1 or m = m. A de-
tailed evaluation of (C.29), using (B.52)(B.55) leads to:
( m + 1)( + m + 1)
( + 1)mC10 | m
= (C.38)
(2 + 1)(2 + 3)
( + m + 1)( + m + 2)
( + 1)(m + 1)C11 | m
= (C.39)
2(2 + 1)(2 + 3)
( m + 1)( m + 2)
( + 1)(m 1)C11 | m
= (C.40)
2(2 + 1)(2 + 3)
( m)( + m)
( 1)mC10 | m
= (C.41)
(2 1)(2 + 1)
( 1 m)( m)
( 1)(m + 1)C11 | m
= (C.42)
2(2 1)(2 + 1)
582 C Matrix Elements
( 1 + m)( + m)
( 1)(m 1)C11 | m
= . (C.43)
2(2 1)(2 + 1)
For the particularly simple case of s p transitions one finds that all allowed tran-
sitions have the same amplitudes:
For evaluating the reduced matrix elements of product tensors angular momentum
algebra provides a useful toolbox. In simple cases, with some algebra and exploiting
6j symbols (B.61), the reduced matrix elements in the coupled scheme |(LS)J M
may be recast into the uncoupled scheme |LML
|SMS
. We communicate here only
(1)
some results. If Rk acts only on the first angular momentum of the coupled scheme
(2)
and Uk 1, we have:
5 (1)
L S J 5Rk LSJ
= S S (1)k+L +S+J (2J + 1) 2L + 1
" #
L L k 5
L 5Rk L
. (C.46)
J J S
Conversely, if Rk
(1)
1 and U acts only onto the second angular momentum in
the coupled scheme, we obtain:
5 (2)
L S J 5Uk LSJ
= L L (1)k+L+S+J (2J + 1) 2S + 1
" #
S S k 5
S 5Uk S
. (C.47)
J J L
C.3 Reduction of Matrix Elements 583
Note: due to the well defined phases of the C LEBSCH -G ORDAN coefficients the se-
quence of the angular momenta is not arbitrary. For the scalar product (rank K = 0)
of two tensor operators of rank k the reduction becomes particularly clear. We first
notice that according to (C.7)
5
j1 j2 J M (Rk Sk )00 |j1 j2 J M
= M M J J j1 j2 J 5Rk Sk j1 j2 J
, (C.48)
where we have used the explicit expression (B.48) for the 3j symbol. Explicitly
(C.48) may be evaluated by inserting on the left side into the scalar product Rk Sk
the quantum mechanical unit operator. Then each part may be reduced according
to (C.46). By applying the orthogonality relations according to R ACAH (1942) (if
one has only to average over the angular momentum part) one finds for the reduced
matrix element of the scalar product:
5
j1 j2 J 5Rk Sk j1 j2 J
= 2j1 + 1) 2j2 + 1 (C.49)
" #
j j k 5 5
(1)J +j1 +j2 1 1 j1 5Rk j1
j2 5Sk j2
.
j2 j 2 J
Most often required are the matrix elements of the spherical harmonics Yk (or of the
renormalized Ck = 4/(2k + 1)Yk ). We have described these in Appendix C.2.2
already in a basis of pure n states. For LS coupled orbital angular momenta with
orbital quantum numbers and the reduced matrix elements Ck
are given
by (C.28).3 In the coupled scheme with a total electron spin S the reduced matrix
element (C.46) is:
3 The following expressions are also valid in multi-electron systems as long as there is only one
active electron (while further electrons have zero orbital angular momentum as it is the case for
alkali atoms, or for He below the first ionization limit). Here and is identical to the total orbital
angular momenta L and L, respectively, while the total spin S 1/2 of the system can in principle
be composed from several particles.
584 C Matrix Elements
5
S J 5Ck SJ
= S S (2J + 1) 2 + 1 (2 + 1) (C.51)
" #
k k
(1)S J .
J J S 0 0 0
For the case S = s = 1/2 (e.g. for an effective one particle system with spin 1/2
i.e. electron in an atom or a single active proton in a nucleus), using (B.70) reduces
(C.51) to:
5 5
5 5
1 5 5 1 j 1/2+k
j j k
j 5Ck 5 j = (1) (2j + 1) . (C.54)
2 2 1/2 1/2 0
For the evaluation of line strengths in E1 transitions, we report here some values for
the reduced matrix element of C1 in the case of a one electron systems:
j j
5 5 1/2 3/2 5/2
5 5 1
1 5 1/2 1/3
j 5C1 5
1/3 0
2 5 2 :
j
3/2 1/3
1/15
2/5
(C.56)
5/2 0 2/5 1/35
C.3 Reduction of Matrix Elements 585
and uses the matrix elements (B.14)(B.17) for the angular momenta. Since the
q act only onto the LML component of the state vector and
operators L Sq only onto
the SMS component one sees that the matrix elements are only nonvanishing for
L
Table C.1 Matrix elements of S/2 for a 2 P state in the uncoupled basis |LML SMS
, with
= ML + 2MS .
ML MS |1 12
|0 12
|1 12
|1 12
|0 12
|1 12
2 1 0 0 1 2
MJ 3/2 1/2 1/2 1/2 1/2 3/2
1 12 | 1/2 0 0 0 0 0
0 12 | 0 0 1/ 2 0 0 0
1 12 | 0 1/ 2 1/2 0 0 0
1 12 | 0 0 0 1/2 1/ 2 0
0 12 | 0 0 0 1/ 2 0 0
1 12 | 0 0 0 0 0 1/2
L
LML SMS | S|LML SMS
z
= LML SMS |L Sz |LML SMS
= 2 ML MS . (C.62)
where in the last step (C.11) was used. The nonvanishing matrix elements of the
components of J may now be written with the W IGNER -E CKART theorem (C.7),
e.g.
J I F MF Jz |J I F MF
(C.64)
F F 1 5
= (1)F 1+MF MF MF 2F + 1 J I F 5J J I F
.
MF MF 0
(a) F = F
J I F MF |Jz |J I F MF
" #
3F +I +J MF (2F + 1)(2J + 1)(J + 1)J J J 1
= (1)
F (F + 1) F F I
(1)2I +2J +1 F (F + 1) + J (J + 1) I (I + 1)
= MF (C.66)
2 F (F + 1)
except for F = 0 where J I F MF |Jz |J I F MF
/ = 0.
(b) F = F + 1
J I (F + 1)MF |Jz |J I F MF
" #
J J 1
= (1)3F +I +J +1
F F +1 I
2(F + MF + 1)(F MF + 1)(2J + 1)(J + 1)J
(2F + 2)
(1)4F +2I +2J +2 MF 2
= 1 (C.67)
2 F +1
(J + I F )(F + J + I + 2)((F + 1)2 + (J I )2 )
.
(2F + 1)(2F + 3)
In the last step we also have inserted the explicit expression for the 6j symbols
(B.72) to obtain these useful, compact relations.
Since C1 may be seen as a tensor of rank 1, this expression contains in principle the
(2 1 + 1)(2 1 + 1) components of the product tensor. One may express these
in irreducible form by tensors of rank 0 (basically the total intensity), rank 1 (so
called orientation) and rank 2 (so called alignment). We shall come back to this in
Appendix C, Vol. 2.
For low Z atoms, LS coupling is characteristic according to the scheme
| LSJ M
and | L S J M
. In this case the selection rules are determined by
the matrix elements S L J M |C1q |SLJ M
. Equation (C.53) factors this matrix
element into its important components. In an electromagnetically induced transi-
tion, the total electron spin does not change, i.e.
S = 0. For E1 transitions we have
k = 1 and with q = 0, 1 one reads in (C.53) from the first 3j symbol the selection
rules for J and M:
M = q and (J 1J ) = 1. Thus M = M 1 or M = M and
J = J 1 or J = J however, 0 0 is strictly forbidden. Finally, the last 3j
symbol recovers the parity selection rule L = L 1.
Selection rules for M1 transitions are derived from the matrix element (5.53)
a|eE0 L + 2
D|b
(M1) b| J +
S|a
B = b| S|a
B , (C.70)
where the index B indicates that the coordinate frame for orbital angular momentum
L and spin
S is chosen here with its z-axis parallel to the magnetic RF field inducing
the transition. This particular polarization dependence can be expressed again by the
W IGNER -E CKART theorem (C.9). Since we discuss a dipole transition with k = 1
we obtain:
L + 2
b| S|a
B = (1)J 1+M 2J + 1 (C.71)
J J 1 5 5
J 5 LJ
+ 2 J 5
SJ
.
M M q
With |
we describe the overlap integral of radial part of the wave function
between initial and final state. The 3j symbol determines the selection rule for the
magnetic quantum numbers M, with q referring in this case to the direction of the
magnetic field vector (rather than the electric field in the E1 case). In spectroscopic
experiments, typically, radio frequency (RF) or microwave radiation (MW) is used
in external, static magnetic field B st , defining the z-axis. If the radiation field vector
B points into this z-direction (i.e. if the polarization vector e and the wave vector k
lie in the xy plane), then q = 0 and transitions with M = M occur. To induce
M =
1 transition (q = 1), the B vector of the inducing field has to be perpendicular
to static field B st .
The two reduced matrix elements in (C.71) determine the selection rules for the
quantum numbers J , L and S. They have to be evaluated (for not too high static
magnetic fields) according to (C.46) and (C.47) in the coupled system. With (C.11)
we have L LL
= L L L(L + 1) and S SS
= S S S(S + 1). Thus we
obtain
5
L S J 5 LLSJ
= S S L L (1)k+L +S +J (2J + 1) 2L + 1
" #
L L 1
L(L + 1) and (C.72)
J J S
5
L S J 5 SLSJ
= S S L L (1)k+L+S+J (2J + 1) 2S + 1
" #
S S 1
S(S + 1). (C.73)
J J L
Hence, the total spin and the total orbital angular momentum do not change during
a M1 transition in LS coupling! In contrast, transitions with J = J 1 or J = J
are possible, provided 0 0. We recall that the radial wave functions for different
principal quantum numbers and otherwise identical configuration are orthogonal.
590 C Matrix Elements
For not too large values of n and one may evaluate (C.75) in closed form,
using standard analytical programs like Maple, Mupad4 or Mathematica on a PC.
We assemble in Table C.2 some characteristic results for n |r|n
in (atomic,
Z scaled) units a0 /Z.
If very high values for n and are required (RYDBERG atoms), there are suitable
alternatives. Typically, it is sufficient to compute n , 1|r|n,
for n = n. Based
on earlier work a closed form has been given e.g. by T OWLE et al. (1996) (again in
units of a0 /Z):
2 F1 , 2 + 1 ; N + 1; x .
with ()t = ( + 1) ( + t 1)
2
x = n n /4n n
= min nr , nr N = max nr , nr ,
= min nr , nr + 2 N = max nr , nr + 2 ,
coefficient A(jj ) corresponding to (4.114)) are found in T OWLE et al. (1996) for H
like RYDBERG atoms. Since for sufficiently high n and all atoms become hydrogen
like, these results are of general relevance.
References
B LUM, K.: 2012. Density Matrix Theory and Applications. Atomic, Optical, and Plasma Physics,
vol. 64. Berlin, Heidelberg: Springer Verlag, 3rd edn., 343 pages.
B RINK, D. M. and G. R. S ATCHLER: 1994. Angular Momentum. Oxford: Oxford University Press,
3rd edn., 182 pages.
C ONDON, E. U. and G. S HORTLEY: 1951. The Theory of Atomic Spectra. Cambridge, England:
Cambridge University Press, 441 pages.
E DMONDS, A. R.: 1996. Angular Momentum in Quantum Mechanics. Princeton, NJ, USA: Prince-
ton University Press, 154 pages.
ISO 21348: 2007. Space environment (natural and artificial) Process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
R ACAH, G.: 1942. Theory of complex spectra II. Phys. Rev., 62, 438462.
T OWLE, J. P., P. A. F ELDMAN and J. K. G. WATSON: 1996. A catalog of recombination lines
from 100 GHz to 10 microns. Astrophys. J. Suppl. Ser., 107, 747760.
Parity and Reection Symmetry
D
The concept of parity conservation plays a key role in modern physics and is used
throughout this book. We present here a short introduction to parity and reflection
symmetry from an atomic and molecular physics perspective.
D.1 Parity
refers to spatial symmetry of states and is defined by:
The parity operator P
P(r) = (r). (D.1)
H
[P, ] = 0.
