Sei sulla pagina 1di 203

 DEGLI STUDI DI PERUGIA

UNIVERSITA
Dottorato di Ricerca in Scienze Chimiche

Sturmian Orbitals in
Quantum Chemistry

Andrea Caligiana
Coordinatore:

Relatore:

Prof. Giovanna Favaro

Prof. Vincenzo Aquilanti

Ottobre 2003

Ringraziamenti
Desidero ringraziare sentitamente numerose persone,
che hanno contribuito allo svolgimento del mio dottorato
ed alla realizzazione di questa tesi: innanzitutto il Prof.
Aquilanti, la cui disponibilita a venirmi incontro e sostenermi nei momenti dicili e stata ancora piu importante
di qualunque competenza scienti ca un relatore possa
trasmettere. Lo stesso dicasi per Simona (Prof.ssa Cavalli), che non ha mai smesso di infondermi ducia e
di aiutarmi a conoscere me stesso. Una parte considerevole dei risultati presentati e dovuta alla preziosa
collaborazione con il Prof. John Avery dell'Universita
di Copenhagen. Il prof. Tarantelli mi ha tante volte
trasmesso parte delle sue conoscenze non solo al Dipartimento di Chimica, ma anche durante lunghi viaggi a
bordo del bus n.9. L'estate del 2002, trascorsa al Dipartimento di Chimica di Cambridge(UK), e stata un periodo stimolante e piacevole, oltre che scienti camente
produttivo: ringrazio il Prof. Handy che mi ha permesso
di lavorare nel suo gruppo di ricerca e di alloggiare al
St. Catharine's College, ed i suoi collaboratori Rudolph
Burcl e David Tew.
Dopo i `senior', passo a menzionare coloro con cui
(oltre al luogo di lavoro) ho condiviso svariate serate
divertenti,... Andrea Beddoni, Andrea Lombardi, An-

drea Simoni, Alessandro Volpi, Leonardo Paci ci, Luca


Benedetti, Massimiliano Bartolomei, Loriano Storchi,
Noelia, Valentina Piermarini...
Voglio in ne ringraziare i miei familiari: e soprattutto
a loro che devo il raggiungimento di questo traguardo,
ed e a loro che dedico questa tesi.

Contents
Introduction
5
1 Hydrogenic Elliptic Orbitals, Coulomb

Sturmian Sets and Recoupling Coef cients among Alternative Bases


9
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . .
1.2 Polar and Parabolic Sturmians . . . . . . . . . . . .
1.3 The conservation of parity in the trasformation between polar and parabolic orbitals . . . . . . . . . .
1.4 Spheroidal Sturmians . . . . . . . . . . . . . . . . .
1.5 Spheroelliptic Sturmians . . . . . . . . . . . . . . .
1.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . .
Bibliography . . . . . . . . . . . . . . . . . . . . . . . .

10
16
24
31
39
48
53

2 Hydrogenic Orbitals in Momentum

Space and Hyperspherical Harmonics. Elliptic Sturmian Basis Sets. 59


2.1 Introduction . . . . . . . . . . . . . . . . . . . . . .
2.2 Background . . . . . . . . . . . . . . . . . . . . . .
2.3 The Canonical Bases of O(4) . . . . . . . . . . . . .
2.3.1 Asymmetric Parametrizations : Spherical and
Zeeman Sets . . . . . . . . . . . . . . . . . .
2.3.2 Symmetric Parametrizations . . . . . . . . .
2.3.3 The Scheme of the Orthogonal Transformations among the Canonical Bases of O(4) . .
2.4 Conservation of parity in the transformations among
momentum space Sturmians . . . . . . . . . . . . .
2.4.1 Spherical and Stark Basis Sets . . . . . . . .
1

60
62
67
69
75
76
80
81

2
2.4.2 Zeeman Basis Sets . . . . . . . . . . . . . .
2.5 The elliptic bases of O(4) . . . . . . . . . . . . . . .
2.5.1 Spheroelliptic Coordinates and Basis sets . .
2.5.2 Elliptic Cylindrical Coordinates of Type I and
II . . . . . . . . . . . . . . . . . . . . . . . .
2.6 Summary and concluding remarks . . . . . . . . . .
Bibliography . . . . . . . . . . . . . . . . . . . . . . . .

82
87
88
90
95
99

3 Many-electron one-center systems 105


3.1
3.2
3.3
3.4

Introduction . . . . . . . . . . . . . . . . . . . . . .
Free atoms . . . . . . . . . . . . . . . . . . . . . . .
Atoms in external elds . . . . . . . . . . . . . . .
Electronic integrals between Sturmian con gurations
3.4.1 Overlap Integrals . . . . . . . . . . . . . . .
3.4.2 Matrix elements of one-electron operators .
3.4.3 Matrix elements of two-electron operators .
3.5 On the choice of the `energy' in the de nition of the
basis set . . . . . . . . . . . . . . . . . . . . . . . .
3.6 Examples . . . . . . . . . . . . . . . . . . . . . . .
Bibliography . . . . . . . . . . . . . . . . . . . . . . . .

106
107
111
112
113
114
116
120
122
124

4 Sturmian approach to one-electron

many-center systems: integrals and


iteration schemes
127

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . 128


4.2 Two-center Nuclear Attraction Integrals S and W . 129
4.3 Solution of one-electron multicenter wave equation;
comparison among performances of di erent iterations134
4.4 Further remarks and conclusions . . . . . . . . . . . 140
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . 142

5 Sturmian Orbitals Applied to Molecules:


an Introduction
147
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . .
5.2 Sturmian basis sets for atomic and molecular calculations . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1 Many-center many-electron systems . . . . .
5.2.2 The integrals. Relationships with Slater-type
orbitals . . . . . . . . . . . . . . . . . . . .
5.3 Further remarks and conclusions . . . . . . . . . . .

148
151
151
156
163

3
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . 166

6 Anharmonic frequencies and Berry


pseudorotation motion in PF5
169
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . .
6.2 Geometry optimization and harmonic frequencies .
6.3 Tunnelling splittings in PF5 . . . . . . . . . . . . .
6.4 Anharmonic vibrations of PF5 . . . . . . . . . . . .
6.5 MULTIMODE with the Reaction Path Hamiltonian
6.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . .
Bibliography . . . . . . . . . . . . . . . . . . . . . . . .

Conclusions

170
171
174
181
187
190
192

197

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . 200

Introduction
The fundamental scope of quantum chemistry is the determination of the wavefunctions for the stationary states of electrons
in the presence of xed nuclei, eventually including the e ect of
external elds and perturbations. Then, according to the BornOppenheimer approximation, the electronic wavefunctions may be
used for the construction of potential energy surfaces, on which
the motion of the nuclei can be studied, by quantum or classical
techniques.
The electronic wavefunctions are obtained solving a time independent Schrodinger equation by methods that include the use
of sets of monoelectronic functions (orbitals) as basis sets for the
approximation of the solutions. The kind of functions used as electronic orbitals a ects strongly the convergence rate and accuracy
of numerical procedures.
The only chemical system whose Schrodinger equation is analytically solvable is the hydrogen atom. In the early days of quantum
chemistry, scientists tried to use hydrogenic functions (i.e. the solutions of the Schrodinger equation for the hydrogen atom) and
5

6
similar functions as orbitals to approximate more complex wavefunctions, incurring in the inconvenient of complicated two-electron
integrals. In the last decades, instead, linear combinations of Gaussian functions have been used as electronic orbitals, replacing all
the other types of basis sets: Gaussians have allowed a faster calculation of two electron integrals, and consequently an extension
of quantum chemistry towards larger systems, permitted also by a
great improvement of computer performances. At the present time,
the main goals of researchers are the implementation of calculations
for larger and larger systems, and also a better understanding of
excited states.
This Doctoral Thesis is devoted to the study of a type of functions, the Sturmian hydrogenic orbitals, which so far have received
very little attention in the eld of quantum chemistry. Hydrogenic
Sturmians are formally similar to the well-known hydrogenic orbitals, but their use has important and peculiar implications that
are worthwhile investigating. We have considered the applications
of Sturmians in ab initio quantum chemistry, without entering the
eld of density functional theory. The Thesis is structured in three
parts.

The rst part (Chapters 1 and 2) is devoted to the study of


the properties of hydrogenic Sturmian orbitals.
In chapter 1 it is shown that there are di erent types of coordinates suitable for the solution of Sturmian Hydrogenic wave-

7
function in con guration space; consequently, there are different types of hydrogenic Sturmians in con guration space.
Symmetry properties and orthonormal transformations among
them are studied.
Chapter 2 deals with the properties of Sturmians in momentum space. In the reciprocal space, Sturmians are strictly
related to hyperspherical harmonics. This property is very
interesting by itself and also because it suggests the application of angular momentum techniques in quantum chemistry.
There are di erent types of hydrogenic Sturmians in momentum space; as in the previous chapter, Symmetry properties
and orthonormal transformations among them are studied.
The relationship between the two reciprocal spaces is also
discussed.

The second part of this thesis (Chapters 3,4 and 5) is dedicated to the applications of Sturmian basis sets in quantum
chemistry.
In chapter 3 the application of Sturmian orbitals as a basis set for atomic structure calculations is investigated, also
including the e ect of external elds.
In chapter 4 there is a treatment of Sturmian formalism applied to one electron molecules according to the MO-LCAO
method.
Chapter 5 deals with an original method for exploiting Stur-

8
mians in a valence bond approach to the study of molecules.
The passage to momentum space is used in order to solve
some multi-center one-electron integrals

Chapter 6 is an excursion in the eld of internal dynamics


of PF5 molecule. The electronic ground state potential energy surface is obtained by density functional calculations.
Then, the nuclear rearrangement dynamics of the molecules
are studied by quantum techniques.

Chapter 1
Hydrogenic Elliptic Orbitals, Coulomb
Sturmian Sets and Recoupling
Coecients among Alternative Bases
Summary

The non-relativistic Schrodinger equation for the Coulomb problem is separable in four di erent coordinate systems in con guration
space: alternative sets of orbitals for the hydrogen-like atoms correspond to each of them and permit to obtain Sturmian sets, useful as
complete orthonormal expansion bases in atomic and molecular calculations. In this chapter the fundamental properties of the already
known hydrogenic orbitals (the familiar polar, the parabolic and the
rarely treated spheroidal sets) are resumed; then we discuss some
properties of the spheroelliptic orbitals, which have been so far practically ignored. We pay particular attention to the symmetries of
the di erent orbital sets and to the relationships between them, and
9

10
we order them in a complete scheme which exhibits passages from
one to the other through explicitly derived orthogonal transformations. Within this context we insert the study on the conservation
of parity in the passage from the polar set to the parabolic one. We
also show that - except for the spheroidal set - all alternatives for
hydrogenic wave functions induce irreducible representations of the
point group D2h , the `quaternion group'. This has relevance for the
discussion of connections between these sets and the corresponding
ones in momentum space, presented in Chapter 2.

1.1 Introduction
The Schrodinger equation for the hydrogen atom was solved in
the early days of quantum mechanics by using polar coordinates
(r; #; '). The other coordinate systems suitable for separating and
solving the equation are less known; each of them leads to hydrogenic orbitals with di erent properties. A systematic search on this
topic was completed in the 70's exploiting the relations between
separation of variables for di erential equations and Lie group theory [1,2]. In spite of the apparent simplicity of the hydrogen atom
as a physical system , its quantum-mechanical treatment turns out
to be wide and complicated; the reason can be found in its exceptional symmetry, which is the symmetry of the (hyper)-sphere
S 3 , embedded in a four-dimensional Euclidean space. Our interest
towards this study is due not only to the scienti c quest for under-

11
standing properties of the only atom of the periodic table which
can be treated exactly by an analytical approach, but also to the
important aspect that the various types of orbitals are candidates
as basis sets for the numerical solution of other atomic and molecular systems, and are therefore basic building blocks for quantum
chemistry.
The physical quantum states for a hydrogenic atom with nuclear
charge Z are described by the wave functions which are solutions
of the Schrodinger equation:
(H^

E)

1 2
r
2

Z
r

=0

(1.1)

where r is the distance nucleus-electron. This is an exactly solvable two-body problem, at variance with all the other atoms of the
periodic table. The hydrogenic hamiltonian operator H^ has a discrete spectrum (bound states, negative E ) and a continuous one
(unbound states, positive E ), and one has to take both to get a
complete set. In view of diculties which arise when working with
continuous basis sets, it is better to associate a discrete set to the
solution of the quantum problem in Eq. (1.1). Such a complete set
of orbitals is called Coulomb Sturmian, the nomenclature originating from the fact that Eq. (1.1) is a Sturm-Liouville problem.
The quest for completeness lead many authors to introduce sets
which are closely related to usual orbitals, but for which completeness holds with no need of continuum states, an early example being
the "Natural Spin Orbitals" by Shull and Lowdin [3]. General Stur-

12
mian functions were introduced by Rotenberg [4, 5], starting from
the Schrodinger equation for the variable r in a radial- eld system
in the three dimensional physical space. The most important example is the Coulomb problem, for which Sturmian functions are
de ned as solutions of the equation
"

1 d 2d
r
2r2 dr dr

l(l + 1)
+ E0
2r2

Zn
R (r) = 0
r n;l

(1.2)

for n = 1; 2; 3; :::: and l = 0; 1; ::::; n 1; E0 is a xed negative


number, a "reference" energy, to which it will be convenient in the
following to associate a "reference" momentum

p0 = ( 2E0 )1=2

(1.3)

and Zn is an eigenvalue, xed by the boundary conditions: Rn;l (0) =


0 and Rn;l (1) ! 0. Eq. (1.2), written in a.u., is very similar to
the radial part of the Schrodinger equation for the hydrogenic atom
separated in polar coordinates: the main di erence is that in the
latter the nuclear charge Z is xed a priori and only a discrete
spectrum of energies turns out to be allowed, while in the SturmLiouville equation the energy E0 is xed, and the "nuclear charge"
Zn is varied until the boundary conditions are ful lled, generating
a complete set of Coulomb Sturmians, which are therefore the radial functions Rn;l (r) multiplied by the usual spherical harmonics
as angular functions: un;l;m(r; #; ') = Rn;l (r) Yl;m(#; ').
It can be veri ed that if for example E0 is 1=2, i.e. the energy of the ground state of hydrogen atom, then one has Z1 = 1

13
and u1;0;0(r; #; ') is the wave function for the orbital 1s; no other
un;l;m(r; #; ') corresponds to a physical wave function, but a complete set can be explicitly de ned. Coulomb Sturmians obey a
peculiar orthonormality relation [6]:
Z

un0;l0 ;m0 (r; #; ')

1
p
un;l;m(r; #; ')dr = 0 nn0 ll0 mm0
r
n

(1.4)

where the momentum p0 (Eq. 1.3) is seen as a crucial parameter.


The number of allowed Zn is unlimited, thus there exists an in nite number of un;l;m(r). Such in nite set is complete, but n is
numerable: therefore there is no continuous spectrum, and this is
the fundamental feature of these sets [3, 6{9]:
X

n;l;m

un;l;m(r0 ; #0 ; '0 ) un0 ;l0 ;m0 (r; #; ') = (r

r0 ) (# #0 ) (' '0 )

(1.5)
In the next section we will further discuss the apparent coincidence
and the di erences between Coulomb Sturmian eigenfunction sets
and the usual hydrogenic orbitals.
The hamiltonian in Eq. (1.1) can be easily formulated, by using
a Cartesian coordinate system centered on the nucleus, but it is
known that these coordinates lead to a non-separable di erential
equation. The problem of the separation of variables is ubiquitous
in quantum mechanics, because in the study of systems with many
degrees of freedom one tries to reduce the original multidimensional
problem in the hope of getting problems of lower dimensionality,
which are simple enough to be solved analytically or numerically,

14
and allowing us to better understand the physical process and to
refer to simpler cases. Separation of variables can be exact or only
approximate. Exact separation can be viewed as symmetry separation, since it is associated to geometrical and dynamical symmetries
of the system [10, 11]; this is the type of variable separation which
applies to the study of the hydrogen atom as a two-body problem
and thus will be under focus in this chapter. Symmetry separation
usually involves the introduction of separation constants to isolate
each variable, and for each of them one gets a di erential equation where only one degree of freedom appears. In other words, a
partial di erential equation becomes a set of (uncoupled) ordinary
di erential equations.
For most quantum systems, which are made of three or more
bodies, an approximate separation can be performed, based on the
assumption that the motion associated to a group of variables can
be treated as the other ones were frozen; the "fast" variables produce an e ective potential for the motion of the "slow" variables,
which will have to be treated accordingly. The semiclassical nature
of this adiabatic separation procedure has been elucidated [12, 13].
The Born-Oppenheimer separation is an example, as well as the
hyperspherical approach to the few-body problem.
In the next sections, we will consider the four coordinate systems
which allow the exact separation of the Schrodinger equation. For
each coordinate system the properties of the corresponding Sturmian basis vary and the ranges of problems for which they are

15
appropriate basis sets di er accordingly.
Sturmian sets in con guration space have hyperspherical harmonics basis sets as their counterparts in momentum space. Care
will be taken to develop the treatment to prepare for the following
chapter, where hyperspherical symmetry is exploited in the momentum space framework. The correspondence is best understood
making reference to symmetry, and thus to group theory. The second order commuting operators which appear as constants of the
motion in the treatment of hydrogen atom in momentum space - our
next chapter will deal with the latter - are invariant with respect to
the inversion operation: as a consequence, momentum space Sturmians induce irreducible representations of the point group D2h [2]
(Schon ies notation is used throughout this thesis, see Ref. [14]).
Thus, only con guration space Sturmians inducing irreducible representation of D2h have a counterpart in the momentum space corresponding to the same set of quantum numbers. For this reason
we pay special attention in the present chapter to the behaviour
of the con guration space Sturmians in the point group D2h . This
correspondence will be treated in the following chapter, where previous problems such as the Coulson's "intractability of spheroidal
orbitals" [15] will be shown to arise from the di erent behaviour of
commuting operators resulting from separation of variables in the
two reciprocal spaces with respect to inversion of parity.
Notations and properties of polar and parabolic Sturmians, having been recently discussed (and also extended to the general d-

16
dimensional hydrogen atom case) [16], are brie y resumed in Sec. 2,
while in Sec. 3 we implement the requirement of parity conservation
in the orthogonal matrix which connects the two sets. Spheroidal
Sturmians are described in Sec. 4 (the corresponding hydrogenic
orbitals have been lucidly illustrated by Sung and Herschbach [17]).
Spheroelliptic Sturmians and their properties are introduced in Sec.
5, where their expansion coecients in terms of the other sets are
given as solutions of a three-term recursion relationship, which also
allows us to obtain the `missing label', i.e. the additional quantum
number characterizing the set (this is typical of elliptic coordinates
and basis sets, an ample variety of which will be encountered in
the next chapter). This work leads to the complete list of Coulomb
Sturmian sets and of the recoupling coecients among alternative
bases: the corresponding connection scheme will be exhibited in
the concluding Section.

1.2 Polar and Parabolic Sturmians


The separation of variables in di erential equations can be associated to the algebraic structure described by the theory of continuous groups [1, 2, 10]. Therefore the Schrodinger equation for a
given system has to be discussed with reference to its dynamical
symmetries. The Coulomb potential is radial, i.e. it depends only
from the distance r between nucleus and electron, and therefore the
hydrogen atom has an evident spherical symmetry and the polar

17
coordinates (see g. 1a) are the most natural choice for solving eq.
(1.1). Writing the polar Sturmian as a product of a radial factor
Rn;l (r) and an angular factor Yl;m (#; ') enables us to separate the
coordinate r from the others. After introducing the separation constant l(l + 1), we obtain two di erential equations: the rst one is
eq. (1.2), and the second one
"

1 @
@
1 @2
sin #
+ 2
Y (#; ') = l(l + 1)Yl;m(#; ')
sin # @#
@#
sin # @'2 l;m
(1.6)
coincides with the equation for the free motion of a particle on
the surface of the three-dimensional sphere S 2 ; it is known that the
wave functions of that system are the spherical harmonics Yl;m (#; ') =
Nl;m Pljmj(cos #) eim' (Nl;m is the normalization factor [18]). The
explicit expression of the polar Coulomb Sturmians in con guration
space is:
"

#
l 1)! 1=2
2n[(n + l)!]3

(n
un;l;m(r; #; ') = (2p0 )3=2

(1.7)

l 2l+1
0 r) Ln l 1 (2p0 r) Yl;m(#; ')

p r (2p
0

where p0 =
2E0 , n = 1; 2; 3; : : :, l = 0; 1; : : : ; n 1 and l 
m  l; L2nl+1l 1 (2p0 r) is a Laguerre polynomial [19]. The use of p0
instead of the more traditional energy E0 serves to emphasize the
connection with the treatment of the same problem in momentum
space, which will be presented in the following chapter. It is important to stress that Eq. (1.7) also represents hydrogenic orbitals:
the di erence between orbitals and Sturmians is that in the former

18

p0 can have any value according to the expression p0 = Z=n, while


in the latter p0 is a xed number and the nuclear charge Z can take
any value Zn according to the relationship Zn = p0 n. The Sturmians can be seen as satisfying the following eigenvalue equations, in
which ^l is the orbital angular momentum of the electron:
1 2 p20
r+2
2

^l2

np0
un;l;m(r) = 0
r
i
l(l + 1) un;l;m(r) = 0
(^lz m) un;l;m(r) = 0

(1.8)
(1.9)
(1.10)

In Eq. 1.8 the product np0 gives the "charge" Zn . To these three
relationships we add

P^ un;l;m(r) =  un;l;m(r)

(1.11)

where P^ is the operator that performs the inversion of all the electronic coordinates with respect to the nucleus, and  = ( 1)l .
The following real linear combination of polar orbitals are conveniently de ned ( = 1):

un;l;jmj = jn; l; m;  i = (2i 1 ) 1=2 (un;l;m  un;l; m)

(1.12)

These functions are symmetric with respect to rotations of  around


all three Cartesian axes ( = 1, according to whether m is even
or odd):

C^2 (x) un;l;jmj =  un;l;jmj


C^2 (y ) un;l;jmj =  un;l;jmj

(1.13)
(1.14)

19
Table 1 - Number of polar (Eq. 1.12) or spheroelliptic functions for
given l,  and  labels.

;  !
l=1

2
3
4
5
6
...
even l
odd l

+1,+1

-1,+1

1
2
2
3
3
4
...
l=2 + 1
(l + 1)=2 (l

+1,-1

-1,-1

0
1
1
1
1
1
1
2
2
2
2
2
2
3
3
3
3
3
...
...
...
l=2
l=2
l=2
1)=2 (l + 1)=2 (l + 1)=2

C^2 (z ) un;l;jmj =  un;l;jmj


P^ un;l;jmj =  un;l;jmj

(1.15)
(1.16)

Therefore the functions un;l;jmj are seen to induce irreducible repre^ C^2 (x); C^2 (y ); C^2(z )g
sentations in the point group D2h = D2 Ci (D2 = fE;
^ P^ g). This is also known as the "quaternion group".
and Ci = fE;
For symmetry operations and groups, we use the notation of Ref.
[14]. Table 1 shows the number of un;l;jmj functions for given l; 
and  values.
The separation of variables in the Schrodinger equation written
in polar coordinates is always possible for the motion in any radial
eld. In the case of Coulomb eld, separation of variables is also
possible in parabolic coordinates (Fig. 1a).
These coordinates have two points of reference: one is on the

20

z
z

) /2
2

1+

P ()

( 1 2 ) /2

P ()

(+) A

(a)

( 1 2)

1/2

A (+)

(b)

x
=cost

Figure 1.1:

The part (a) of this gure shows polar and parabolic coordinates. The de nition of the polar coordinates with respect to the cartesians
is x = r sin # cos '; y = r sin # sin '; z = r cos #. The relatioships of parabolic
coordinates with the cartesians and the polars are x = (1 2 )1=2 cos '; y =
(1 2 )1=2 sin '; z = 12 (1 2 a nd 1 = r(1 + cos #); 2 = r(1 cos #); ' = '.
The part (b) serves for the de nition of spheroidal coordinates:  = a+ b where
1    1 and  = a  b where 1    1. ' is the angle between the plane
ABP and a xed plane through AB; 0  ' < 2 . Therefore if the y axis is s
een as pointing downward in the plane of the drawing, ' is as in the part (a).

21
nucleus, and the other is at in nite distance along a direction that
in Fig. 1a coincide with the z axis. The explicit expression for the
parabolic sturmians is, according to the notation in [20]:

Un;;m (1 ; 2 ; ') =

1
2

3
0

 p0p12 e


Lm(n+

p [(+n m 1)=2]![(n  m 1)=2]!


 n f[(+n+m 1)=2]![(n +m 1)=2]!g
m
p  

1+
2


(1.17)

m
im'
m 1) (p0 1 ) L (n  m 1) (p0 2 ) e
1
2

The parabolic orbitals satisfy the following eigenvalue equations:


1 2 p20
r +2
2

np0
Un;;m (r) = 0
r


K^ z  Un;;m (r) = 0


^lz m Un;;m (r) = 0

(1.18)
(1.19)
(1.20)

The quantum numbers n; ; m can have the following values: n =


1; 2; 3;. . . , (n 1)  m  (n 1) and (n jmj 1)   
(n jmj 1). For alternative notations for parabolic quantum
numbers and generalized Laguerre polynomials, and their relationships with the present ones, see Ref. [21]. In eq. (1.19) the operator
K^ z corresponds to the component along the z axis of the reduced
Runge-Lenz vector [2,22,23], which classically gives the direction of
the major axis of an elliptic orbit; it is a costant of the motion and
therefore commutes with the hamiltonian operator. The operator
K^ is de ned as:

K^ =

1 ^
l  p^
2

p^  ^l

Zr 
=p0
r

(1.21)

22
The vectors K and l are orthogonal in classical mechanics, and for
^ = 0. Since ^l is invariant
the quantum mechanical operators ^l  K
with respect to inversion, while p^ and r change sign, eq. (1.21)
^ also changes sign under inversion; thus K^z doesn't
implies that K
commute with P^ and the parabolic functions [Eq. (1.17)], being
eigenfunctions of K^ z , do not have a well de ned parity. We will
come back to this point in the following section, both because this
is crucial in establishing relationships with the momentum space
treatment and also because for applications it is often useful to
have a basis of de nite parity.
The trasformation matrix between the polar set jn; l; mi [Eq.
(1.7)] and the parabolic set jn; ; mi [Eq. (1.17)] has to conserve the
quantum numbers n and m, then it is made of blocks of dimension
n jmj because for xed n and m, there exist n jmj polar orbitals
and n jmj parabolic orbitals. Within a phase factor, the matrix
elements - which coincide with the overlaps of the functions of the
two sets - are particular Clebsch-Gordan (CG) coecients [18, 22,
24, 25] in which the two coupling vectors have the same modulus
[26]:

jn; l; mi

Pn jmj

= (n jmj 1)

( ) (n +m 1) cn;m
l; jn; ; mi
1
2

(1.22)

1); 2 + 2 2
2 + 2 j l; mi
This relationship shows that the connection between the polar and
parabolic basis can be seen as a change of coupling schemes [17],
^ ).
involving the angular momentum operators k^ = 12 (^l  K

cn;m
l;

= h 21 (n

m ; 1 (n

1); 

23
An alternative view of these matrix elements stems from the
Regge symmetry [25, 27] for the Clebsch-Gordan coecients:
h 12 (n 1); 12 (m ); 12 (n 1); 21 (m + )jl; mi =
h 12 (n + m 1); 2 ; 12 (n m 1); 2 jl; 0i
(1.23)
The last form (a Clebsch-Gordan coecient where the projection
of the orbital angular momentum is zero) suggests that  can be
interpreted as a helicity quantum number [21]. Indeed the choice
of the reference frame for which the component of the orbital angular momentum vector ^l is zero amounts to change the quantization
axis on the plane of the orbit, where the Runge-Lenz vector lies.
This choice is reminiscent of analogous ones which occur in different contexts: we recall the transformation between space- xed
and body- xed frames in molecular collisions [28], the transformation between the Hund's cases (e) ! (c) in molecular spectroscopy
and atomic scattering [29, 30] and the passage between symmetric
and asymmetric coordinates in the hyperspherical treatment of the
three-body problem [31, 32].
In general, the CG coecients enjoy two basic three-term recurrence relationships with respect to angular momentum and to
the projection quantum numbers [33, 34]. For the CG coecients
in Eq. 1.22, we have with respect to l:
v
u
u
t

v
u

u [(n
[n2 l2 ][l2 m2 ] n;m
1)2 l2 ][l2 m2 ] n;m
t
cl 1; cn;m
+
cl+1; = 0
l;
(2l 1)(2l + 1)
(2l + 3)(2l + 1)
(1.24)

24
and with respect to :
1q 2
[n
4

(m +  + 1)2 ] [n2 (m  1)2 ] cn;m


l;+2 +


1 2
1
(n 1) + (m2 2 ) l(l + 1) cn;m
l; +
2
2
1q 2
[n (m  + 1)2 ] [n2 (m +  1)] cn;m
l; 2 = 0
4

(1.25)

Eqs. 1.24 and 1.25 can be seen from a matricial point of view:
the CG coecients that appear there are eigenvectors of tridiagonal
matrices. Eq. 1.24 corresponds to the matrix representation of K^ z
by the set jn; l; mi, the eigenvalues being ; eq. 1.25 corresponds to
the matrix representation of ^l2 by the set jn; ; mi, the eigenvalues
being l(l + 1). Here we see a manifestation of properties which can
be viewed from the relationships between CG coecients and Hahn
polynomials of a discrete variable [35]. For these polynomials and
their duals, there is an exchange in the role of the discrete variable
and their degree, and Equations like (1.24) and (1.25) can be also
seen as the nite di erence equations respectively either in l or .