Thus, the eigenfunctions of the Hamiltonian are also eigenfunctions of the parity
operator. We distinguish states with positive and negative parity (alternatively also
has eigenvalues Pat = 1:
called odd and even), i.e. P
P(r) = (r) = (r).
The product of two functions with unequal parity has negative parity, while the
product of two functions with the same parity has positive parity. We note that the
integral over all space disappears for a function F (odd) (r) with odd parity:
2
F (odd) (r)d 0. (D.2)
0 0
Y m ( , + ) = (1) Y m (, ),
so that the parity of the states n m (r) is Pat = (1) , hence, spherical harmonics
which differ by
= 1 have opposite parity.
Specifically, the relation
= r
Pr (D.3)
holds, and we may say that the dipole operator D = er has odd parity. As trivial
as that statement may seem, it has significant consequences in the context of selec-
tion rules for optical transitions. As discussed in detail in Chap. 4, for an optical
2
transition to occur, the integral 0 0 Y b mb (, )rY a ma (, )d must not disap-
pear. Thus, with (D.2) and (D.3) the product Y b mb (, )Y a ma (, ) must have odd
parity and transitions are thus only allowed if the parity of initial | a ma
and final
state | b mb
is different. A dipole transition thus connects only states with opposite
parity. The selection rule for single electron transitions is
= 1.
Somewhat more general expressing the particle point of view the photon is
said to have odd parity, Pph = 1. The selection rule (4.86) for dipole transitions
may thus be inferred directly and elegantly from a general parity conservation rule
for the whole system including atom and photon. Since the photon with its negative
parity disappears during absorption and appears in emission, absorption or emission
of a photon must be accompanied by a change of parity in the atomic system, hence
the selection rule
= 1.
At this point some general remarks about the choice of basis states or wave functions
and their symmetry properties are in order. For simplicity we first discuss again a
single electron and ignore its spin.
To study different atomic problems some choices of basis states may be more
suitable than others. The one most commonly used in atomic physics for the angu-
lar part of the wave functions (states) is the representation by (complex) spherical
harmonics.
There are, however, a variety of cases where another symmetries may be more
adaptable to a given problem, such as the interaction with an external electric field
(see Sect. 8.2). In that case the eigenstates are linear combinations of states with
M and |M| remains a good quantum number, while the sign of M is indefinite.
Similarly, in optical transitions the polarization of the exciting light may lead to
particular linear superpositions of different basis states as discussed in Chap. 4. All
these linear combinations are per definition also pure states, and in principle, one
can find an infinite number of such states which, properly orthonormalized, could
also serve as a basis set. In standard quantum mechanical textbooks the | m
(or
|LM
) states are somewhat overemphasized and one is led to believe that these are
the only suitable basis states for atomic systems. That is only true if a system is
characterized by the magnitude of angular momentum z .
2
L and its z-component L
In addition, the measurement to be described must select states which diagonalize
both operators simultaneously, so that
2
L |LM
= L(L + 1)2 |LM
(D.4)
z |LM
= M|LM
.
L
i(p1)/2 i(p1)/2
Tkqp = (1)q Tkq + p Tkq = (1)q Tkq + p Tkq
(D.5)
2 2
while for q = 0: Tk0+ = Tk0 and Tk0 = 0.
596 D Parity and Reection Symmetry
i(p1)/2
|J Mp
= (1)q |J M
+ p|J M
for 0 < M J (D.6)
2
|J 0+
= |J 0
and |J 0
= 0 for M = 0.
Note that the relations between real and complex unit polarization vectors (4.7) are
also constructed according to that scheme. And by inserting (B.20) into (D.5) we
obtain the real spherical harmonics
2k + 1
Ykqp (, ) = Ckqp (, ) where
4
2(k q)! q
Ckq+ (, ) = P (cos ) cos(q) for 0 < q k (D.7)
(k + q)! k
2(k q)! q
Ckq (, ) = P (cos ) sin(q) for 0 < q k
(k + q)! k
Rotation of the real tensor operators, real states and real spherical harmonics is
achieved by (E.25) using the real rotation matrices (E.26)(E.28).
D.3 Reection Symmetry of Orbitals Real and Complex Basis States 597
2
L Ykqp = k(k + 1)2 Ykqp
2z Ykqp = q 2 2 Ykqp
L with q = 0, 1, . . . , k, (D.9)
v (xz)Ykqp = pYkqp with p = 1, (D.10)
Often one wants to express (D.7) in Cartesian coordinates. We apply a very use-
ful relation communicated by C AOLA (1978). In our notation the so called solid
harmonics Ykq (x, y, z) are (renormalized):
Ckq (x, y, z)
4
= Ykq (x, y, z) = r k Ckq (, ) (D.11)
2k + 1
(kq)/2
(1)j (x + iy)j +q (x iy)j zkq2j
= (1)q (k q)!(k + q)! .
22j +q (j + q)!j !(k q 2j )!
j =max(q,0)
Inserting this into (D.5) leads to explicit expressions for the renormalized real solid
harmonics Ckqp . Table D.2 summarizes the most important cases.
Figure D.1 illustrates graphically the angular dependence of the three real basis
orbitals for k = = 1: |px
, |py
, and |pz
(short: px , py and pz orbital, respec-
tively) and compares them to the usual complex basis orbitals |1 1
and |10
. Ob-
viously, the real spherical harmonics are represented by dumbbell like states parallel
to the x-, y-, and z-axes of the atom, respectively. In contrast the complex basis con-
sists of two doughnut like orbitals (|1 1
with opposite sense of orientation) while
the |10
orbital is also dumbbell like and identical to |pz
.
One may think of preparing these basis states by optical dipole excitation.
However, since electrons have also a spin, and m are typically not
good quantum numbers. Thus, a complete optical preparation of the orbitals shown
in Fig. D.1 is only possible in very special cases. If e.g. one can make use of a
598 D Parity and Reection Symmetry
z z
Fig. D.1 Angular part of
atomic p orbitals (spherical p =+1
harmonics) in the complex z
(top row) and real (bottom |+1 |0 |-1
row) basis. Plotted are the
squared moduli |Y1q |2 and
|Y1q |2 , respectively. The
colour shading indicates x y x y x y
regions of different signs of
the functions. The and
$ signs indicate positive and
negative reflection symmetry
in respect of the xy plane,
z p =+1
z
p =+1 z p = 1
| px | pz | py
while reflection symmetry in
respect of the xz plane
according to (D.10) is
characterized by red p = +1
and p = 1 x y x y x y
(a) The complex basis set requires circularly polarized light for its preparation: left
and right hand circularly polarized, propagating along the z-axis to excite the
|1 +1
(equivalent to Y1+1 ) and the |1 1
(equivalent to Y11 ) states, respec-
tively. Linearly polarized light propagating perpendicularly to the z-axis and
with its E vector aligned parallel to the z-axis excites the |10
state (equivalent
to Y10 ).
(b) The real basis set requires linearly polarized light for its preparation: the |pz
state (identical to |10
) is excited as just described, while the preparation of |px
D.4 Reection Symmetry in the General Case 599
and |py
requires linearly polarized light with its E vector parallel to x and y,
respectively.
Generally speaking, the real basis functions are the basis of choice when the sign
of the angular momentum is irrelevant, i.e. whenever there is no specific preferred
orientation. Thus, they are relevant if external electric fields are present. They play
a particularly important role in molecular physics and quantum chemistry as dis-
cussed in detail in Chaps. 34 of Vol. 2, but also in collision problems. There, the
electrostatic field between different, neighbouring atoms defines a suitable z-axis
and is typically very strong so that spin quantum numbers can be treated as a mere
afterthought. Thus, the orbital angular momentum quantum numbers and the abso-
lute value of their projection onto this axis become good quantum numbers. Real
atomic orbitals (AOs), as depicted in the bottom row of Fig. D.1, and linear com-
binations of them attached to different atoms are the basis for a quantitative
treatment of electronic wave functions in molecules.
Reflection symmetry is also an important property in the context of collision
problems. As we shall discuss in Chaps. 68, Vol. 2, reflection symmetry in respect
of the scattering plane is conserved during a binary collision for the combined wave
function of the interacting particles. Thus, either reflection symmetry of both parti-
cles remains constant during the collision, or both particles change their reflection
symmetry in respect of the collision plane.
J
v (xy)|J M
= (1) DJM M (, 0, 0)J M = eiM (1) |J M
M
= (1)+M |J M . (D.12)
We note here in passing that with this relation the |+1
and |1
p states in Fig. D.1
have both positive reflection symmetry in respect of the xy plane, while the |0
state
has negative reflection symmetry as intuitively obvious.
Reflection at the xz plane leads to
v (xz)|J M
= (1) DJM M (0, , 0)J M
M
= (1) dMM
J
()|J M
= (1)+M+J |J M
. (D.13)
Here we have used the symmetry relation (E.9) for the reduced rotational matrix.
From (D.12) and (D.13) we see that the standard complex |J M
representation are
eigenstates of the xy plane reflection operator but not of the operator for reflection
in the xz plane. In analogy to (D.5) one may, however, construct also for the general
case (J integer or half integer) states of specified reflection symmetry in respect of
the xz plane:
|J Mp
= v(p, M) (1)M |J M
+ (1)+J p|J M
with (D.14)
1/2 for M = 0
v(p, M) = 1/ 2 for 0 < M J and p = 1 or +i (D.15)
i/ 2 for 0 < M J and p = 1 or i.
v (xz)|J Mp
= p|J Mp
.
(D.16)
1. The states defined by (D.14) are eigenstates of Jz2 . One easily derives
"
+1 for p = +1 or +i
Jz |J Mp
= iM J M(p)
1 for p = 1 or i
so that Jz2 |J Mp
= M 2 2 |J Mp
. (D.17)
2. Jz changes the reflection symmetry from +(1)J to (1)J and vice versa, and
so does Jx :
Jx |J Mp
= a(M)J (M + 1)(p) + a(M)J (M 1)(p) (D.18)
as one verifies for J 1 and M > 0 easily with some algebra. In contrast Jy
does not change reflection symmetry:
Jy |J Mp
= a(M)J (M + 1)p a(M)J (M 1)p (D.19)
a(M) = J (J + 1) M(M + 1).
2
3. As specific examples, we have already discussed the p orbitals |px
= |11+
,
|py
= |11
and |pz
= |10+
depicted in the lower part of Fig. D.1. Without
proof we note here in passing the relations:
4. The simplest, and somewhat exceptional case with half integer J is a pure spin
state with = 0, J = 1/2, describing e.g. an electron in a 1 S1/2 state. The basis
states in the |J M
representation are |1/2 1/2
and |1/2 1/2
, with spin point-
ing in the +z- and z-direction, respectively. Reflection on the xy plane gives
according to (D.12)
1 1 1 1 1 1 1 1
v (xy)
= i and
v (xy) = i . (D.21)
22 22 2 2 2 2
We see that the up and down spin states have opposite reflection symmetry in
respect of the xy plane. The corresponding states with defined xz reflection sym-
metry are obtained from (D.14) ( = even) with p = i:
1 These formulas hold for all J 1 and M 1. The other cases are easily derived individually
1 1 1 1 1
|+i
= i (D.22)
2 22 2 2
1 11 1 1
|i
= i
+ . (D.23)
i 2 22 2 2
These states are identical to (E.17) and (E.18), respectively, obtained by coor-
dinate rotation for a spin state pointing into y direction. Reflection at the xz
plane gives with (D.13):
In this particular case (and only in this one), these xy reflection symmetric states
are also eigenvectors of Jy . With (B.4) and (B.14)(B.17) one easily verifies that
Jy |i
= |i
. (D.25)
2
We finally apply the concept of reflection symmetry to optical transitions treated
in detail in Chap. 4. In Sect. 4.1.2 we have introduced linear (ex , ey ) and circular
(e+1 , e1 ) polarization vectors. We may write |ex
, |ey
and |e+1
, |e1
as the real
and complex representation of the corresponding photon states, respectively. Then
the definitions (4.4)(4.7) are in complete agreement with (D.14). Since for the pho-
ton J = = 1, M = 1, we have p = +1 for |ex
and p = 1 for |ey
, i.e. |ex
and |ey
have positive and negative reflection symmetry in respect of the xz plane.
Since the (fully quantized) Hamiltonian for the system photon + atom is invariant
under reflection at any plane through the centre of the atom, the reflection symmetry
of the state vectors describing the combined system must conserve reflection symme-
try. Depending on whether we use light which is polarized in the x- or y-direction,
the atom changes or conserves its reflection symmetry in an optical transition as
indicated by the following scheme:
|J Mp; ex
|J Mp; . . .
|J Mp; ey
J M(p); . . . .
to the transition has +i (1) = i reflection symmetry, thus the final state must
also have i symmetry, which corresponds to |Jb
pointing into +y-direction. In
contrast, for excitation with ex polarized light, the combined initial system |Ja
|ex
has +i (+1) = i symmetry, which must be conserved and corresponds to reversing
the orientation of the spin into |Jb
.
Circularly polarized photons |e+1
or |e1
propagating into z-direction have
both positive reflection symmetry in respect of the xy plane (in analogy to the |+1
and |1
p states in Fig. D.1) and cannot change this reflection symmetry of the
atomic states. Formally one sees this from (D.12), since in a dipole transition both
and M change by 1 in respect of the photon frame where z k. Thus, (1)+M
remains constant during the excitation process and reflection symmetry in respect
of the xy plane does not change.
good quantum number: Quantum number for eigenvalues of such observables that
may be measured simultaneously with the H AMILTON operator (see Sect. 2.6.5).
References
C AOLA, M. J.: 1978. Solid harmonics and their addition theorems. J. Phys. A, Math. Gen., 11,
L23L25.
Coordinate Rotation
E
We compile here the most important formulas for rotation of coordinate systems
and their consequences on angular momentum states and tensor operators without
going into details of the derivation.
(xyz) x y z x y z x # y # z# .
First, one rotates the system around its original z-axis (z = z) through an angle .
The following rotation through an angle around the new y axis (y = y ) brings
z into the direction z and x into direction x . The final rotation through an angle
is around the thus tilted z = z# axis (x x # and y y # ).
The operator achieving this rotation is derived from the angular momentum opera-
tor
J:
D( ) = exp(i Jz ) exp(i Jy ) exp(i Jz ).
The corresponding rotation matrix elements, more precisely the matrix elements
J of the rotation group, may be derived with
of the irreducible representation D
some angular momentum algebra (following the notation of B RINK and S ATCHLER,
1994),1 and are written as
DJMN ( ) = J M|D( )|J N
= exp(iM)dMN
J
() exp(i N ), (E.1)
Special cases:
4
DJM0 ( ) = CJ M (, ) = Y (, ) (E.3)
2k + 1 J M
(J M)! M
J
dM0 () = (1)M P (cos ) if M > 0 (E.4)
(J + M)! J
J
and d00 () = PJ (cos ). (E.5)
1 Here the coordinate system is rotated. An alternative convention, also found in the literature, is to
2
d22 = d22
2
= cos4 (/2) (E.13)
2 = d2 = d2
20 = d02
2 = d 2
12 0 = d21 = d12
d20 2 2 2
02 d21
= ( 3/8) sin2 = sin (1 + cos )/2
2
d21 = d12
2 0 = d 2
21 = d12
2
2
d22 = d22
2 = sin4 (/2)
= sin (1 + cos )/2
2
d11 = 2
d11 2
d11 = d11
2
The components Tkq ( # , # ) of tensor operators of rank k (e.g. Ykq or a vector with
k = 1) can be expressed by tensor components in the old system
Tkq # , # = Tkq (, )Dkqq ( )
q
= exp i q k
Tkq (, ) exp(iq)dqq () (E.15)
q
with , and # , # being the angles in the old and new system, respectively.
We illustrate the procedure by a particularly simple and important example,
a spin s = 1/2 state |1/2 1/2
# pointing into the = , = direction in the orig-
1/2
inal coordinate frame xyz. Inserting for this case dMN () from (E.11) into (E.14)
we obtain:
608 E Coordinate Rotation
(a)
(b)
Fig. E.2 Rotation to describe a spin s pointing into (a) +y-direction, (b) +x-direction
#
1 i i 1 i 1
= exp exp cos + exp sin .
2 2 2 2 2 2 2 2
(E.16)
(For simplicity of writing we have dropped here J = 1/2.) The third rotation angle
around the final z# axis is somewhat arbitrary and just contributes a free phase
factor.2 Figure E.2(a) indicates the rotation angles for one specific case where the
spin s points into the +y-direction of the (xyz) coordinate system. Thus, the oblig-
atory rotation angles are = = /2 and = = /2. We choose = 3/2,
make use of (E.11)(E.13), and thus obtain from (E.16) for this particular spin state
1 #(y) 1 1 1
+ = i+ . (E.17)
2 2 2 2
If the spin s points into the y-direction, we make exactly the same rotation and
find from (E.14)
1 #(y) 1 1 1
= i + + (E.18)
2 i 2 2 2 .
Somewhat more compact, we designate the spin states parallel and antiparallel to
the y-axis with by |
(y) and |
(y) , respectively, and those which refer to the z-axis
with |
and |
. Equations (E.17) and (E.18) may then be written:
1 1
|
(y) = i|
|
and |
(y) = i|
+ |
. (E.19)
2 i 2
For completeness we also describe a state where the spin s points into the x-
direction. As illustrated in Fig. E.2(b) the E ULER angles in this case are (0, /2, )
and we obtain
1 #(x) 1
= |
|
. (E.20)
2 2
All these states are of course eigenstates of the squared spin angular momentum
S 2 , with spin quantum number s = 1/2. The eigenvalue of S 2 are 2 s(s + 1) =
3 /4. And specifically |
and |
are also eigenstates of Sz , the spin projection
2
onto the z-axis, with eigenvalues /2 and /2, respectively. What about the states
|1/2
(x,y) ?
Here is a little exercise for the active reader, which can be solved with a few
lines of algebra (for complex quantities): Using (B.4), (B.16) and (B.17), show that
the spin states |
(y) and |
(y) are indeed eigenstates of the spin component in y
direction with eigenvalues /2:
Sy |
(y) = |
(y) and
Sy |
(y) = |
(y) . (E.21)
2 2
And the same holds of course for the corresponding spin states pointing in x-
direction.
For many-particle systems we have to construct spin functions which account for
the fact that identical particles are indistinguishable. We have discussed this in the
main text on several occasions, see e.g. Table 6.1 in the context of fine structure and
Sect. 7.3 where specifically the He atom is treated. Let us take the latter example,
i.e. the two electrons in the He ground state (1s)2 1 S0 , or any singlet state of two
electrons in the continuum after ionization (with a symmetric spatial wave function).
The two spins of the electrons are anti-parallel and since we cannot distinguish
which of these two fermions has which spin we have to construct an antisymmetric
description of their joint state, according to (7.36)
0
= 1 (1)(2) (1)(2) , (E.22)
0
2
where the numbers in brackets refer to electron (1) and electron (2), respectively.
Obviously, this state of the two electrons is inseparable, i.e. it cannot be written
as product of two wave functions of the individual electrons. One calls such states
entangled.
Equation (E.22) refers to a particular coordinate frame and |
and |
reflect
the spin components with respect to its z-axis. The total (spin) angular momentum
610 E Coordinate Rotation
operator is
S = S1 +
S 2 , and the singlet state (E.22) is constructed such that the
respective eigenvalues of S and
2
Sz are zero:3
2
Sz 00 = (
S2z )00 = 0 and
S1z + S 00 = 0. (E.23)
One may now be curious what happens if we define this entangled, antisymmetric
spin state with respect to another axis, say the y-axis. This may be done in complete
analogy to (E.22) from spin states with angular momentum parallel and antiparallel
to the y-axis:
0
= 1 (1)(2) (y) (1)(2) (y) . (E.24)
y
2
In full analogy to (E.23) we have now Sy |y0
= 0, i.e. the projection of the total spin
onto the y-axis is also zero. The relation (E.19) between the |
(y) and |
(y) states
and the |
and |
states parallel to the z-axis may now be inserted into (E.24):
with just a few lines of algebra (in which many terms cancel) one recovers again
(E.22), i.e. the singlet state with respect to the z-axis. Thus, the definition of this
entangled state does not depend on the coordinate axis and its angular projection is
zero in all coordinate systems.
So much about coordinate rotation for entangled states. Entanglement is a very
active field of current research with far reaching consequences, not only for funda-
mental conceptual and philosophical reasons (dating back to the famous E INSTEIN ,
ROSEN , P ODOLSKY 1935 paper, quoted by now over 5000 times), but also in view
of a number of potential future applications such as quantum cryptography or quan-
tum computation. We cannot enter here in a detailed discussion of this wide field
but refer the reader to some relatively recent reviews and references quoted there
(T ICHY et al., 2011; H ORODECKI et al., 2009; R AIMOND et al., 2001), see also the
respective chapters in B LUM (2012).
For reference we also communicate the rotation procedure (FANO, 1960) for the
real harmonics, real states and real tensor operators as defined in Appendix D. In
this case, the transformation is achieved by
# #
Tkq p , = Tkqp (, )Dkqp,q p ( ) (E.25)
qp
3 The first relation is obvious, since each of the operators S1z and
S2z acts only onto one of the
states which have the eigenvalues /2, i.e. they compensate each other. The latter relation may
easily be verified as an exercise with the help of (B.8) and (B.17).
References 611
with p, p = 1. The real rotation matrices are obtained by applying the complex
rotation (E.15) onto the defining equations (D.5) and finally back-transformation to
the real basis. One obtains:4
For k q, q > 0:
(E.26)
Dkqp,q p ( )
k
k
(1)q dqq () cos(q + q ) + p dqq () cos(q q )
if p = p
= (1)q
(1)q p dqq
k () sin(q + q ) + d k
qq () sin(q q )
if p = p
For q or q = 0: (E.27)
"
k cos(q + q ) if p = p
Dkqp,q p ( ) = (1)q+q 2dqq ()
sin(q q ) if p = p
specifically Dkqp,0+ ( ) = Ckqp (, ) (see Eq. (D.7))
References
B LUM, K.: 2012. Density Matrix Theory and Applications. Atomic, Optical, and Plasma Physics,
vol. 64. Berlin, Heidelberg: Springer Verlag, 3rd edn., 343 pages.
B RINK, D. M. and G. R. S ATCHLER: 1994. Angular Momentum. Oxford: Oxford University Press,
3rd edn., 182 pages.
E INSTEIN, A., B. P ODOLSKY and N. ROSEN: 1935. Can quantum-mechanical description of
physical reality be considered complete? Phys. Rev., 47, 777780.
FANO, U.: 1960. Real representations of coordinate rotations. J. Math. Phys., 1, 417423.
H ERTEL, I. V. and W. S TOLL: 1978. Collision experiments with laser excited atoms in crossed
beams. In: Adv. Atom. Mol. Phys., vol. 13, 113228. New York: Academic Press.
H ORODECKI, R., P. H ORODECKI, M. H ORODECKI and K. H ORODECKI: 2009. Quantum entan-
glement. Rev. Mod. Phys., 81, 865942.
R AIMOND, J. M., M. B RUNE and S. H AROCHE: 2001. Colloquium: Manipulating quantum en-
tanglement with atoms and photons in a cavity. Rev. Mod. Phys., 73, 565582.
T ICHY, M. C., F. M INTERT and A. B UCHLEITNER: 2011. Essential entanglement for atomic and
molecular physics. J. Phys. B, At. Mol. Phys., 44, 192001.
4 See also H ERTEL and S TOLL (1978), Eq. (15) (some small typos there have been corrected here).
Multipole Expansions and Multipole Moments
F
One often needs suitable approximations for potentials which at large distances are
nearly symmetric around a centre, but not completely. For scalar as well as for vector
potentials, multipole expansions may be derived. We treat here the important case
of an electrostatic potential arising from a charge distribution localized in a small
region around the origin as e.g. encountered when describing the interaction of
atomic nuclei with the electron charge cloud of an atom (see Fig. 9.12 in Chap. 9).