1.3 The conservation of parity in the


trasformation between polar and parabolic
orbitals
Let us consider explicitly, at given n and m values, the matrix
elements of ^l2 on the basis jn; ; mi: those on the row labeled by 
are the coecients appearing in Eq. (1.25). We observe that such

25
a matrix is centrosymmetric (a matrix A of dimensions d  d is
centrosymmetric if Aj;k = Ad+1 j;d+1 k ), implying the existence of
a simple similitude transformation which factorizes it in two blocks
[36]. Accordingly, the set jn; ; mi transforms into a new set of
eigenfunctions of the parity operator P^ , which commutes with ^l2 .
Block-diagonalization occurs because matrix elements of ^l2 between
functions which have opposite parity vanish. The explicit form of
the new basis of de nite parity, which will turn out to be connected
by a simple relationship with the originary basis, can be obtained
by projecting out the even and odd components of the parabolic
Sturmians in Eq. (1.17).
As we can see from Figure 1, inverting the position vector r
of the electron amounts, in parabolic coordinates, to exchanging
1 and 2 and turning ' into ' +  . By summing or subtracting
Un;;m (1 ; 2 ; ') and Un;;m (2 ; 1 ; ' +  ), we get the even (U + ) and
the odd (U ) components of the latter ( = 1, as before):
1
U  (1 ; 2 ; ') = p [Un;;m (1 ; 2 ; ')  Un;;m (2 ; 1 ; ' +  )] =
2
(1.26)
1
= p [Un;;m (1 ; 2 ; ')  ( )m Un;;m (2 ; 1 ; ')] =
2
 q
m
h


1
= p Nn;;m p0 1 2 e p

Lm(n+ m 1) (p0 1 )
2
i
Lm(n  m 1) (p0 2 )  ( )m Lm(n  m 1) (p0 1 )Lm(n+ m 1) (p0 2 ) eim'
0

1
2

1
2

1+
2

1
2

1
2

The factor Nn;;m can be written explicitly by comparing Eq. (1.26)


and Eq. (1.17). The comparison also shows that these de nite-

26
parity parabolic orbitals are linear combinations of two orbitals
which have opposite values of :
1
Un; jj;m = jn; ; m; i = p [Un;;m  ( )m Un;
2

;m ]

(1.27)

Therefore, the new de nite-parity parabolic functions continue to


be eigenfunctions of H^ and ^lz , but they are so now also for K^ z2 and
P^ (and not for K^ z ):

K^ z2 Un; jj;m = 2 Un; jj;m


P^ Un; jj;m =  Un; jj;m

(1.28)
(1.29)

Compatibility of these two equations follows from the fact that K^ z2


and P^ commute.
Having de ned the parity basis set, we derive the matrix elements Cl;n;m;
jj for its transformation into the polar set:

jn; l; mi =

n X
jmj 1
=0 or 1

( ) (n +m 1) Cl;n;m;
jj jn; ; m; i
1
2

(1.30)

where the step of the sum over  is two units; it starts from 0 or 1,
according to whether n jmj 1 is even or odd: indeed, in the ket
jn; ; m; i of de ned parity,  can only be positive or zero. The
desired orthonormal transformation matrix has to be factorized in
two blocks distinguished by the sign of the parity , at xed values
of n and m. We have
1 + ( )l n;m
q
Cl;n;m;
=
cl;jj
jj
2(1 + ;0 )

(1.31)

27
For each parity, columns are now labeled by jj and , instead of 
as in Eq. (1.22), and the quantities Cl;n;m;
jj are combinations of CG
coecients in Eq. (1.22) which di er only for the sign of :

Cl;n;m;
jj =

h
i
1
m cn;m
cn;m
+

(
)
l;
l; 
2(1 + ;0 )

(1.32)

This matrix element is zero when l is an even (odd) number and


p
 = 1 ( = +1). For  6= 0, it takes the value 2cn;m
l;jj , and in
the particular case  = 0, it is simply the CG coecient cn;m
l;=0 .
The matrix elements cn;m
l;=0 constitute an eigenvector of parity  =
( 1)m ; accordingly, when l and m have opposite parity, they are
zero.
A parity factor which is analogue to the one in Eq. 1.31 had
already appeared in the literature in di erent contexts, for example, in the study of the passages among the various angular momentum coupling schemes in the quantum mechanical treatment of
molecular [28] and open-shell atomic [29] collisions. Let us show an
example; for n = 4 and m = 0 we have:

0
B
B
B
B
@

u4;0;0
u4;1;0
u4;2;0
u4;3;0

C
C
C
C
A

B
B
B
B
B
@

3
2
1
2

1
2
q
1

2
q

1
5
1
5

1
2
q

1 1
2 5
1
2
q
3 1
2 5

1
q2
1 1
2 5
1
2q
3 1
2 5

1
q2
3 1
2 5
1
q2
1 1
2 5

10
CB
CB
CB
CB
C@
A

U4; 3;0
U4; 1;0
U4;1;0
U4;3;0

1
C
C
C
C
A

(1.33)

28
Block-diagonalizing according to the conservation of parity, one has:
0
B
B
B
B
@

u4;1;0
u4;3;0
u4;0;0
u4;2;0

1
C
C
C
C
A

B
B
B
B
B
B
@

3q 101
0
0

1
10

1
q 10
3 101

0
0

0
0

p1
2
p1
2

0
0

p1
2
p1
2

0
C
CB
CB
CB
CB
C@
A

U4;3;0
U4;1;0
U4+;1;0
U4+;3;0

1
C
C
C
C
A

(1.34)

Eq. (1.31) de nes analogues of the CG coecients which conserve the parity; we will refer to them as parity-conserving ClebschGordan coecients (pCG). In the following we will show how to
work out the three-term recurrence relationships enjoyed by pCG
coecients. Indeed, the eigenvectors of the matrix of K^ z2 in the
basis jn; l; mi are eigenfunctions also of H^ and ^lz : in fact since
[K^ z2 ; H^ ] = [K^ z2 ; ^lz ] = 0, the matrix elements of functions which differ in n and/or m are zero, and n and m are preserved as good
quantum numbers. Therefore, the matrix is diagonalized by the
coecients of the orthogonal transformation between jn; l; mi and
jn; ; m; i: those coecients are exactly pCG-coecients. We will
see that the matrix is tridiagonal, and therefore the pCG coecients
enjoy a three-term recurrence relationship.
As we have seen, [K^ z2 ; P^ ] = 0, and the parity of jn; l; mi is ( 1)l ,
so we expect that only matrix elements hn; l0 ; mjK^ z2 jn; l; mi with l
and l0 having the same parity can be di erent from zero. This
implies that the matrix is factorized in two blocks, for even and

29
odd l. In the l-th row only three elements are di erent from zero:


m ][n (l+1) ]
[l m ][n l ]
hn; l; m j K^ z2 j n; l; mi = [(l+1)(2l+1)(2
+
l+3)
(2l 1)(2l+1)
2

m ][n (l+1) ][(l+1)


hn; l; m j K^ z2 j n; l +2; mi = (2l1+3) [n (l+2) ][(l+2) (2l+1)(2
l+5)
r
[
n
(
l
1)
][(
l
1)
m
][
n
l ][l m ]
hn; l; m j K^ z2 j n; l 2; mi = (2l1 1)
(2l 3)(2l+1)
(1.35)
Consequently, we have a recurrence relationship among the Cl;n;m;
jj
which di er by two units in the value of l:
v
u
u
t

[n2

m]
2

1)2 ] [(l 1)2 m2 ] [n2 l2 ] [l2 m2 ] n;m;


Cl 2;jj +(1.36)
(2l 3)(2l 1)2 (2l + 1)
(
)
[(l + 1)2 m2 ] [n2 (l + 1)2 ] [l2 m2 ] [n2 l2 ]
2
+
 Cl;n;m;
jj +
(2l + 1)(2l + 3)
(2l 1)(2l + 1)
v
u 2
u [n
(l + 2)2 ] [(l + 2)2 m2 ] [n2 (l + 1)2 ] [(l + 1)2 m2 ] n;m;
t
Cl+2;jj = 0
(2l + 1)(2l + 3)2 (2l + 5)
Eq. (1.36) relates the coecients which belong to the same parity
block in the transformation matrix between the basis sets jn; l; mi
and jn; ; m; i.
In order to work out the second recurrence relationship for the
pCG coecients, we have to consider the matrix of ^l2 in the basis
jn; ; m; i. Since [^l2; P^ ] = 0, only the matrix elements between
functions of the same parity can di er from zero, so we have again
a factorization in two blocks. The e ect of ^l2 on jn; ; m; i is
identical to its e ect on jn; ; mi, thus the recurrence over  of the
pCG coecients
1q 2
[n (m +  1)2 ] [n2 (m  + 1)2 ] Cl;n;m;
j 2j +
4
(l

30



1
1 2
(n 1) + (m2 2 ) l(l + 1) Cl;n;m;
jj +
2
2
(m +  + 1)2 ] [n2 (m  1)2 ] Cl;n;m;
j+2j = 0 (1.37)

1q 2
[n
4
is identical to the recurrence for the CG (Eq. 1.25), apart from the
range of .
Finally, we complete this section by building-up explicitly real
parabolic orbitals which enjoy the symmetry properties of the group
D2h : they are

Un;;jj;jmj = jn; ; m; ;  i = (2i 1 ) 1=2 Un; jj;m  Un; jj;

(1.38)

^ K^ z2 ; ^lz2 , and to induce irreThey are seen to be eigenfunctions of H;


ducible representations for the group D2h , being eigenfunctions of
the following operators ( = 1 corresponds to even or odd m):

C^2 (x) Un;;jj;jmj =  Un;;jj;jmj


C^2 (y ) Un;;jj;jmj =  Un;;jj;jmj
C^2 (z ) Un;;jj;jmj =  Un;;jj;jmj
P^ Un;;jj;jmj =  Un;;jj;jmj

(1.39)
(1.40)
(1.41)
(1.42)

For xed n and jmj, there exist n jmj functions of type Un;;jj;jmj.
Table 2 indicates how many of these functions there are for each
couple of  and  labels at a given n jmj value. From the fact
that eq. (1.22) is invariant with respect to the sign of m, it follows that the coecients of the transformation between jn; l; m;  i
and jn; ; m; ;  i coincide with the pCG coecients, i.e. with the
coecients of the transformation between jn; l; mi and jn; ; m; i.

31
Table 2 - Number of D2h parabolic functions (Eq. 1.38) for given
 and  labels and n jmj xed.

;  !

even m
odd m

+1,+1

-1,+1

(n

jmj)=2 (n jmj)=2

+1,-1

(n

0
jmj)=2 (n

-1,-1

0
jmj)=2

1.4 Spheroidal Sturmians


A basic study of the hydrogen atom in spheroidal coordinates (Fig.
1b) is due to Coulson and collaborators [15, 37]. This two-center
coordinate system is very suitable for studying the motion of the
electron in H+2 within the Born-Oppenheimer approximation, because it renders the Schrodinger equation separable, though not
analytically solvable. In this way, "diatomic orbitals" - i.e. orbitals
for one electron between two nuclei - can be computed [38], and also
exploited for quantum calculations on two-electron problems [39].
In the case of hydrogen atom, one of the two centers coincides with
the nucleus and the other is set at a distance  along a reference
axis. A brilliant exposition, whose spirit we follow, concerning both
H and H+2 can be found in Ref. [22]. The spheroidal orbitals are the
best zero-order functions to treat the hydrogen atom perturbed by
a point charge [40]. Analytical solutions for an electron interacting with two xed nucleus are known only for special values of the
internuclear distance and of the ratio of nuclear charge [37]. More
recent applications are cited in Refs. [17, 41, 42], and Ref. [43] is a

32
monograph on spheroidal functions.
In order to solve the Coulomb problem in spheroidal coordinates one writes the spheroidal Sturmians T (; ; ') as a product
( ) N ( )('), and introduces two separation constants, m2 e f .
De ning

p
Z
E0 1=2
= 0 = n

2
2
2n
we one can write the separated equations as:
d2 
2
2 = m  (1.43)
d'
"
#
"
#
d
d
m2
2
2
2
(1  )
+ f +  2n 
 = 0 (1.44)
d
d
1 2
The equation in  is formally identical to the one in  , but the
interval of de nition of the variable is di erent (see Fig. 1a). Eq.
(1.43) also occurred in the separation in the polar variables # and
' (see Eq. 1.6); its normalized solutions are again (') = p12 eim' ,
and the T (; ; ') are eigenfunctions of ^lz .
Let us consider Eq. (1.44). Since for  ! 1 the asymptotic
solution is e  , we demand for the unknown function ( ) the
form:
nX1
( ) = e 
cl Pljmj( )
(1.45)
l=jmj

where the Pljmj( ) are associated Legendre polynomials [19] in which


the variability interval is [1; 1[, not [ 1; 1]. In order to calculate
the unknown coecients of Eq. (1.45), we have to put it in Eq.
(1.44) and after some passages omitted for brevity, taking into account some properties of the Pljmj( ) [19,44], a three-term recurrence

33
relationship for the cl is obtained [17, 22]:
(n + l + 1)(l + m + 1)
(n l)(m l) f + 2 l(l + 1)
+cl
cl+1
=0
(2l 1)
2
(2l + 3)
(1.46)
This amounts to say that, after xing n and jmj, it is necessary to
diagonalize a tridiagonal matrix of order n jmj to get the n jmj
spheroidal orbitals; the eigenvalues are the separation constants
f associated to each orbital. As a consequence, at variance with
what happens for the polar and parabolic basis sets, neither the
spheroidal eigenfunctions nor the eigenvalues f can be written as
a "closed" expression. Symmetrizing the recurrence relationship of
the cl (see Appendix) one has [22]:

cl 1

v
u
u
t

h
[n2 l2 ][l2 m2 ]
gl 1 + l(l + 1) f
(2l 1)(2l + 1)

2 gl +

v
u
u
t

[(n 1)2 l2 ][l2 m2 ]


gl+1 = 0
(1.47)
(2l + 3)(2l + 1)
Comparison with Eq. (1.24) shows that Eq. (1.47) is exactly what
one would get when solving the equation
2

(^l2 + 2 K^ z ) = (f + 2 )
by expanding

(1.48)

in the polar basis:


=

nX1
l=jmj

gl jn; l; mi

Now the equivalence of the two problems is clear: to nd spheroidal


orbitals means to nd eigenfunctions of the operator ^l2 + 2 K^ z ,

34
besides obviously those of H^ and ^lz . The classical observable which
corresponds to the former operator is a constant of the motion in
the case of a two-centre Coulomb potential [45].
We list the eigenvalue equations which completely characterize
the spheroidal basis:
!

1 2 p20 np0
r + 2 r Tn;f;m (r) = 0
n = 1; 2; 3 : : : (1.49)
2
h
i
(^l2 + 2 K^ z ) (f + 2 ) Tn;f;m (r) = 0
(1.50)


^lz m Tn;f;m (r) = 0
jmj = 0; 1; : : : ; n 1(1.51)
So, in order to obtain spheroidal orbitals without solving Eq. (1.44)
we can calculate the coecients of the expansion of the unknown
functions in an appropriate basis, by diagonalizing the matrix of
the operator ^l2 + 2 K^ z in such a basis. Looking at this operator,
one understands how spheroidal orbitals are related to those we
have encountered before. As the distance  tends to zero, does
too, therefore the operator becomes ^l2 , which is associated to the
jn; l; mi basis: as  ! 0 the jn; f; mi set coincides with the jn; l; mi
set. On the other hand, as the distance between the two centers
goes to in nity, the contribution from K^ z becomes dominant, thus
the jn; f; mi coincide with the jn; ; mi. One could work out the
same result by a more complicated study of the asymptotic limits
of the functions jn; f; mi [37]. Considering the eigenvalues (f + 2 ),
what we have just said implies that
lim (f + 2 ) = l(l + 1)
!0

and

lim
!1

f + 2
=
2

(1.52)

35
This con rms the role of spheroidal coordinates as a bridge between
the parabolic and the polar coordinates [17].
Eq. (1.47) allows the determination of the coecients of the
expansion of the basis jn; f; mi by the basis jn; l; mi; we propose
another way to get the jn; f; mi, calculating the coecients of their
expansion in parabolic orbitals jn; ; mi:

Tn;f;m (r) =

n X
jmj 1
= (n jmj 1)

hn;m
f; Un;;m (r)

(1.53)

The coecients hn;m


f; are the elements of the eigenvectors which
diagonalize the matrix of ^l2 + 2 K^ z on the basis jn; ; mi. Such
a matrix consists of blocks of dimension n jmj, where parabolic
functions having the same n and m interact; in fact the matrix
elements of two functions jn; ; mi that di er for n and/or m are
zero for the non-combination theorem, since ^l2 + 2 K^ z commutes
with H^ and ^lz . In the -th row of the matrix, only three elements
are di erent from zero:

hn; ; m j ^l2 + 2 K^ z j n;  2; mi =

1 2
[n (m  + 1)2 ] [n2 (m +  1)]
4
hn; ; m j ^l2 + 2 K^ z j n; ; mi =
i
1h 2
(n 1) + (m2 2 ) + 2 
2
hn; ; m j ^l2 + 2 K^ z j n;  +2; mi =
1q 2
[n (m +  + 1)2 ] [n2 (m  1)2 ]
4

(1.54)

(1.55)
(1.56)

36
Thus the coecients in Eq. (1.53) satisfy a three terms recurrence:
1q 2
[n (m  + 1)2 ] [n2 (m +  1)] hn;m
f; 2 +
4



1 2
1
(n 1) + (m2 2 ) + 2  (f + 2 ) hn;m
f; +
2
2
1q 2
[n (m +  + 1)2 ] [n2 (m  1)2 ] hn;m
f;+2 = 0 (1.57)
4

For ! 0, Eq. (1.57) is seen to tend to the recurrence (1.25) in 


of the coecients cn;m
l; , because the matrix of the eigenvectors tends
to the trasformation matrix between jn; l; mi and jn; ; mi; in the
opposite limit the matrix elements in Eqs. (1.54) and (1.56) become
negligible with respect to the one in Eq. (1.55), since the latter
contains ; i.e., the basis jn; ; mi tends to diagonalize ^l2 + 2 K^ z
when gets larger and larger.
The operator ^l2 +2 K^ z can be seen, in the neighbourhood of its
asymptots, as a sum of a main part and a small perturbation: for
small the perturbation is 2 K^ z , and spheroidal functions spring
from the e ect of such perturbation on the zero-approximation
functions jn; l; mi ; on the other hand, when is large, the perturbation is ^l2 , and the jn; f; mi di ers very little from the zeroapproximation functions jn; ; mi. Then we are able to use a perturbation approach at both of the extremes of our problem. For
example, for small the second-order formula for the eigenvalues
stemming from the perturbation of a level jn; l; mi is, by applying
the Rayleigh-Schrodinger Perturbation Theory [23]:
(f + 2 )[2]
l =

(1.58)

37
= l(l+1) 4 2

"

1 [(n 1)2 l2 ][l2 m2 ]


2(l + 1) (2l + 3)(2l + 1)

1 [n2 l2 ][l2 m2 ]
2l (2l 1)(2l + 1)

and the corresponding rst-order spheroidal eigenfunction is:


v
u
u
t

[(n 1)2 l2 ][l2 m2 ]



un;l+1;m+
(l + 1)
(2l + 3)(2l + 1)
(1.59)
v
u 2
u [n l2 ][l2 m2 ]
u
+ t
l (2l 1)(2l + 1) n;l 1;m

(Tn;f;m )[1]
l = un;l;m

In the case of large , spheroidal eigenstates come from corrections


on the parabolic eigenstates; the eigenvalues are found to be
(f + 2 )[2]
 =
1
2  + (n2
2

1
1) + (m2
2

2 ) +

 2
(n
8

(1.60)

2 + m2

1)

while the rst order eigenfunctions are


(Tn;f;m )[1]
 =
1q 2
= Un;;m
[n (m +  + 1)2 ] [n2 (m  1)2 ] Un;+2;m
8
1q
+ [n2 (m  + 1)2 ] [n2 (m +  1)2 ] Un; 2;m (1.61)
8
Fig. 2 illustrates an application of these formulas to the approximation of the eigenvalues f for the case n = 5; m = 1. It is seen that
increasing the quantum number f , the approximation improves on
both extremes, encouraging its use for Rydberg states.

(f+ )/(2+n)

38

-2

-2

/4

arctan

Figure 1.2:

Correlation diagram according to second order perturbation


theory at both sides for the case n = 5; m = 1. The continuous lines show
^ z . The
the dependence on of the eigenvalues f of the operator ^l2 + 2 K
approximations t o f calculated by perturbation theory are shown by dashed
lines.

/2

39

1.5 Spheroelliptic Sturmians


Besides the familiar spherical harmonics Yl;m (#; '), there exists a
di erent set of functions which solves the Schrodinger equation for
free motion on the S 2 sphere. We will call these functions elliptic harmonics since, as we will see soon, they arise from the
parametrization of the sphere by elliptic coordinates (referred to
also as "conical", e.g. in Ref. [20]) and Jacobi elliptic functions [19]
appear in their study. From a group-theoretical point of view,
there are two di erent function sets which induce irreducible representations of the continuous group of three-dimensional rotations
O(3) [1]. The close relationship between the hydrogen atom theory and the sphere S 2 (see the beginning of Section 2) implies that
elliptic harmonics can be used for building up a new type of hydrogen orbitals, which we will call spheroelliptic according to Ref. [2].
They are given as the product of the radial part of Eq. (1.7) and an
elliptic harmonic. There exist alternative equivalent parametrizations which render the motion on S 2 separable, and lead to elliptic
harmonics. One of them is [1]:

x = sn( ; k) dn( ; k0); y = cn( ; k) cn( ; k0 ); z = dn( ; k) sn( ; k0 )


k2 + k02 = 1;
0  k  1;
0  k0  1
(1.62)
We see that Jacobi elliptic functions cn,sn and dn [19] appear. They
contain a continuous parameter (k or k0 ), called modulus, which can
assume any value between zero and one. Useful is consideration of
the modular angle , de ned as sin = k [46] which can be put into

40
relationship with the distance between the "foci" of the ellipses (see
Fig. 3).
The parametrization in Eq. (1.62) leads to the elliptic harmonics:

Yl;k ( ; ) = Ml;k ( ) Nl;k ( )

(1.63)

k ( ) and N k ( ) solve Lam


where Ml;
e equations in Jacobian form
l;
[1]; l and  are the quantum numbers which characterize these
k ( ; ) depends parametrically on k , so it has
harmonics. The Yl;
one degree of freedom more than the spherical harmonics.
The discussion in the previous Section demonstrates that it is
not necessary to solve explicitly the Schrodinger equation in order
to obtain the basis corresponding to a given parametrization: one
can calculate the coecients of the expansion of the unknown wavefunction on a suitable known basis. To do that, one has to know
the set of operators which characterize the unknown basis; for the
k ) [1]:
elliptic harmonics, they are (jl;  i = Yl;

^l2

l(l + 1) jl;  i = 0


E^  jl;  i = 0

(1.64)

E^ = ^lx2 + k02 ^ly2

(1.65)

It is natural to expand the harmonics jl;  i in terms of spherical


harmonics jl; mi, because they are solutions for the same physical
problem, and are both eigenfunctions of ^l2 . In the transformation
between the two basis sets the quantum number l is conserved,
thus the transformation matrix is factorized in blocks of dimension

41

k =1.00

= /2

y
x
z

k =0.866

= /3

y
x
z

k =0.707

= /4

y
x
z

k =0.500

= /6

y
x
z

k =0.00

= 0.00

Figure 1.3:

y
x

Representation of spheroelliptic coordinates and orbitals. Left


panels: various values for the modulus k and the corresponding modular angle
(Sec. 5) are given. In each closed curve on the surface of a sphere, one elliptic
coordinate is xed and the other coordinate is allowed to vary in its interval
of de nition. At the limiting values of the modulus, spheroelliptic coordinates
are seen to coincide with the polar ones, as we expected. It is also seen that
the modular angle is in relationship with the distance between the foci of
the ellipses. On the right panels, the spheroelliptic Sturmian jl = 2; 1 ;  =
+1;  = +1i is represented for di erent values of k . One can see the gradual
passage from the dz2 to the dy2 z2 orbital.

42
2l + 1:

jl; i =

l
X
m= l

xl;m jl; mi

(1.66)

By computing explicitly the matrix elements of the operator E^ on


the basis jl; mi, one sees that the transformation coecients obey
the following three-term recurrence [1]:

Am 2 xl;m 2 + (Bm

 ) xl;m + Am xl;m+2 = 0

(1.67)

where
n
o1=2
1
(1 k02 ) (l m)[l2 (m + 1)2 ](l + m + 2)
4
h
i
1
= (1 + k02 ) l(l + 1) m2
2

Am =
Bm

Eq. (1.67) has two important symmetry properties. The rst


one is that it relates coecients for which m has the same parity.
In other words Eq. 1.67 corresponds to a matrix factorized in two
blocks corresponding to even and odd m, thus some jl;  i are linear
combinations of jl; mi with even m, while others are linear combinations of jl; mi with odd m. Along a line already pursued in Secs.
2 and 3, we also exploit a second less evident symmetry property,
namely that the matrix is centrosymmetric, so it can be further
factorized in two blocks [36]; this amounts to changing the basis
set from jl; mi to jl; m;  i = p12 (jl; mi  jl; mi) [compare with the
angular part of eq. (1.12)]. All this means that the jl;  i can be
labeled by two further quantum numbers ( = 1 and  = 1),
respectively associated with the parity of m ( = 1) and with the

43
relationship between xl;m and xl; m : xl;m =  xl; m . From now
on, we will represent the elliptic harmonics by the ket jl; ; ;  i.
Indicating the re ection operators with respect to the planes
yz and xz as X^ and Y^ , one has X^ jl; mi = jl; mi and Y^ jl; mi =
( )m jl; mi. As a consequence, one has: X^ jl; ; ;  i =  jl; ; ;  i
and X^ Y^ jl; ; ;  i =  jl; ; ;  i. The value of the parameter k (or
k0 ) a ects  , not the other quantum numbers, therefore the symmetry with respect to the re ection planes is not altered by varying
k (or k0 ). The spheroelliptic orbitals induce irreducible representations of the group D2h [47] [ = ( 1)l ]:

C^2 (x) jl; ; ;  i =  jl; ; ; si


C^2 (y ) jl; ; ;  i =  jl; ; ; si
C^2 (z ) jl; ; ;  i =  jl; ; ; si
P^ jl; ; ;  i =  jl; ; ; si

(1.68)
(1.69)
(1.70)
(1.71)

When k0 tends to its limiting values, zero or one, the elliptic


harmonics coincide with simpler functions. As k0 ! 0, one has
E^ ! ^lx2 , then  has to tend to m002 : the jl; ; ;  i coincide with the
linear combination of spherical harmonics p12 (jl; m00 ijl; m00 i) for
which the quantization axis is x. Thus the matrix elements of the
transformation between the latter functions and the jl; mi ful ll a
recurrence which is the limit of Eq. (1.67) as k0 ! 0; in this limit,
the eigenvalue becomes m002 and the eigenvectors can be identi ed

44
with suitable linear combinations of Wigner rotation functions [25]:
i
o1=2
1 nh 2
l (m 1)2 (l m + 2)(l + m)

4

 
 
Dml 2;m00 (0; ; )  Dml 2; m00 (0; ; ) +
2 2
2 2
i
1h
2
l(l + 1) m
m002 
2

  
 
l
l
Dm;m00 (0; ; )  Dm; m00 (0; ; ) +
2 2
2 2
i
o1=2
1 nh 2
l (m + 1)2 (l m)(l + m + 2)

4

 
 
Dml +2;m00 (0; ; )  Dml +2; m00 (0; ; ) = 0
2 2
2 2

(1.72)

On the other hand, as k0 ! 1, E^ ! ^lx2 + ^ly2 , so that by linearly


combining E^ and ^l2 one sees that the jl; ; ;  i become eigenfunctions of ^lz2 ; they coincide with spherical harmonics quantized on the
z axis, namely jl; m;  i = p12 (jl; mi  jl; mi). Indeed, in this limit
the matrix of E^ on the basis jl; mi is diagonal.
Looking at the two limits of the operator E^ , one might try to
associate it to a rotation of the quantization axis by =2. However,
the operator describing the rotation by an angle is sin ^lx +cos ^lz .
Its matrix in the basis jl; mi is tridiagonal, and therefore the elements of the matrix that rotates the jl; mi by radians from the z
axis towards to x axis enjoy a three-term recurrence:
q
1
sin (l + m)(l
2

m + 1) dlm 1;m00 ( ) + (m cos m00 ) dlm;m00 ( )


q
1
+ sin (l m)(l + m + 1) dlm+1;m00 ( ) =(1.73)
0
2

45
where we have explicitly made the identi cation with Wigner reduced d-matrix elements for which eq. (1.73) is a well-known property [25]. Comparing Eqs. (1.73) and (1.67) we can see that they
are not equivalent, and consequently one cannot associate a variation of k0 with a rotation of the quantization axis. To con rm this,
we note that for any value of k0 any function jl; ; ;  i maintains
its symmetry with respect to the re ection on xed planes ( and 
are not a ected by variation of k0 ), and this would not be possible if
the variation of k0 were associated to a rotation of the quantization
axis.
The reasonings about the relationship between spherical and elliptic harmonics can be extended to the corresponding parametrizations. In other words, the Jacobi elliptic functions must tend - as
the modulus takes limiting values - to the ordinary circular functions. Indeed the function sn( ; k) can be de ned as sn( ; k) 
sin ' where ' is given by [19]:

Z '

1
d
0 (1 k2 sin2  )1=2

(1.74)

!0 '
For k ! 0, the integrand function tends to one, therefore k=
!0 sin . In Figure 3 we show a representation of the
and sn( ; k) k=
spheroelliptic coordinates on the surface of a sphere, for di erent
values of the modulus.
In order to clarify some of the previous concepts, it will be useful
to show an example concerning elliptic harmonics: the case l = 2
will be considered [1]. The elements of the matrix in eq. (1.75) are

46
;k0 = hl = 2; m j^
2
02 l2 jl = 2; mb i, where 2  ma ; mb  2:
Eml=2a ;m
a lx + k ^
y
b

El;k0 =

(1.75)
1

6
02
0
0
(1 + k02 )
0
2 (1 k )
C
B
5
3
0
2
0
2
C
B
0 p 2 (1 + k )
0
(1 k ) p 0
2
C
B
B
6 (1 k02 ) C
02 )
02 ) 6
C
B (1
0
3(1
+
k
0
k
2
2
C
B
3
5
C
B
0
2
0
2
0
(1
k
)
0
(1
+
k
)
0
A
@
2
2
p
6 (1 k02 )
02 )
0
(1
+
k
0
0
2
By appropriately changing the order of rows and columns it is easy
to show that this matrix if made of two blocks, which correspond to
the symmetry classes  = 1. Moreover the matrix is centrosymmetric, thus using as a basis the harmonics jl; m;  i we get a further
factorization [36]:
J 1El;k0 J =
(1.76)
0

(1 + k02 )
0
3(1 k02 )
0
0
B
C
0
2
B
C
0
(4 + k )
0
0
0
B p
C
B
C
0
2
0
2
B
C
3(1
k
)
0
3(1
+
k
)
0
0
B
C
0
2
B
C
0
0
0
(1
+
k
)
0
@
A
0
0
0
0
(1 + 4k02 )
where
0
1
1 0 0
1 0
B
C
B 0 1
C
0
0
1
B
C
p
1 B
J = p2 BB 0 0 2 0 0 CCC
(1.77)
B 0 1
C
0 0 1 A
@
1 0 0 1 0
Therefore the matrix is factorized in four blocks (three 1x1 and one
2x2) corresponding to the four symmetry classes ( = 1;  = 1)
of the elliptic harmonics.

47
Note that one can follow a more straightforward procedure by
directly calculating the matrix of E^ on the basis jl; m;  i, instead of
jl; mi: a recurrence relationship identical to Eq.(1.67) is obtained,
except for the matrix element Bm=1; =1 which now reads
1
4 (1

Bm=1; =1 = 12 (1 + k02 ) [l(l + 1) m2 ] 

(1.78)

k02 ) [l(l + 1) m(m 1)] [l(l + 1) (m 1)(m 2)]

Diagonalizing the matrix in Eq. (1.76), one nds [1]:

1;2 = 2[1 + k02  (k04

k02 + 1)1=2 ]

p
j2; 1;2; +1; +1i = ( 3=2)(1 k0 2)(k04 k02 + 1) 1=4

[2(k04 k02 + 1)1=2  (1 + k02 )] 1=2
"
#
(1 + k02 )  2(k04 k02 + 1)1=2
p
(j2; 2i + j2; 2i)
 j2; 0i +
02
6(1 k )

3 = 4 + k02

4 = 1 + k02
5 = 1 + 4k02

j2; 3; +1; 1i = p12 (j2; 1i + j2; 1i) = dxz


j2; 4; 1; +1i = p12 (j2; 2i j2; 2i) = dxy
j2; 5; 1; 1i = p12 (j2; 1i j2; 1i) = dyz

We indicate the proper identi cation with the d orbitals according


to their standard de nition [48]. The right panel in Fig. 3 refers
to the orbital j2; 1 ; +1; +1i in the above formulas. Varying k0 its
shape varies too, but the symmetries with respect to the planes
are conserved. On the formula for 1 , we see that it tends to 6 as
k02 ! 1, as it should since the corresponding eigenfunction tends
to jl = 2; m2 = 0i, which is the harmonic dz :
2

l(l + 1) m2 = 6 0 = 6

48
. As k02 ! 0, 1 ! 4, the corresponding eigenfunction is indeed a
harmonic dy z (m002 = 4), i.e. the equivalent of dx y , but taking
x instead of z as the quantization axis.
For the case l = 3 (a 7x7 matrix), the  -symmetry (even and
odd m) and the  -symmetry lead to a factorization into four blocks:
one is 1x1 and three are 2x2. For higher l's, three-dimensional or
larger blocks may result from the factorization. In Table 1 we show
the number of functions in each of the four symmetry classes for
the lower values of l.
2

1.6 Conclusion
In this chapter we have resumed the most important features of
polar, parabolic and spheroidal hydrogenic orbitals, especially emphasizing their use as Sturmian basis sets; we have also studied
their behavior with respect to the inversion and in general to the
symmetry operations of the point group D2h . Spheroidal Sturmians do not belong to the D2h point group, while the other types of
Sturmians are found to induce irreducible representations of that
group, or to do that after an easy transformation, as the one in
Eq. (1.38). Within this context we have given [eq. (1.31)] a general expression for the coecients which conserve the parity in the
transformation between the polar and the parabolic sets. Then we
have presented the features of a lesser known set, the spheroelliptic
one.

49

20

15

10

5
1

0.2

0.4

0.6

Figure 1.4:

0.8

0.2

0.4

0.6

0.8

Representation of spheroelliptic coordinates and orbitals. Left


panels: various values for the modulus k and the corresponding modular angle
(Sec. 5) are given. In each closed curve on the surface of a sphere, one elliptic
coordinate is xed and the other coordinate is allowed to vary in its interval
of de nition. At the limiting values of the modulus, spheroelliptic coordinates
are seen to coincide with the polar ones, as we expected. It is also seen that
the modular angle is in relationship with the distance between the foci of
the ellipses. On the right panels, the spheroelliptic Sturmian jl = 2; 1 ;  =
+1;  = +1i is represented for di erent values of k . One can see the gradual
passage from the dz2 to the dy2 z2 orbital.

50

,,
|m >

|lm>

|lm>

|lm>

,
|m >

|m >

|m,, >
|fm>
|lm>

,,
|l >

,
|l >
|lm>
|fm>

|m,>

Figure 1.5:

|l >

|lm>
|fm>

|m >

The connections among polar and parabolic Sturmians are


shown in the upper panel. The quantum number n has not been indicated
explicitely since it is conserved in all transformations. The functions of type
jn; l; mi (polar Sturmians) correspond to eq. (1.7): they can be expressed as
products of a Laguerre polynomial and a spherical harmonic, and are eigen^ ^l2 ; ^lz . The functions of type jn; ; mi (parabolic Sturmians)
functions of H;
correspond to eq. (1.17): they can be expressed as products of two Laguerre
^ K
^ z ; ^lz .
polynomials and an exponential function, and are eigenfunctions of H;
In the lower panel we have inserted also spheroidal and spheroelliptic Sturmians. The functions of type jn; f; mi (spheroidal Sturmians) are described
in Section 4: they can be expressed as products of two spheroidal Coulomb
^ ^l2 + 2 K
^ z ; ^lz . The
functions and an exponential, and are eigenfunctions of H;
functions of type jn; l;  i (spheroelliptic Sturmians) are described in Section 5:
they can be expressed as products of a Laguerre polynomial and two Lame
^ ^l2 ; ^lx2 + k 02 ^ly2 .
polynomials, and are eigenfunctions of H;

51
We have paid a great deal of attention to the relationships and
the recoupling coecients among the alternative basis sets. Thus,
now we are able to show (Fig. 5) a complete connection scheme for
alternative hydrogenic Sturmians in con guration space, including
the orthogonal transformations among them. Each connection line
in these gures corresponds to an orthogonal transformation between two alternative basis sets at the extremes of the line. Dotted
lines are associated to the Wigner rotation matrices [18, 24] which
perform the change of the polar axis:

jn; l; mi =

l
X

 
l
Dm;m
0 (0; ; ) jn; l; m0 i
2 2
m0 = l

Solid lines represent CG coecients linking the polar and parabolic


sets [Eq. (1.22)]. Dashed lines correspond the transformations of
type

jn; f; mi =

nX1

l=jmj

gln;m jn; l; mi

whose coecients ful ll Eq. (1.47). Dashed-dotted lines show the


passage between jn; ; mi and jn; f; mi functions, described by Eqs.
(1.53) and (1.57). Double lines correspond to the transformation
between polar Sturmians, such as jn; l; mi and spheroelliptic Sturmians jn; l; ; ; si = Rn;l (r)jl; ; ;  i, described by Eqs. (1.66) and
(1.67). An orthonormal transformation between any two Coulomb
Sturmian sets can be performed following the corresponding path
on our graphs.
In the next chapter we will extend the study of hydrogen atom

52
in momentum space presented in ref. [49], in order to consider also
elliptic Sturmians and to implement the conservation of parity in
the transformations between any alternative basis sets. We will
compare the results with the ones of the present chapter, nding
analogies and di erences of the treatment of hydrogen atom in the
two reciprocal spaces.

Appendix: symmetrization of three-term


recurrence relationships
Let us consider a secular equation c 1Fc = f in the case that F
is tridiagonal. The elements of the kth eigevector of F ful ll the
system:
8
>
>
>
>
>
<
>
>
>
>
>
:

c1;k (F1;1 fk ) + c2;k F1;2


c1;k F2;1 + c2;k (F2;2 fk ) + c3;k F2;3
...

...

cn 1;k Fn;n 1 + cn;k (Fn;n

=0
=0

fk ) = 0
(1.79)
Therefore the elements of eigenvectors of a tridiagonal matrix enjoy
a three-term recurrence relationship: ci 1;k Fi;i 1 + ci;k (Fi;i fk ) +
ci+1;k Fi;i+1 = 0.
This reasoning can run backward: if we have a transformation
P
between two sets of functions i = k di;k k whose coecients
di;k exhibit a three-term recurrence in k, the passage to the set
i corresponds to the diagonalization of a certain operator whose
matrix calculated on the set k is tridiagonal.

53
The hermiticity requirement for the operators associated to physical observables implies that if the matrix F is real, it has to be
symmetric with respect to the principal diagonal. Therefore if we
have a non-symmetric three term recurrence

ci 1 Xi 1 + ci Yi + ci+1 Zi+1 = 0

(1.80)

we have to symmetrize it in order to associate it to a Hermitian


operator. The procedure that we adopt (e.g. in Sec. 4) is the
following. Let us demand ci = Qi gi , where the gi are the elements
that will enjoy the symmetric three term recurrence relationship.
Substitution in Eq. (1.80) and division by Qi lead to:
!

Q
Q
Xi 1 i 1 gi 1 + Yi gi + Zi+1 i+1 gi+1 = 0
Qi
Qi

(1.81)

Now let us impose the symmetry with respect to the principal diagonal:
Q
Q
Xi 1 i 1 = Zi i
Qi
Qi 1
Consequently we get the ratio Qi 1 =Qi , which is enough to get an
p
p
explicit symmetrized recurrence: Xi 1 Zigi 1 +Yigi + Xi Zi+1 gi+1 =
0. By such a procedure the Qi are determined to within a constant
that can be xed imposing normalization and a convention for an
overall phase factor.

Bibliography
[1] J. Patera and J. Winternitz. J. Math. Phys., 14:1130, 1973.

54
[2] E.G. Kalnins, W. Miller, and P. Winternitz. S.I.A.M. J. Appl.
Math., 30:630, 1976.
[3] H. Shull and P.-O. Lowdin. J. Chem. Phys., 30:617, 1959.
[4] M. Rotenberg. Ann. Phys. (N.Y.), 19:262, 1962.
[5] M. Rotenberg. Adv. At. Mol. Phys., 6:233, 1970.
[6] J. Avery. Hyperspherical Harmonics, Applications in Quantum
Theory. Kluwer Academic, Dordrecht, The Netherlands, 1989.
[7] J. Avery. Hyperspherical Harmonics and Generalized Sturmians. Kluwer Academic Publishers, Dordrecht, The Netherlands, 2000.
[8] V. Aquilanti, S. Cavalli, C. Coletti, D. Di Domenico, and
G. Grossi. Int. Rev. Phys. Chem., 20(4):673, 2001.
[9] V. Aquilanti and J. Avery. Adv. Quantum Chemistry, 39:71,
2001.
[10] W. Miller Jr. Symmetry and Separation of Variables. AddisonWesley Publishing Company, Reading, Massachusetts, 1977.
[11] W. Miller Jr. Lie Theory and Special Functions. Academic
Press, New York and London, 1968.
[12] V. Aquilanti, S. Cavalli, and M.B. Sevryuk. J. Math. Phys.,
34:3351, 1993.

55
[13] V. Aquilanti, S. Cavalli, and M.B. Sevryuk. J. Math. Phys.,
35:556, 1994.
[14] M. Hamermesh. Group Theory and its Application to Physical
Problems. Dover Publications, Inc., New York, 1989.
[15] C.A.Coulson and A.Joseph. Proc. Phys. Soc. London, 90:887,
1967.
[16] V. Aquilanti, S. Cavalli, and C. Coletti. Chem. Phys., 214:1,
1997.
[17] Stella M. Sung and Dudley R. Herschbach. J. Chem. Phys.,
95:7437, 1991.
[18] R. N. Zare. Angular Momentum. Wiley & Sons, New York,
1988.
[19] M. Abramowitz and I.A. Stegun. Handbook of Mathematical
Function. Dover, New York, 1964.
[20] P. M. Morse and H. Feshbach. Methods of Theoretical Physics.
McGraw-Hill, New York, 1953.
[21] V. Aquilanti, S. Cavalli, C. Coletti, and G. Grossi. Chem.
Phys., 209:405, 1996.
[22] B. R. Judd. Angular Momentum Theory for Diatomic
Molecules. Academic Press, New York, 1975.

56
[23] L.D. Landau and E.M. Lifshitz. Quantum Mechanics. Nonrelativistic Theory. Addison-Wesley Publishing Co., Reading
Mass., 1958.
[24] M.E. Rose. Elementary Theory of Angular Momentum. Wiley,
New York, 1957.
[25] D.A. Varshalovich, A.N. Moskalev, and V.K. Khersonskii.
Quantum Theory of Angular Momentum. World Scienti c,
Singapore, 1988.
[26] D. Park. Z. Phys., 36:155, 1960.
[27] T. Regge. Nuovo Cimento, 10:544, 1958.
[28] V. Aquilanti, L. Beneventi, G. Grossi, and F. Vecchiocattivi.
J. Chem. Phys., 89:751, 1988.
[29] V. Aquilanti and G. Grossi. J. Chem. Phys., 73:1165, 1980.
[30] V. Aquilanti, S. Cavalli, and G. Grossi. Z. Phys. D., 36:215,
1996.
[31] V. Aquilanti, S. Cavalli, and G. Grossi. Theor. Chim. Acta,
79:283, 1991.
[32] V. Aquilanti and S. Cavalli. Few-Body Systems. Suppl.6: 573,
1992.
[33] K. Schulten and R.G. Gordon. Comp. Phys. Comm., 11:269,
1976.

57
[34] K. Schulten and R.G. Gordon. J. Math. Phys., 16:1961, 1975.
[35] V. Aquilanti, S. Cavalli, and D. De Fazio. J. Phys. Chem.,
99:15694, 1995.
[36] A.R. Collar. Quart. Journ. Mech. and Applied Math., 15:265,
1962.
[37] C.A.Coulson and P.D.Robinson. Proc. Phys. Soc. London,
71:815, 1958.
[38] M. Aubert, N. Bessis, and G. Bessis. Phys. Rev. A, 10:51,
1974.
[39] M. Aubert, N. Bessis, and G. Bessis. Phys.Rev. A, 10:61, 1974.
[40] P.D. Robinson. Proc. Phys. Soc. London, 71:828, 1958.
[41] T. Levitina and E. J. Brandas. Int. J. Quantum Chem., 65:601,
1997.
[42] T. Levitina, E. J. Brandas, and B. Larsson. Int. J. Quantum
Chem., 85(4-5):392, 2001.
[43] I. V. Komarov, L. I. Ponomarev, and S. Yu. Slavyanov.
Spheroidal and Coulomb Spheroidal Functions. Nauka,
Moskow, 1976.
[44] I.S. Gradshteyn and I.M. Ryzhik. Table of Integrals, Series
and Products. Academic Press, New York, 1980.

58
[45] C.A. Coulson and A. Joseph. Int. J. Quantum Chem., 1:337,
1967.
[46] E. T. Whittaker and G. N. Watson. A Course of Modern
Analysis. Cambridge University Press, London, 1935.
[47] J. Patera and P. Winternitz. Journal of Chemical Physics,
65:2725, 1976.
[48] L. Pauling and E.B. Wilson Jr. Introduction to Quantum Mechanics. McGraw-Hill, New York and London, 1935.
[49] V. Aquilanti, S. Cavalli, and C. Coletti. Phys. Rev. Lett.,
80:3209, 1998.

Chapter 2
Hydrogenic Orbitals in Momentum
Space and Hyperspherical Harmonics.
Elliptic Sturmian Basis Sets.
Summary

Momentum space hydrogenic orbitals can be regarded as orthonormal and complete Sturmian basis sets, and can be explicitly given in terms of (hyper)-spherical harmonics on the fourdimensional hypersphere S 3 . Among the alternative coordinate systems which allow separation of variables, the usual ones involving
parametrizations of the sphere S 3 by circular functions correspond
to canonical subgroup reduction chains; we also investigate harmonic "elliptic" sets (as e.g. obtained by parametrizations in terms
of Jacobi elliptic functions). In this chapter we list the canonical
hydrogenic Sturmian sets and the orthogonal transformations connecting them. The latter enjoy very useful three-term recurrence
59

60
relationships which allow their ecient calculations even for large
strings. We also consider modi cations needed when the conservation of the symmetry of Sturmians with respect to parity. Finally
we discuss some properties of elliptic hydrogenic Sturmians and
their relations with canonical Sturmians. Since elliptic Sturmians
cannot be expressed in closed form, it is important to nd expansions in a suitable basis set and to calculate the transformation
coecients. We derive three-term recursion relationships ful lled
by the coecients of the transformation between elliptic Sturmians
and canonical Sturmians. A concluding discussion on the connections between con guration space and momentum space hydrogenic
Sturmians completes this chapter.

2.1 Introduction
The investigation of the non-relativistic Schrodinger equation for
hydrogenic atoms in momentum space has great relevance not only
for the obvious importance of the physical problem, but also because it allows the construction of basis sets enjoying di erent symmetry properties useful in atomic and molecular structure calculations and for the description of atoms in elds.
Hydrogenic orbitals in momentum space, rst obtained by Podolski and Pauling [1] by direct Fourier transform, can indeed be identi ed, within a normalization factor, with hyperspherical harmonics of a 4-dimensional sphere, as shown by Fock [2] through his

61
famous stereographic projection. Modernly they can be referred
to as momentum space Sturmians [3{7], being the counterpart of
the so called "natural spin" [8] or con guration space Sturmian
orbitals [9, 10], which are nding increasing applications in atomic
and molecular quantum chemistry. For recent reviews, see Refs. [11]
and [12]. The reciprocity existing between con guration and momentum space has to be applied with care when trying to extend all
the symmetry properties which can be found in momentum space,
to the corresponding con guration space Sturmians. The symmetry
and completeness properties of these sets make them in fact adapt
to solve quantum mechanical problems where the hyperspherical
symmetry of the kinetic energy operator is broken by the interaction potential, but the corresponding perturbation matrix elements
can be worked out explicitly, as in the case of Coulomb interactions.
In this chapter, after explaining the connection between the
hydrogen atom and the four-dimensional hypersphere (section 2),
we review and update the classi cation of hydrogenic Sturmians
in momentum space given in ref. [13] (section 3), requesting the
conservation of parity (section 4). In addition elliptic Sturmians in
momentum space are introduced (section V), in which the modulus
of Jacobi elliptic functions can be seen as an additional `degree of
freedom', making them more exible with respect to the canonical
Sturmians. As also stressed in the concluding Sec. 6, the aim of
this chapter has been to show that they are less `intractable' than
previously believed [14].

62

2.2 Background
In 1935 Fock obtained the wavefunctions of hydrogen atom (p)
in momentum space, by solving the Schrodinger Equation related
by Fourier transform to the one in con guration space:
1 Z
(p2 + p20 ) (p) = 2


(p0 )

jp p0j2 dp

(2.1)

where p0 corresponds to the energy E0 = p20 =2 (atomic units will


be used throughout this chapter). The wavefunctions (p) can be
related to 4-dimensional spherical harmonics, i.e. eigenfunctions
of the Laplace operator on S 3 . The momentum p with Cartesian
coordinates:
px = p sin # cos '

py = p sin # sin '

(2.2)

pz = p cos #
where p is the modulus of p and # and ' its polar angles, is projected onto a 4-dimensional hypersphere of unit radius:
2p p
x = 2 0 x2 = sin  sin # cos '
p0 + p
2p p
y = 2 0 y2 = sin  sin # sin '
p0 + p
2p p
z = 2 0 z2 = sin  cos #
p0 + p
p2 p2
w = 02 2 = cos 
p0 + p

(2.3)

63
where 0     . Eq. 2.3 exhibits the relationship between the
Euclidean space (px ; py ; pz ) and the S 3 surface (x2 + y 2 + z 2 + w2 =
1), and also gives an explicit parametrization in polar coordinates
(; #; ') of the S 3 hypersphere. Fig. 1 illustrates Fock's stereographic projection, but in order to make it readable we have drawn
it considering one dimension lower, i.e. projecting a 2-dimensional
plane on the 3-dimensional sphere S 2 , instead of the 3-dimensional
space on the 4-dimensional hypersphere S 3 .
The eigenfunction for the hydrogen atom in momentum space
is [6, 7, 12]:
n;l;m (p) =

4p50=2
 Y (; #; ')
(p20 + p2 )2 n;l;m

(2.4)

where Yn;l;m(; #; ') is a four-dimensional hyperspherical harmonic,


the wave function for the free motion of a particle on the S 3 surface. Thus the principal quantum number n (which labels the energy spectrum) can appropriately be interpreted as a hyperangular
momentum quantum number, manifesting that the hidden symmetry giving rise to the accidental degeneracy emerging in the threedimensional con guration space treatment is actually a four dimensional symmetry, which has been analyzed and discussed in various
papers [15], reviews [16] and books [17]. The invariance group for
the hamiltonian of the hydrogen atom is thus O(4), the continuous
group of rotations in the four-dimensional Euclidean space. The
actual symmetry of the hydrogen atom Schrodinger equation has
therefore become apparent only in the momentum space approach:

64

u2

,
u1

u3

p1

p2

Figure 2.1:

The stereographic projection establishes a relationship between


points on a sphere S d 1 and a Euclidean space Rd . The case d = 3 is the
one used by Fock for the hydrogen atom (see Section 2). The familiar "geographic" case d = 2 is sho wn here for illustration: stereographic projection of
a point p in a plane with coordinates p1 = p sin ' and p2 = p cos ' onto the
surface of a sphere having unitary radius with coordinates u1 = sin # sin '; u2 =
sin # cos '; u3 = cos #, where cos # = (p20 p2 )=(p20 + p2 ). The shaded zone represents the projection of the points of the plane onto the part of spherical bowl
comprised between the equator and the axes u1 and u2 . The points ful lling
the condition p = p0 are projected onto the equator.

65
the degeneracy existing for a given n gives rise to n2 Sturmians
which belong to the same irreducible representation of O(4).
The hydrogenic hamiltonian can be expressed as a function of
^
the orbital angular momentum ^l and of the Runge-Lenz vector A
^ [15, 18]:
in its reduced form K
1
^ 2 + 1) 1
H^ = (^l2 + K
2
where

(2.5)


1 ^
^l + np0r =p0
l

p
^
p
^

(2.6)
2
r
^ can be identi ed with the
The Cartesian components of ^l and K
generators of four-dimensional rotations, in virtue of the isomorphism of the hydrogen atom with the sphere S 3 . Let us consider
the Euclidean four-dimensional space: there are six possible distinct
bidimensional planes identi ed by the various couples of Cartesian
axes; by indicating one of those planes as ef , a rotation of an angle
upon it is performed by the operator ei J^ef , where J^ef is one of
the six generators of O(4) [17, 19]:

K^ = A^ =p0 =

@
1
J^ef = ue
i
@uf

@
uf
@ue

(2.7)

Their commutation relations are:




[J^ef ; J^gh ] = i fh J^eg

fg J^eh

eh J^fg + eg J^fh

(2.8)

Eq. 2.8 enables us to associate the generators of O(4) with the


^ , on the basis of their commutation relations
components of ^l and K

66
[19]:

^lx = J^23 ^ly = J^31 ^lz = J^12


(2.9)
K^ x = J^14 K^ y = J^24 K^ z = J^34
The generalized Beltrami-Laplace operator for the motion in d dimensions can be written as [19]:
X

@2
1 @ d 1@
= d 1
r
2
@ue r @r
@r

1 X ^2
J
r2 f>e ef

(2.10)

A d-dimensional hyperspherical harmonic solves the eigenvalue equation on the d-dimensional sphere S d 1 parametrized by the set of
angles
d 1 :
X

b>a

J^ab2 Yj; (
d 1 ) = j (j + d 2) Yj; (
d 1 )

(2.11)

where is a set of d 2 labels (quantum numbers). For the case


d = 4 one has a six-term sum:
X

b>a

J^ab2 Yj; (
3 ) = j (j + 2) Yj; (
3 )

(2.12)

where denotes two quantum numbers. Use of eq. (2.9) shows that
^ 2, and therefore
the operator in eq. (2.12) coincides with ^l2 + K
momentum space Sturmians (which, as we have seen, essentially
coincide with four-dimensional harmonics) can be identi ed with
basis sets for irreducible representations of the rotational group
O(4), as suggested by eq. (2.4). The hyperangular momentum
quantum number j = n 1 labels the energy levels.
That of eq. (2.3) is only one of the possible parametrizations
separating the Hamiltonian for the motion on S 3 (for it, = l; m):

67
starting from other suitable parametrizations alternative momentum space orbitals can be worked out. A complete classi cation of
the bases of O(4), including symmetry properties and connections
among them will be given in the following sections.

2.3 The Canonical Bases of O(4)


As mentioned in the previous section, the hyperspherical harmonics
in eq. (2.4) are not the only possible choice for four-dimensional
harmonics. Such a variety of choices is illustrated by the alternative ways of building the four-dimensional S 3 hypersphere starting
from lower dimension (hyper)-spheres. In particular there exist two
basically distinct ways which have been depicted in g. 2: the rst
one ( g. 2a) consists of starting from the S 2 sphere (i.e. the physical three-dimensional sphere) and adding a rotation in the fourth
dimension. The "inverse" of this construction in group theory corresponds to the following chain reduction of O(4) into its subgroups:
O(4)  O(3)  O(2).
An alternative approach ( g. 2b) starts from two S 1 circles,
each described by an angle, belonging to orthogonal R2 spaces and
then forms their product R2  R2 = R4 ; a third angle is needed
to cover S 3 . In group theory this corresponds to the subgroup reduction chain O(4)  O(2)  O(2). The related parametrizations
will be given the nicknames of asymmetric and symmetric for the
rst and second cases, respectively. Fig. 2 illustrates graphically

68

0

S (,,)

O(4)  O(3)  O(2)

x = sin  sin # cos '


y = sin  sin # sin '
z = sin  cos #
w = cos 

m
l
(a)
n-1

Yn;l;m (; #; ')

0    =2

3
S (,,)

O(4)  O(2)  O(2)

x = sin  cos '


y = sin  sin '
z = cos  sin 
w = cos  cos 

z x

n-1

Yn;;m (; ; ')

The two di erent constructions of S 3 sphere corresponding to


the asymmetric (a) and symmetric (b) parametrization. The relationship of
cartesian coordinates with spherical coordinates is obtained by labeling with
angular variables the intersections (nodes) between two tree "branches" and
considering that the branch converging to the node from the left (right) represents the cosine (sine) of the involved angle. Thus, starting from a leaf and
going down to the root of the tree, through the various nodes, we establish
a relation between cartesian coordinates and angles. Each of these angles is
related to rotations in a space whose dimension is equal to the number of the
upper free ends (or leaves), and thus to the corresponding rotation operator,
so that every node can also be labeled by its eigenvalue (quantum number).
Thus, in both types of parametrizations, according to the particular subspaces
involved in t he construction of S 3 we will have alternative basis sets of the
canonical type, i. e. corresponding to di erent subgroup reduction chains (see
also Ref. [43]).

Figure 2.2:

(b)

69
the two schemes for the construction of S 3 , and shows connections
between Cartesian and hyperspherical coordinates and the representation of the corresponding hyperspherical harmonics according
to the tree method. This graphical method, a full account of which
can be found in refs. [20{23], permits a visualization not only for the
connection between coordinates, but also for the rotation operators
for which the harmonics are eigenfunctions.
The following paragraphs will deal with the classi cation and
description of the various sets corresponding to the di erent parametrizations. Obviously the choice of a particular basis set is completely
pointless when the symmetry of the problem is the original S 3
one (as in the case of hydrogenoid atoms), but it becomes crucial when such a symmetry is broken, and yet the perturbation
hamiltonian can still be written in terms of the generators of the
four-dimensional rotation group, as in the case for instance for hydrogenoid atoms in electric and/or magnetic elds. Important in
quantum chemistry are molecular orbitals which describe symmetry
breakings occurring when atoms join to form molecules.
2.3.1

Asymmetric Parametrizations : Spherical


and Zeeman Sets

The harmonics in g. 2a correspond, as sketched before, to a subgroup reduction chain O(4)  O(3)  O(2), where a fourth dimension is added to the Euclidean tridimensional space [x; y; z ] whose
generators are ^lx , ^ly and ^lz , or according to the previous identi -

70
cation, J^12 , J^23 , J^13 (see Eq. 2.9). Explicitly these harmonics can
be written as the product between the usual tridimensional spherical harmonic Yl;m (#; '), eigenfunctions of the quadratic "orbital
angular momentum" operator ^l2 = ^lx2 + ^ly2 + ^lz2 with eigenvalues
l(l + 1), and a properly normalized Gegenbauer polynomial [24] of
the additional variable cos  (Eq. 2.3):

Yn;l;m(; #; ') = Bn;l sinl  Cnl+11 l (cos ) Yl;m (#; ')

(2.13)

where the normalization factor Bn;l is de ned as follows:


"
#
2n(n 1 l)! 1=2
(2l + 2)
Bn;l = l+1
2 (l + 3=2)
(n + l + 1)

(2.14)

The harmonics Yn;l;m(; #; ') are also eigenfunctions of ^lz with eigenvalue m; this corresponds to the choice of the (xy ) plane where the
rotation subgroup O(2) acts and thus of z as the polar axis. Continuing to identify O(3) as acting on the tridimensional space [x; y; z ]
- the fourth dimension being not involved yet - when the (yz ) or
(xz ) planes are chosen for the O(2) subspaces one has the harmonics Yn;l;m0 (; #; '0 ) and Yn;l;m00 (; #; '00 ), eigenfunctions of ^ly (y
polar axis) and ^lx (x polar axis) respectively. Interchanges among
the polar axes, illustrated for example in g. 3, can be performed
by means of a Wigner rotation matrix:

Yn;l;m0 (; #0 ; '0 ) =

m0

l
Dm;m
0 (0; =2; =2) Yn;l;m (; #; ')

(2.15)

Let us now consider a three-dimensional subspace where the rotation group O(3) acts, and let w be one of the axes: two of the

71

(a)

n-1

m
Wigner rotation matrix

n-1

Figure 2.3:

Representation of the change of quantization axis for the spherical basis sets from m (z axis) to m0 (y axis), corresponding to Eq. (2.15).

generators involved in the tridimensional rotation operator are now


^ according to eq. (2.9). For example
components of the vector K
the basis [(xy )w]z described by tree (b) in g. 4 is diagonal with
respect to the operators K^ x2 + K^ y2 + ^lz2 and ^lz . Let the eigenvalue
corresponding to the rst operator be now denoted as ( + 1),
Labarthe's quantum number  having the same range as l. This
basis set, which we call as a Zeeman basis, has been recently introduced [25{27]. It is useful for the treatment of the diamagnetism in
hydrogen atom and for the description of the behavior of Rydberg
atoms in magnetic elds. Note that, analogously to what has been
just shown for the spherical basis one can choose K^ y or K^ x as diagonal operators thus leading to Yn;;0 (0 ; 0 ; 0 ) and Yn;;00 (0 ; 00 ; 00 )

(b)

72
harmonics, where 0 and 00 denote the corresponding eigenvalues.
The connection between these bases and the Zeeman basis (compare
with the illustration in g. 3 of the case of the spherical basis) is
given by a Wigner rotation matrix, similar to eq. (2.15). Beside the
Zeeman basis set (and those basis sets with a di erent projection
axis, i.e. Yn;;0 (0 ; 0 ; 0) and Yn;;00 (0 ; 00 ; 00)), there are six other
bases of the same kind which can be identi ed: three of them are
eigenfunctions of the quadratic operator K^ x2 + ^ly2 + K^ z2 (with eigenvalue 0 (0 +1)) and of K^ x , ^ly and K^ z respectively (with eigenvalues
00 , m0 and ), while the last three bases are diagonalized by the
quadratic operator ^lx2 + K^ y2 + K^ z2 and by the operator ^lx , K^ y and
K^ z respectively (with eigenvalues m00 , 0 and ). A summary of the
twelve bases of O(4) corresponding to asymmetric parametrizations
in given in Table 1.
Orthogonal transformations which lead to a change in the tridimensional angular momentum quantum number (transformations
among l; ; 0 ; 00 ) can be performed by means of the Z matrices introduced in [13] (see also [28]). For example the orthogonal transformation between theP spherical (eq. 2.4) and the Zeeman basis
n;m
sets [Yn;l;m(; #; ') =  Zl;
Yn;;m(0 ; ; ')] is depicted in g. 4.
In ref. [13] it was shown that the harmonic superposition between
the two bases, which had previously been de ned indirectly through
the steps: Spherical basis ! Stark basis ! Zeeman basis and calculated as a sum on two vector coupling coecients [25,29] (see Eq.
2.22), can be written as a single sum of the Racah type:
n;m = ( )l+ [C (l)C ()]
Zl;
X

1
2

(2.16)

( )r ( m+l+2p(l)+1 + r)( n m p(l2) p() 1 r)! ( m++2p()+1 + r)


r!( l m2 p(l) r)!(  m2 p() r)!( n+m+p(l2)+p()+1 + r)!(m + r)!