In the final subsection we shall also introduce a more general definition of multipole
tensor operators and multipole moments.1
We first note that the inverse distance between two points r and r in space may be
expanded for r r into a power series of x = r /r:
1 1 1/2
= 1 2x cos + x 2 (F.1)
|r r | r
1 3 1 5 3
= 1 + x(cos ) + x 2 cos2 + x3 cos3 cos
r 2 2 2 2
where is the angle between r and r . With (B.24) we identify the angle dependent
terms in brackets as L EGENDRE polynomials. With this result in mind, we commu-
nicate the L APLACE expansion which holds for r < r:
r k
1
= Pk (cos )
|r r| r k+1
k=0
1 The term multipole moments is used in the literature for a variety of slightly different defini-
tions. As far as necessary for distinction, we shall add some specification, indicated in (brackets).
r k r k (r) (r )
k
= C kq (, )C kq , = C Ck (F.2)
r k+1 r k+1 k
k=0 q=k k=0
r k
k
= Ckqp (, )Ckqp , .
r k+1
k=0 q=0,p=1
In the second line of (F.2) we have made use of the addition theorem (C.22) for
the renormalized spherical harmonics, with , and , being the angles of the
position vectors r and r , respectively. In the following equality we have made use
(r) (r )
of (C.21) and the scalar product notation (C.20) for the two tensors Ck and Ck .
The third line of (F.2) introduces the real (renormalized) spherical harmonics as
defined in Appendix D.3.
With these relations we write the electrostatic interaction energy dV12 (r, r ) of a
charge dq, located at a position r , with a point charge e located at r
1 edq 1 e (r) k (r )
dV12 r, r =
= Ck r dq Ck . (F.3)
40 |r r| 40 r k+1
k=0
For an extended electric charge with density (r ), localized around the origin, the
charge element is dq = (r )d3 r , and the total interaction energy with the point
charge e is obtained from
e (r ) 3
V (r) = dV12 r, r = d r. (F.4)
40 |r r|
With (F.3) we may write this as a multipole expansion
e 1
k
V (r) = Qkq Ckq (, ) or (F.5)
40 r k+1
k=0 q=k
e 1
k
in real form = Qkqp Ckqp (, ), (F.6)
40 r k+1
k=0 q=0,p=1
where we have used in the second row for abbreviation the renormalized real solid
harmonics Ckqp as introduced in Appendix D.3.
These multipole expansions, derived from (F.3), are most appropriate if r r.
Thus, it is suitable e.g. for describing the field of an atomic nucleus as seen by an
electron far away from the nucleus (i.e. for 1), or for the far field of an atom in
a well defined polarization state (with 1).
The total charge is given by
q = Q00 = r d3 r . (F.9)
If only positive or only negative charges are involved, the dipole moments are made
to disappear by suitable choice of the origin, thus defining a charge centre. They
also disappear if both charges have the same centre (characteristic for atoms, unless
they are placed in an electric field, see Sect. 8.2.8). In contrast, if positive and neg-
ative charges with different centres are to be described (typical for polyatomic and
heteronuclear diatomic molecules) the dipole moment is finite.
With (F.8) and Table D.2 the real quadrupole moments are:
1 2
Q20+ = 3z r 2 r d3 r Q20 = 0
2
3 3
Q21+ = 3 z x r d r Q21 = 3 zy r d r (F.11)
3 2 3
Q22+ = x y 2 r d3 r Q22 = x y r d3 r .
2 2
616 F Multipole Expansions and Multipole Moments
e q D r 1 C2qp (x, y, z)
V (r) = + 3 + Q2qp + . (F.13)
40 r r 2 qp r5
The components (F.10) of the electric dipole moment D are zero for charges cen-
tred at the origin, while the quadrupole moments Q2qp according to (F.11) are fi-
nite for an anisotropic charge distribution. The renormalized real solid harmonics
C2qp (x, y, z) are found in Table D.2.
Closely related to the (charge) multipole moments just discussed, but without any
reference to the charge of the particles, one defines irreducible spatial multipole
tensor operators of rank k. They are constructed from the position vector and have
the components
using again the (renormalized) solid harmonics (D.11). They may also be written as
real quantities according to (D.5):
For explicit expressions we refer to Table D.2. General rules for tensor operators
have been laid out in Appendix C.
2 E.g.
following JACKSON (1999), in classical electrodynamics one usually expands in terms of the
standards spherical harmonics, multiplying the terms in the sum (F.7) by 4/(2 + 1), so that the
(complex) multipole moments are given as
q m = Y m (, )R (R)d3 R.
F.3 Multipole Tensor Operators 617
The expectation values Qkq
, i.e. the matrix elements J M|Q2q | J M
aver-
aged over the population probabilities w(M) of | J M
states, are called (spatial)
multipole moments:
Qkq
= r 2 w(M)J M|Ckq ( )|J M
. (F.18)
M
They characterize the particle distribution (e.g. for the protons in an atomic nucleus,
for the electrons in an atom, or for electrons and nuclei in a molecule).
For a given angular momentum J of a quantum state the above definitions (F.14)
and (F.16) allow to construct multipole tensor operators and multipole moments of
rank 0 k 2J . They all have a well defined parity
kq (r) = (1)k Qkq (r)
PQ (F.19)
but also a reflection symmetry in respect of the xy plane (see also Appendix D.4):
Most commonly used is the quadrupole tensor Q2 with its components Q2q =
C2q (r, , ) according to Table D.2. Higher moments become increasingly com-
plicated. If the quantum state |J M
of an object is known exactly, the components
of the quadrupole moment are obtained from the respective matrix elements of the
quadrupole tensor Q2 . These matrix elements can be derived from the W IGNER -
E CKART theorem (C.9):
J M
J J 2
J M Q2q | J M
= (1) 2J + 1 J Qk J
(F.22)
M M q
with J Q2 J
= J C2 J
J |r 2 | J
. (F.23)
and if the radial wave functions Rj (r) are known, the radial matrix element r 2
is obtained from
2
r = J |r | J
= r 4 |Rj |2 dr.
2
The reduced matrix element J C2 J
still depends on the coupling scheme of the
states under discussion. For LS coupling detailed expressions for the renormalized
spherical harmonics have been derived in Sect. C.3.1.
Specifically, for a quasi-one-fermion system with spin s = 1/2 (one active atomic
electron or nuclear proton) we use (C.55) to obtain the diagonal matrix elements:
1 1
j mQ20 j m
2 2
1 1
= j mC20 j m j |r 2 |j
2 2
2j m1/2 j j 2 j j 2 2
= (1) (2j + 1) r
m m 0 1/2 1/2 0
1 3m2 j (j + 1) 2
= r . (F.28)
4 j (j + 1)
For the last equality we have used (B.58) for the 3j symbols and the fact that j is
half integer, so that (1)2j = 1. The highest (positive) value is found for m = 0,
the lowest (negative) value for m = j :
1 1
j 0Q20 j 0 = r 2 /4 (F.29)
2 2
1 1 1 2j 1 2
jj Q20 jj = r . (F.30)
2 2 4 j +1
F.3 Multipole Tensor Operators 619
The expectation values of the quadrupole tensor, i.e. the (spatial) quadrupole
moments according to (F.18), may be related to the (charge) quadrupole moments
defined in Sect. F.2 by
Q2q = qQ2q
= r 2 J C2 J
w(M)J M20|J M
. (F.31)
For a prolate (cigar shaped) charge distribution, dominated by low M states, (F.31)
leads to Q > 0, while an oblate (pancake like) distribution is characterized by large
|M| and Q < 0 (see e.g. Fig. 9.15).
In nuclear physics, the quadrupole moment of a nucleus (assuming cylindrical
symmetry around the z-axis) is
Note the factor 2eZ (overall charge eZ) as already introduced in (F.12): averaging
J M|Q20 | J M
over state populations w(M) is equivalent to integration over
the charge distribution (r) = eZ|(r)|2
. The probability density |(r)|2 refers
to the total wave function for the charged particles of interest, while the population
of the states w(M) depends on the specific experiment.3 For a single active spin 1/2
particle system (i.e. a proton in the nucleus) in a pure |jj
state, instead of (F.30)
the nuclear quadrupole moment thus reads (see also B OHR and M OTTELSON, 1998,
Eq. (3-27))
1 1 2j 1
Qsp = 2 jj Q20 jj = eZ j |r 2 |j
.
2 2 2j + 2
3
Tkq = (Ji1 Ji2 . . . Jik )i1 i2 . . . ik Ckq (r), (F.32)
i1 i2 ...ik =1
where ij = 1 to 3 stands for x, y and z. The angular momenta are given in a.u., i.e. in
units of . Applying this operation to the explicit expression for the (renormalized)
solid harmonics Table D.2 we obtain Table F.1.
Table F.1 Multipole tensor operators constructed from angular momenta for k = 1 to 3 (only the
zero component is given for the octupole moment k = 3); for reference we also communicate the
relations with the FANO and M ACEK (1973) orientation (O1 ) and alignment parameters (A1+
and A2+ )
k p = +1 p = 1
0 T00+ = 1 T00 = 0
1 T10+ = Jz T10 = 0
1 T11+ = Jx T11 = Jy = J (J + 1)O1
T20+ = (3Jz2 T20 = 0
2
2 J )
2 T21+ = 3(Jz Jx + Jx Jz ) = J (J + 1) 3A1+ T21 = 3(Jz Jy + Jy Jz )
2 T22+ = 3(Jx2 Jy2 ) = J (J + 1) 3A2+ T22 = 3(Jx Jy + Jy Jx )
T30+ = (15Jz3 9
J Jz + 3Jz )
2
3 etc.
The Tk0 (= Tk0+ ) components are obviously diagonal in J and M. Explicitly, the
matrix elements for k = 1 are
M
J M|T10 |J M
= M = J T1 J
(F.33)
J (J + 1)
and describe the orientation of the state (nonvanishing angular momentum). The
latter equality follows from the W IGNER -E CKART theorem (C.9) and (B.55). The
matrix elements of T20 characterize the alignment (quadrupole moment or the
anisotropy) of the charge density distribution in a given state:
J M|T20 |J M
= 3M 2 J (J + 1) (F.34)
2[3M 2 J (J + 1)]
= J T2 J
.
(2J + 3)(2J + 2)2J (2J 1)
The latter follows again from the W IGNER -E CKART theorem, this time using
(B.58).
The reduced matrix elements for rank k = 1 and k = 2 follow directly from (F.33)
and (F.33), respectively:
J T1 J
= J (J + 1) and (F.35)
J T2 J
= (2J + 3)(J + 1)J (2J 1). (F.36)
For an arbitrary rank k one finds (see e.g. M ACEK and H ERTEL, 1974, and refer-
ences given there; note, however, that the factor (2J + 1)1/2 is specific for the
B RINK and S ATCHLER notation of the reduced matrix elements):
k! (2J + k + 1)!
J Tk J
= k . (F.37)
2 (2J + 1)(2J k)!
We point out that these (general) multipole tensor operators and their expectation
values, the (general) multipole moments, may be finite also for odd rank in contrast
Acronyms and Terminology 621
J Tk J
J M Tkq |J M
= J M Qkq |J M
, (F.38)
J Qk J
with J Qk J
defined by (F.23). We emphasize that (F.38) also holds for the re-
spective expectation values (the multipole moments).
Often one is only interested in the expectation values (multipole moments) of
these tensors in an orbital angular momentum basis with sharp = (see e.g. Ap-
pendix D.2 in Vol. 2). We may then use the explicit expression (C.33) for the reduced
matrix elements of the renormalized spherical harmonics. These are nonvanishing
only for even k and we note that their sign is always negative in contrast to (F.37)!
Thus, the multipole moments Tkq
(constructed from angular momenta) have the
opposite sign from Qkq
(which are built from spatial coordinates)!
Expectation values of these multipole tensor operators are widely used to char-
acterize anisotropy and orientation of atomic or molecular systems in particular
if these cannot be described by pure states. Strictly speaking, they are measured in
units of k but one usually drops the unit. We shall refer to them as multipole mo-
ments as long as no confusion can arise with the (spatial) multipole moments con-
structed from position vectors.4 More details will be discussed in the context of the
density matrix in Chap. 9 and Appendix C, Vol. 2. In addition, we shall introduce
there yet another kind of irreducible tensor operators to characterize anisotropic
state populations, the so called statistical tensor operators. Their expectation values
are the state multipoles which again can be related by simple numerical factors with
any other irreducible representation of tensor operators, in analogy to (F.38).
is important to keep in mind that the sign of Tk0
is opposite to that of Qk0
. When speaking
4 It
about shapes of charge distributions we have to note that oblate implies Q20
< 0 and T20
> 0,
while a prolate shape is encountered if Q20
> 0 and T20
< 0.
622 F Multipole Expansions and Multipole Moments
References
B OHR, A. and B. R. M OTTELSON: 1998. Nuclear Structure, vol. 1: Single-Particle Motion. Sin-
gapore: World Scientific, reprint from 1969 edn., 471 pages.