73

Table 1 - The 12 bases corresponding to the reduction of the group


O(4) to the subgroup chain O(4)  [O(3)]i;j;k  [O(2)]i;j . The
bases are classi ed according to the diagonal operator acting on
the hyperplane [i; j; k] and to that acting on the plane (i; j ).

d = 3-subspace

d = 2-subspace
(x; y ) ^lz
(y; z ) ^lx
(z; x) ^ly

basis set

[x; y; w]: K^ x2 + K^ y2 + ^lz2 (Zeeman sets)

(x; y ) ^lz
(y; w) K^ y
(x; w) K^ x

jn; ; mi
jn; ; 0i
jn; ; 00i

[x; z; w]: K^ x2 + K^ z2 + ^ly2

(z; x) ^ly
(z; w) K^ z
(x; w) K^ x

jn; 0; m0 i
jn; 0; i
jn; 0; 00i

[y; z; w]: K^ y2 + K^ x2 + ^lx2

(y; z ) ^lx
(y; w) K^ z
(y; w) K^ y

jn; 00; m00 i


jn; 00; i
jn; 00; 0i

[x; y; z ]: ^lx2 + ^ly2 + ^lz2 (spherical sets)

jn; l; mi
jn; l; m00i
jn; l; m0i

74

(a)

/2

Z matrix

n-1

(b)

n-1

Figure 2.4:

The transformation from the spherical to the Zeeman basis set,


which implies the change of the tridimensional quantum number from l to .

where
( n+2k 1 + p(k))!( k+2m q (k))!( k 2m q (k))! ( n 2 k + p(k))
C (k)= n+k+2
( 2 + p(k)) ( m+2k+1 + q (k)) ( k m2 +1 + q (k))( n 2k 1 p(k))!
n;m
is zero
p(k) = 1+( 41) and q (k) = ( 1) 4 1 . Note that Zl;
n;m
n; m
when n + l +  + m is even and shows the symmetries Zl;
= Zl;
n;m
n;m
and Zl;
= Z;l
. The sum in 2.16 [28] is a hypergeometric function
4 F3 of unit argument and can be connected with Racah polynomials
[30, 31], although it cannot be reduced to the ordinary Racah's
or 6-j coecient which performs angular momentum recoupling.
Indeed, like a Racah's recoupling coecient it is orthogonal with
respect to summation on two angular momentum quantum numbers
(l and ), but contains a projection quantum number. As shown
n;m
in Ref. [13], the Z;l
coecient can be compactly written as a 6-j
symbol extended to allow not only multiples of 1/2, as ordinary
k n

k m

75
vector recoupling coecients, but also multiples of 1/4:
n;m = (
Zl;

l+
2

+1+p(l)+p() (l + 1 )( + 1 )
2
2

The symbol




1
4
1
4

p()
p(l)

n 1
2
m 1


2
l

1
4
1
4

(2.17)

i enjoys most properties of ordinary 6-j symbols

such as several recurrence relationships, but some simmetries fail,


so caution is needed in its use. Additional properties and relations
involving the Z matrix will be given in subsection 3.3.
2.3.2

Symmetric Parametrizations

We will deal now with the symmetric parametrizations, corresponding to the subgroup reduction chain O(4)  O(2)  O(2). The best
known of these basis sets is that corresponding through a Fourier
transform to the Sturmian orbitals written in parabolic coordinates
in con guration space (see chapter 1). Due to its importance in the
treatment of the hydrogen atom in an electric eld, we call the corresponding harmonic set as the Stark basis. For its use for building
molecular orbitals and multidimensional expansions, see [3,4,12,32].
This parametrization is illustrated in g. 2b and can be written as

x = sin  cos ' y = sin  sin '


z = cos  sin  w = cos  cos 

(2.18)

This coordinate set is referred to as cylindrical in Ref. [15]. The


corresponding hyperspherical harmonics

4p50=2
n 
n
1
m
n;;m (p) = 2 2 2 ( ) ({)
D  (nm ;1) m (  '; 2; ' )
2
(p + p0 )
2
(2.19)
1
2

1
2

76
simultaneously diagonalize ^lz and K^ z (with eigenvalues m and ),
where we give to the z axis the role of a privileged direction in
physical space. This explains why the Hamiltonian for a hydrogen
atom in a weak electric eld directed along the z axis is still diagonal
in this basis. Similarly other two basis sets of this kind can be
constructed when the x and y axes are taken as quantization axes
(see Table 2). Thus we can have a basis diagonalizing ^lx and K^ x ,
and another one diagonalizing ^ly and K^ y , leading to a total of three
di erent basis sets corresponding to symmetric parametrizations.
The orthogonal transformation between spherical ( g. 2a) and
Stark basis ( g. 2b), analogously to what is found in con guration
space (see Chapter 1), is given by a Clebsch-Gordan coecient [33]:

cn;m
l;  n;l;m (p)

(2.20)

1
m  1
m+
; (n 1);
jl; mi
cn;m
l; = h 2 (n 1);
2 2
2

(2.21)

n;;m (p) =

( )

1+m
2

From this formula, by interchange of the quantum number l with 


we also have the connection between Zeeman and Stark basis sets.
Similarly to the treatment in con guration space, also in this case
it will be of importance to explicitly introduce parity conservation.
This will be done in Section 4.
2.3.3

The Scheme of the Orthogonal Transformations among the Canonical Bases of

O(4)

We have just shown that the number of distinct canonical basis sets
for S 3 amounts to 15. Actually the total number of solutions to

77
Laplace equation on S 3 (harmonics) is 120, however due to the fact
that for the identi cation of a plane or of a 3-dimensional space it is
only necessary to specify the involved axes and not their sequence,
to di erent systems of hyperspherical coordinates, identifying the
same plane, there correspond the same harmonics (or harmonics
di ering for a phase factor only, i.e. still eigenfunctions of the
same rotation operator), so the number of di erent bases reduces
to 15. Ref. [13,34] provides a discussion on the classi cation of the
120 coordinate systems and of how this reduction is achieved. The
15 basis sets can be depicted together with their connections by
the graph in Fig. 5. This is a projective plane RP 2 representation,
which in turn can be obtained from a graph, associated to the
icosahedral group, illustrating the 120 coordinate systems (see ref.
[13]). The connections between di erent basis sets, which are placed
at the vertices of 10 triangles and 6 pentagons, are identi ed by
various types of segments: (i) vector coupling coecients (eq. 2.20)
are drawn by solid lines, (ii) rotation matrix elements (eq. 2.15) by
dotted lines and (iii) Z coecients (eq. 2.17) by zig-zag lines. Fig. 5
is also an "abacus" to obtain relationships among ordinary elements
of angular momentum algebra, augmented by the Z coecients:
following the sides of the plane gures one can write interesting
sum rules. For example following the side of the triangle where m
n;m
is conserved Zl;
can be written as a sum involving two ClebschGordan coecients
n;m =
Zl;
X

( ) h

(2.22)

n 1+m  n 1 m 
1+m  n 1 m 
;
;
; jl; 0ih
;
;
; j; 0i
2
2
2
2
2
2
2
2

(the phase = n 1 2m  + l +  is an integer: if it were omitted, the sum would be l ). We recognize on the right hand side
the transformation coecient used for the de nition of the Zeeman
basis [25]. An alternative sum rule can be obtained from g. 5,

78
following the sides of a pentagon:
 n;m X
0

dlm;m0 ( )Zl;
= ( 1)m+ dm;0 ( )
2
2

(2.23)

0

0
0
h n 12+ m ; 2 ; n 12

m0 0
n
; ; l; 0ih

1 + 0 m0 n 1 0 m0
;
;
; ; ; 0i
2
2
2
2

belongs to the family of well-known \pentagonal" relationships of


angular momentum theory [35], such as the Biedenharn and Elliott
identity (but also the Clebsch-Gordan series itself).
Other sum rules can be derived from g. 5 (some of them are
well known, as the addition formula for rotation matrices, obtained
following the sides of the triangle where l is conserved). Besides
those obtained by the graph in g. 5 the Z coecient enjoys other
interesting relations; here three-term recurrence relationships will
be given where only one of the quantum numbers is varied. Explicitly we have a three term recurrence relationship involving n:
n + 1q

[(n 1)2

l2 ][(n

1)2

2 ][(n

(l + 1)2 ]

1)2

n 2;m +
( + 1)2 ] Zl;

[(n 1)2

2 l + l + 2 l l2 + l2 + 2 l2 +
n;m
n2 (1 +  + 2 + l + l2 2m2 n2 ) + 2m2 ] Zl;
+
n[ 
n
q

[(n + 1)2

1q

[(n + 1)2

l2 ][(n + 1)2

(l + 1)2 ][(n + 1)2

2 ] 

(2.24)

n+2;m = 0
( + 1)2 ] Zl;

or involving l (which is completely symmetric to the relation involving ):


v
u

2
2l + 3 u
t (l
2

m2 )[(l

1)2

m2 ](n2 l2 )[n2
(2l 3)

(l

1)2 ]

Zln;m2;

79

000
111
000000000
111111111
000
111
000000000
111111111
000
111
00000000
11111111
00000000
11111111
000000000
111111111
000
111
11111111
00000000
00000000
11111111
000000000
111111111
00000000
11111111
000000000
111111111
00000000
11111111
000000000
111111111
00000000
11111111
000000
111111
000000000
111111111
00000000
11111111
000000000
111111111
00000000
11111111
000000
111111
000000000
111111111
00000000
11111111
00000000
11111111
000000
111111
000000000
111111111
00000000
11111111
00000000
11111111
000000
111111
000000000
111111111
00000000
11111111
00000000
11111111
000000
111111
000000000
111111111
00000000
11111111
000000000
111111111
000000
111111
000000000
111111111
00000000
11111111
000000000
111111111
000000
111111
000000000
111111111
00000
11111
00000000
11111111
000000000
111111111
000000
111111
000000000
111111111
0000011111111
11111
00000000
000000000
111111111
000000
111111
000000000
111111111
00000
11111
00000000
11111111
000000000
111111111
000000
111111
000000000
111111111
0000011111111
11111
00000000
000
111
000000000
111111111
000000
111111
000000000
111111111
0000011111111
11111
00000000
0000
1111
000
111
000000000
111111111
000000
111111
000
111
00000
11111
00000000
11111111
0000
1111
000
111
000000000
111111111
0000000
1111111
000
111
00000
11111
0000
1111
000
111
000000000
111111111
0000000
1111111
000
111
00000
11111
0000
1111
0000000001111
111111111
0000000
1111111
000
111
00000
11111
0000
000000000
111111111
0000000
1111111
000
111
00000
11111
0000
1111
000000000
111111111
0000000
1111111
000
00000
11111
0000111
1111
0000000001111
111111111
0000000
1111111
000
111
00000
11111
0000
000000000
111111111
0000000
1111111
000
111
00000
11111
0000
0000000001111
111111111
0000000
1111111
000
00000
11111
0000111
1111
0000000001111
111111111
0
1
0000000
1111111
000
111
00000
11111
0000
000000000
111111111
0
1
0000000
1111111
000
111
00000
11111
0000111
1111
000000000
111111111
0
1
0000000
1111111
000
00000
11111
0000
1111
0000000001111
111111111
0
1
0000000
1111111
000
111
00000
11111
0000
000000000
111111111
0
1
0000000
1111111
000
111
0000111
1111
000000
111111
000
111
000000000
111111111
0000000000
1111111111
000
000000000000
111111111111
0000
1111
000000
111111
000
111
00000000
11111111
0000000001111
111111111
0000000000
1111111111
000
111
000000000000
111111111111
0000
000000
111111
000
111
00000000
11111111
0000000000
1111111111
000
111
000000000000
111111111111
0000
1111
000000
111111
000
111
00000000
11111111
0000
1111
0000000000
1111111111
000
111
000000000000
111111111111
000000
111111
000
111
00000000
11111111
0000
1111
0000000000
1111111111
000000000000
111111111111
000000
111111
000
111
00000000
11111111
0000
1111
0000000000
1111111111
000000000000
111111111111
000000
111111
000
111
00000000
11111111
0000
1111
0000000000
1111111111
000000000000
111111111111
000000
111111
000
111
00000000
11111111
0000
1111
0000
1111
0000000000
1111111111
000000000000
111111111111
000000
111111
000
111
00000000
11111111
0000
1111
0000
1111
0000000000
1111111111
000000000000
111111111111
000000
111111
000
111
00000000
11111111
0000
1111
0000
1111
0000000000
1111111111
000000000000
111111111111
000000
111111
000
111
00000000
11111111
0000
1111
0000
1111
0000000000
1111111111
000000000000
111111111111
000000
111111
000
111
00000000
11111111
0000
1111
0000
1111
0000000000
1111111111
000000000000
111111111111
000000
111111
000
111
00000000
11111111
0000
1111
0000
1111
0000000000
1111111111
000000000000
111111111111
000000
111111
000
111
00000000
11111111
0000
1111
0000
1111
000000
111111
000000000000
111111111111
000000
111111
00000000
11111111
0000
1111
0000
1111
000000
111111
000000000000
111111111111
000000
111111
000000000
111111111
00000000
11111111
0000
1111
00000000
11111111
0000
1111
000000
111111
000000
111111
000000000
111111111
00000000
11111111
0000
1111
00000000
11111111
0000
1111
000000
111111
000000000
111111111
00000000
11111111
0000
1111
00000000
11111111
0000
1111
000000
111111
000000000
111111111
00000000111111111
11111111
0000
1111
00000000
11111111
0000
1111
000000
111111
000000000
00000000
11111111
0000
1111
00000000
11111111
0000
1111
000000
111111
000000000
111111111
00000000
11111111
00000000
11111111
000
111
000000
111111
000000000
111111111
00000000
11111111
00000000
11111111
000000000
111111111
000
111
000000
111111
000000000
111111111
00000000
11111111
00000000
11111111
000000000
111111111
000
111
000000
111111
000000000
111111111
00000000
11111111
00000000
11111111
000000000
111111111
000
111
000000
111111
000000000
111111111
00000000
11111111
000000000
111111111
000000
111111
000000000
111111111
00000000
11111111
000000000
111111111
000000
111111
000000000
111111111
00000000
11111111
000000000
111111111
000000
111111
000000000
111111111
00000000
11111111
000000000
111111111
000
111
000000
111111
000000000
111111111
00000000
11111111
000000000
111111111
000
111
00000000
11111111
000000000
111111111
000000000
111111111
000
111
00000000
11111111
000000000
111111111
000000000
111111111
00000000
11111111
000000000
111111111
00000000
11111111

| l m>

| >

| m>

| >

| l m>
|l m>
| m>

|m>
|m>

|> |
> |>

|>

|m>

| m>

Figure 2.5:

Orthogonal transformations among orthogonal bases of O(4).


Each basis is labeled by two quantum numbers in the graph, while the quantum
number n has not been inserted explicitly, because it labels all the bases.

80
1

pn + 2

[2 l 3l3 4l3 2l4 + 3 4l 4l2 + 32


2l + 1
4l2 4l2 2 + 2m2 (l2 + l 1) + 2n2 (l2 + l 1 + 2m2 )] Zln;m2;
+

2l

1 [(l + 1)2

m2 ][(l + 2)2

m2 ][n2
2l + 5

(l + 1)2 ][n2

Zln;m
+2; = 0

(l + 2)2 ]
(2.25)

or m:

m + 1q
( m + 2)(l m + 2)[(m 1)2 2 ]
2
q
n;m 2
(l + m)( + m)(l + m)[(m 1)2 l2 ] Zl;
+ m[ 

2

l + l + l2

l2 + l2 + l2 2

n;m
+m2 (1 +  + 2 + l + l2 m2 2n2 ) + 2n2 ] Zl;
m 1q
( m)(l m)[(m + 1)2 2 ]
(2.26)
2
q
 [(m + 1)2 l2 ]( + m + 2)(l + m + 2) Zl;n;m+2 = 0

Note that these relations are extremely useful for calculation of full
strings of coecients.

2.4 Conservation of parity in the transformations among momentum space


Sturmians
An issue of great relevance is the conservation of the parity with
respect to the inversion of electronic coordinates in the transformations among O(4) bases: such a symmetry is present in many

81
physical systems and thus it is important to have Sturmian bases
with de nite parity.
2.4.1

Spherical and Stark Basis Sets

Here we exploit the correspondence with the polar and parabolic


Sturmian orbitals in con guration space (see previous chapter).
Since the momentum space Sturmian in eq. (2.4) is the Fourier
transform of polar orbitals in the con guration space, it must have
the same symmetry properties: thus for the inversion one has

P^ n;l;m(p) = ( 1)l n;l;m(p)

(2.27)

where the operator P^ performs the inversion of the vector p =


(px ; py ; pz ). As we did in the previous chapter for the parabolic
basis in con guration space, because of the complete reciprocity
between the two spaces, Stark Sturmians with the correct parity
n;jj;m (p) can be built by the formula ( = 1 is the parity with
respect to inversion)
n;jj;m(p) = jn; ; m; i =

p1

[ n;;m(p)  ( )m n;

p)]

;m (

(2.28)
and the conservation of parity in the passage to the set n;l;m(p)
involves the parity-conserving Clebsch-Gordan coecients (pCG):

hn; l; m j n; ; m; i = (

1+m
2

Cl;n;m;
jj = ( )

1+m
2

1 + ( )l n;m
cl;jj
2(1 + ;0 )
(2.29)

82
which have been de ned in the previous chapter Eq. (1.31), Sec.
1.3.
2.4.2

Zeeman Basis Sets

As we have seen in the previous sections, there exist nine di erent


Zeeman sets. They do not correspond to any Sturmian orbitals in
con guration space. In order to study their behavior with respect to
the inversion, let us consider the tree in g. 4b; the corresponding
parametrization is:
2p p
x = 2 0 x2 = cos 0 sin  cos '
p0 + p
2p p
(2.30)
y = 2 0 y2 = cos 0 sin  sin '
p0 + p
2p p
z = 2 0 z2 = sin 0
p0 + p
p2 p2
w = 02 2 = cos 0 cos 
p0 + p
As a consequence one has the following relationships among the
coordinates (px ; py ; pz ) in momentum space and the angles (0 ; ; ')
of the Zeeman parametrization of the 4-dimensional sphere S 3 on
which the 3-dimensional momentum space has been projected:
p2 + p2
px = 0
cos 0 sin  cos '
2p0
p2 + p2
cos 0 sin  sin '
(2.31)
py = 0
2p0
p2 + p2
pz = 0
sin 0
2p0

83
The corresponding momentum space Sturmian can be expressed as
(see eq. 2.13):
n;;m (p) =

4p5=2
( )n 1 jmj ({)+m jmj 2 0 2 2 Bn; cos 0 Cn+11  (sin 0 ) Y;m(; ')
(p0 + p )
(2.32)
where Bn; had been de ned in eq. (2.14). Taking into account
that inversion changes the sign of p = (px; py ; pz ) but leaves p2
unchanged, from eqs. (2.30) and (2.31) one sees that such an operation corresponds to the transformation of angles (0 ; ; ') into
( 0 ; ; ' +  ). Considering that

Cn+11  (sin( 0 )) = Cn+11  ( sin 0 ) = ( )n


and that
one has:

 1 C +1 (sin 0 )
n 1 

eim('+) = ( )m eim'
P^ n;;m (p) = ( )n

 1+m

n;;m(p)

(2.33)

Thus we have shown that the basis n;;m (p) is the set of eigenfunctions of the operator which performs the inversion of p, and the
same proof applies to the bases n;0 ;m0 (p) and n;00 ;m00 (p), thus
the three matrices Z which - as in g. 4 - perform the transformation to the functions n;l;m(p), n;l;m0 (p) and n;l;m00 (p) should
be factorized in two blocks, in order to conserve the parity. These
conclusions are consistent with the observation that the matrix eln;m
ements Zl;
are zero when n + l +  + m is even [13, 36].

84
The transformation between Stark and Zeeman basis sets (eq.
2.20 with  instead of l) involves the pCG coecients, which have
been amended by requesting the conservation of parity [compare
with eq. (2.29)]. One has to consider that the Zeeman basis set has
parity ( )n  1+m :

p)

n;jj;m(

n X
jmj 1

( )

jj=0 or 1

C;n;m;
jj =

1+m
2

C;n;m;
jj n;;m (p)

1 +q( )n  1+m n;m


c;
2(1 + ;0 )

(2.34)

We will now consider the transformations that involve rotation


matrices. In Fig. 5 there are four triangles whose three sides are
represented by dotted segments. In one of these, l remains a good
quantum number and so does parity, which is ( 1)l for spherical
bases. For the other three triangles, the situation di ers because of
those vertices which correspond to Zeeman Sturmians that diago^ . Since K^ and P^ do not
nalize one of the Cartesian components of K
commute, those functions cannot be simultaneously eigenfunctions
of P^ . For de niteness, let us consider the Sturmian n;;00 (p); the
corresponding parametrization is:
2p p
x = 2 0 x2 = cos 0 sin 0 cos '00
p0 + p
2p p
y = 2 0 y2 = cos 0 cos 0
p0 + p
2p p
z = 2 0 z2 = sin 0
p0 + p

(2.35)

85

p2 p2
w = 02 2 = cos 0 sin 0 sin '00
p0 + p
thus one can write
p2 + p2
cos 0 sin  cos '00
px = 0
2p0
p2 + p2
py = 0
cos 0 cos 
(2.36)
2p0
p2 + p2
sin 0
pz = 0
2p0
Therefore the inversion of p in the tridimensional space corresponds
to the transformation of angles (0 ; ; '00 ) ! ( 0 ;  ;  '00 ).
The expression of the Sturmian is, apart from a normalization factor
which is not needed for symmetry considerations:
p50=2
 0 C +1 (sin 0 ) P j00 j (cos  ) ei00 '00

cos
n 1 

2
(p0 + p2 )2
(2.37)
Because of the change in the sign of the argument of the exponential
function as a consequence of the inversion, it is evident that the
Sturmian in eq. (2.37) cannot be eigenfunction of P^ . Applying
the projection procedure  (p) = p12 [ n;;00 (p)  n;;00 ( p)] one
can work out:
n;;00 (p) =

n;;j00 j (p) =

p1

n;;00 (p)  ( )n

00 1

p)

n;; 00 (

(2.38)

 = 1, according to the sign of the linear combination. This result


is completely analogous to the one we had in the construction of
de nite parity Stark Sturmians (eq. 2.28): in order to obtain the

86
parity components of a Sturmian which was originally eigenfunction
for one of the K's Cartesian components, one has to combine two
Sturmians having opposite values of  (or 0 or 00 ). Thus, the
orthogonal transformation
n;;m (p) =

00

 (0; =2; =2)


Dm;
00
n;;00 (p)

becomes, requesting the conservation of the inversion symmetry,


n;;m(p) =

00

;

Dm;
j00 j (0; =2; =2) n;;00 (p)

where
;
Dm;
j00 j (0; =2; =2) =

1+
( )n  1+m 
q
Dm;00 (0; =2; =2) (2.39)
2(1 + 00 ;0 )

The form of eq. (2.39) is due to the fact that matrix elements
 00 (0; =2; =2) di ering only in the sign of 00 can only have
Dm;
equal or opposite values.
In summary, the basis sets which have  (or 0 or 00 ) as a
quantum number can be turned into "parity Sturmians" performing
a linear combination with the Sturmians labeled by  (or 0 or
00 ). These "parity Sturmians" will no longer be labeled by 
(or 0 or 00 ), but by jj (or j0 j or j00 j) plus , the eigenvalue
of the inversion operator P^ . As a consequence, matrix elements of
orthogonal transformations where jj (or j0j or j00 j) is conserved as
a quantum number have to be premultiplied by the suitable parity
factor, which appears in eqs. (2.29),(2.34) and (2.39).

87

2.5 The elliptic bases of O(4)


Besides the canonical (i.e. canonical) bases described in previous
sections - deriving from parametrizations of the hypersphere S 3 in
terms of the well known circular function sin and cos - there exist, as is described in Refs. [15, 37], bases of O(4) resulting from
elliptic parametrizations of S 3 . Such parametrizations can be set
up by Jacobi elliptic functions sn(a; k), cn(a; k) and dn(a; k) [24],
and represented by ternary forks (see Fig. 6) [38]. Trigonometric
alternatives are also useful, but a crucial fact is that the value one
gets for such functions depends also on a number, k, called modulus,
that can have any value in the range [0; 1]. Thus, it is understood
that the Sturmians coming from elliptic parametrizations will have
an additional "degree of freedom" (the modulus) with respect to
the canonical Sturmians: the former are more " exible", but their
manipulation is more complicated - for example they can not be
expressed in a closed form - and in fact, as far as we know, they
have never been used for the calculation of properties of atomic and
molecular systems other than the hydrogen atom. In the following,
we will analyze the symmetry properties of elliptic sturmians, in order to clarify their relation with canonical Sturmians and to insert
them in the overall scheme of the alternative hydrogenic Sturmians. Since elliptic Sturmians cannot be expressed in closed form,
one has to expand them in some suitable basis set and calculate
the coecients of the expansion: the three-term recursion relation-

88
ships ful lled by the transformation coecients among canonical
and elliptic Sturmians will be given.

2.5.1

Spheroelliptic Coordinates and Basis sets

The simplest kind of elliptic Sturmians can be derived from the


parametrization of the hypersphere S 3 in spheroelliptic coordinates
(see g.6a):

x = sin  sn( ; k) dn( ; k0 )

y = sin  cn( ; k) cn( ; k0 )

z = sin  dn( ; k) sn( ; k0 )

w = cos 

(2.40)

The relationship between the modulus k and k0 is k2 + k02 = 1.


The constants of motion for the corresponding harmonics are ^l2 +
K^ 2, ^l2 and ^lx2 + k02^ly2 [15, 37]. The momentum space Sturmian set
resulting from the parametrization (2.40) is the Fourier transform of
the corresponding counterpart in con guration space, introduced in
ref. [37] and analyzed in Chapter 1, as it can be seen by comparing
operators and coordinates of the two reciprocal spaces.
We will now deal with elliptic parametrizations, deriving from
elliptic cylindrical coordinates of type I and II, which don't have
a counterpart in con guration space (the answer to the intriguing
question of why some elliptic Sturmians in momentum space do not
have counterparts in con guration space will be given in Section 6).

89

l, (k)

(a)

n-1

(b)

n-1, b(k)

m
,

(c)

n-1, q(k)

Ternary forks for setting up elliptic coordinate systems for S 3


[spheroelliptic (a), elliptic cylindrical of type I (b), elliptic cylindrical of type II
(c)]. Although several alternatives are possible, we found it useful [38], exploiting results from Ref. [37] for S 2 and Ref. [15] for S 3 , to adopt the following
conventions: in the de nitions of coordinates, a left branch corresponds to
sn( ; k )dn( ; k 0 ), a middle branch corresponds to cn( ; k )cn( ; k 0 ), and a right
branch corresponds to dn( ; k )sn( ; k 0 ). The arguments and of the elliptic
functions are the variables; the modulus parameter k [and k 0 = (1 k 2 )1=2 ,
with 0  k; k 0  1] adds a degree of freedom to these coordinate systems.
The labels b; q;  are separation constants depending on k and in general are
to be determined as eigenvalues of secular equations. For S 3 , the most general separable coordinates (the ellipsoidal [15]) would accordingly be put into
correspondence with "quaternary" forks (they depend essentially on two independent parameters which span the projective plane of Fig. 5). The procedure
can be extended to S d 1 for which trees involving forks of up to the d-th order can be de ned (using results from Ref. [44], where di erent but equivalent
graphs were introduced).

Figure 2.6:

90
2.5.2

Elliptic Cylindrical Coordinates of Type I


and II

Elliptic cylindrical coordinates of type I parametrize S 3 as (see g.


6b):

x = sn( ; k) dn( ; k0) cos '


z = sn( ; k) dn( ; k0)

y = sn( ; k) dn( ; k0) sin '


w = cn( ; k) cn( ; k0 )

(2.41)
(

cos m'
The corresponding hyper-harmonics are given as L( )L0( )
sin m'
0
where L( ) and L ( ) are solutions of equations related to the socalled associated Lame equation [15]: these solutions are not available in closed form, but information on a certain basis set and on
its connections with the others can be obtained by studying the
corresponding commuting operators.
By setting F^j = 2i (^lj + K^ j ) and G^ j = 2i (^lj K^ j ) (j = x; y; z )
such operators are [15]:

^ 2 + G^ 2)
C^ = 2(F

(2.42)
!

k2 + 1 ^ ^
^
^
^
^
^
B = Fx Gx + Fy Gy + 2
FG
k 1 z z
R^ = (F^z + G^ z )2

(2.43)
(2.44)

These expressions do not give much information in this form, but


we will obtain equivalent commuting operators which can be more
^ 2 = G^ 2 =
readily interpreted. Considering that l  K = 0, one has F
1 ^2 ^ 2
^ 2
^2
^
4 (l + K ). From eq. (2.44) it follows R = ( {lz ) = lz . So,

91
the meaning of C^ and R^ is clear: they are absolutely equivalent to
^ 2 and ^lz2. Now we have to explain the
the familiar operators ^l2 + K
meaning of the operator B^ . Eq. (2.43) can be rewritten as

B^ = F^x G^ x + F^y G^ y + F^z G^ z +

F^ G^ =
k2 1 z z

1  (^lz + K^ z )(^lz K^ z )
^  G^
=F
k2 1
2
^
^
The product F  G can be expressed as:
^2 ^
F^  G^ = 12 (F^ + G^ )2 12 (F^ 2 + G^ 2) = l2 + C4
Therefore the elliptic cylindrical harmonics of type I diagonalize
the operator
^l2 C^  1  (^lz + K^ z )(^lz K^ z )
+
2 4
k2 1
2
Multiplying by -2 and expanding, the latter is found to be equivalent to
^l2 + 1 2 K^ z2 + 2 1 ^lz2 C^
(2.45)
1 k
k 1
2
Consequently, the basis can be characterized by the set of operators
^l2 + K^ 2, ^l2 +  1 1k  K^ z2 and ^lz2. In the limit k ! 0, the second one
tends to ^lx2 +^ly2 +^lz2 + K^ z2 , which by an appropriate linear combination
with C^ and ^lz2 gives K^ x2 + K^ y2 + ^lz2 : this is tantamount to say that
the elliptic cylindrical harmonics of type I tend, in this limit, to
the Zeeman set Yn;;m . In the opposite limit, the operator tends to
K^ z2 , thus the set tends to that of de nite parity Stark harmonics
Yn; jj;m.
2

92
The elliptic cylindrical coordinates of type II are (see g. 6c):

x = cn( ; k) cn( ; k0 ) cos ';

y = cn( ; k)cn( ; k0 ) sin ('2.46)

z = sn( ; k) dn( ; k0 );

w = dn( ; k) sn( ; k0) (2.47)

The operators which characterize the corresponding set of hyperharmonics are the ones in eqs. (2.42) and (2.44), plus [15]


Q^ = F^x G^ x + F^y G^ y + 1 2k2 F^z G^ z

(2.48)

By some passages similar to the ones we made for the elliptic base
of type I, Q^ can be written as:
^l2 k2 ^ ^ ^ ^ C^
+ (l + Kz )(lz Kz ) +
Q^ =
2 2 z
4
A more perspective equivalent expression for this operator is
^l2 + k2K^ z2 k2^lz2 C^
(2.49)
2
Therefore this basis can be characterized by the set of operators
^l2 + K^ 2, ^l2 + k2K^ z2 and ^lz2. As k ! 0 it goes into the spherical basis,
since ^l2 + k2 K^ z2 ! ^l2 , while as k ! 1 one has the operator ^l2 + K^ z2 ,
which by linear combination with C^ and R^ is seen to correspond to
the operator which characterizes the Zeeman basis.
Now we can insert the elliptic cylindrical bases of type I and
II in a scheme ( gure 7) where their relationships with the other
bases are shown. The basis I, as the modulus k varies, changes
continuously from Yn;;m to Yn; jj;m, and the basis II from Yn;l;m to
Yn;;m .

93

y w

n1, q(k)

n1, b(k)

Ell. Cylindrical
type II
k
1

lm

nm

Z l
z

Ell. Cylindrical
type I
k
k

C ||

m
n1

n1
z

n1

Figure 2.7:

n1

|| m

nm

Relationships among elliptic and canonical Sturmians. For limiting values of the modulus k , the elliptic Sturmians coincide with the canonical
ones, i.e. those corresponding to binary trees.

94
It is possible to get the unknown elliptic cylindrical eigenfunctions by calculating the coecients of their expansion in suitable
basis sets. As we diagonalize the matrix of ^l2 + k2 K^ z2 on the basis
^ 2 and ^lz2, therefore
Yn;l;m, the eigenvectors diagonalize also ^l2 + K
they are elliptic cylindrical functions of type II. By calculating the
matrix of the operator ^l2 + k2 K^ z2 on the basis Yn;l;m we can prove
P


that the coecients in Yn;q;m
= l bn;m;
q;l Yn;l;m - where Yn;q;m is the
elliptic cylindrical basis set of type II and q is the eigenvalue of
^l2 + k2K^ z2 - are related by the three-term recursion relationship
r

1) m ][n l ][l m ] bn;m; +


k2 [n (l 1)(2][(l l3)(2
q;l 2
l 1) (2l+1)

(l+1) m ][n (l+1) ]
l m ][n l ]
l(l + 1) + k2 [ (2l+1)(2
+ k2 [ (2l 1)(2
l+3)
l+1)
2

m ][n (l+1) ][(l+1)


k2 [n (l+2) ][(l+2)
(2l+1)(2l+3) (2l+5)
2

q bn;m;
q;l +

m ] n;m;
bq;l+2
2

=0
(2.50)

The step of the recurrence is two because ^l2 + aK^ z2 ; P^ = 0 (a can


be any number) thus matrix elements hYn;l;mj^l2 + aK^ z2 jYn;l0;m i are
always zero if l and l0 have di erent parity, and elliptic cylindrical
functions have de nite inversion symmetry , indicated by  = 1.
As k ! 0, eq. (2.50) is diagonal, while as k ! 1 it becomes a
n;m
recursion in l for the coecients Zl;
(see Eq. 2.24) where the
eigenvalue q corresponds to (n2 1) ( + 1) + m2 , since the
operator tends to ^l2 + K^ z2 which can be written as 4C^ (K^ x2 +
K^ y2 + ^lz2 ) + ^lz2 .
As another example of application of this method, let us calh

95
culate the overlaps, i.e. the transformation coecients, between

Yn;b;m
(the elliptic cylindrical basis set of type I, where b is the


eigenvalue of the operator ^l2 + 1 1k K^ z ) and Yn; jj;m. Putting
P

Yn;b;m = jj an;m;
b;jj Yn;jj;m one has, by calculating the matrix of


^l2 + 1 1k K^ z2 on the basis Yn; jj;m:
2

1 2
[n (m +  1)2 ] [n2 (m  + 1)2 ] an;m;
b;j 2j +
4




1 2
1
1
(n 1) + (m2 2 ) +
2 b an;m;
b;jj +
2
2
2
1 k
1q 2
[n (m +  + 1)2 ] [n2 (m  1)2 ] an;m;
b;j+2j = 0 (2.51)
4
According to the scheme of gure 7, the latter recurrence relationship is diagonal as k ! 1, while as k ! 0 eq. (2.51), it coincides
with the recursion in  for the coecients C;n;m;
jj , and the eigenvalue b becomes (n2 1) ( + 1) + m2 . In g. 8 we show the
correlation diagrams for the separation constants b and q of the elliptic cylindrical orbitals of type I and II in the case n = 3; m = 0.
q

2.6 Summary and concluding remarks


The scope of this chapter has been to o er a complete view of
hydrogenic orbitals in momentum space, which essentially coincide
with harmonics on the hypersphere S 3 . From the alternative sets of
coordinates for which the Laplace equation on the hypersphere separates, di erent sets of orbitals originate: we have classi ed them
and studied some of their properties, particularly investigating the

96

4
Elliptic Cylindrical Basis, Type II

q
n+k

Elliptic Cylindrical Basis, Type I

0
2

Arctan k

Figure 2.8:

/4

b
2
n+1/(1-k )

0
/2

Arctan 1/(1-k )

Correlation diagram showing the variation of the quantum


numbers b and q as a function of k 2 , for the case n = 3; m = 0. Left panel:
case of elliptic cylindrical basis set of type II. Right panel: case of elliptic
cylindrical basis set of type I. On the left extreme, q coincides with l(l + 1); at
the opposite side, b coincides with 2 . In the middle of the diagram, both of
them coincide with (n2 1) ( 1) + m2 (Zeeman basis set).

97
connections between them. Within the framework of information
already available in the literature [4, 13] we have introduced explicitly the requirement of the conservation of parity with respect
to inversion and have obtained three-term recurrence relationships
for the transformation coecients, including the Z -matrix elements
recently de ned [25, 26, 39]. Our detailed investigation has lead to
three-term recurrence relationships which provide a simple calculation scheme for the development of cylindrical elliptic orbitals of
rst and second type both in the familiar spherical and in the Stark
orbitals. This also has allowed to elucidate the relationships with
respect to the `canonical' bases.
Finally, amplifying the results of the previous chapter which
focused on orbitals in con guration space, we can discuss analogies and di erences of the obtained basis sets in the two reciprocal
spaces. Of the four types of orbitals in con guration space, three
of them (polar, parabolic, spheroelliptic) have a counterpart in momentum space (spherical, Stark, spheroelliptic) respectively. The
fourth type of orbitals (spheroidal), characterized by being eigenfunction of the operator ^l2 + 2 K^ z , has no counterpart in momentum space. On the other hand, three types of orbitals among those
available in momentum space (Zeeman, elliptic cylindrical of type
I, elliptic cylindrical of type II) characterized respectively by being
eigenfunctions of the operators K^ x2 + K^ y2 + ^lz2 , ^l2 + 1=(1 k2 )K^ z2
and ^l2 + k2 K^ z2 , have no counterpart in con guration space. The
reason for this "missing correspondence" (one manifestation being

98
the intractability of spheroidal orbitals according to Coulson [14])
^ operator: according
can be attributed to the di erent role of the K
to its de nition (Eq. 2.6), it can be seen that it is a second order
di erential operator in con guration space, while it acts as a rst
order di erential operator on eigenfunctions in momentum space.
This is at variance with the role of ^l, which is of the rst order in
both spaces. Since the con guration space Schrodinger equation
is of the second order, as it is its counterpart describing motion
on S 3 , separability of variables implies the action of second order
operators [15, 40]. On the other hand in con guration space there
^ 2 (or any of its cartesian
are no diagonal operators which contain K
components).
The fact that each alternative type of orbitals in momentum
space provides eigenfunctions of cartesian components of ^l2 and
K^ 2 or of their linear combinations, also implies that these orbital
sets induce irreducible representations in the group D2h [15]: the
correspondence holds for the reciprocal con guration space only
in the case of polar, parabolic and spheroelliptic orbitals, which
also induce irreducible representations in D2h , while the spheroidal
orbitals do not.
A detailed description of Sturmian basis sets and of their counterpart in momentum space is extremely important for the implementation of the remarkable plane wave expansion:
exp(ip  r) = (2 )3=2

p2 + p2
un;l;m(r)  0 2 n;l;m (p)
2p0
n;l;m
X

(2.52)

99
which is seen to be essentially an expansion in four-dimensional
spherical harmonics and Sturmian polar orbitals (their explicit expression can be found in the previous chapter. Eq. (2.52) is a key
formula in momentum space quantum chemistry and was introduced by Shibuya and Wulfman [41] in 1965 who used it to expand
one-electron molecular orbitals and to obtain secular equations for
multicenter problems. A detailed mathematical analysis for such
an expansion can be found in ref. [42].
Previously, we had introduced an extension of such a formulation to Stark Sturmians [3], which has proved to have competitive convergence properties when applied to the calculation of the
ground state energy of the hydrogen molecular ion. This research
has emphasized the importance of exploiting the diverse alternative
separable coordinate systems. The extension of such expansions to
the di erent types orbitals treated in this chapter seems extremely
promising and stimulating, not only in the three-dimensional case,
but also in the d-dimensional wave expansion, as already done [4]
for spherical and Stark Sturmians. The solution of many-body
Coulomb problems would in fact require multidimensional plane
wave expansions [12].

Bibliography
[1] B. Podolsky and L. Pauling.
[2] V. Fock.

Phys. Rev., 34:109, 1929.

Z. Phys., 98:145, 1935.

100
[3] V. Aquilanti, S. Cavalli, C. Coletti, and G. Grossi.
Phys., 209:405, 1996.
[4] V. Aquilanti, S. Cavalli, and C. Coletti.
1997.

Chem.

Chem. Phys., 214:1,

[5] V. Aquilanti and J. Avery. Chem. Phys. Letters, 267:1{6, 1997.


[6] J. Avery. Hyperspherical Harmonics, Applications in Quantum
Theory. Kluwer Academic, Dordrecht, The Netherlands, 1989.
[7] J. Avery. Hyperspherical Harmonics and Generalized Sturmians. Kluwer Academic Publishers, Dordrecht, The Netherlands, 2000.
[8] H. Shull and P.-O. Lowdin.

J. Chem. Phys., 30:617, 1959.

[9] M. Rotenberg.

Ann. Phys. (N.Y.), 19:262, 1962.

[10] M. Rotenberg.

Adv. At. Mol. Phys., 6:233, 1970.

[11] V. Aquilanti and J. Avery.


2001.

Adv. Quantum Chemistry, 39:71,

[12] V. Aquilanti, S. Cavalli, C. Coletti, D. Di Domenico, and


G. Grossi. Int. Rev. Phys. Chem., 20(4):673, 2001.
[13] V. Aquilanti, S. Cavalli, and C. Coletti.
80:3209, 1998.

Phys. Rev. Lett.,

101
[14] C.A.Coulson and A.Joseph.
1967.

Proc. Phys. Soc. London, 90:887,

[15] E.G. Kalnins, W. Miller, and P. Winternitz. S.I.A.M. J. Appl.


Math., 30:630, 1976.
[16] M. Bander and C. Itzykson.

Rev. Mod. Phys., 38:330, 1966.

[17] M.J. Engle eld. Group Theory and


Wiley-Interscience, New York, 1972.

the Coulomb Problem.

[18] L.D. Landau and E.M. Lifshitz. Quantum Mechanics. Nonrelativistic Theory. Addison-Wesley Publishing Co., Reading
Mass., 1958.
[19] B. R. Judd. Angular Momentum Theory
Molecules. Academic Press, New York, 1975.
[20] Yu.F. Smirnov and K.V. Shitikova.
1976.

for Diatomic

Sov. J. Part. Nucl., 8:344,

Fonctions Speciales et Theorie de la


Representation des Groupes. Dunod, Paris, 1969.

[21] N.J. Vilenkin.

[22] N.Ya. Vilenkin, G.I. Kuznetsov, and Ya.A. Smorodinskii. Sov.


J. Nucl. Phys., 2:645, 1966.
[23] V. Aquilanti, S. Cavalli, C. Coletti, D. De Fazio, and G. Grossi.
In C.A. Tsipis, V.S. Popov, D.R. Herschbach, and J.S. Avery,

102
editors, New Methods in
1996.

Quantum Theory, page 233. Kluwer,

[24] M. Abramowitz and I.A. Stegun. Handbook


Function. Dover, New York, 1964.
[25] J.J. Labarthe.

J. Phys. B, 14:L{467, 1981.

[26] D.R. Herrick.

Phys. Rev. A, 26:323, 1982.

of Mathematical

[27] D. Delande. These de Doctorat. PhD thesis, Universite Pierre


et Marie Curie, Paris, 1988.
[28] C. Coletti.

Tesi di Dottorato. Universita di Perugia, 1997.

[29] F. Penent, D. Delande, and J.C. Gay.


1988.

Phys. Rev. A, 37:4707,

[30] A.F. Nikiforov, S.K. Suslov, and V.B. Uvarov. Classical Orthogonal Polynomials of a Discrete Variable. Springer-Verlag,
Berlin, 1991.
[31] V. Aquilanti, S. Cavalli, and C. Coletti.
344:587, 2001.

Chem. Phys. Lett.,

[32] V. Aquilanti, , S. Cavalli, C. Coletti, D. Di Domenico, and


G. Grossi. Quantum Systems in Chemistry and Physics. Vol
I: Basic Problems and Model Systems, page 289, 2000.
[33] D. Park.

Z. Phys., 36:155, 1960.

103
[34] V. Aquilanti and C. Coletti. Chem. Phys. Lett., 344:601, 2001.
[35] V. Aquilanti, S. Cavalli, and G. Grossi.
1996.

Z. Phys. D., 36:215,

Sov. J. Nucl. Phys., 15:113, 1972.


[37] J. Patera and J. Winternitz. J. Math. Phys., 14:1130, 1973.
[38] V. Aquilanti, S. Cavalli, and G. Grossi. J. Chem. Phys.,
[36] M. Kil'dyushov.

85:1362, 1986.

Europhys. Lett., 5:303, 1988.


[40] W. Miller Jr. Symmetry and Separation of Variables. Addison[39] D. Delande and J.C. Gay.

Wesley Publishing Company, Reading, Massachusetts, 1977.

[41] I. Shibuya and C.E. Wulfman.


1965.
[42] E. J. Weniger.

Proc. Roy. Soc., A286:376,

J. Math. Phys., 26:276, 1985.

[43] A. A. Izmest'ev, G. S. Pogosyan, A. N. Sissakian, and P. Winternitz. J. Math. Phys., 40:1549, 1999.
[44] E. G. Kalnins and Jr. W. Miller.
1986.

J. Math. Phys., 27:1721,

104

Table 2 - The 3 bases corresponding to the reduction of the group


O(4) to the subgroup chain O(4)  [O(2)]i;j  [O(2)]k;l. The bases
are classi ed according to the diagonal operator acting on the planes
(i; j ) and (k; l).

d = 2-subspace d = 2-subspace
(y; z ) ^lx
(x; w) K^ x

basis set
jn; 00; mi

(z; x) ^ly

(y; w) K^ y

jn; 0; mi

(x; y ) ^lz

(z; w) K^ z

jn; ; mi

Chapter 3
Many-electron one-center systems
Summary

This chapter deals with the application of polar Sturmian orbitals to the solution of the Schrodinger equation for free atoms
and atoms in external elds. We de ne a N -electronic generalized
Sturmian basis set as a product of one-electron Sturmian orbitals.
Expanding the electronic wavefunction by such a generalized Sturmian basis set, we obtain a secular problem in which the matrix
elements do not contain any kinetic energy operator. The solution of the secular problem yields accurate values for the energies
of electronic ground and excited states. In the Sturmian method
framework, the e ect of strong external elds can be taken into
account with little additional e ort; we provide the formalism for
the treatment of external perturbation and show the results of a
preliminary calculation for static polarizability of Lithium.
105

106

3.1 Introduction
Chapters 1 and 2 contain a deep insight into the properties of the
di erent species of hydrogenic Sturmian orbitals. In this chapter
we will see the application of polar Sturmian orbitals (Eq. 1.7),
to the calculations of physical properties for isolated atoms. Illustrating the approach to atomic structure as a many-electron
one-nucleus problem, we discuss here the procedure of building a
polyelectronic basis set as a product of "monoelectronic" Sturmian
orbitals [1{3, 7, 9, 11, 12]: an alternative method, which de nes N electronic Sturmians as solutions of the 3N -dimensional hydrogen
atom can be elegantly formulated and also has a momentum space
counterpart in terms of hyperspherical harmonics [4{6], but still
lacks of developments in the algebraic and computational implementation in order to be computationally competitive.
Sturmian approach is quite di erent from the procedure used
commonly in modern ab initio quantum chemistry:

The one-electron orbitals are Sturmian orbitals. Their analytic expression is known a priori, except the exponent of
the radial part, whose numerical value is obtained after the
calculation of the properties of the system.
(In the most common ab initio procedures for quantum mechanical calculations, Gaussian functions are used as basis
set. An Hartree-Fock iterative calculation is performed to
combine those basis functions into one-electron orbital).

107

Con gurations are built exploiting Sturmian orbitals, and


then a con guration interaction (CI) is performed.

The upper eigenvalues converge quickly towards the energies


p
2E ,
of the excited states (actually the calculation provides
not the energy E ).

The e ect of strong external perturbations can be included by


a small computational e ort and with no need of perturbative
methods.

3.2 Free atoms


The SE for a N -electron free atom is:
"

where

1 2
p2
r
+ V0 (x) + V 0 (x) + 0 (x) = 0
2
2
N

X
r2 = r2j
j =1

V0 (x) =
and

V 0 (x) =

(3.2)

N
X

Z
j =1 rj

(3.3)

N
X

1
i>j rij

The electronic energy has been indicated as


charge as Z .

(3.1)

(3.4)
p

2 , and the nuclear


2
0

108
The wavefunction is expanded by a linear combination of Slater
determinants
(x) =

where

C  (x)

 (x) =

un ;l ;m ;ms (x1 ) un ;l ;m ;ms (x1 )


p1 un ;l ;m ..;ms (x2 ) un ;l ;m ..;ms (x2 )
N!
.
.
un ;l ;m ;ms (xN ) un ;l ;m ;ms (xN )










(3.5)

x
x

   un
   un

( 1)
N ;lN ;mN ;msN
N ;lN ;mN ;msN ( 2 )

   un

;l ;mN ;msN ( N )

..
.

N N

..
.

(3.6)
The one-electron functions un;l; m; ms (xj ) are the polar Sturmian orbitals given in Eq.(1.7), multiplied by the spin function (ms =
1=2); x stands for fx1; x2 ; : : : ; xN g and fxj g = frj ; #j ; 'j g.
Our basis set is not completely de ned yet, because we have
to establish a rule for assigning a value to the exponents kj . We
demand our basis functions  (x) to be a solution to the di erential
Equation [1, 8, 9, 11, 12]:
"

1 2
p20
r +  V0(x) + 2  (x) = 0
2

(3.7)

This is true if we impose two subsidiary conditions for the exponents:

k12 + k22 + :::kN2 = p20

(3.8)

n1 k1 = n2 k2 = ::: = nN kN = Z 

(3.9)

and

109
Eqs.(3.8) and (3.9) imply that the exponents' value is:

kj =

p0
1
1
nj n + n +    + n1N
q

2
1

2
2

(3.10)

The solutions of Eq.(3.7) enjoy a potential weighted orthonormality


relationship [8, 12]:
Z

dx 0 (x)V0 (x) (x) =  0 ;

2E


(3.11)

The de nition of the generalized N -electron Sturmian basis set


 (x) is analogue to the de nition of one-electron hydrogenic Sturmian functions: the energy ( p20 =2) is constant for all the functions,
while the `charge'  plays the role of an eigenvalue.
We notice that the same Sturmian orbital may have di erent
exponents if it appears in two di erent con gurations. For example,
p
in  = 1s2 one has k1s = p0 = 2, while in  = 1s2s one has
q
k1s = p0 = 5=4.
We have employed the unknown energy p20 =2 in the de nition
of our N -electron basis set (Eq. 3.7). The exponents kj depend
on the value of p0 , i.e. on the value of the electronic energy (p0 =
p
2E ); consequently it looks like we are following an iterative
approach, in which we have to guess an initial value for p0 and then
iterate to get improved values for p0 until convergency is achieved.
Now we are going to show that actually using the basis set de ned
in Eq.(3.7) doesn't imply any iteration: one has to calculate just
one matrix, and diagonalize it. Substitution of expansion (3.5) into

110
Eq.(3.1) yields:
X

"

p2
1 2
r
+ V0 (x) + V 0 (x) + 0 C  (x) = 0
2
2

(3.12)

Then, taking into account that the functions  (x) obey Eq.(3.7),
we can eliminate the kinetic energy operator from Eq.(3.12):
X

[  V0 (x) + V0 (x) + V 0 (x)] C  (x) = 0

(3.13)

Multiplying Eq.(3.13) by a function  0 (x) and integrating over the


coordinates, exploiting Eq.(3.11) we obtain:
X Z

d(x) 0 (x)[V0 (x) + V 0 (x)] (x)

2E 0 ;

C = 0 (3.14)

Then if we divide the latter equation by p0 and introduce the


de nitions
1Z
T00 ; 
dx  0 (x)V0 (x) (x)
(3.15)
p0
1Z
0
dx 0 (x)V 0 (x) (x)
(3.16)
T 0 ; 
p0
T 0 ;  T00 ; + T0 0 ;
(3.17)
we arrive to a concise expression for our secular problem:
X

[T 0 ;

p0  0 ; ] C = 0

(3.18)

The integral (3.15) can be easily expressed taking into account


Eqs.(3.11), (3.8) and (3.9):
0

11=2

X
2E
p
1
T00 ; =
 0 ; = 0  0 ; = = @ kj2 A
p0 


j 

11=2

1A
 0 ; = Z @
2
j  nj
(3.19)
X

 0 ;

111
The integral T00 ; doesn't depend on p0 's value, and this is true also
for the electronic repulsion integral T0 0 ; : in the radial part of the
integral, one can perform a change of integration variable from r to
p0 r, so that p0 's value is not needed to solve the integral. In order to
calculate energies and wavefunction for the electronic states of an
isolated atom, one has just to compute and diagonalize the matrix
T 0 ; .

3.3 Atoms in external elds


The extension of the method exposed in the previous section to
include the e ect of a strong electric eld is straightforward. We
have just to add a Coulomb interaction term in Eq.(3.1). If we
consider a constant electric eld of intensity E oriented along the z
axis, the interaction has the expression:
N

X
V 00 (x) = E rj cos #j

(3.20)

j =1

So the treatment performed in the previous section would give,


including in the Hamiltonian the potential in Eq.(3.20):
Xh

T00 ; + T0 0 ; + T000 ;

p0  0 ; C = 0

(3.21)

where

T000 ; 

dx 0 (x)

N
X
j =1

p0 rj cos #j  (x)

(3.22)

(3.23)

 2
p0

112
The value of integral (3.22) is not a function of p0 . The procedure
for solving Eq.(3.21) is the following:
1. Establish a value for  ;
2. Compute the matrix elements;
3. Diagonalize the matrix, getting p0 ;
4. The electric eld intensity is E = p20 ;
5. The energy of the atom in the electric eld is p20 =2.

3.4 Electronic integrals between Sturmian con gurations


In this section, the procedures for computing electronic integrals
involving Sturmian basis set are exposed. As we are dealing with
atomic systems there is no multicenter integral, consequently all
the formulas are analytical. At variance with Hartree-Fock orbitals,
Sturmian orbitals are not orthogonal to each other, so we have to
use generalized Slater-Condon rules [?]. In order to expose these
rules, we introduce the antisymmetrizer
X
A^ = ( 1)p P^

(3.24)

where the sum is extendend to all possible permutations of electrons, and p is the permutations parity.

113
3.4.1

Overlap Integrals

Let us consider an overlap integral h


electron con gurations










f1 (1) f2 (1)
f1 (2) f2 (2)
1
1
 (x) = p F (x) = p
..
..
N!
N!
.
.
f1 (N ) f2 (N )

 0 i between two N -

   fN (1)
   fN (2)
..
.



..
.
fN (N )

(3.25)

and  0 (x) = p1N ! G(x) similarly de ned. The overlap can be expressed as:

h j 0 i = N1 ! hF j Gi

1 ^
hA[f1 (1) : : : fN (N )] j A^[g1 (1) : : : gN (N )]i
N!
= hf1 (1) : : : fN (N ) j A^[g1 (1) : : : gN (N )]i


hf jg i hf jg i    hf jg i

1
1
1
2
1
N


hf2 jg1 i hf2 jg2 i    hf2 jgN i



= ..
..
..
..

.
.
.
.




hfN jg1 i hfN jg2 i    hfN jgN i

The overlap between two con gurations is the determinant of the


matrix whose elements are the overlaps between orbitals. In this
chapter we deal with atoms and exploit polar Sturmians so these
overlaps can be computed analytically. We have to solve the integral:
Z

un0 ;l0 ;m0 (r; #; ') un;l;m(r; #; ')r2 dr sin #d# d' =

= 23+2l (k

 k

0 )l+3=2

"

(n l 1)!(n0 l 1)!
2n[(n + l)!]3 2n0 [(n0 + l)!]3

#1
2

(3.26)

114
Z

exp (k +k0 )r r2l+2 L2nl+1l 1 (2k r)L2nl0+1l 1 (2k0 r)dr


Z

Yn0 ;l0 (#; ')Yn;l (#; ') sin #d#d'

The result of the angular integral is nl;n0 l0 , while the radial integral can be solved considering the explicit expression for Laguerre
polynomials:
Z

exp (k +k0 )r r2l+2 L2nl+1l 1 (2k r)L2nl0+1l 1 (2k0 r)dr =


2

exp (k +k0 )r r2l+2 4(n + l)!


2

(n0 + l)!

n0X
l 1(

1)k
k!

k=0
nX
l 1 n0X
l 1

nX
l 1(
j =0

1)j
j!

n+l
n l j 1

n0 + l
n0 l j 1

(2k r)j 5

(2k r)k 5 dr =

( 2k )j ( 2k0 )k Z 1
dr exp (k +k0 )r r2l+2+j +k
j
!
k
!
0
j =0 k=0
The latter radial integral is easy to solve:
Z 1
(2l + j + k + 2)!
dr exp (k +k0 )r r2l+2+j +k =
(k + k0 )2l+j +k+3
0
Considering Eq.(3.10), one would expect the result of Eq.(3.26) to
be a function of p0 , but a deeper analysis indicates that a change of
variable could be performed inside the integral, from r to p0 r; this
implies the result to be independent from p0 's numerical value.

(n+l)!(n0 +l)!