B RINK, D. M. and G. R. S ATCHLER: 1994. Angular Momentum. Oxford: Oxford University Press,
3rd edn., 182 pages.
FALKOFF, D. L. and G. E. U HLENBECK: 1950. On the directional correlation of successive nu-
clear radiations. Phys. Rev., 79, 323333.
FANO, U.: 1960. Real representations of coordinate rotations. J. Math. Phys., 1, 417423.
FANO, U. and J. H. M ACEK: 1973. Impact excitation and polarization of emitted light. Rev. Mod.
Phys., 45, 553573.
JACKSON, J. D.: 1999. Classical Electrodynamics. New York: John Wiley & Sons, 3rd edn., 808
pages.
M ACEK, J. and I. V. H ERTEL: 1974. Theory of electron-scattering from laser-excited atoms.
J. Phys. B, At. Mol. Phys., 7, 21732188.
Convolutions and Correlation Functions
G
In this appendix all distribution functions f (x) are normalized such that
+
f (x)dx = 1. (G.1)
The mathematical operation called convolution quantifies the overlap between two
mathematical functions f1 (x) and f2 (x) as one is reversed and shifted in respect of
the other. A convolution is defined as
(f1 f2 )(x) := f1 ( )f2 (x )d, (G.2)
Thus, both procedures are identical if one of the functions is even, f (x) = f (x),
which is the case in most of the physical applications.1 If f1 f2 one calls G(x) =
(f * f )(x) an autocorrelation function.
1 An alternative method to solve (G.3) is given by F (f1 * f2 ) = 2 F (f1 )F (f2 ) where F (f ) de-
scribes the F OURIER transform of f (see Appendix I).
S1
S2
x,
Fig. G.1 Convolution scheme of two rectangular profiles as manifested by two slits S1 and S2 of
unequal size. Their respective transmission profiles are shown as full black and red dashed lines.
As slit S2 is shifted over slit S1 in x, direction, the overlap (grey area) changes. The resulting
total transmission is shown as heavy red line (convolution)
the emission or absorption process and add to the broadening. The combination of
both effects is described by a convolution of the respective line profiles.
The physics involved in all these processes and examples will be discussed in
some detail in Sect. 5.1.1, and Chaps. 1, 2 in Vol. 2, and in other parts of these vol-
umes. In the present appendix we summarize a few useful mathematical expressions
which are needed to describe such measurements quantitatively. At the end one aims
of course at an intelligent de-convolution of the signal detected as a function of x,
i.e. one tries to extract the original profile f1 ( ) form the convoluted signal. This is
no trivial task unless the shapes of the two profiles involved are both well known.
Often one is content with just measuring the convolution and gleaning the width of
the profile f1 ( ) from known convolution formulas.
The angle brackets . . .
indicate averaging, the second equality giving one possible
recipe how to accomplish this averaging for the stationary case (e.g. for a CW light
beam or for continuous noise). Of course, integration has to be over a sufficiently
long time Tav so that all statistical short term fluctuations are smoothed out.
If EA and EB are square integrable functions, instead to average one has to inte-
grate over all times as in (G.3) and the 1/Tav factor becomes redundant. In normal-
ized form, the correlation function is called first-order degree of coherence:
EA (t)EB (t + )
g (1) () = . (G.5)
[|EA (t)|2
|EB (t)|2
]1/2
2 The further treatment is independent of whether EA and EB relate actually to two different fields
or represent the same field at different positions in space.
626 G Convolutions and Correlation Functions
Correlation functions are e.g. used in state-of-the-art ultrafast science for the
determination of shape and duration of short laser pulses as to be detailed in Chap. 1,
Vol. 2. In such a measurement, the electric field amplitude EA (t) (or the intensity
|EA (t)|2 ) of the pulse to be characterized is convoluted with a second light pulse
of well known profile EB (t) (or intensity |EB (t)|2 ). This is achieved by delaying
one pulse for a variable time in respect of the other. Both signals are multiplied
and integrated over a long time Tav . The signal is then recorded as a function of the
delay time and the result may finally be de-convoluted.
If EA (t) and EB (t) are identical (or just displaced in time) the first-order degree
of coherence (G.5) simply becomes the normalized autocorrelation function
E (t)E(t + )
E (t )E(t)
g (1) () = = = g (1) (). (G.6)
|E(t)|2
|E(t)|2
The important symmetry with respect zero follows directly from the definition (G.4)
for A = B. If the shape of a pulse is already well known, it allows one to determine
directly the pulse duration (see below for special examples).
Note that here the 1/Tav factor cancels out in (G.5) and (G.6): these definitions
hold for CW fields as well as for a single pulses or a pulse sequence. In practice, the
averaging may not be that trivial, especially if E(t) is not given in closed analytical
form. For CW fields the average . . .
can be performed over times at a given po-
sition in space or alternatively over the whole ensemble in space at a given fixed
time. One calls a system ergodic if the temporal average is equal to the average over
a representative ensemble which is generally true for reasonable physical systems.
Usually the temporal average is more convenient to perform while determining the
ensemble average (averaging over the whole space) would be a rather difficult task.
According to the ergodicity theorem,both averages are identical.
the convoluted FWHM thus being the geometric mean of both widths
x = 1.177w = 1.177 w12 + w22 = (
x1 )2 + (
x2 )2 . (G.13)
The hyperbolic secant squared is another, often used line profile, in particular for
laser pulse intensities. Normalized according to (G.1) it is
sech2 [(x x0 )/w] 1
fH (x; w, x0 ) = = (G.15)
2w 2w cosh [(x x0 )/w]
2
2
1 2
= ,
2w e(xx0 )/w + e(xx0 )/w
with a FWHM of
x = 2 ln( 2 1) w = 1.763w. (G.16)
628 G Convolutions and Correlation Functions
The convolution is a not completely trivial integral. For the autocorrelation function
one finds (properly normalized)
1 (x/w) cosh(x/w) sinh(x/w)
(fH * fH )(x; w) = (G.17)
w [sinh(x/w)]3
The first equality is exact, the sech4 gives an excellent approximation, and the sech2
is still a good approximation for small |x| 1.7; in the far wings, however, it is
somewhat too high.
Another profile often encountered is the L ORENTZ profile, describing e.g. the natu-
ral line profile or collision broadening (see Sects. 5.1.1 and 5.1.3, respectively). We
write the L ORENTZ profile
2 2 /4
fL (x; , x0 ) = (G.19)
(x x0 )2 + 2 /4
with a FWHM
x = , (G.20)
2 2 /4
(fL * fL )(x; 2 , x2 ) = (G.21)
(x (x1 x2 ))2 + 2 /4
with the FWHM of the convolution
x = = 1 + 2 .
Convolution of a L ORENTZ with a G AUSS profile is required e.g. for optical line
shapes of atoms or molecules in the gas phase if D OPPLER broadening and collision
broadening are on the same order of magnitude (or at low temperatures D OPPLER
and natural linewidth). This so called VOIGT profile
+
e2 /w
2 2
2 1
fV (x; , ) = (fG * fL )(x; , 0) = d, (G.22)
w 2 ( x)2 + 2 /4
Acronyms and Terminology 629
References
NIST-DLMF: 2013. Digital library of mathematical functions: 7.19 Voigt functions, NIST.
http://dlmf.nist.gov/7.19, accessed: 9 Jan 2014.
SWP 5.5: 2005. Scientific work place, Poulsbo, WA 98370-7370, USA: MacKichan Software,
Inc. http://www.mackichan.com/, accessed: 9 Jan 2014.
W EISSTEIN, E. W.: 2011. Convolution, Wolfram Research, Inc., Champaign, IL, USA. http://
mathworld.wolfram.com/Convolution.html, accessed: 9 Jan 2014.
W IKIPEDIA CONTRIBUTORS: 2014. Voigt profile, Wikipedia, The Free Encyclopedia. http://en.
wikipedia.org/wiki/Voigt_profile, accessed: 9 Jan 2014.
Vector Potential, Dipole Approximation,
Oscillator Strength H
In view of its clearness and compactness we use in this book apart from a few ex-
ceptions in general the dipole length approximation for describing the interaction
of electromagnetic waves with atoms and molecules. It is thus appropriate to justify
this approach in some detail. In addition, we shall provide the basis for some useful
concepts related to the dipole approximation.
p 2field 1
=
H + V (r) = p + eA)2 + V (r)
( (H.1)
2me 2me
p2 pA
e e2 2
= + V (r) + + A (H.2)
2me me 2me
where the so called C OULOMB gauge of the field has been used:
A = 0. (H.3)
For simplicity we have restricted the discussion to a single active electron. One
easily generalizes (H.2) to multi-electron systems by replacing momentum p and
position vector r with p i and r i , respectively, for each electron i, and summing
over all electrons.
The vector potential A relates to the electric and magnetic field vectors
E(r, t) = A(r, t) + Vext and B(r, t) = A(r, t) (H.4)
t
with Vext (r, t) describing an additional, external potential. If such a field is present,
one has to add Vext to V (r) in (H.1) and (H.2). Not included in this description are
of course all interactions related to the electron spin which follow from the D IRAC
equation, including spin-orbit interaction, as briefly discussed in Chap. 6.
In vacuum, we write the vector potential, electric field, and magnetic field for a
plane electromagnetic wave as:
A0 i(krt)
A(r, t) = ee + e ei(krt) = A0 e cos(kr t) (H.5)
2
iE0 i(krt)
E(r, t) = ee e ei(krt) = E0 e sin(kr t) (H.6)
2
iB0 k i(krt) k
B(r, t) = ee e ei(krt) = B0 e sin(kr t). (H.7)
2 k k
The relations among the amplitudes are derived from (H.4)
E0 = A0 , B0 = kA0 = A0 , and B0 = E0 /c. (H.8)
c
H.1.2 Intensity
In principle, the amplitude E0 (r, t) (and A0 and B0 ) as well as the averaged intensity
I (r, t) may still depend slowly on time t and position r. This is called SVE.1 More
about cycle averaged intensities is discussed in Sect. 1.3.1, Vol. 2. The electric field
is measured in units [E0 ] = V m1 , the intensity in [I ] = W m2 .
Note that all field quantities used here represent quasi-monochromatic waves:
their bandwidth is assumed to be much narrower than any atomic or molecular ab-
sorption line studied. The transition to a continuous spectrum is made by replacing
I I()d, where the spectral intensity distribution I() (or the spectral energy
density u() = I()/c) refer to the unit
of angular frequency interval. Finally, one
has to integrate over all frequencies: . . . d.
For a static, homogeneous external magnetic field B one derives the vector potential
by inverting (H.4):
1 1
A = r B = B r. (H.12)
2 2
This may be verified by inserting this expression into (H.4).2
For such a constant B field the Hamilton operator (H.2) becomes
=
H
p2
+ V (r) +
e
LB +
e 2 A2
, (H.13)
2me 2me 2me
e 2 A2 e2 2 2
= r B sin , (H.14)
2me 8me
where is the angle between the external B field and the position vector r of the
electron in the atom. We estimate for, say, 30 T and r = a0 a maximum value of ca.
5 108 eV. Thus, in spectroscopy this term only plays a role if extreme accuracy
is asked for or very high magnetic fields and very large orbital radii are involved
(i.e. high lying RYDBERG states). On the other hand, it is this very term which is
responsible for all diamagnetism of matter.
For the following considerations we derive an important relation between the matrix
elements of momentum and position. For electrons i and j the standard commuta-
tion rules for canonically conjugated observables (momentum and position) hold:
[xi , p
yj ] = 0 and [xi , p
xj ] = iij , etc. (H.15)
we can write
] = i p
[xi , H xi and [r, H ] = i p, (H.17)
me me
where r = r i and p= p i . Thus, for the matrix elements of
p between eigen-
states |a
and |b
of the Hamiltonian we derive
me ]|a
= me (Wa Wb )b|r|a
.
b|
p |a
= b|[r, H (H.18)
i i
With (Wa Wb )/ = Wba / = ba the sought-after relation between the matrix
elements of
p and r is:
b|
p |a
= ime ba b|r|a
. (H.19)
Before actually evaluating the matrix elements of the interaction in the Hamiltonian
(H.2) in detail, we have a second look at the term proportional to A2 , now for an
atom or molecule in an electromagnetic wave. Clearly, this term is a time dependent
additional energy in the Hamiltonian. By averaging A2 over one period we obtain
with (H.11) an estimate for its dependence on intensity I :
e2 A2
e2 I
Up = = . (H.20)
2me 20 cme 2
H.1 Interaction of the Field of an Electromagnetic Wave with an Electron 635
This is identical to the expression (8.139) discussed in Sect. 8.5.1 for the pondero-
motive potential Up . There, a completely classical picture of an electron oscillating
in the electric field of the wave is used to derive it.