3.4.2

Matrix elements of one-electron operators

Here we consider matrix elements of one-electron operators of the


form
V = v (1) + v (2) + : : : + v (N )
(3.27)

115
We want to obtain a general expression for the matrix element
h j V j 0 i. Considering the formulas given above, it is straightforward to show that

h j V j 0 i = N1 ! hF j V j Gi = NN! hF j v(1) j Gi

(3.28)

Then we de ne the symbol FiN 1 as the minor obtained from a


N N determinant F after deleting the rst row and the ith column.
We can expand the determinants F and G in terms of its minors:

F=
G=

N
X

( 1)i+1 fi (1)FiN 1

(3.29)

( 1)j +1gj (1)GNj 1

(3.30)

i=1
N
X
j =1

Substituting these expansions into Eq. (3.28), we obtain:

hF j v(1) j Gi =

N X
N
X

( 1)i+j hfi j v j gj ihFiN 1 j GNj 1 i (3.31)

i=1 j =1

Then, considering that A2N 1 = (N 1)!AN 1 , the matrix element


(3.28) can be made explicit in a very elegant form:

h j V j 0 i =

N X
N
X
i=1 j =1

( 1)i+j hfi j v j gj ijSij j

(3.32)

where jSij j is the minor obtained from the determinant of the overlap matrix by deleting the ith row and the j th column.
If the operator in Eq.(3.27) is the Coulomb potential for an atom
embedded in a static electric eld (see section 3.3), then the matrix

116
element in Eq.(3.32) corresponds to the integral
Z

un0 ;l0 ;m0 (r; #; ')  p0 r cos #  un;l;m(r; #; ')  r2 dr sin #d# d' =
Z

#1

"

(n l 1)!(n0 l0 1)!
0


2n[(n + l)!]3 2n0 [(n0 + l0 )!]3
0
0
exp (k +k0 )r rl+l +3 L2nl+1l 1 (2k r)L2nl0 +1l0 1 (2k0 r)dr

0
0
= 23+l+l kl+3=2 kl 0+3=2 p
Z

Yn0 ;l0 (#; ') cos #Yn;l (#; ') sin #d#d'
(3.33)

The radial part of this integral is very similar to the one in Eq.(3.26):
Z

0
0
exp (k +k0 )r rl+l +3 L2nl+1l 1 (2k r)L2nl0 +1l0 1 (2k0 r)dr =
2

0
exp (k +k0 )r rl+l +3 4(n + l)!
2

4(n0 + l0 )!

n0 X
l0 1 (
k=0

1)k
k!

nX
l 1

( 1)j
j!
j =0

n0 + l 0
n0 l 0 j

(2k

r )j 5

n+l
n l j
(2k

r)k 5 dr

nX
l 1 n0 X
l0 1

( 2k )j ( 2k0 )k Z 1
0
dr exp (k +k0 )r rl+l +3+j +k
j !k!
0
j =0 k=0
The solution of the latter integral is:

(n+l)!(n0 +l0 )!

3.4.3

(l + l0 + j + k + 3)!
0
dr exp (k +k0 )r rl+l +3+j +k =
(k + k0 )l+l0 +j +k+4
0
1

Matrix elements of two-electron operators

In this paragraph we are going to demonstrate generalized SlaterCondon rule for matrix elements between determinants involving

117
two-electron operators. The typical example of a two-electron operator is the Coulomb repulsion among N electrons, corresponding
to the sum of N (N 1)=2 terms 1=rij . If we de ne the potential as

V0 =

1
j>i rij

(3.34)

then the following passage is straightforward:

h j V 0 j 0 i = N (N2

1)

h j r1 j 0 i = 2(N 1


12

1
h
Fj
j Gi
2)!
r
12

(3.35)
We now expand the determinants F and G in terms of their double
minors:

F=

N X
N
X

( 1)i+j +1 [fi (1)fj (2)

fj (1)fi (2)]FijN 2

(3.36)

( 1)k+l+1 [gk (1)gl (2) gl (1)gk (2)]GNkl 2

(3.37)

i=1 j =i+1

and

G=

N X
N
X

k=1 l=k+1

In Eq.(3.36), FijN 2 is the double minor obtained from the determinant F by deleting the rst and second rows and the ith and j th
columns. Considering that A2N 2 = (N 2)!AN 2 , and combining
Eqs.(3.35), (3.36) and (3.37), we get
N

X
h j V 0 j 0 i =

N X
N X
N
X

( 1)i+j +k+l Cij ;kljSij ;klj (3.38)

i=1 j =i+1 k=1 l=k+1

where

Cij ;kl =

dx1 dx2 fi (1)fj (2)

1
[g (1)g (2)
r1 2 k l

gl (1)gk (2)] (3.39)

118
and jSij ;klj is the determinant of the matrix obtained by deleting
ith and j th row, and kth and lth column from overlap matrix S .
From Eq.(3.39), we understand that in order to calculate the
matrix element given in Eq.(3.35), we have to solve integrals of the
tipe:
Z

1
dx1 dx2 un ;l ;m (x1 )un ;l ;m (x1 ) un ;l ;m (x2 )un ;l ;m (x2 )
r12
(3.40)
The operator can be expanded as [10]:
1

l
1 X
r<
4 X
1
=
r12 r> l=0 m= l r>

!l

1
Y  (
)Y (
)
2l + 1 l;m 1 l;m 2

(3.41)

where Yl;m (
i ) are normalized spherical harmonics and
i = f#i ; 'i g.
Inserting the latter expansion into Eq.(3.40), and separating radial
and angular integrals (un;l;m(xj ) = Rn;l (rj ) Yl;m (
j )), we obtain :
1 Z
 (
)
4
d
1 Yl;m (
1 )Yl ;m (
1 )Yl;m
1
l;m 2l + 1
X




(3.42)

d
2 Yl;m (
2 )Yl ;m (
2 )Yl;m (
2 )
3

r< l 1
r> r>
The integral of the product of three spherical harmonics is:
dr1 r12 Rn ;l (r1 ) Rn ;l (r1 )
1

v
u
u
t

dr2 r22 Rn ;l (r2 ) Rn ;l (r2 )


3

 (
)Y 0 0 (
)Y 00 00 (
) =
d
Yl;m
l ;m
l ;m

(2l0 + 1)(2l00 + 1)
hl; 0jl0; 0; l00; 0ihl0; m0; l00 ; m00jl; mi
4 (2l + 1)

(3.43)

119
So we can use Eq.(3.43) to solve the angular integral in Eq.(3.42),
that becomes:
4

l=max(jl

l j;jl l

min(l X
+l ;l +l )
2

(2l1 + 1)(2l + 1)
1
hl2 ; 0jl1 ; 0; l; 0ihl1 ; m1; l; mjl2 ; m2i
2
l
+
1
4

(2
l
+
1)
2
j)
(3.44)

(2l4 + 1)(2l + 1)
hl3 ; 0jl4 ; 0; l; 0ihl4 ; m4 ; l; mjl3 ; m3 iRn ;l
4(2l3 + 1)
1

;n ;l ;n ;l ;n ;l ;l
2

where Rn ;l ;n ;l ;n ;l ;n ;l ;l is the radial part of the integral, that


can be explicitly written as:
1

Z 1


Z r1

+1
1
0

rl

1 r12 Rn1 ;l1 (r1 )Rn2 ;l2 (r1 )

dr

Z 1

+2
l
dr2 r
2 Rn3 ;l3 (r2 )Rn4 ;l4 (r2 ) + r1
l

(3.45)

1
dr2 r
2

3 3 (r2 )Rn4 ;l4 (r2 )

Rn ;l

The latter integral can be solved analytically; the result is


16(f

1;3 f3;1 f2;4 f4;2 )3=2

4
Y
k

l3 1
3X

c
l1 1
1X

2f

1;3 )a

a!

=0
2+l+lX
3 +l4 +c+d

=0
1 l+lX
3 +l4 +c+d
i

=0

c
)

lk

c!

=0

2f

3;1

=1

(2fk )

1 + l1
n1
1 a
n
l

(nk

lk

nk (nk

l4 1
i1=2 n4X

lk )!

3 + l3
l3
c

2;4

2f

b
)

b!

=0

4 ;2 ) d

2f

d!

=0

h

l2 1
 n2X

1)!

4 + l4
n4
l4
d

2 + l2
l2
b

1
(3.46)

(2 + l + l3 + l4 + c + d)!(1
l + a + b + l1 + l2 )!
3;1 + f4;2 )3+c+d+l+l3 +l4 (f1;3 + f2;4 )2 l+a+b+l1 +l2

(f

(2 + l + l3 + l4 + c + d)!(1
l + a + b + l1 + l2 + i)!
3 1 + f4;2 )3+c+d+l+l3 +l4 i (f1;3 + f2;4 + f4;2 + f3;1 )2 l+a+b+l1 +l2 +i

i!(f ;

3 4
j !(f3;1 + f4;2 )2 + 3 + 4 + +
(1

1 + l2 + a + b)!
4 2 + f3;1 )3+l+j+l1 +l2 +a+b

l + l + l + c + d)!(2 + l + j + L
l l l
c d j (f
; +f ; +f ;

13

24

In the latter equation, we have introduced the notation:

fp;q 

np n1q + n1p
2

(3.47)

120

3.5 On the choice of the `energy' in the


de nition of the basis set
We have chosen to use the energy of the atom, p20 =2, in the de nition of the basis set (Eq. 3.7). It would be interesting to understand
whether this is the best possible choice for the de nition of the basis set; perhaps a di erent value for the energy in Eq.(3.7) would
give a basis set that has better convergency properties. Let us try
and de ne a basis set by the equation:
"

1 2 ~
p~2
r
+  V0 (x) + 0 ~  (x) = 0
2
2

(3.48)

The orbital exponents be de ned according to Eqs.(3.8) and (3.9).


In principle p~0 can take any value. If p~0 = p0 , the basis set ~  (x)
coincides with  (x). Expanding the wavefunction by the basis set
de ned in Eq.(3.48) and eliminating the kinetic energy operator as
in Eq.(3.13), we can transform the SE into:
X

"

p~20
2

~ V0 (x) + V0 (x) + V 0 (x) E C ~  (x) = 0

(3.49)

Multiplying on the left by ~  0 (x), integrating over the electronic


coordinates, dividing by p~0 and considering Eq.(3.11), we get:
X

"

T; 0

p~
S 0
p~0 ; 0 + 0 S; 0 + ; E C = 0
2
p~0

(3.50)

The symbol S; 0 stands for the overlap between basis functions:

S; 0

 dx~ 0 (x)~  (x)

(3.51)

121
Eq.(3.50) is a generalized eigenvalue problem, where the eigenvalue
is the energy E . We have to investigate which is the best value for
p~0 , and whether it coincides with p0 .
In the particular case of a single basis function ~  (x) (a trivial
1x1 matrix) our problem becomes:

p~0 E
+ =0
2 p~0

T;

(3.52)

So we get the energy as a function of the parameter p~0 :

p~2
E= 0
2

p~0 T;

(3.53)

2 =2, when p~0 = T; .


The energy assumes the minimum value, T;
Exploiting Eq.(3.18) would give the same value for the energy, so
we conclude that in the particular case of a basis set made of one
function, the best value for p~0 is p0 .
For the more interesting case of a larger basis set, we are not
able to give any general proof, but considering a speci c case we
demonstrated that the best value for p~0 is not always p0 . Let us
approximate the ground state of Helium atom by a linear combination of two functions: ~ 1 (x) = 1s2 (1 S ) and ~ 2 (x) = 1s2s(1 S ).
According to Eq.(3.50), the approximated ground state energy is
the lowest solution of the second order equation:







p~

E
2 + p~

T1;1

p~

T1;2 + 2 S1;2 +
0

S;
p~

1 2
0

T2;2

p~

2 + p~
0

=0

(3.54)

122
If we assign p~0 a value of 2.47 a.u., Eq.(3.54) gives E = 2:8557
a.u., while exploiting Eq.(3.18) one obtains p0 = 2:389 a.u. and
thus E = 2:8533 atomic units. This means that in general p0 is
not the best choice for the de nition of the energy of the basis set.

3.6 Examples
When calculations over free atoms are performed using Eq.(3.18)
convergence rate toward the ground state energy is good, and moreover the upper eigenvalues give very good approximations to the
excited state energies. We performed a calculation over Helium (1 S
states), using a basis set made of 1 S bielectronic Sturmian functions with quantum numbers up to n1;max = 4; l1;max = 2; n2;max =
10; l2;max = 2. Accurate 1 S states for helium are available since
long time [13], and continue to be a benchmark for alternative basis sets in variational calculations [14,15]. The ground state energy
we have obtained is 2.902690 a.u., to be compared with the best
variational value 2.903724377 a.u. (error = 0:036%). Considering
also upper eigenvalues, we have been able to reproduce carefully
by our ab initio calculation the Helium 1 S electronic transition
energies between the ground state and several excited states (Table
3.1). In the literature there are some recent examples of accurate
calculations on Helium excited states [16, 17] performed with various methods, but they seem to be not suitable for extensions to
more complex atoms.

123
1 S Levels

1s2

1s 2s
1s 3s
1s 4s
1s 5s
1s 6s
1s 7s

Table 3.1: The 1 S levels of the helium atom.


Calculated(a)

(a.u)

-2.902690
-2.147417
-2.061402
-2.033607
-2.021181
-2.014546
-2.010585

(a) This work.


(b) NIST Atomic
bin/AtData/main

Transition energies
1s ns 1s2 (cm 1 )

165763
184641
190742
193469
194925
195794

Spectra Database
asd).

Experimental(b)

166277.547
184641.369
190741.633
193468.726
194924.965
195794.328

(http://physics.nist.gov/cgi-

The capability of the Sturmian method to reproduce atomic


electronic transition energies can be shown to be a general feature,
and was tested also on atoms having more than two electrons [8,12].
We have tested the Sturmian method also for a preliminary calcualation of the static polarizability of Lithium 2 S ground state,
working with a minimal basis set (1s2 2s, 1s2 2p, 1s2 3s, 1s2 3p). We
calculated the energy E of the free atom and the energy E 0 of the
atom in the presence of a 10 5 a.u. electric eld (E ). Then exploiting the approximated relationship that links the energy di erence
to the static polatizability zz
1
E0 = E
E
(3.55)
2 zz
we obtained for zz the result of 25:0  10 24 cm3 , to be compared
with the experimental value of 24:3(0:5)  10 24 cm3 given in the

124
literature [18].

Bibliography
[1] O. Goscinski, Preliminary Research Report No. 217, Quantum
Chemistry group, Uppsala University (1968).
[2] J. Avery,

J. Mol. Struct. (Theochem), 458, 1 (1999).

J. Math. Chem., 21, 285 (1997).


[4] J. Avery and D.R. Herschbach, Int. J. Quantum Chem., 41,
[3] J. Avery,

673 (1992).

[5] V. Aquilanti, S. Cavalli, and C. Coletti,


(1997).

Chem. Phys., 214, 1

[6] V. Aquilanti, S. Cavalli, C. Coletti, D. Di Domenico and G.


Grossi, Int. Rev. Phys. Chem., 20, 673 (2001).
[7] J.P. Gazeau and A. Maquet,

J. Chem. Phys., 73, 5147 (1980).

[8] J. Avery, Hyperspherical Harmonics and Generalized Sturmians. Kluwer Academic Publishers, Dordrecht, The Netherlands
(2000).
[9] V. Aquilanti and J. Avery,

Chem. Phys. Lett., 267, 1 (1997).

[10] B. R. Judd. Angular Momentum Theory for Diatomic


Molecules. Academic Press, New York, 1975.

125

Adv. Quantum Chem. , 31, 201 (1999).


[12] V. Aquilanti and J. Avery, Adv. Quantum Chem., 39,
[11] J. Avery,

71

(2001).

Phys. Rev., 126, 1470 (1962).


[14] A.J. Thakkar and T. Koga, Phys. Rev. A, 50, 854 (1994).
[15] S.P. Goldman, Phys. Rev. A, 57, 677 (1998).
[16] G.W.F. Drake and Zong-Chao Yan, Chem. Phys. Lett., 229,
[13] C.L. Pekeris,

486 (1994).

[17] J.J. De Groote, M. Masili and J.E. Hornos


4755 (1998).

J. Phys. B, 31,

[18] R.W. Molof, H.L. Schwartz, T.H. Miller and B. Bederson Phys.
Rev. A, 10, 1131, 1974.

126

Chapter 4
Sturmian approach to one-electron
many-center systems: integrals and
iteration schemes
Summary

In this chapter we give exact formulas and computational recipes


for nuclear attraction integrals involving both the polar and the
parabolic (or Stark) Sturmian set (see Sec. 1.2). These integrals are
needed for the matrix elements entering the secular problem for the
construction of molecular orbitals within the Born-Oppenheimer
scheme. Such a secular problem can be formulated at di erent iteration levels for the one-electron Schrodinger equation, and their relative performances against truncation are quantitatively assessed.
127

128

4.1 Introduction
Shibuya and Wulfman rst proposed the use of suitably normalized
Coulomb orbitals as basis sets for one-electron molecules in quantum chemistry, exploiting the momentum space approach [1], where
these orbitals are related to hyperspherical harmonics. Later, this
matter was further developed by other researchers [2, 3, 8, 10{13,
19,23{27] who when working in con guration space generally refer
to these orbitals as Sturmians. In the literature there are also recent and extensive reviews and books [4, 20, 30]. By means of this
method, the Schrodinger equation is trasformed into an algebraic
eigenvalue problem, whose matrix elements are linear combinations
of nuclear attraction integrals (NAIs) involving Sturmians. Actually, it is possible to formulate the algebraic problem according to
di erent levels of iterations, which are not equivalent when using
a nite-size basis set, as necessary when truncation is applied in
explicit calculations. The most extensive results on the electronic
energy of H+2 were obtained by Koga and Matsuhashi, who demonstrated the extreme accuracy of the Sturmian method [13].
In Sec. 4.2 we reconsider the basic integrals S and W [1], for
polar and parabolic Sturmians; in Sec. 4.3 we show examples of calculations for H+2 in order to demonstrate that the Sturmian method
yields satisfactory accuracy also in the case of a small-size basis set,
and then we report a quantitative comparison among the performances of the di erent iterations of the wave equation.

129

4.2 Two-center Nuclear Attraction Integrals S and W


We recall that Sturmians orbitals are de ned in con guration space
as the set of functions f (x) which solve the equation


1 2 1 2
r + 2 p0
2

np0 
f (x) = 0
r

(4.1)

where p0 is a xed parameter and is a set of three hydrogenic


quantum numbers: the usual principal quantum number n is the
rst of the set, and the others depend on the coordinates chosen to
separate variables in Eq.(4.1). This can be done in four types of
coordinate systems (polar, parabolic, spheroidal, spheroelliptic; see
Chapter 1) [31]. Further alternatives are available in momentum
space (chapter 2); see [28,32] for a classi cation. In this chapter we
only deal with NAIs involving polar (unlm) and parabolic (Unm )
Sturmians, which permit closed form treatments. De nitions and
relationships between these two alternative Sturmian basis sets can
be found in Ref. [21]. These Sturmian hydrogenic sets obey a peculiar orthonormality condition:

nZ
1
dx f 0 (x) f (x) = 0 ;
p0
jxj

(4.2)

and are complete with no need for inclusion of the continuum.


An explicit expression for the integral [4]

nZ
R) = p0 dx unlm(x) jx1j un0l0m0 (x

n0 l0 m0 (
Snlm

R)

(4.3)

130
(R is the distance between the two centers) can be obtained working
in momentum space and exploiting angular momentum quantum
theory [1]. The following form for the integral is essentially due to
Dunlap [2]; we have rederived it according to our phase conventions
for the orbitals [21]:
n0 l0 m0 (
Snlm

R) = (

0
)l l+M 2


p30

8 0
n
>
>
< 2

hl0; m0; l; mjL; M i >>


:

v
u

1
n0 1
2

l0

N;L

[nn0 N (2l + 1)(2l0 + 1)]1=2 (4.4)


9

n 1
2
n 1

N 1 >
=
2 >
N 1
[

>
>
;

R)

NLM (

L)(N + L + 1)
1u
t (N
N +1;LM (R)
2
N (N + 1)
v
u
u
t

1 (N + L)(N L
2
N (N 1)

Here,
M
8
9

1)

N 1;LM (R)5

 m + m0 , h:: ; :: j ::i is a vector coupling coecient and

   >>=
   > is a 9j symbol.
>
>
:  >
;
>
>
<

The other two-center nuclear attraction integral,


n l m (R) =
Wnlm

00 0

1Z
1
dx unlm (x)
p0
jx Rj un0l0 m0 (x)

(4.5)

is often mentioned in the literature [1,8,10{13,16], but actually no


explicit expression has been published yet. Exploiting the expansion [4]

jx

1 r
1 X
<
=
x0 j r> l=0 r>

!l

4 X  0
Y (^x )Ylm (^x)
2l + 1 m lm

(4.6)

131
n0 l0 m0 (R):
we have obtained the following expression for the integral Wnlm
"

0 0
0 0
R) = 2 (n + l)!(n + lnn)!(0n(2l +l 1)(21)!(l00n+ 1)l

n0 l0 m0 (
Wnlm

l00

#
1)!(2l0 + 1) 1=2

^ )hl; 0 j l0; 0; l00; 0ihl0; m0; l00; m00 j l; mi


Yl00 m00 (R

(4.7)

nX
l 1

( 1)j
j =0 j !(2l + j + 1)!(n l

n0 X
l0 1

( 1)k
0
0 l0
k=0 k !(2l + k + 1)!(n

1)!

(l + l0 + l00 + j + k + 2)!
1)!
(2s)l00 +1

2l00
a l00 1
2s (l + l0 + l00 + j + k + 2)! X (2s)
+ e 2s 
a
!
a=0

00 +j +k+1 "
l+l0 lX
b=0

(l + l0

l00 + j + k + 1)!
b!

where s  p0 R. It is worthwile to notice that matrix elements


depend on s, not on R: this means that the we have to pick a value
for s, solve the secular equation to get p0 , and eventually obtain
the physical distance R as s=p0 . For R = 0 both (4.3) and (4.5)
reduce to (4.2). Furthermore, it can be shown that [1, 8]:
0

W =

00

=
(l + l0 + l00 + j + k + 2)
b+l00
(2
s
)
;
(b + 2l00 + 1)!

00 1
00
S 00 S 0
n

(4.8)

where the sum over 00 is unbounded, making Eq.(4.8) unsuitable


for direct calculation of W 0 (see however next section).
When using parabolic Sturmians as a basis set for the oneelectron multicenter systems, we have to evaluate S and W integrals

132
involving those functions. In Ref. [21], the question of the evalun0 0 m0 was tackled expanding parabolic functions by the
ation of Snm
polar ones, and then exploiting the result in Eq.(4.4). Here we show
n0 0 m0 is studied directly
the result that is obtained when integral Snm
in momentum space, according to Shibuya and Wulfman's spirit:
n  m (R) =
Snm

0 0 0

1
nZ 3
dx Unm (x) Un0 0 m0 (x
p0
jxj

R) =

(4.9)

!
nn0 1=2 X n 1  + m n0 1 0 + m0 n00 1 00 + m00
=2
h 2 ; 2 ; 2 ; 2 j 2 ; 2 i
p30
n00

n 1  m n0 1 0 m0 n00 1 00 m00  1 1=2


h 2 ; 2 ; 2 ; 2 j 2 ; 2 i n00 fUn00 00 m00 (R)
2v
u

00
1 4u
00 )2 m002
t (n + 1
Un00 +1;00 1;m00 (R)
4
n00 (n00 + 1)
v
u
u
t

v
u
u
t
v
u
u
t

(n00

1 + 00 )2 m002
Un00 1;00 1;m00 (R)+
n00 (n00 1)

(n00 + 1 + 00 )2 m002


Un00 +1;00 +1;m00 (R)
n00 (n00 + 1)

(n00

39

=
1 00 )2 m002
5
00
00
00
U
(
R
)
n
1
;
+1
;m
;
n00 (n00 1)

where m00  m + m0 0and


00   + 0 .
0 m0
n

The integral Wnm in the parabolic basis set is introduced
in this work for the rst time. For its evaluation, the potential is
expanded according to Eq.(4.6), and then the parabolic coordinates
appearing as variables in the Sturmians Unm are transformed into
n0 0 m0 is solved in
polar coordinates r; ; , so that the integral Wnm
polar coordinates. The integrations over r and  are easy; for the
integration over , the spherical harmonic Yl00 ;m00 (; ) is expressed

133
in terms of a Jacobi polynomial (Ref. [17], page 138); then the
integration over  is performed by Eq. 7.391(2) of Ref. [18], yielding
a generalized hypergeometric series 3 F2 (1); the latter is replaced by
a Clebsch-Gordan coecient, according to Eq.(22) at page 241 of
Ref. [17]. The nal expression is easily programmable in Fortran
or in Mathematica language:
Z

0 0 0R

n  m
Wnm

n+

nn0

= 2

 0
n + 0

n  m

2
X
a

0 +0

2
X
c

1)

c!(c +

m0

)!

^
Yl00 ;m00 R

1)

"

0 +0 1
2

00

 (2l

00 +1

(2s)l

( +

00

+ 2)!

 a
;

b+c

2s

+ 00 +2

X
l

=0

( +

00
i!

+ 0
2

b+d+ m m

00

00

m0

 i1=2

1)

n m 

1)

)!

+ 0 +m00 ! a + c + m+m0
2
2
( + l00 + 1)!

( + 1)(

d!(d +

a+c+ m m

+ 1)

+ 0 +m00 !
2

+m

0 1

=0

hl00 ; m00

2
X

(4.10)

 0
n

b!(b + m)!

0 0

=0



2
X

+ 1
2

"

n++m

n m 

a!(a + m)!

( ) =

 0
n + 0 + m0

b+d+ m m



x jx Rj Un00 m0 x

n  m
(

=0

00

0 1


X

=0

dx Unm (



 0
n

) =

0 0 1
2

00 

00  #1=2

)!

 a
;

#
+ 2)!

(2s)

00

b+c

d+m
2

X
l00 +1
00
s l e 2s
j =0

(2 )

(

00

j!

)
+ 1)!

(2s)

where   a + b + c + d + m + m0 . The explicit expressions for


the simplest cases of the integral in Eq.(4.10) are given in Table
n0 l0 m0 integrals (Eq.4.5) can be found in
4.1. The analogue for Wnlm
Refs. [4, 16].

134
Table 4.1: Nuclear attraction integrals W involving parabolic Sturmians, Eq.(4.10).
n; ; m n0 ; 0 ; m0

1,0,0
1,0,0
1,0,0
2,-1,0
2,1,0
2,-1,0

1
n0 0 m0 (R) = 1 R dx3 U
Wnm
nm (x) jx Rj Un0 0 m0 (x)
p
1 e s (1 + s)
s
s
0

1,0,0
2,-1,0
2,1,0
2,-1,0
2,1,0
2,1,0

2 s+e

(2+5
p s+6s +4s )
2 s 2s
2 s e p(2+3s+2s )
2 2s
3+3s+2s e s (3+9s+14s +13s +8s +4s )
2s
3 3s+2s e s (3+3s+2s +s )
2s
3+e s (3+6s+6s +4s +2s )
2s
2s

4.3 Solution of one-electron multicenter wave equation; comparison among


performances of di erent iterations
Let us consider the Schrodinger equation for one electron in the
presence of several xed nuclei, that will be labeled by index a. We
require that all functions in the set have the same exponent p0 , and
constrain this exponent to the (unknown) electron energy p20 =2.
Expanding the wavefunction by Sturmian orbitals centered on the
nuclei, one has:
X

"

p2
1 2
r
+ v (x) + 0 c f = 0
2
2

(4.11)

where  = fa; g and

v (x) =

va (x) =

Za

jx Xaj

(4.12)

135
Exploiting Eq.(4.1), and dividing Eq.(4.11) by p0 , one obtains the
generalized eigenvalue problem [4, 9]:
X

(A; 0

p0 C; 0 ) c = 0

(4.13)

where
1Z 3 
dx f 0 (x)v (x)f (x)
p0
n Z 3  va (x)
C; 0 =
dx f 0 (x)
f (x)
p0
Za 

A; 0 =

(4.14)
(4.15)

These matrix elements are linear combinations of NAIs W and S . In


the example discussed in this chapter, applications of the Sturmian
method are referred to a one-electron molecule with nuclear charges
Z1 and Z2 . In this case, one has (subscripts 1 and 2 refer to the
two nuclei) [4]:

Z1
0
; 0 + Z2 W (R)
0
n

Z1 Z2  0
+
S (R)
Af1; g;f2; 0 g =
n n0
Cf1; g;f1; 0 g = ; 0
0
Cf1; g;f2; 0 g = S (R)
Af1; g;f1; 0 g =

(4.16)
(4.17)
(4.18)
(4.19)

These matrix elements are sucient for Sturmian calculations on


any two-center one-electron molecule. We have considered H+2 , for
which Z1 = Z2 = 1. Sturmian orbitals have the correct exponential
behavior on the nuclei, thus noticeable accuracy can be achieved
using a small basis set, as can be seen in g.(4.1).

136

-0.6

-0.8
{1s}
-1

{1s, 2s}

-1.2

E ( = - p0 / 2 )

{1s, 2s, 2pz}

-1.4

-1.6

-1.8

-2

R ( = s / p0)

Figure 4.1: Energy of the electron ground state in H+2 . The circles

stand for the 10- gure accuracy data by Koga et al. [13]. The three
curves represents the results of Eq.(4.13) for three di erent polar basis
sets: f1sg; f1s; 2sg; f1s; 2s; 2pz g.