In respect of standard laser spectroscopy, the ponderomotive potential is usually
negligible. As described in Sect. 8.5.1, however, it leads to very interesting phe-
nomena if the objects studied are exposed to very intense electromagnetic radiation
which can be generated without difficulties by todays short pulse lasers.
eikr ieikr
D=
pe= e (H.22)
me me
(with the dimension of a force MLT2 ). For electromagnetic waves in the IR, VIS,
UV, and VUV spectral range the wavelength is usually very large compared to the
objects studied. We thus can assume k r 1 and expand the exponential function
in (H.22):
1
D= (1 + ik r + )
p e. (H.24)
me
In the electric dipole approximation (short: dipole approximation) only the first
term is taken into account and one speaks of E1 transitions. Thus, the dipole transi-
tion matrix elements of D between two eigenstates |a
and |b
is:
1
Dab = a|
p |b
e = i a||b
e = ia|r|b
e. (H.25)
ba me ba me
For the last equality we have made use of (H.19). Strictly mathematical, both forms
of
Dab are completely equivalent if the wave functions used are exact solutions of
636 H Vector Potential, Dipole Approximation, Oscillator Strength
the unperturbed Hamiltonian. However, since we usually know the wave functions
only approximatively (except for some special cases such as the H atom) both for-
mulas (H.26) and (H.27) lead to slightly different results. Both variants are used in
the literature: they are called dipole velocity approximation (since
v = p /me ), and
the dipole length approximation, respectively.
The matrix elements of the interaction potential (H.21) are thus
i e 2I ()
ab (t) = a|U
U (r, t)|b
= a||b
eeit + e e+it (H.26)
2 me c0
N N
Dab = i a| i |b
e = ia| r i |b
e. (H.29)
ba me
i=1 i=1
The first and second equality refer to the dipole velocity and dipole length form
of the transition operator, respectively. They are identical if (and only if) the wave
functions are exact. The sum has to be carried out in principle over all electrons i,
with N being the total number of electrons in the system.
Several, slightly differing, quantities are used in the literature to characterize the
overall strength of a dipole transition between levels a and b with angular mo-
mentum quantum numbers ja ma and jb mb , and degeneracies ga = 2ja + 1 and
H.2 Line Strength and Oscillator Strength 637
gb = 2jb + 1, respectively. In principle, j may refer to the orbital (L) or to the to-
tal angular momentum (J ), whichever is appropriate. Note that we abbreviate the
quantum numbers a ja ma := a, but we also use a and b for designating the respec-
tive energy levels as long as no confusion can arise.
We essentially3 follow C ONDON and S HORTLEY (1951) and define the line
strength symmetric in respect of initial and final energy levels and as a sum over
all polarizations:
S(jb ja ) S(ja jb ) := |r ba |2 b jb mb |r|a ja ma
2
mb ma mb ma
(H.30)
= b|x|a
2 + b|y|a
2 + b|z|a
2 .
mb ma
Using the W IGNER -E CKART theorem in the form (4.82) and the orthogonality re-
lations of the 3j symbols (B.42) we obtain the compact relations:4
2
S(jb ja ) = (2jb + 1)b |r|a
jb C1 ja
2
2 (H.32)
= (2ja + 1)a |r|b
ja C1 jb
2 S(ja jb ).
where the equality sign holds in so far as the radial matrix element does not depend
on the individual fine structure levels J but only on the orbital quantum numbers L
and n which is a good approximation for light atoms.
3 The line strength S CS (jb ja ) used by C ONDON and S HORTLEY (1951) is related by S CS (jb ja ) =
eS(jb ja ) to the quantity used here which leads to more compact formulas for the A and B
(opt)
coefficients and for the oscillator strength fba .
4 Notethat a factor 3 arising here from the sum over all polarizations q is compensated by a factor
1/(2 1 + 1) = 1/3 from the orthogonality relations of the C LEBSCH -G ORDAN coefficients when
summing over ma and mb .
638 H Vector Potential, Dipole Approximation, Oscillator Strength
Here Dba is the dipole transition matrix element (H.29). In the second equality we
use explicitly the dipole length form for a single electron system, with r ba as de-
fined by (4.79). The oscillator strength is thus proportional to the transition proba-
bility (4.65). And since it is also proportional to the absorbed (or emitted) photon
energy Wb Wa = Wba = ba , its value is positive for absorption and negative for
emission.
For the particular case of linear polarization with elin z this gives a simple ex-
pression for the oscillator strength:
(opt) Wba zba 2 me ba
fjb mb ja ma =2 =2 |zba |2 . (H.35)
E h a0
It may be further evaluated with (4.75) and the W IGNER -E CKART theorem in the
form (4.82):
2
|zba |2 = b jb mb |r0 |a ja ma
2 (H.36)
2 ja 1 jb
= (2jb + 1)b |r|a
jb C1 ja
2 .
ma 0 mb
2 Wba 2 2
1
= |aq |2 b |r|a
jb mb |C1q |ja ma
. (H.38)
ga E h m m
a b q=1
H.2 Line Strength and Oscillator Strength 639
Using (4.82), the orthogonality relation (B.42) for the 3j symbols, and q |aq |2 = 1
(unit vector e) the sums can be carried out:
2
Wba gb r
1
1
(opt)
fjb ja =2 | |
jb C1 ja
2 |aq |2 (H.39)
E h ga
b a
a0 3
q=1
2
(opt) 2 Wba gb r
fjb ja = b | | a jb C1 ja
.
2
(H.40)
3 E h ga a0
where we have exploited that here q = 0, and thus only contributions from ma = mb
arise. For an initial singly degenerate ns state (H.42) is reduced to (H.35). Equivalent
expressions are derived from (H.41) with (H.30) and the dipole moment D = er:
2me ba 2
b jb mb |r|a ja ma
(opt)
fjb ja = (H.43)
3ga m m
2
a b
2
2 Wba r
8 2 me Wba
= j m
b b b | | a a a =
j m
|D ba |2 .
3ga Eh m m a0 3ga h2 e2 m m
a b a b
(H.44)
Note the factor 3 in the denominator in comparison to (H.42), compensating for the
equal contributions from the components of |r|2 = |x|2 + |y|2 + |z|2 .
We sum the oscillator strength according to (H.35) over all existing final states |b
=
|b jb mb
and rewrite the expression suitably:
640 H Vector Potential, Dipole Approximation, Oscillator Strength
(opt)
2me ba 2me ba
fjb mb ja ma = |zba |2 = a|z|b
b|z|a
(H.45)
b b b
1
= ime ba a|z|b
b|z|a
ime ab a|z|b
b|z|a
.
i
b
The oscillator strength is often used in atomic and molecular physics to characterize
dipole transitions. It allows one to compare the strength of different transitions in
(opt)
different atoms or molecules. For systems with one active electron fjb ja 1 strictly
holds, the classical reference being an oscillating electron which has an oscillator
strength = 1.
For systems with Ne active electrons one has to replace in (H.45) the single
e (i)
electron coordinate z by N i=1 z and obtains the sum rule:
(opt)
fjb ja = Ne . (H.48)
b
Finally, we point out that the summation just discussed must include the ion-
ization continuum if the basis is to be complete. In this context, continuum states
are usually normalized in the energy scale (see Appendix J). Correspondingly one
defines the optical oscillator strength density (OOSD), df(opt) /d, with the dimen-
sion Enrg1 , where is the energy of the emitted electron in the continuum. Thus,
(H.48) includes summation over discrete states and integration for energies beyond
the ionization potential WI :
discrete (opt)
dfja
(opt) (opt)
fjb ja fjb ja + d. (H.49)
WI d
b b
References
C ONDON, E. U. and G. S HORTLEY: 1951. The Theory of Atomic Spectra. Cambridge, England:
Cambridge University Press, 441 pages.
ISO 21348: 2007. Space environment (natural and artificial) process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
FOURIER Transforms and Spectral Distributions
of Light I
Various, slightly different notations can be found in the literature. We use the stan-
dard notations of modern physics with the F OURIER transform1
)
X() = F X(t) = X(t)eit dt (I.2)
of the time dependent function X(t). Conversely, X(t) is recovered by the inverse
F OURIER transform
1
)
X(t) = F 1 X() = )
X()e it
d. (I.3)
2
1 In
other notations the factor 1/2 is applied symmetrically as 1/2 , or completely avoided by
using frequency instead of angular frequency = 2.
and correspondingly
1 ) ()eit dt.
X (t) = X (I.5)
2
By rewriting (I.1) as
1
X(t) = X( ) e i(t )
d d
2
one identifies the expression in square brackets as the D IRAC delta function:
1
(t ) = F 1 ei = ei(t ) d (I.6)
2
1 ic t 1
and equivalently (c ) = F e = ei(c )t dt. (I.7)
2 2
Thus, the carrier frequency just shifts the F OURIER transform F[X(t) exp(ic t)]
towards positive frequencies in respect of F[X(t)]. This formally also resolves the
problem one might envisage with negative frequencies in the spectral distribution
)
X(): since for any physically relevant distribution with a FWHM= one expects
c , there will be no contributions from negative frequencies in (I.2). An ana-
logue relation applies to a shift in time.
Another set of useful relations concerns F OURIER transforms of convolutions
(see Appendix G). The proof is similar to that for P LANCHERELs theorem.
I.1 Short Summary on FOURIER Transforms 645
it
F (X Y )(t) = X t Y t t dt e dt
it
= X t e dt Y t t ei(tt ) dt
= X t eit dt Y t eit dt .
1 i t 1 i t it
= )
X e d )
Y e d e dt
2 2
1 i t
= )
X e d
2
1
) ei( )t d eit dt
Y
2
1 1
= ) Y
X ) d d ei( )t dt
2 2
1
= X ) Y ) d d
2
1
= ) Y
X ) d .
2
2
)X() = F X ( )X( + t)d . (I.17)
c t k c r c t .
2 Where necessary, we shall simply reintroduce the wave vector by replacing
Similarly, we keep the r dependence of the envelope in the back of our mind and set E (t)
E (r, t) when required.
I.2 How Electromagnetic Fields are Written 647
3 Even )+ ()) = E
though (E + (t)) = E (t), note that (E ) (), since
+ + it
) () =
E E + (t)eit dt = E (t) e dt
= ) ().
E (t)eit dt = E
648 I FOURIER Transforms and Spectral Distributions of Light
)
Note that in this terminology E0 (t) is a real quantity. E() will also be real if and
only if E0 (t) is symmetric around its maximum, while E )+ () is usually complex.
However, with these definitions E )+ () typically has a pronounced resonance at
> 0 so that contributions from negative frequencies to the inverse FT (I.22) can
safely be neglected. In other words: artificial truncations of the limits of integrals
are not needed, since anyhow only positive frequencies contribute when working
with E)+ (). These relationships will be illustrated below for specific examples.
If one inserts E (t) according to (I.22) into expression (I.19) for the cycle averaged
intensity and integrates over all times, one recovers P LANCHERELs theorem (I.8)
0 c 1 + 2
I (t)dt =
E () d = I()d = F, (I.25)
2 2
leading to the fluence F (i.e. the total pulse energy Wtot per unit area). With the
second equality we have defined the spectral intensity distribution:
0 c )+ 2 0 c ) 2
I() = E () = E( c ) . (I.26)
4 4
The assumption that this indeed describes the spectrum of the radiation is by
no means as trivial as commonly assumed. Note, that this expression is not the
F OURIER transform of the intensity I (t)!
One may, however, rationalize this definition by remembering how optical (or
other) spectra of electromagnetic radiation are measured: Assume e.g. an interfero-
metric measurement (FABRY-P ROT interferometer, diffraction grating etc.), where
the signal S is generated by superposition of two (or many) amplitudes. A typical
interference pattern emerges from, say two wave fields E + (t) and E + (t + ), which
have traversed different optical path length c (corresponding to a delay time ). The
signal is then
S E (t) + E (t + ) E + (t) + E + (t + ) (I.27)
= 2I0 + 2 Re E (t + )E + (t) ,
and only the interference term [in square brackets] is relevant for the measurement
of the spectrum. If such a measurement is made with short pulses, one obviously
integrates over all times (or sufficiently long times), thus an autocorrelation function
of the field is recorded. To obtain the spectrum one has to F OURIER transform this
signal which according to (I.17) leads to
1 )+ 2
F E (t) * E + (t) = E () I ().
2
I.4 Examples 649
If a prism spectrograph is used a similar argument can be based on the wave optical
interpretation of the index of refraction.
Hence, the definition of the spectrum according (I.26) is justified. Proper normal-
ization is provided by the P LANCHEREL theorem in the form (I.25), which simply
states energy conservation: the fluence F is independent of whether we integrate in-
tensity over all times or spectral intensity distribution over all angular frequencies.
The units are [F ] = J m2 , [I (t)] = W m2 , and [I()] = J s m2 , respectively.
It is important to note that the above use of P LANCHERELs theorem is only pos-
sible for square integrable functions, i.e. for pulses of electromagnetic radiation with
finite duration. If we want to describe CW light sources the integration over all times
has to be replaced by an average over a sufficiently long time Tav . P LANCHERELs
theorem (I.25) must then be replaced by
Tav /2
1 0 c Tav /2 + 2
I (t)dt = E (t) dt = I()d = I, (I.28)
Tav Tav /2 2Tav Tav /2
where I() is now the spectral intensity of the ensemble, measured in units [I()] =
W s m2 , while I is the average intensity with [I ] = W m2 . More about station-
ary, quasi-monochromatic or chaotic light sources, their spectra and their coherence
properties will be found in Sect. 2.1, Vol. 2.
I.4 Examples
In the following we shall present the most commonly used temporal and spectral
profiles of light pulses. The F OURIER transforms collected here can be derived
by analytic integration, where necessary with the help of a suitable computer pro-
gramme (e.g. SWP 5.5, 2005). They are also found in standard textbooks or at
the web (e.g. W EISSTEIN, 2012; W IKIPEDIA CONTRIBUTORS, 2014). We write the
field envelope as
E0 (t) = E0 h t/|t0 | with h(0) = 1 and (I.29)
0 c 2
I0 = E (I.30)
2 0
where E0 is the maximum field amplitude, |t0 | a characteristic width, and I0 the
cycle averaged maximum intensity.
We point out that the spectral intensity distribution as defined in (I.26) refers to a
single pulse. It has the dimension [I()] = J s m2 . Integrated over all frequencies
it gives the fluence F per pulse, i.e. the total pulse energy per unit area:
F= I (t)dt = I0 h2 (t/t0 )dt. (I.31)
are characterized by individual time constants i and their individual spectra Ii ().
One then has to average these spectra and to renormalize in respect of the average
duration c of the pulses. One obtains an overall spectrum:
1
I() = Ii () = w(i )Ii ()di . (I.32)
c 0
This spectrum has now the dimension W s m2 and the integration over all frequen-
cies gives the average, stationary intensity I of the beam.
Gaussian pulse shapes are probably the most popular ones, mainly due to their con-
venient mathematics. Starting from (I.18) we write the field envelope
2
t
E0 (t) = E0 h(t) = E0 exp 2 (I.33)
G
with a FWHM= 2 ln 2G = 1.665G . The corresponding cycle averaged intensity
(I.19) is now
t2
I (t) = I0 exp 2 2
G
with a FWHM
t1/2 = 2 ln 2G = 1.177G . (I.34)
1 2 2
)
E() = G exp G and (I.36)
4
E ) c )ei = G exp 1 G2 ( c )2 i .
)+ () = E( (I.37)
4
Thus, with (I.26) the (experimentally measurable) intensity spectrum becomes
2
G2 I0 c 2
I() = I0 exp G ( c )2 = 2 exp , (I.38)
2 2 G G
with G = 2/G and a FWHM
1/2 = 2 ln 2G = 2 2 ln 2/G = 2.355/G . (I.39)
I.4 Examples 651
One easily verifies that I()d = F as in (I.35). In frequency space
1/2 =
2 ln 2
t1/2
1/2 = = 0.441. (I.40)
Another, quite popular distribution function for describing the time dependence of
short light pulses is the hyperbolic secant
t E0 2E0
E0 (t) = E0 sech = = (I.41)
s cosh(t/s ) et/s + et/s
with a FWHM of 2 ln(2 3)s = 2.634s . The corresponding intensity distribu-
tion is
2
2
I (t) = I0 sech (t/s ) = I0 t/
2
, (I.42)
e s + et/s
with a FWHM
t1/2 = 2 ln(1 + 2)s = 1.763s . (I.43)
) s 2E0 s
E() = E0 s sech = /2 or (I.45)
2 e s + es /2
)+ () = E0 s eic sech s ( c ) .
E (I.46)
2
The spectral intensity profile for a pulse with the carrier frequency c is
0 c 2 s
I () = 2 2
E0 s sech ( c ) . (I.47)
4 2
In angular frequency space it has a FWHM
1/2 = 4 arcsech(1/ 2)/(s ) = 1.122/s (I.48)
The intensity distribution has the same temporal shape with I0 = 0 cE02 /2. It is
convenient to directly evaluate the F OURIER transform (I.24) of Ei+ (t):
t0 +r
)+ () = E0
E exp i(c t c ti i t) dt
i
t0
exp[i(c )i ] 1
= E0 r exp i(ti + i ) (I.51)
i(c )i
)i () = E0 i exp[iti ] 1 exp[ii ] .
E (I.52)
ii
The spectral intensity distribution (I.26) is thus
0 c )+ 2 I0 i2 i ( c )
Ii () = Ei () = sinc2 , (I.53)
4 2 2
with the sinc x = (sin x)/x function. The overall fluence is
F= I (t)dt = I()d = I0 i .
Another important case is a rectangular spectrum. Such a situation may arise, e.g.
if one spectrally filters a broad band light CW source with a sharp band-pass filter
(e.g. in stellar interferometry). The spectrum would then be
I.4 Examples 653
*
I /
for c 12
< < c + 12
I() = (I.55)
0 else,
where I is the average stationary intensity of the beam. The integral (I.3) can easily
be carried out in this case and one obtains the inverse F OURIER transform:
I ic t
t sin x
F 1 I() = e sinc with sinc(x) = . (I.56)
2 2 x
This formula turns out to be useful e.g. for understanding stellar interference spec-
troscopy (see Sect. 2.1.4, Vol. 2).
*
0 if x < 0
h(x) = (I.57)
exp(x) if x 0
is used to describes an exponentially decaying field (oscillating at a carrier fre-
quency c ) and its cycle averaged intensity, respectively:
t
E(t) = E0 h and
2e
t
I (t) = I0 h with FWHM
t1/2 = e ln 2 = 0.693e . (I.58)
e
) 2e E0
E() = , and
1 + 2ie
)+ () = ei 2e E0
E .
1 + 2ie ( c )
1 I0 e2 I0 2 /4
I() = = , (I.59)
1 + [2e ( c )]2 2 2 /4 + ( c )2
with a FWHM
1/2 = = 1/e . (I.60)
654 I FOURIER Transforms and Spectral Distributions of Light
In frequency space this gives
1/2 = 1/(2e ) = 0.159/e , so that the time-
bandwidth product becomes
) 4ee E0
E() = , or alternatively (I.66)
1 + (2ee )2
)+ () = E0 ei
E , (I.67)
2 /4 + ( c )2
with an FWHM = 1/ee . The intensity spectrum is given by
2
I0
I () = (I.68)
2 2 /4 + ( c )2
with a FWHM
1/2 = 2 1 = 0.644 = 0.643/ee , (I.69)
We point out that (I.68) is not a L ORENTZ profile, and may, e.g. not easily be convo-
luted. We emphasize that in any case this is not a typical profile for characterizing a
real light pulse: it has a cusp at time zero and extremely wide tails, both in the time
and in the frequency domain.
I.5 Fourier Transform in Three Dimensions 655
1
L(, )= (I.71)
2 /4 + ( c )2
2
with = 1. The inverse FT is given by (I.63) as
L(, )d
1 1 |t|/2+ic t
F 1 L(, ) =
L(, )eit d = e (I.72)
2 2
with = 1/ee . We have set here = 0.
We finally can now supplement the proof for the additivity of linewidths when
convoluting two L ORENTZ profiles, as introduced by (G.21). According to (I.12) the
inverse FT of a convolution of two L ORENTZ profiles is proportional to the products
of their inverse FT:
F 1 L 1 (, 1 ) L 1 (, 2 ) (I.73)
1 1 |t|/2i1 t 1 1 |t|/2+i2 t 1 (1 +2 )|t|/2+i(2 1 )t
= 2 e e = e .
2 2 2
Now we just transform this back, compare with (I.72), and obtain:
1 (1 +2 )|t|/2+i(2 1 )t
L1 (x, 1 ) L1 (x, 2 ) = F e (I.74)
2
2 (1 + 2 )2 /4
= .
(1 + 2 ) (x (2 1 ))2 + (1 + 2 )2 /4
Thus, the convolution of one L ORENTZ profile with another one leads again to a
L ORENTZ profile. The FWHM of the two profiles are just added:
= 1 + 2 . (I.75)
(x x )
x x = 2 . (I.81)
x sin
For problems with some kind of spherical symmetry (e.g. in atomic physics)
it is useful to rewrite (I.76) in spherical polar coordinates, with standard position
coordinates (r, r , r ) and coordinates in wave vector space (k, k , k ):
2
) k , k ) =
X(k, X(r, r , r )eikr r 2 sin r drdr dr . (I.82)
0 0 0
X(r) = X(r, r , r ) = X m (r)Y m (r , r ), where (I.83)
=0 m=
2
X m (r) = X(r, r , r )Y m (r , r ) sin r dr dr . (I.84)
0 0
By expanding also the plane wave (I.82) according to (J.13) one obtains the 3D
F OURIER transform in k space
+
) = X(k,
X(k) ) k , k ) = 4 (i) X m (k)Y m (k , k ), (I.85)
=0 m=
Acronyms and Terminology 657
References
BADDOUR, N.: 2009. Operational and convolution properties of two-dimensional Fourier trans-
forms in polar coordinates. J. Opt. Soc. Am. A, 26, 17671777.
BADDOUR, N.: 2010. Operational and convolution properties of three-dimensional Fourier trans-
forms in spherical polar coordinates. J. Opt. Soc. Am. A, 27, 21442155.
SWP 5.5: 2005. Scientific work place, Poulsbo, WA 98370-7370, USA: MacKichan Software,
Inc. http://www.mackichan.com/, accessed: 9 Jan 2014.
W EISSTEIN, E. W.: 2012. Fourier transform, Wolfram Research, Inc., Champaign, IL, USA.
http://mathworld.wolfram.com/FourierTransform.html, accessed: 9 Jan 2014.
W IKIPEDIA CONTRIBUTORS: 2014. Fourier transform, Wikipedia, The Free Encyclopedia.
http://en.wikipedia.org/wiki/Fourier_transform, accessed: 9 Jan 2014.
Continuum
J
In the following we use the radial functions in the form u(r) = rR(r). In the con-
tinuous spectrum, following B ETHE and S ALPETER (1957) Eq. (4.11ff), they are
normalized in T scale by demanding
T +
T
druT (r) uT (r)dT = 1. (J.2)
0 T
T
Here T (k) is any function of the electron wavenumber k in the continuum, possibly
k itself. In the theoretical treatment of photoionization normalization in scale is
frequently used, with being the energy of the electron in the continuum. The in-
tegration over dT needs only to be over a small interval 2
T around T , since all
other contributions cancel.1
With such normalized continuum functions one may represent any wave function
including the continuum as partial wave expansion (for simplicity written here only
for one electron systems):
1 Summation of (J.1) over all n states in the discrete spectrum would be the equivalent to this second
integration in (J.2). Since only one state contributes, the result is also =1.
(r, , ) (J.3)
+ ,
1
= Y m (, ) an m un (r) + dT (k)a m (T )uT (r) .
r k=0
m n= +1
we obtain
(for a system with reduced mass M)
d me dk M 1 M 1
= Eh a02 k and = = . (J.6)
dk M d me Eh a02 k me a0 (2Eh )1/2
d/dk = k. (J.7)
To determine the dimension of uk (r), we write the units in the expression under
the integral in (J.2) [udrdk] = 1, or as dimensional equation u2 L1 L1 = 1 so that
u = 1, i.e. uk (r) in k scale normalization is dimensionless. For clarity it might be
useful to explicitly write out the a.u., i.e. to replace r r/a0 and Z/k Z/(ka0 )
in (J.9), so that the expression becomes independent of the units used.