137
As it is clearly demonstrated in Ref. [4], there is an interesting
relationship between matrix elements A; 0 and C; 0 , which could
not be gured out at a rst glance but is connected to Eq.(4.8):

A; 0 =

 00

Z 00
C; 00 00 C 0 ; 00
n

(4.20)

This allows a simpler form for obtaining eigenvalues and eigenvectors of the one-electron system [4, 9]
X

0s
@

Za Za0
A 0
nn0 ;

; 0 p20 A c = 0

(4.21)

An even simpler form was given by Novosadov [23] (see also [2]):
X

0s
@

Za Za0
C 0
nn0 ;

; 0 p0 A c = 0

(4.22)

Eqs.(4.21) and (4.13) can be regarded as, respectively, the second


and third iteration of Eq.(4.22). In the limit of an in nite basis set, Eqs.(4.22), (4.21) and (4.13) would be equivalent, while
when a nite expansion is exploited, Eq.(4.13) is more accurate
than Eq.(4.21), which in turn is more accurate than Eq.(4.22). The
reason is that the A matrix elements can be computed directly and
analytically, as linear combinations of S and W : otherwise, if we
were going on to utilizing Eq.(4.20) for the calculation of A truncating the summation according to a chosen dimension for the basis set,
the three equations would be exactly equivalent. It is interesting
to provide a quantitative comparison of the di erent performances
of equations (4.13), (4.21) and (4.22) over H+2 . For the solution of

138
eigenvalue problems and generalized eigenvalues problems, we have
used Lapack libraries written in Fortran, available on the Internet.
Fig.(4.2) clearly demonstrates that Eq.(4.13) is much more convenient than the others. In order to obtain more accurate results
than those reported in Fig.(4.2) for Eq.(4.13) by enlarging the basis set, we encountered problems of a computational nature, due to
excessive closeness to zero of some elements of the C matrix. These
diculties can be interpreted as an indication of the tendency of
the basis set towards overcompleteness [16], when its dimension is
increased: in other words, since C matrix can be viewed as an overlap matrix in a Sobolev space [33], zero eigenvalues are signatures
for linear dependence among functions in the basis set.
The most direct way for avoiding such a problem was proposed
by Lowdin and others [14, 15], who suggested to proceed with an
expansion of the molecular orbital in terms of a complete set centered on a single point in the space, which will be the midpoint in
the H+2 case. It is interesting to explore the suggestion, inserting
such a single-point expansion in Eq.(4.11): one obtains the secular
problem:
Xh

nl

00

0 0i

n l m (s=2) k n l c
2Wnlm
 nl
nlm; = 0

(4.23)

where m is 0; 1; : : : for states ; ; : : :. The secular equation turns


out to be split in two blocks because of the symmetry of H+2 : indeed, the sum involves only even l's for gerade states, and only odd
l's for ungerade states. Results obtained by solving Eq.(4.23) are

139

-1

1.0010

1.0010

-2

Err. (atomic units)

first iteration

1.0010

1.0010

-3

second iteration

-4

1.0010

1.0010

-5

third iteration

-6

50

100

cpu time (arb. units)

Figure 4.2: Comparison among the performances of three di erent


iterations of the wave equation [ rst order, Eq.(4.22); second order,
Eq.(4.21); third order, Eq.(4.13)]. Data refer to H+2 , for an internuclear
distance of 2 a.u.; polar Sturmians were used as basis set. The absolute
error in atomic units is computed with respect to the benchmark results
by Koga et al. [13].

140
given in Fig.(4.3). As the internuclear distance increases, a larger
and larger basis set would be necessary to approximate correctly
the H+2 wavefunction, and one should face the problem of computn0 l0 m0 matrix elements for high quantum numbers, but the
ing Wnlm
accuracy obtained at such an elementary level is most encouraging.

4.4 Further remarks and conclusions


This chapter has considered the calculation of integrals which appear on matrix elements for one electron multicenter problems in
the Sturmian approach. A brief investigation has also been carried out to assess the relative eciencies of alternative iteration
schemes, where these integrals play a role. Relative merits of the
polar versus parabolic basis need to be studied more extensively
(for some preliminary results, see [21]).
In the perspective extension of the use of Sturmian sets to the
solution of molecular problems of increasing complexity, it is important to develop ecient algorithms for the computation of bielectronic multicenter repulsion integrals, treated in the next paper.
To this aim, we note that the Sturmian orbitals can be written as
linear combination of Slater-type orbitals (STO): therefore, results
and codes developed for the calculation of integrals between STO
(for example see Refs. [5{7]) can be usefully employed for carrying
out calculations by the Sturmian approach: test cases for molecules
are discussed elsewhere [35].

141

Energy of the electron in the H2 ground state (1g)


-0.6

-0.8

-1.2

E = - k /2 (a.u.)

-1

-1.4

-1.6

-1.8

-2

2
R = s / k (a.u.)

Figure 4.3: The circles stand for the 10- gure accuracy data by Koga

et al. [13]. The three curves represents the results of Eq.(4.23) for three
di erent basis sets: fnmax ; lmax g = f3; 2g; f6; 5g; f8; 6g.

142

Bibliography
[1] T. Shibuya and C.E. Wulfman,
(1965).
[2] Brett I. Dunlap

Proc. Roy. Soc., A286, 376

Chem. Phys. Lett., 30, 39 (1975).

[3] B. R. Judd, Angular Momentum Theory


Molecules. Academic Press, New York (1975).

for Diatomic

[4] J. Avery, Hyperspherical Harmonics and Generalized Sturmians. Kluwer Academic Publishers, Dordrecht, The Netherlands
(2000).
[5] J.F. Rico, R. Lopez and G. Ramrez,
(1988).

J. Comp. Chem., 9, 790

[6] J.F. Rico, R. Lopez, M. Paniagua and G. Ramrez,


Phys. Commun., 64, 329 (1991).
[7] J.F. Rico, R. Lopez, I. Ema and G. Ramrez,
Commun., 105, 216 (1997).
[8] T. Koga,

Comp.

Comp. Phys.

J. Chem. Phys., 83, 2328 (1985).

[9] T. Koga and R. Kawa-ai

J. Chem. Phys., 84, 5651 (1986).

[10] T. Koga and T. Matsuhashi,

J. Chem. Phys., 87, 1677 (1987).

[11] T. Koga and T. Matsuhashi,

J. Chem. Phys., 87, 4696 (1987).

143

J. Chem. Phys., 88, 1110 (1988).


[13] T. Koga and T. Matsuhashi, J. Chem. Phys., 89, 983 (1988).
[14] H. Shull and P.-O. Lowdin, Svensk Kem. Tidskr., 67, 373
[12] T. Koga and T. Matsuhashi,

(1955).

J. Chem. Phys. , 30, 627 (1955).


[16] J. Avery and R. Shim, Int. J. Quantum Chem., 83, 1 (2001).
[15] K.M. Howell and H. Shull,

[17] D.A. Varshalovich, A.N. Moskalev, and V.K. Khersonskii.


Quantum Theory of Angular Momentum. World Scienti c,
Singapore, 1988.
[18] I.S. Gradshteyn and I.M. Ryzhik. Table of Integrals, Series
and Products. Academic Press, New York, 1980.
[19] J. Avery, Hyperspherical Harmonics, Applications in Quantum
Theory. Kluwer Academic Publishers, Dordrecht, The Netherlands (1989).
[20] J. Avery,

Adv. Quantum Chem., 31, 201 (1999).

[21] V. Aquilanti, S. Cavalli, C. Coletti, and G. Grossi,


Phys., 209, 405 (1996).
[22] V. Aquilanti, S. Cavalli and C. Coletti,
(1997).
[23] B.K. Novosadov

Chem.

Chem. Phys., 214, 1

Opt. Spectrosc., 41, 490 (1976).

144
[24] Ch. Duchon, M.Cl. Dumont-Lepage and J.P. Gazeau, J. Phys.
A, 15, 1227 (1982).
[25] Ch. Duchon, M.Cl. Dumont-Lepage and J.P. Gazeau, J. Chem.
Phys., 76, 445 (1982).
[26] H.J. Monkhorst and B. Jeriorski,
(1979).
[27] V. Aquilanti and J. Avery,

J. Chem. Phys., 71, 5268

Chem. Phys. Lett., 267, 1 (1997).

[28] V. Aquilanti, S. Cavalli and C. Coletti


3209 (1998).

Phys. Rev. Lett., 80,

[29] V. Aquilanti, S. Cavalli, C. Coletti, D. Di Domenico and G.


Grossi, Int. Rev. Phys. Chem., 20, 673 (2001).
[30] V. Aquilanti and J. Avery,
(2001).

Adv. Quantum Chem., 39,

[31] V. Aquilanti, A. Caligiana and S. Cavalli,


Chem., 92, 99, (2003).

Int. J. Quantum

[32] V. Aquilanti, A. Caligiana, S. Cavalli and C. Coletti,


Quantum Chem., 92, 212, (2003).
[33] E.J. Weniger

71

Int. J.

J. Math. Phys., 26, 276 (1985).

[34] URL: http://bong.chm.unipg.it/downloads/download.html

145

Sturmian Orbitals in Quantum Chemistry: An Introduction in "Fundamental world of

[35] V. Aquilanti and A. Caligiana,

quantum chemistry", Vol. I, 297-316, E. J. Brandas and E. S.


Kryachko (Eds.), 2003 Kluwer Academic Publishers.

146

Chapter 5
Sturmian Orbitals Applied to
Molecules: an Introduction
Summary

The Sturmian basis set approach to the solution of quantum


chemical problems is the con guration space counterpart of hyperspherical analysis in momentum space and leads to promising algorithms alternative to the usual SCF-plus-con guration interaction
hierarchies. In this chapter we introduce the basic steps of a computational scheme for molecular structure problems, where Sturmian
orbitals appear as building blocks in a valence-bond type of approach. The example is the hydrogen molecule. The mathematical
ingredients are considered, particularly the possibility of exploiting
relationships with Slater-type orbitals in order to circumvent the
bottleneck of calculation of matrix elements.
147

148

5.1 Introduction
Being the Schrodinger equation (SE) for the non-relativistic hydrogen atom analytically soluble in a closed form, the use of hydrogenoid orbitals as basis sets to approximate the wave functions
of atoms and molecules has been extensively considered since the
early days of quantum mechanics. Slater-type orbitals (STO) and
other variants have been proposed and amply tested as alternatives. In order to tackle the obstacles encountered for the ecient
computation of matrix elements (speci cally the bielectronic integrals), preference has been given to the Gaussian orbitals (GTO),
particularly convenient for such purposes.

The current software available for ab initio calculations of atomic


and molecular properties, typically exploiting Gaussian basis set
expansions, almost invariably initiate with Self Consistent Field
(SCF) or Multicon guration Self Consistent Field (MC-SCF) calculations yielding molecular orbitals as linear combinations of GTO
centered on the various nuclei of the molecule, and then proceed
with post-Hartree-Fock techniques (Con guration Interaction, Coupled Clusters, ...). Among the disadvantages of such a procedure,
we mention the diculties encountered in the calculation of longrange interactions and of the properties of excited states, and the
slow convergence in reproducing the cusps in the electronic densities
at nuclei.
The Sturmian orbitals can be straightforwardly expressed as

149
linear combinations of Slater-type orbitals. The use of the latter
has been superseded by the GTO, for which the slower convergence is often amply balanced by the much simpler algorithm for
the calculation of integrals. Apart from this, the employed techniques are identical. The approach involving Sturmian orbitals will
be shown to proceed along a route fully alternative to modern ab
initio quantum chemistry: it will be seen that there is no need of
iterative Hartree-Fock calculations, because the introduction of the
Sturmian expansion of the electronic wave function leads after algebraic manipulations to a generalized eigenvalue problem (GEP)
where the matrix elements contain mono- and bi-electronic integrals
among Sturmian orbitals and the eigenvalue is not the electronic
energy E but the momentum wave number p0 such that E = p20 =2
(atomic units are implied). The typical integrals of quantum chemistry depend explicitly on the nuclear coordinates Ri , while in the
Sturmian approach parameters si = p0 Ri appear. The basis feature
is therefore that the Born-Oppenheimer separation is not settled before the calculations. The si parameters are instead xed and only
after the solution of the GEP from the relationships Ri = si =p0 one
nds out what is the point in the potential energy surface which
corresponds to an energy p20 =2. Another unusual property of this
approach is that the eigenvalues obtained by solving the GEP come
out as correct solutions not only for ground states, but also for the
excited states corresponding to the same symmetry.
We consider the application of the Sturmian orbital procedure

150
to the ab initio quantum chemistry of molecules as many-electron
many-nuclear systems. The recent proposal, as reviewed in Avery's
book [10, 32], involves a MO-LCAO approach where one-electron
molecular orbitals are computed rst, and then their products are
employed to build-up N -electron con gurations. Such an approach
is currently being implemented, one diculty being the recovery
of the Born-Oppenheimer separation, because, as indicated above,
the matching of si parameters and Ri coordinates occurs through
scaling by the eigenvalues p0 .
This chapter presents an account of a promising method for the
application of the Sturmian orbital approach to molecular structure calculations. The new proposal (Sec.5.2.1) for the extension of
generalized Sturmian sets to molecular structure is founded on concepts and formalisms previously elaborated. In the framework of
quantum chemistry jargon, it will be recognized as a valence-bond
type of approach: the calculation of the potential energy curve for
the H2 molecule provides encouraging results already at the level
of minimal basis sets. Sec.5.2.2 is devoted to a discussion of the
practical obstacles to the implementation of this approach, the calculation of integrals, which can be in part circumvented through
the angular momentum algebra applicable in the momentum space
perspective, and in part tackled by exploiting the connections with
STO and the existence of programs dealing with integrals involving
the latter. Conclusion are in Sec.5.3.

151

5.2 Sturmian basis sets for atomic and


molecular calculations
There are four di erent coordinate systems (polar, parabolic, spheroidal
and spheroelliptical) which allow separation and solution of the partial di erential equation (4.1): we can associate to each of them a
Sturmian basis set corresponding to alternative symmetry properties [21] (Chapter 1). For de niteness, here we will only deal with
the most common one of these sets, which employs polar coordinates (r; #; ').
5.2.1

Many-center many-electron systems

In this section we propose an extension of the Sturmian formalism


for the study of systems having many nuclei and electrons. The
SE for a molecule coincides with Eq.(3.1), provided that we de ne
V0 (x) according to
N
X
XX
Za
(5.1)
V0 (x) = V0;a (x) =
a i=1

jxi Xaj

We expand the electronic wavefunction as in Eq.(3.5), and the basis


functions  (x) are de ned by an equation analogous to Eq.(3.7):
"
#
1 2
p20
z
r +  Vp (x) + 2  (x) = 0
(5.2)
2
The potential Vpz (x) is the sum of N contributions, each of them
representing the interaction of one electron with one of the nuclei:
N
X
Zp;j
(5.3)
Vpz (x) =
j =1

jxj Xp;j j

152
This means that actually there is not only one equation de ning the
basis set, because there are many choices for Vpz (x): the basis set
can be divided into several subsets, each of them associated to an
equation like (5.2) having a di erent Vpz(x). A function  (x) is a
determinant of Sturmian orbitals u (x Xa ) like Eq.(3.5), provided
that two subsidiary conditions are satis ed: the rst one is Eq.(3.8),
and the second is a modi cation of Eq.(3.9) taking into account
that in general Sturmian orbitals appearing in the determinants
are centered on the di erent nuclei of the molecule:

nk n0 k0 n00 k00


=
=
=    = 
Za0
Za0
Za00

(5.4)

A combination of Eqs.(5.4) and (3.8) leads to the relationship:

 =

p0

( Zna )2 + ( Zna00 )2 + ( Zna0000 )2

+

(5.5)

This approach can be interpreted as a Valence Bond (VB) method,


since we perform CI over con gurations which are products of nonorthogonal atomic orbitals. In order to clarify the concepts exposed
and show how this method works in practice, we will show an application to the simplest molecule: H2 . The crudest approximation
to the ground state 1 g wavefunction is:
g (x1 ; x2 ) =
1

p1 [1sA(1)1sB (2) + 1sB (1)1sA(2)] 


2

 p12 [ (1) (2) (1) (2)]

(5.6)

153
In this case the basis set is constituted by two bielectronic functions;
1 (x) = 1sA(1)1sB (2) corresponds to the potential

V1z (x) =

(5.7)

jx1 XAj jx2 XB j

in Eq.(5.2), and 2 (x) = 1sB (1)1sA(2) corresponds to

V2z (x) =

(5.8)

jx1 XB j jx2 XAj

We can express the overall interaction between electrons and


nuclei as:
V0 (x) = V1z (x) + V2z(x)
(5.9)
According to Eq.(1.7), 1s Sturmian orbitals are de ned as
!
k3 1=2
1s(1) =
e


k r

(5.10)

The exponent k can be computed combining Eqs.(5.4) and (5.5):


p
it is p0 = 2. In general, the exponent of an orbital will depend
on the principal quantum numbers of all the orbitals forming the
con guration  (x) in which it appears.
Now we will calculate an approximation to H2 electronic energy
exploiting the wave function (5.6). Entering Eq.(5.6) inside (3.1)
and writing the potential V0 (x) as in Eq.(5.9) one can obtain the
1x1 secular problem:
Z
"

[1sA (1)1sB (2) + 1sB (1)1sA(2)]


#

(5.11)

1 2
1 p20
z
z
r + V1 (x) + V2 (x) + r + 2 1sA(1)1sB (2)d1d2 = 0
2
12

154
The latter expression can be simpli ed with the aid of Eq.(5.2):
Z

[1sA (1)1sB (2) + 1sB (1)1sA(2)]

(5.12)

1
(1 1 )V1z (x) + V2z(x) +
1sA (1)1sB (2)d1d2 = 0
r
12

Taking into account that the basis set functions associated to the
same operator Vpz (x) enjoy potential weighted orthonormality relationships
Z
2E
dx 0 (x)Vpz(x) (x) =  0 ;
(5.13)

we can modify Eq.(5.12) as follows:

Z
1
2E
z
2E + 1sA(1)1sB (2) V2 (x) +
1s (1)1sB (2)d1d2 (5.14)
1
r12 A


+ 1sB (1)1sA(2) (1

1
z
z
1 )V1 (x) + V2 (x) +
1sA (1)1sB (2)dx = 0

r12
If we write down all the integrals explicitly and exploit Eq.(5.5),
we can rewrite Eq.(5.14) as:

2p0 +p0

1s2B (2)d2

1s2A (1)

jx1 XB j

d1

1s2A(1)d1

+(1

"

"

1sA (1)1sB (1)d1


Z

jx2 XAj d2

1s2A (1)1s2B (2)


d1d2
+
r12
Z
Z
1sA (1)1sB (1)
1sA (2)1sB (2)d2
d1
Z

pp0 )
2

1s2B (2)

1sA (2)1sB (2)d2

jx1 XAj

1sA (2)1sB (2)

jx2 XB j d2

1sA (1)1sB (1)

jx1 XB j d1

(5.15)

155
Z

1sA (1)1sB (1)d1

1sA(2)1sB (2)

jx2 XAj d2

1sA (1)1sB (1)1sA(2)1sB (2)


d1d2 = 0
r12
Eventually, we divide the latter by p0 and write the NAI's in a
more compact form, exploiting the de nitions (4.3) and (4.5). The
expression for the eigenvalue is:
+

"

p0 =

2+

100 (s) + 1
2W100
p0

1sA (1)1sB (1)1sA(2)1sB (2)


d1d2+
r12
(5.16)
#

h
i
p
1 Z 1s2A (1)1s2B (2)
100 (s) = 1 + M100 (s)S 100 (s)
d1d2 2 2M100
(
s
)
S
100
100
100
100
p0
r12
The M's are two-center overlap integrals between Sturmians:
Z

0 l 0 m0
Mnnlm
 un0l0m0 (x)unlm(x)dx

For example,

1
M100
100 = 3 e

s (3 + 3s + s2 )

(5.17)
(5.18)

We can show (see Appendix) that these M integrals can be evaluated in momentum space exploiting angular momentum algebra,
analogously to the case of the S integrals encountered in Chapter
4.
Although p0 is in the de nition of the exponent k , mono- and
bi-electronic integrals in Eq.(5.16) actually do not depend on the
value of p0 : rather, they depend on the product s = k  R (R
being the internuclear distance). Therefore, we pick up a value for
s, compute the matrix elements and then we can get the eigenvalue

156

p0 ; eventually, the physical internuclear distance is obtained as R =


p
2s=p0 .
Results on the potential energy curve for the ground state of
H2 , calculated by this procedure, are reported in Fig.(5.1). Even
if really minimal, the basis set of Eq.(5.6) gives a good qualitative
picture of the chemical bond: the equilibrium distance is close to the
real one, and the asymptotic behavior is correct. A more elaborate
calculation is also shown in Fig.(5.1). We have tested that even
modest size calculations employing larger basis sets yield accurate
results also for excited states.
5.2.2

The integrals. Relationships with Slatertype orbitals

The evaluation of integrals is always an important issue of quantum


chemical methods, because it can involve large e orts in terms of
computer times and therefore may limit the concrete possibility of
extending the calculations to cover complex systems. This discussion will pertain speci cally to the mono- and bi-electronic integrals
encountered in Sec.5.2.
Generally, the use of a basis set as de ned in Eq.(5.2) leads to
monoelectronic integrals S ; W and M containing two orbitals which
can di er in the exponents appearing in Eq.(1.7). This represents
a limit for the application of momentum space harmonic analysis, at least according to its present developments. To circumvent
this bottleneck and yet be able to test the perspective power of

157

H2 ground state ( g) potential energy curve

-1

V(R) (a.u.)

-1.05
(c)

-1.1

(b)
-1.15

(a)

3
R (a.u.)

Figure 5.1: Comparison between the benchmark

ab initio potential
energy curve for H2 denoted as (a), and the Sturmian approach. Curve
(c) corresponds to the single con guration function in Eq.(5.6) and curve
(b) has been obtained including in the expansion all con gurations from
all Sturmian orbitals up to 3d. In the latter case, the variance in the
energy minimum with the one given in [28] is +0.3%.

158

Table 5.1: Data corresponding to curve (a)


R (a.u.)
1
1.2
1.3
1.39
1.4
1.4011
1.5
1.6
1.8
2
2.2
2.4
2.6
2.8
3
3.2

V(R) (a.u.)
-1.1245388
-1.1649344
-1.1723462
-1.174452
-1.1744748
-1.174475
-1.1728541
-1.1685821
-1.1550675
-1.1381316
-1.1201304
-1.1024201
-1.0857874
-1.0706776
-1.0573174
-1.0457865

159

Table 5.2: Data corresponding to curve (b)


R (a.u.)
1.0821753
1.2020791
1.3254692
1.4521964
1.5820863
1.7149417
1.8505432
1.9886483
2.1289914
2.2712856
2.4152266
2.560499
2.7067849
2.8537741
3.0011745
3.1487233
3.2961955
3.4434107
3.590235
3.7365797

V(R) (a.u.)
-1.1423625
-1.1611956
-1.1694278
-1.1701995
-1.1657679
-1.1577794
-1.1474508
-1.1356935
-1.1231972
-1.1104875
-1.0979637
-1.0859263
-1.0745945
-1.0641206
-1.0545992
-1.046076
-1.0385553
-1.0320072
-1.0263758
-1.0215866

160

Table 5.3: Data corresponding to curve (c)


R (a.u.) V(R) (a.u.) R (a.u.) V(R) (a.u.)
0.7290924 -0.90468312 2.6097341 -1.0278317
0.80695213 -0.97216731 2.7124688 -1.0231117
0.88703841 -1.0204927 2.8151786 -1.0188752
0.96934499 -1.0543036 2.9178118 -1.0151008
1.0538422 -1.0770535 3.020328 -1.0117613
1.1404765 -1.0913646 3.1226974 -1.0088259
1.2291706 -1.0992606 3.2248993 -1.0062621
1.3198246 -1.102322 3.3269203 -1.0040368
1.412318 -1.1017926 3.4287536 -1.0021176
. 1.5065128 -1.0986553 3.5303976 -1.000473
1.6022571 -1.0936868 3.6318546 -0.99907337
1.6993889 -1.0874984 3.7331303 -0.99789094
1.7977411 -1.0805681 3.8342327 -0.9969001
1.8971455 -1.0732652 3.9351713 -0.99607738
1.9974369 -1.0658709 4.0359568 -0.99540153
2.0984569 -1.0585944 4.1366007 -0.99485338
2.2000568 -1.0515868 4.2371146 -0.99441577
2.3021
-1.0449529
4.33751 -0.99407342
2.4044637 -1.0387603 4.4377982 -0.99381281
2.5070395 -1.0330478 4.5379902 -0.99362199

161
the method, we explored the possibility of making recourse to the
relationship between Sturmian orbitals and STO.
In practice, since Rico and collaborators have proposed, implemented and made available an ecient program for the calculation
of two-center monoelectronic integrals involving STO, based on the
recurrence relationships which they enjoy [29, 30], and since the
Sturmian orbitals can be given as simple linear combinations of the
STO (explicitly, indicating the latter as nlm :

u1;0;0
u2;0;0
u2;1;m
u3;0;0
u3;1;m
u3;2;m
u4;0;0
u4;1;m
u4;2;m

= 1;0;0 p
= 1;0;0
3 2;0;0
= 2;1;m p
p
= q
2;0;0 + 10 3;0;0
1;0;0 2 3 p
= 2 23 2;1;m
5 3;1;m
(5.19)
= 3;2;m p
p
p
= 1;0;0 3 3 q2;0;0 + 3 10 3;0;0
35 4;0;0
p
p
3
= 5 2;1;m p
5 2 3;1;m + 21 4;1;m
3
= p2 3;2;m
7 4;2;m

etc.) we found it expedient to use the program in order to calculate


the bicenter monoelectronic integrals S and M. The W integrals
can instead be calculated analytically even though the exponents
di er.
However, this program doesn't calculate the three-center monoelectronic integrals
Z

dx unlm (x)

jx Xa j un0l0m0 (x Xa0 )

which can occur in molecules with three or more nuclei.

(5.20)

162
The most demanding e ort for our valence-bond Sturmian approach appears, as usual in quantum chemistry, when the calculation of the bielectronic multicenter integrals is needed:
Z Z

dxdx0 uA (x)uB (x)

0
0
jx x0 j u (x )u (x )
C

(5.21)

Our test cases presented in Sec.5.2.1 were tackled in a rather tortuous manner, making recourse to another program due to Rico
and coworkers [31] which in turn in order to obtain integrals for
STO expands them as linear combinations of GTO. The resulting
Gaussian integrals are nally calculated using available techniques.
The ensuing disadvantage of this inelegant procedure is that, since
a satisfactory t of an STO requires about 10 Gaussians, an integral involving four STO's requires the calculation of 104 integrals
among GTO. Moreover our integrals such as (5.21) are linear combinations of integrals among STO: for example an integral involving
four 3s-Sturmians requires a sum over 34 integrals of the STO program, and therefore 34  104 integrals over Gaussians. Therefore
this program, which recommends itself for the accuracy which can
be adjusted due to exibility in the choice of the number Gaussians required to approximate a STO and satisfactorily covers the
range of permitted quantum numbers (nmax = 6; lmax = 3), although appropriate for test calculations, must be superseded by
more direct approaches to tackle integrals such as Eq.(5.21). Alternatives should also be explored, involving for example elliptic or
other suitable coordinates.

163

5.3 Further remarks and conclusions


The account presented in this chapter on developments regarding
the use of Sturmians orbitals in quantum chemistry demonstrates
how fruitful was the idea put forward long ago by Shull and Lowdin
of "natural spin orbitals". The rst steps of the proposed approach
show perspectives of advantages over conventional methods, particularly for excited states. Not considered in this chapter have been
the alternatives which open by exploiting the various ways available
to separate variables for Eq.(4.1): they have been recently classi ed
both in con guration [21] and in momentum space [22]. Since the
computation of integrals is a feature which will eventually decide
the crucial potentialities and scope of the Sturmian approach, the
alternative basis sets and the relationships among them are an important goal to further study; some progress have been made for
the parabolic set [12, 13].
Needless to say, these orbitals might nd a speci c role also
whenever in quantum chemistry there is a need for an expansion
in a complete basis set. Further extensions which are of relevance
for the application to heavier systems are those involving the inclusion of relativistic e ects: see Chapter 8 in Ref. [10] and recent
work by Smitkowski [34{37]. For the particular scheme introduced
in Sec.(5.2.1) the numerical results on H2 indicate that the development of a direct algorithm for the calculation of bielectronic
multicenter integrals would greatly contribute to practical imple-

164
mentations to systems of increasing complexity. Our current e ort
on this problem shows the importance of using momentum space
techniques, which allow the exploit of powerful angular and hyperangular momentum algebra.

Appendix: Analytical solution for twocenter overlap integrals

Let us consider the overlap integral between two polar Sturmians


centered on two di erent nuclei a and b, separated by a distance R:
b;n0 l0 m0
a;nlm

 ua;n0l0m0 (x)ub;nlm(x)dx

(5.22)

Working in momentum space, the latter is equivalent to:


Z

a;n0 l0 m0 (p) b;nlm (p)dp =

ipR

nlm (

p) n0l0m0 (p)dp

(5.23)

Inserting in Eq.(5.23) the plane wave expansion (Eq. 2.52), the


relationship between nlm (p) and hyperspherical harmonics (Eq.
2.4), and the volume element in p-space:
!
k2 + p2 3
d

dp =
2k

(5.24)

we get an integral involving hyperspherical harmonics:


0 l0 m0
3=2
Mb;n
a;nlm = (2 )
Z

n00 l00 m00

un00 l00 m00 (x)

4 k 5= 2
d
2  2 2 Ynlm (
)Yn0 l0 m0 (
)Yn00 l00 m00 (
)
(p + k )

(5.25)

165
One can easily transform the fourth relationship in Eq.(2.3) into
(1 + cos )2 =

4k5=2 3=2
k
(k2 + p2 )2 

(5.26)

Then the integral becomes:


!
2 3=2 X
=
un00 l00 m00 (x)
k
n00 l00 m00

b;n0 l0 m0
a;nlm

M
Z

(5.27)

d
(1 + 2 cos  + cos2 )Ynlm (
)Yn0 l0 m0 (
)Yn00 l00 m00 (
)

The product of two hyperspherical harmonics can be expressed


as (Eq. 2.32 in Ref. [6]):

Ynlm (
)Yn0 l0 m0 (
) =

N;L;M

"

#
nn0 N (2l + 1)(2l0 + 1) 1=2
2 2

8 0
n
>
>
< 2

hl0; m0 ; l; mjL; M i >>

1
0
n 1
2

l0

n 1
2
n 1

(5.28)

N 1 >
=
2 >
N 1
Y

NLM
>
>
;

(
)

l
L
From a three-term recurrence relationship involving Gegenbauer
polynomials [38], we have obtained the following expression:
:

cos  YNLM (
) =

v
u
u
t

1 (N L)(N + L + 1)
YN +1;LM (
) (5.29)
2
N (N + 1)

v
u

L 1)
1u
t (N + L)(N
YN 1;LM (
)
2
N (N 1)
Inserting Eqs.(5.28) and (5.29) into Eq.(5.27), we get a solution
for the integral:
b;n0 l0 m0
a;nlm

!
"
#
2 3=2 X nn0 N (2l + 1)(2l0 + 1) 1=2
=
k
2 2
NLM

(5.30)

166
8 0
> n
>
< 02

hl0; m0 ; l; mjL; M i >>


:

("

1+

l0

n 1
2
n 1

N 1 >
=
2 >
N 1

>
>
;

L)(N + L + 1) (N + L)(N L 1)
+
(1 N;1 )
4N (N + 1)
4N (N 1)

(N

v
u
u
t

uNLM (x)
v
u
u
t

v
u

4(N

(N

L)(N + L + 1)
uN +1;LM (x)
N (N + 1)

(N + L)(N L
N (N 1)

1 u
t (N
+
4(N + 1)
+

1
n 1
2

1)

v
u
u
t

1)

uN 1;LM (x)(1 N;1 )

L)(N + L + 2)[(N + 1)2


N (N + 2)

L2 ]

(N + L)(N

1)2

L 2)[(N
N (N 2)

uN +2;LM (x)
L2 ]

uN 2;LM (x)uN 2;LM (x)(1 N;1 )(1 N;2 )g

Bibliography
[1] H. Shull and P.-O. Lowdin,
[2] V. Fock,

J. Chem. Phys.

, 30, 617 (1959).