For normalization in scale we use (J.4)(J.7) and obtain in a.u.:
2 dk Z
u (r) = sin kr /2 + ln(2kr) + +
d k
2 Z
= sin kr /2 + ln(2kr) + + or explicitly (J.10)
k k
2 me Eh /2 1/4 Z
= sin kr /2 + ln(2kr) + + .
Eh a0 M ka0
(J.11)
The latter expression is independent of the units used. The prefactor (Eh a0 )1/2
is again obtained from a dimensional analysis: normalized in scale from (J.2)
we have [u2 drd] = 1, so that u2 = Enrg1 L1 and u = Enrg1/2 L1/2 . Note that
for electron scattering the mass factor me /M 1 disappears. Different authors use
slightly different notations and energy units (often without mentioning which).2
(k) (r) = eikr = 4 i j (kr) Y m (k , k )Y m (r , r ), (J.13)
=0 m=
Here, = 2/Eh , i.e. energies are measured in RYDBERG units = Eh /2, lengths in a0 .
662 J Continuum
with the polar coordinates (r, r , r ) representing r, while (k, k , k ) gives the mag-
nitude and direction of the wave vector k, and j (kr) = u (kr)/(kr) are spherical
B ESSEL functions. They are solutions of the radial S CHRDINGER equation (2.110)
for vanishing potential:
2
1 d2 u k ( + 1)
+ u (r) = 0. (J.14)
2 dr 2 2 2r 2
As one easily verifies, the most simple ones are:
sin x sin x cos x
j0 (x) = sinc x = and j1 (x) = . (J.15)
x x2 x
All others may in principle be derived from the recursion formula:
2 + 1
j +1 (x) = j (x) j 1 (x). (J.16)
x
Asymptotically the following relations hold:
" #
x /[(2 + 1)(2 1)(2 3) ] for x
j (x) = . (J.17)
sin(x /2)/x for x
With the addition theorem (C.22) and the angle between k and r one may also
write (J.13) as:
(k) (r) = eikr = (2 + 1)i j (kr)P (cos ). (J.18)
=0
The probability density | (k) |2 of the plane waves (J.13) or (J.18) is given per
wavenumber to the third power (dimension L3 ) and per volume (L3 ), so they are
overall dimensionless. They may be normalized in k scale (see e.g. OVCHINNIKOV
et al., 2004):
1
)(k) (r) =
eikr = r|k
. (J.19)
(2)3/2
To verify this normalization we first integrate over position space using (I.78):
k 2 |k 1
k = )(k) (r)
)(k) (r)d3 r a03 ka0 k a0 .
2 1
of states. Finally, with (I.80), integration over k spaces confirms the correct normal-
ization of (J.19):
3
k k k d k = a03 ka0 k a0 d3 k 1.
holds, and the 3D plane wave (J.19) normalized in energy scale becomes
ka0 ikr
) (r) = (2)
(k) 3/2
e . (J.22)
a03 Eh
)(k) (r)|2 is
Its dimension is now Enrg1/2 L3/2 so that the probability density |
correctly given per energy and volume.
References
B ETHE, H. A. and E. E. S ALPETER: 1957. Quantum Mechanis of One- and Two-Electron Atoms.
Berlin, Gttingen, Heidelberg: Springer Verlag, 369 pages.
B URKE, P.: 2006. Electron-atom, electron-ion and electron-molecule collisions. In: G. W. F.
D RAKE, ed., Handbook of Atomic, Molecular and Optical Physics, 705729. Heidelberg, New
York: Springer.
C OOPER, J. W.: 1962. Photoionization from outer atomic subshells. A model study. Phys. Rev.,
128, 681693.
OVCHINNIKOV, S. Y., G. N. O GURTSOV, J. H. M ACEK and Y. S. G ORDEEV: 2004. Dynamics of
ionization in atomic collisions. Phys. Rep., 389, 119159.
Index of Volume 1
theorem for F OURIER transform, 645 Density of states, 2124, 238, 240
VOIGT profile, 628, 629 particle in a box, 105, 106
C OOPER minimum, 260, 529 Detailed balance
Coordinate rotation, 575, 605611 principle of, 184
exercise, 207 Detuning, 229
Coordinate system Diamagnetism, 396398
atomic vs. photon, 174 Dielectric function, 57
cartesic and polar, 108 Diffraction
electron vs. nucleus, 295 D EBYE-S CHERRER, 58, 59
Core electron, 148, 151 experiment, 60
Core potential, 149 He scattering, 60
alkali atoms, 165 image, 42
Correlation function LEED, 59
1st order, 625 Dimensional analysis, 119, 192, 554, 556, 661
Correspondence principle, 127 Dipol vector, 183
C OULOMB gauge, 631 Dipole approximation, 189, 631640
C OULOMB integral, 356, 359, 365, 506 electric, 635
C OULOMB law, 50, 59 magnetic, 250254
C OULOMB potential, 119, 121, 123 Dipole excitation, linear combination of states,
H atom, 121 217225
screened, 148, 149 Dipole matrix element, 193
C OULOMB wave, 262 length approximation, 193, 636
Coupling
velocity approximation, 193, 636
jj , 301, 354, 360
Dipole operator, 594
LS, 301, 307, 354, 360, 361
magnetic field, 386
LS vs. jj , 512514
multi-electron system, 264, 363
RUSSEL -S AUNDERS, 301, 354, 360, 361,
Dipole oscillator, classical, 182
508
Dipole radiation
break down, 362
angular characteristic, 203212
spin-orbit, see Spin-orbit
Dipole transition
Crystal lattice, 40
C URIE constant, 396 amplitude, 193
C URIEs law, 396 E1 transitions in the H atom, 201203
Cyclotron frequency, 54, 332 in He, 362365
electron, 79 length approximation, 636
Cylindrical capacitor, 52 matrix element, 190, 635
operator, 190, 193, 263
D selection rules for E1 transitions,
Damping constant, 229 196203
DARWIN term, 305 selection rules for E1-transition, 202
D E B ROGLIE, 57, 89 velocity approximation, 636
wavelength, 5759 D IRAC delta function, 644
Decay constant, 15 D IRAC equation, 79, 93, 296, 303, 333
Decay rate, 15 Direct product, 578
Degeneracy, 21, 22 Dispersion, 421
, 123 anomalous, 422
m, 111 close to resonance, 232, 233
, removal of, 137146 normal, 420, 422
m, removal of, 130, 131 Dispersion relation, 131134
Degree of coherence matter waves, 94
1st order, 625 Displaced terms
Degrees of freedom, 19 alkaline earth term schemes, 372
Density distribution, 125 C atom, 515
Density functional theory, 510512 Ne atom, 518
668 Index of Volume 1
319, 342, 347, 349, 351, 371, 373, Nuclear quadrupole moment, 447, 449, 616
431, 514, 522, 525, 529, 551, 629 oblate or prolate, 478
NMR spectroscopy, 487491 Nuclear radius, 29, 121
apparatus, 488 liquid drop model, 476
CW spectrum of ethanol, 488 Nuclear spin, 449
magic angle spinning (MAS), 490 eigenvalue equations, 449
occupation probability of levels, 489 Nuclear spin resonance, see NMR
N OBEL prize in chemistry Nucleons, 48
Richard R. E RNST (1991), 490
KOHN and P OPLE (1998), 510 O
F ENN, TANAKA, W THRICH (2002), 490 O atom, G ROTRIAN diagram, 516
N OBEL prize in physics Oblate, 477, 478, 619
Wilhelm C. RNTGEN (1901), 530 nuclear shape, 478
L ORENTZ and Z EEMAN (1902), 377 Observable, 9799, 118
Joseph J. T HOMSON (1906), 50 commuting, 101
Albert A. M ICHELSON (1907), 278 non-commuting, 101
Max K. E. L. P LANCK (1918), 31 simultaneous measurement, 100, 101
Johannes S TARK (1919), 399 One electron cyclotron oscillator, 334
Albert E INSTEIN (1921), 27, 31 One particle problem, 117134
Niels H. D. B OHR (1922), 64 One sided exponential distribution, 653
C ORNELL, K ETTERLE, W IEMAN (1925), One-loop QED effects, 323, 324
25 Operator, 96, 97, 100
Arthur H. C OMPTON (1927), 28 energy, 101
Louis DE B ROGLIE (1929), 57 momentum, 101
Werner K. H EISENBERG (1932), 100 position in space, 101
S CHRDINGER and D IRAC (1933), 90 simultaneous measurement, 101
Otto S TERN (1943), 70 Optical path difference, 40
Isidor I. R ABI (1944), 482 Orbital angular momentum, 71, 295, 299,
Wolfgang PAULI (1945), 22, 138 302
Max B ORN (1954), 88, 89 components, 110, 111
L AMB and K USCH (1955), 317 eigenfunctions, 109113
T OMONAGA, S CHWINGER, F EYNMAN square, 111, 112
(1965), 79, 324, 534 Orbital energies, 509
R AMSEY, D EHMELT, PAUL (1989), 55, Orders of magnitude, 59
225, 288, 332, 482 energy scales, 7
G LAUBER, H ALL, H NSCH (2005), 248, length scales, 5
319 time scales, 7
M ATHER, S MOOT (2006), 6, 7 Orthonormality relation, 96
E NGLERT and H IGGS (2013), 46 Oscillator strength, 238, 239, 636640
N OBEL prize in physiology or medicine density, 256
L AUTERBUR, M ANSFIELD (2003), 490 sum rule, 239, 639, 640
Non-crossing rule, 391, 394
Non-local potential, 507 P
Non-stationary problems Pair production, 526
dipole excitation (E1), 169225 Paramagnetism, 394396
light matter interaction, 227270 Parity, 593, 594
Non-stationary states, 186, 187 conservation in E1 transitions, 202
normal Z EEMAN effect, 128131, 382386 multi-electron systems, 594603
Nuclear gN factor, 447, 449 Parity violation, 249
Nuclear magnetic moment, 447, 449, 485, 487, Particle detection, 75
488 Particle diffraction, 5861
Nuclear magnetic resonance, see NMR C60 , 60, 61
Nuclear mass, 119 He atoms, 59, 60
energy correction, 68, 69 Particles and waves, 57, 64
674 Index of Volume 1
Symbols Benzene
12, 6 potential, 149 D6h point group, 262
H CKEL orbitals, 277285
A (anisotropy) parameter, 356, 360, 551, 557
A BBE sine law, 679 B ETHE formula, 528530, 534, 537, 538, 546
Adiabatic representation, 478480 ionization, 537
A IRY diffraction pattern, 25 B ETHE integral, 526
A IRY disc, 26 B ETHE ridge, 546, 548, 549, 553
Alignment, 596, 604, 616, 619, 698700 B ETHE surface, 545
angle, 36, 44, 605, 610, 611 Binary peak in (e, 2e) process, 553
parameter, 622, 623, 698 Birefringence, 38, 40, 41
Alkali halide potentials, 220224 B IRGE -S PONER plot, 177
Ammonia maser, 1, 252 Black body radiation, 105
Amplification profile, 15 B OLTZMANN
Amplified spontaneous emission, 11 distribution, 155, 173, 335, 389, 591
Anharmonicity constant, 162 statistics, 158
Anti-S TOKES lines, 334 Bond order, 195
Anti-symmetrization, 521 Bonding orbital, 183
Antibonding orbital, 183 B ORN approximation
Ar2 TPES spectrum, 369 first order, elastic, 444448
Arn clusters, 400 generalized oscillator strength (GOS),
Atomic form factor 530534
inelastic, 527 inelastic collisions, 460, 461, 522, 525530
AUGER electron, 533 integral inelastic cross sections, 534
AUTLER -T OWNES effect, 637639 B ORN phase shift, 447
Autocorrelation function, 5360 B ORN series, 553
Avoided crossing, 220224, 263 B ORN -O PPENHEIMER approximation,
139151
B collision processes, 458, 474
Baseline in interferometry, 87, 91 BPP, see Beam parameter product
Beam divergence, 21, 22 Branching ratio, 653
Beam parameter, 1721
product, 26, 34, 35 C
Beam radius, 1735, 650 C6 H +
4 absorption spectra, 328
Beam waist, 1721 C2 DFWN spectrum, 354
and coherence volume, 87 Carrier envelope, 46
and lateral coherence radius, 91 phase, 46, 55, 188