, 98,145 (1935).

Z. Phys.

[3] H. Shull and P.-O. Lowdin,

, 67, 373 (1955).

Svensk Kem. Tidskr.

[4] O. Goscinski, Preliminary Research Report No. 217, Quantum


Chemistry group, Uppsala University (1968).
[5] K.M. Howell and H. Shull,

J. Chem. Phys.

[6] B. R. Judd, Angular Momentum Theory


Academic Press, New York (1975).

, 30, 627 (1955).


.

for Diatomic Molecules

167
[7] V. Aquilanti, , S. Cavalli, C. Coletti, D. Di Domenico, and
G. Grossi. Quantum Systems in Chemistry and Physics. Vol I:
Basic Problems and Model Systems, page 289 (2000).
[8] J. Avery, Hyperspherical Harmonics, Applications in Quantum
Theory. Kluwer Academic Publishers, Dordrecht, The Netherlands
(1989).
[9] T. Shibuya and C.E. Wulfman, Proc. Roy. Soc., A286, 376 (1965).
[10] J. Avery, Hyperspherical Harmonics and Generalized Sturmians.
Kluwer Academic Publishers, Dordrecht, The Netherlands (2000).
[11] J. Avery,

Adv. Quantum Chem.

, 31, 201 (1999).

[12] V. Aquilanti, S. Cavalli, C. Coletti, and G. Grossi,


209, 405 (1996).
[13] V. Aquilanti, S. Cavalli, and C. Coletti,
(1997).
[14] J.P. Gazeau and A. Maquet,
[15] B.K. Novosadov
[16] Brett I. Dunlap

, 214, 1

Chem. Phys.

J. Chem. Phys.

Opt. Spectrosc.

Chem. Phys.

, 73, 5147 (1980).

, 41, 490 (1976).


, 30, 39 (1975).

Chem. Phys. Lett.

[17] Ch. Duchon, M.Cl. Dumont-Lepage and J.P. Gazeau,


15, 1227 (1982).

J. Phys. A

[18] Ch. Duchon, M.Cl. Dumont-Lepage and J.P. Gazeau,


Phys., 76, 445 (1982).

J. Chem.

[19] H.J. Monkhorst and B. Jeriorski, J. Chem. Phys., 71, 5268 (1979).
[20] V. Aquilanti and J. Avery,

Chem. Phys. Lett.

[21] V. Aquilanti, A. Caligiana and S. Cavalli,


92, 99, (2003).

, 267, 1 (1997).
,

Int. J. Quantum Chem.

168
[22] V. Aquilanti, A. Caligiana, S. Cavalli and C. Coletti, Int. J. Quantum Chem., 92, 212, (2003).
[23] P. M. Morse and H. Feshbach,
McGraw-Hill, New York (1953).

Methods of Theoretical Physics

[24] T. Koga and T. Matsuhashi,

J. Chem. Phys.

[25] T. Koga and T. Matsuhashi,

J. Chem. Phys.

[26] J. Avery and R. Shim,

, 87, 1677 (1987).


, 89, 983 (1988).
, 83, 1 (2001).

Int. J. Quantum Chem.

[27] V. Aquilanti and J. Avery,

, 39, 71 (2001).

Adv. Quantum Chem.

[28] W. Kolos and L. Wolniewicz,

J. Chem. Phys.

[29] J.F. Rico, R. Lopez and G. Ramrez,


(1988).

, 49, 404 (1968).


, 9, 790

J. Comp. Chem.

[30] J.F. Rico, R. Lopez, M. Paniagua and G. Ramrez,


Commun., 64, 329 (1991).
[31] J.F. Rico, R. Lopez, I. Ema and G. Ramrez,
mun., 105, 216 (1997).

Comp. Phys.

Comp. Phys. Com-

[32] J. Avery and S. Sauer, in Quantum Systems in Chemistry


Physics, Vol. 1, Kluwer Academic Publishers (1998).
[33] J. Avery,

this book, and private communication

[34] R. Szmytkowski,

J. Phys. B

[35] R. Szmytkowski,

J. Phys. A

[36] R. Szmytkowski,

Phys. Rev. A

[37] R. Szmytkowski,

J. Phys. B

and

(2002).

, 30, 825 (1997).


, 31, 4963, (1998).
, 65, 012503 (2001).

, 35, 1379 (2002).

[38] M. Abramowitz and I.A. Stegun. Handbook of Mathematical Function. Dover, New York, 1964.

Chapter 6
Anharmonic frequencies and Berry
pseudorotation motion in PF5
Summary

We study the dynamics of PF5 , using a density functional potential energy surface. The full surface has twenty D3h minima
connected by thirty C4v transition states. We use the theory of the
molecular symmetry group to understand the tunnelling splittings
of the vibrational levels. We estimate a value for the splitting by
studying the motion between two equivalent minima. This splitting is very small for the ground vibrational state. Therefore we
then study the vibrations of PF5 in one of its minima, using normal
coordinates with anharmonic perturbation theory and variational
theory. Our predicted fundamental frequencies agree well with observation. Finally we use the theory of the Reaction Path Hamiltonian to study the interaction of the vibrational modes with a large
169

170
amplitude motion connecting two minima.

6.1 Introduction
In this chapter, we shall study the Berry pseudorotation [1] dynamics of PF5 . Phosphorus penta ouride has twenty equivalent
D3h minima which are connected by thirty C4v transition states.
Firstly we shall look at the structure of the vibrational energy levels including the e ects of the multiple tunnellings, using the theory of the molecular symmetry group. We shall estimate a value
for the splittings which turns out to be very small for the ground
state. Then we shall look at the dynamics in the region of one
minimum, generating a quartic normal coordinate potential energy
surface using density functional theory. We use both perturbation
theory and variational theory to predict fundamentals, overtones
and combination bands.
We draw attention to some earlier studies on PF5 . Tschumper,
Fermann and Schaefer report structures, thermochemistry and electron anity using a variety of density functionals [2]. Wasada and
Hirao have studied the minimum and the transition state of the
pseudorotation using ab initio quantum chemistry [3]. They report
a barrier which is estimated between 1480 cm 1 and 1770 cm 1 .
Breidung and Thiel have used SCF theory to calculate an anharmonic force eld for PF5 . This reference provides a careful analysis
of the existing experimental data. The level of agreement they

171
obtained for the fundamentals (with MP2 for the quadratic force
constants) is similar to the accuracy we shall obtain using density
functional theory.

6.2 Geometry optimization and harmonic


frequencies
We use density functional theory (DFT) for all our calculations of
the potential energy surface, speci cally employing three di erent
functionals (B97-1, B3LYP and B3P91), with a TZ2P basis set.
Our experience is that these hybrid calculations give potential energy surfaces with good structural features. The calculations were
done using CADPAC6.5 [4].
We located the D3h minimum and the C4v transition state. Geometrical parameters (which are de ned in Figure 1) and the energy
barrier are given in Table 6.1.
The three computations show excellent agreement with the electronic di raction values [5] for the two bond lengths of the minimum (within the experimental error). The transition state barrier is uniformly predicted to be 1280  12 cm 1 . This agrees
well with other computations in the literature (for example Daul et
al. have reported a B3LYP/6-311++G(3df,3pd) calculation which
gives 1300 cm 1 [6]), but it is at variance with the earlier calculations of Wasade and Hirao (who only used a DZP basis in their
correlated studies). In Tables 6.2 and 6.3 we give the harmonic

172

l eq

l eq

l ax

l pi

l ot

l ax

l eq

l ax

l ot

l eq

l ot

2
x

D 3h

l eq
4

4
y

l eq

l ot

l ax
3

C 4v

D 3h

Figure 6.1: Berry pseudorotation for a molecule XY5 , connecting two


equivalent D3h minima by a C4v transition state.

173
Table 6.1: DFT optimization of PF5 ground state and the Berry
pseudorotation saddle point. (a Ref. [5]).

Ground state (D3h )


lax (
A)

leq (A)
Transition state (C4v )
lpi (
A)

lot (A)
(C4v ) (deg.)
Barrier to interconversion
E (cm 1 )

Expt.a

B97-1 B3LYP B3P91


1.579
1.535

1.580
1.535

1.575
1.531

1.535
1.579
102.4

1.535
1.580
102.4

1.531
1.575
102.4

1279

1292

1271

1.577  0.005
1.534  0.004

frequencies for the minimum and for the transition state.

We discuss the accuracy of these predictions in Section 5. Note


that the one imaginary frequency of the transition state has B1
symmetry, and inspection shows that it corresponds to motion towards the D3h structure. The barrier at 1280 cm 1 is quite high
and so we must expect any ground state tunnelling features to be
very small; however we now proceed to study them.

174
Table 6.2: Calculated harmonic frequencies (cm 1 ) at the D3h minimum, compared with observed fundamentals from a Ref. [22] and
references therein.
Mode (D3h ) B97-1 B3LYP B3P91 Expt.a
1 (A01 )
792
783
792
816
0
636
626
634
648
2 (A1 )
00
3 (A2 )
934
920
933
946.4
00
4 (A2 )
559
557
560
575
0
5 (E )
1001
988
1000 1025.7
0
511
509
511
533
6 (E )
0
7 (E )
157
156
157
174
00
496
493
496
512
8 (E )

6.3 Tunnelling splittings in PF5


Let us consider the topology of the ground state potential energy
surface of PF5 . In this 12-dimensional hypersurface there are
!
5
 2 = 20
2
equivalent minima, each one corresponding to a trigonal bipyramid (D3h point group) structure di ering from the others only
for the labeling scheme of the nuclei. The PF5 molecule can undergo large amplitude motions: a well-known concerted mechanism,
called Berry pseudorotation (BP) [1], involves the lowest energy
barrier [7{9] for interconversion of equivalent minima. Such a motion is shown in Fig. 2: the transition state for the BP has C4v
symmetry, during the motion C2v symmetry is conserved (Fig. 3),

175
Table 6.3: Calculated harmonic frequencies in cm 1 for the C4v
transition state.
Mode (C4v ) B97-1 B3LYP B3P91
1 (A1 )
993
981
994
784
775
783
2 (A1 )
501
498
503
3 (A1 )
4 (B1 )
630
618
628
5 (B1 )
113i
117i
114i
6 (B2 )
558
554
558
966
949
964
7 (E )
8 (E )
552
547
552
356
354
357
9 (E )
and one of the three equatorial bonds (we call it `pivot') will be
almost una ected by the BP. Since there are three possible choices
for the pivot, each minimum (D3h ) is connected to the three nearest
equivalent minima on the surface by three equivalent paths. Thus
we are dealing with a potential energy surface having 20 equivalent
D3h minima and (20  3)=2 = 30 equivalent C4v transition states
(TS) connecting them in pairs. In Fig. 3 we report the Balaban
graph corresponding to our surface [10]. In the literature there
are several isomorphic versions of this graph, and we have tried to
choose an easily readable one.
We can determine the splitting patterns, i.e. the number and
the degeneracy of the splitting levels, by group theoretical considerations. If we neglected the presence of multiple minima, we could

176

C3 S 3

v
v

v
v

C2

C2

C2

C2

C4

C2

D3h
trigonal bypiramid

C 2v
intermediate configuration

C 4v
transition state

Figure 6.2: Conservation of C2v symmetry during the Berry pseudorotation.

177

12
5

34

45
4

2
1

35
13
25
23

15
5

14

3
2

24
4

24

14
5

15

23
25
13
35

1
2

45

34
3

12

Figure 6.3: Balaban graph representing Berry pseudorotation of a

molecule XY5 . The D3h isomers are labeled by the numbers of the two
axial ligands, putting rst the one from which the equatorial ligands are
seen in clockwise sequence. If the two numbers are in decreasing order,
we invert and oversign them; such an isomer is the enantiomer of the not
oversigned one. The pivot which allows the passage between two isomers
has the number that does not appear in the labels of the isomers.

178
label the vibrational wavefunctions according to the irreducible representations of the D3h group. But when we want to consider the
interactions among several equivalent structures, we have to exploit
the molecular symmetry (MS) group to predict the degeneracy of
splitting levels and to label them [11]. The complete nuclear permutation inversion (CNPI) group of PF5 has 240 elements, being
the product of the complete permutation group (5! elements) and
the inversion group (2 elements). The MS group is a subgroup of
the CNPI group, obtained from the latter after deleting all the unfeasible elements, i.e. all the elements that interconvert numbered
equilibrium forms of the molecule separated by an insuperable barrier in the potential energy surface. For our molecule there is no
unfeasible element in the CNPI group, so this coincides with the MS
group. Then in order to predict the splitting patterns we have to
obtain the correlation between the A01 irreducible representation of
the D3h group and the corresponding irreducible representations in
the CNPI group G240 of PF5 . This was achieved using a code written by D. J. Wales [12]. The result is that the twenty ground state
vibrational levels span 6 irreducible representations: 2 of them are
non-degenerate, 2 are 4-fold degenerate and 2 are 5-fold degenerate.
In order to provide a quantitative estimation of the tunnel splittings, we exploit an approximate method described by Wales [12,
13]. This approach is an analogue to the LCAO method for the
construction of molecular orbitals. If there are n distinguishable
versions of a given structure, then there are n degenerate wave

179
functions, each localized on one particular potential energy well,
for any given vibrational state of the molecule. When rearrangements occur only between these n distinguishable versions, we may
approximate the true wave function as a linear combination of localized wave functions. We then construct the secular equations. In
the case of PF5 , there is just one type of degenerate rearrangement
which links 20 versions of the same structure. All the diagonal elements of the Hamiltonian H^ are the same ( ), and the o -diagonal
elements of H^ are denoted by , the tunneling matrix element, if
the corresponding minima are directly linked by the rearrangement
(adjacent), and are set zero otherwise. Then the splitting levels
are the eigenvalues of a 20x20 matrix which in each row has 3 o diagonal elements equal to and 16 equal to zero. Diagonalization
yields the splitting pattern for the ground state of PF5 , and the result, represented in Fig. 4, is seen to be consistent with the group
theoretical predictions.
Next we have to provide a numerical (approximated) value for
. Let us consider the equation
!

1 d2
+ V (Q) Ei 'i = 0
2 dQ2

(6.1)

where Q is the length of the vector whose components are the 18


mass-weighted Cartesian coordinates of PF5 , and V (Q) is one of
the 30 equivalent reaction paths connecting two minima by a saddle
(see Fig. 5). Each point of the path has been calculated with the
following method: a) enforce C2v symmetry; b) enforce atoms 1

180

(1)

+3

(4)

+2

A1

(5)

D3h

(5)

(4)

(1)

(20)

240

Figure 6.4: The splitting pattern for the vibrational ground state of

PF5 , taking into account the multiple equivalent minima. The numbers
between parentheses indicate the degeneracy of levels.

181
and 5 to be on the yz plane, and atoms 2 and 4 to be on the xy
plane (referring to Fig. 1); c) x a value for ; d) optimize all the
other internal degrees of freedom (we use the functional B3P91 and
basis set TZ2P). Eq.(6.1) is solved numerically, and the di erence
between the two lowest-lying levels is used to evaluate (Fig. 5):

E
' 2

E1

= 0:06 cm 1

(6.2)

Our estimated value of is indeed very small, and in fact no splitting due to Berry pseudorotation has been detected experimentally. Takami and Kuze have studied the 3 fundamental and the
3 + 7 7 hot band using infrared absorption spectroscopy. They
state that they do not observe any splitting attributable to BP with
their 10 MHz resolution [14]. Of course our calculation gives only
a crudest estimate for the splitting.
However, we nd that there will be a substantial splitting for
the higher vibrational levels. Table 2 shows that the magnitude of
the frequency of vibration which corresponds to interconversion is
160 cm 1 , i.e. seven quanta, suce to reach the top of the barrier. Therefore if it were possible to create such a highly excited
vibration, then the e ects of BP would be observable.

6.4 Anharmonic vibrations of PF5


We have therefore con rmed that the e ects of BP are negligible for
the lowest vibrations of PF5 . We now consider one of the minima

182

Energy (Eh)

-840.72

-840.73

C4v

E2

-840.74

D3h
-400

E1

-200

D3h
0

200

Q (a.u.)

Figure 6.5: A C2v reaction path connecting a D3h minimum and a C4v

saddle. The representation of the splitting of the lower vibrational levels


due to the interaction of the two minima is qualitative. The numerical
values for the levels E1 and E2 are the two lowest eigenvalues of Eq.
(6.1).

400

183
in isolation, using the B3P91 functional and the TZ2P basis set
to determine the anharmonic force eld. We expand the potential
energy surface through quartic terms:
1 X
1X 2 1X
!i qi +
ijk qi qj qk +
 qqq q
(6.3)
2 i
6 ijk
24 ijkl ijkl i j k l
In this standard notation !i are the harmonic wavenumbers, qi
are the dimensionless normal coordinates, and ijk , ijkl are the
third and fourth order force constants, respectively. If Eq. (6.3)
corresponds to a Taylor series, then we use this identi cation:

V =

@ 3 V
ijk =

@qi @qj @qk 0

(6.4)


@4V

ijkl =

(6.5)
@qi @qj @qk @ql 0
As usual the third and semi-diagonal fourth order (iijk ) force
constants were evaluated by nite di erence calculations from analytical second derivatives obtained at the equilibrium geometry
and at the geometries displaced from equilibrium along the normal
modes:

jk (+qi ) jk ( qi )


(2qi )

(6.6)

jk (+qi ) + jk ( qi ) jk (0)


(qi )2

(6.7)

ijk =
iijk =

184
We have discussed elsewhere how the magnitude of the displacements are chosen, and subsequent averaging [15]. Essentially the
displacement was chosen to be 100 Eh over the equilibrium geometry.
To obtain the vibrations on this potential energy surface, we
may use either perturbation theory or variational theory. Secondorder perturbation calculations were carried out using the program
`SPECTRO' [16]. In a simplest form, SPECTRO obtains the energy levels of an asymmetric top i.e. it determines the anharmonic
constants xkl

E (v) =

1
1
1 X
!k (vk + ) + xkl (vk + )(vl + );
2 kl
2
2

(6.8)

using standard formulae [17]. The program includes the e ects of


Fermi type I resonances and Fermi type II resonances. When they
occur the divergent part is removed and a resulting 2x2 matrix
is diagonalized. (for more details see Ref. [16]). The asymmetric
top formulae can be used for PF5 , all that is necessary is to very
slightly change (+7  10 6 amu) the mass of an equatorial uorine
to lower the symmetry from D3h to C2v . Results of the application
of perturbation theory are in Table 6.4.
In this case Fermi resonances were not found to be signi cant,
the largest resonance occurring for 5 and 1 + 7 and producing a
shift of 0.03 cm 1.
It is well known that the second order perturbation theory ap-

185
Table 6.4: Fundamental frequencies (cm 1 ) from a quartic force
eld (DFT B3P91/TZ2P, Eq. 6.3), calculated by b perturbation
theory (SPECTRO) and cvariational theory (MULTIMODE). a See
Ref. [22] and references therein.
Mode (D3h )
1 (A01 )
2 (A01 )
3 (A002 )
4 (A002 )
5 (E 0 )
6 (E 0 )
7 (E 0 )
8 (E 00 )

PTb
786
629
926
557
991
512
157
493

VTc a Expt.
787 816
630 648
927 946.4
557 575
992 1025.7
512 533
157 174
493 512

proach produces exact vibrational eigenvalues (at least in 1D case)


only for Morse potential. Therefore, this approach in e ect \Morsi es" the potential of Eq. (1). This can be viewed as replacing
the polynomial expansion in normal coordinates Qi by polynomial
expansion in the Morse coordinates yi = 1 exp( i Qi ) so that the
rst through fourth derivatives of potential with respect to Qi stay
the same. However, this introduces higher-order terms into the potential, which in many cases may bring the vibrational levels closer
to reality. To nd the true eigenvalues of Eq. (1) we use variational
theory. The variational normal coordinate code `MULTIMODE' is
based on the Watson representation for the kinetic energy operator [18]. It was rst used to perform vibrational self-consistent eld

186
(SCF) calculations, which were followed by con guration interaction (CI) using the orthonormal virtual expansion functions of the
SCF zero-point level. The `MULTIMODE' package has been described previously [19]. `MULTIMODE' uses numerical integration
for the matrix elements. We use the parameter Icoupl = 3 which
means that the code properly considers all terms in our potential
V . The Coriolis terms are evaluated up to the case when 3 normal
coordinates are simultaneously nonzero.
The SCF vibrational states are formed from a basis of NVF=6
optimised one-dimensional functions in the 12 normal modes. These
are in turn obtained from an initial set of 12 primitive harmonic oscillator basis functions, integrated over 15 points by Gauss-Hermite
quadrature, in matrix elements constructed from one-dimensional
cuts of the potential. For each required state the basis is diagonalised for each mode in turn in the eld of the remaining modes
in their respective states until self-consistency is reached.
For improved descriptions of the vibrational states we perform
CI. The CI basis is itself built from 1-mode, 2-mode, etc. expansion
sets in all normal modes of the molecule, i.e. from sets allowing
simultaneous excitation of one, two etc. modes. Each category is
subject to a maximum quantum Q for each mode and a maximum
sum S over all modes. We report calculations using four modes,
with Q=6 and S=6, in Table 6.4.
Table 4 shows a remarkable agreement between the perturbation
theory and the variation theory predictions for the fundamentals.

187
This is hardly surprising; when using the data from Table 2, it is
seen that the e ects of anharmonicity are very small (less than 10
wavenumbers for each mode). The anharmonic e ects are smaller
than the errors arising from the use of DFT, but even those do not
exceed 35 wavenumbers. Therefore the fundamentals of PF5 are
well reproduced by considering it as a rigid isolated molecule.

6.5 MULTIMODE with the Reaction Path


Hamiltonian
In the previous section the vibrations of PF5 were studied around
one of its minima, in complete isolation from its other minima.
At this particular minimum, there must be three large amplitude
motions which lead to the three transition states towards the three
neighbouring minima. The remaining motions can be described by
normal coordinates. Therefore for a complete study of PF5 we need
a representation with 9 normal modes and 3 large amplitude modes.
Such a mathematical formulation does not exist. However we have a
version of MULTIMODE which treats one large amplitude motion
using the theory of the Reaction Path Hamiltonian (RPH) [20].
We are presently developing a two large amplitude program, with
the Reaction Surface Hamiltonian [21] ; we can see how to treat the
extension to the Reaction Volume Hamiltonian, although a program
formulation is a long way o .
In this section we report studies using RPH. This requires the

188
determination of the `Reaction Path' connecting two adjacent minima (using DFT). Ideally we should nd the steepest descent path
from the TS to the minima (it is symmetric about the TS); for
simplicity we determined a closely related path which is parameterised by the angle  (see g 1), enforcing C2v symmetry. The DFT
energy is minimised with respect to all other internal coordinates.
This gives the same reaction path as obtained in section 3. At each
point on the path (which is symmetric about the transition state),
the projected force constant matrix is determined, giving the harmonic frequencies and the associated vectors, which are required for
the RPH implementation. Our representation of the reaction path
through the angle  allows an automatic continuation of it beyond
the minima by further reducing the angle . The greatest diculty
with the application of RPH is that the vibrational modes must
continuously change along the path, in particular at an accidental
crossing. This was carefully checked, and we also ensured that all
frequencies and vectors were symmetrical across the TS.
E ectively this entire calculation is an attempt to model the role
of the multiple minima on the vibrational dynamics. From section
3 we know the e ects on vibrations along the paths connecting the
multiple minima, with levels at E0  ; E0  2 ; E0  3 . Here we
shall only obtain E0  . We therefore expect the zero point energy
to be split in this way, and indeed we expect all the vibrational
levels to be split, i.e. the number of levels will be doubled. Because
the reaction path motion is degenerate at the minima, we expect

189
the this motion to be split into 4, with two near pairs, corresponding
to motion along the path and motion perpendicular to the path. If
we were able to study the full problem then each original level E0
would be split into 6 levels, according to the theory of section 3.
We believe that the magnitude of the splittings should be in the
ratios given by section 3.
The results of the RPH calculations are given in Table 6.5.
The original MULTIMODE calculations are also given (as VT)
for comparison, in order to demonstrate the splittings discussed
above. The VT calculations use only the full quadratic force eld
at the minima (i.e. no anharmonicity); the RPH calculations use
only a quadratic force eld perpendicular to the path (i.e. only
anharmonicity along the path). The table shows the predicted levels for the zero point, the fundamentals, the overtone 27 and the
combination bands 7 + k (k=1-6,8). 7 corresponds to the (degenerate) motions at the minima, one of which corresponds to the
reaction path. All the expected splittings are present. Highlights
are
(i) the zero point motion is split by 0.13 cm 1 .
(ii) the degenerate mode 7 is split into two motions: (162.17,
162.54), which is perpendicular to the path (this is a fundamental vibration), and (195.75,197.40), which corresponds to the rst
excited vibration along the path, with a consequent larger splitting.
(iii) a nondegenerate fundamental, such as 1 is split (798.79,798.92)
by 0.13 cm 1 .

190
(iv) the combination 8 + 7 , both originally degenerate, now
correspond to 8 levels, easily identi able as 4 related pairs.
We therefore believe that these RPH calculations have gone
some way towards providing a numerical demonstration of the expected vibrational levels for PF5 .

6.6 Conclusion
In this chapter we have studied the dynamics corresponding to the
low lying vibrations of PF5 . We have used density functional theory
to determine the important aspects of the potential energy surface,
a theory which we believe is entirely adequate for these purposes,
bearing in mind that we do not study parts of the surface which
involve the breaking of bonds. We have calculated the barrier to
interconversion between equivalent minima to be 1280 cm 1 .
Firstly we studied the e ects of tunnelling motion on the vibrations around the twenty minima which are connected by thirty
transition states. We used the Molecular Symmetry group to predict the degeneracies of the vibrations; we also used a Huckel type
analysis for this purpose. We predicted that the splittings of these
vibrations are very small.
Secondly we studied the anharmonic motion of PF5 around one
of its minima, using both perturbation theory and variational theory. We compared our predictions for the fundamentals with those
available in the literature, agreement to 15 cm 1 being obtained.

191
Finally we use the Reaction Path Hamiltonian to study the vibrations with one large amplitude motion connecting two adjacent
minima. We were able to interpret the resultant splittings of the vibrations and would also predict the nature of the vibrational levels
if it were possible to include the e ect of all the minima simultaneously.
To sum up, the principal weakness of the work presented in this
chapter is the use of Density Functional Theory to generate the
potential energy surfaces. We were interested in a study of the dynamics, and methods for this, and we believe that our methods are
completely appropriate to a more accurate potential energy surface.
From the calculations of Wasada and Hirao [3] one possibility is that
our transition state is too low. However a much greater criticism
is that we have only been able to calculate the interactions of one
minima with another, and not the interactions of one with minima
with three other minima. The e ect of including all the interactions simultaneously will be to substantially reduce the tunnelling
splitting, but at present we have no numerical way of determining
this, which should require that the large amplitude motion be occurring on at least a two-dimensional surface, to accomplish for the
trifurcation of the reaction path from each minimum towards its
neighbours.

192

Bibliography
[1] R. S. Berry,

, 32, 933, 1960.

J. Chem. Phys.

[2] G. S. Tschumper, J. T. Fermann, H. F. Schaefer III,


Phys., 104, 3676, 1996.
[3] H. Wasada, K. Hirao,

J. Chem.

, 114, 16, 1992.

J. Am. Chem. Soc.

[4] R.D. Amos, I.L. Alberts, J.S. Andrews, S.M. Colwell, N.C. Handy,
D. Jayatilaka, P.J. Knowles, R. Kobayashi, G.J. Laming, A.M. Lee,
P.E. Maslen, C.W. Murray, P. Palmieri, J.E. Rice, E.D. Simandiras,
A.J. Stone, M.-D. Su, and D.J. Tozer, Cadpac6.5, The Cambridge
Analytic Derivatives Package (University of Cambridge, UK), 2000.
[5] L. S. Bartell, K. W. Hansen,

, 4, 1775, 1965.

Inorg. Chem.

[6] C. Daul, M. Frioud, O. Schafer, A. Selloni, Chem. Phys. Lett., 262,


74, 1996.
[7] P. Russenger, J. Brickmann,

Chem. Phys. Lett.

, 30, 276, 1975.

[8] P. Russenger, J. Brickmann,

J. Chem. Phys.

, 62, 1086, 1975.

[9] P. Russenger, J. Brickmann,

J. Chem. Phys.

, 66, 1, 1977.

[10] A. T. Balaban, Chemical Applications of Graph Theory, Academic


Press, New York, 1976.
[11] P. R. Bunker, Molecular
Press, New York, 1979.

Symmetry and Spectroscopy

[12] D. J. Wales,

J. Am. Chem. Soc.

[13] D. J. Wales,

J. Am. Chem. Soc.

, Academic

, 115, 11191, 1993.

[14] M. Takami, H. Kuze,

, 115, 11180, 1993.


, 80, 2314, 1984.

J. Chem. Phys.

[15] A. Miani, E. Cane, P. Palmieri, A. Trombetti, and N.C. Handy, J.


Chem. Phys., 112:248, 2000.

193
[16] J.F. Gaw, A. Willetts, W.H. Green, and N.C. Handy, in "Advances
in Molecular Vibrations and Collision Dynamics", ed. J.M. Bowman, JAI Press, Greenwich, CT (1990).
[17] D. Papousek, M.R. Aliev, Molecular Vibrational-Rotational Spectra, Academia, Prague (1982).
[18] J. K. G. Watson,

, 15, 479, 1968.

Mol. Phys.

[19] S. Carter, J.M. Bowman, and N.C. Handy,


100, 191, 1998.

Theor. Chem. Acc.

[20] D. P. Tew, N. C. Handy and S. Carter Mol. Phys., 99, 393 (2001).
[21] D. P. Tew, N. C. Handy and S. Carter J. Chem. Phys., submitted
[22] J. Breidung, W. Thiel,

, 320, 39, 1994.

J. Mol. Struct.

194

Table 6.5: Fundamental frequencies, and overtones and combination bands involving 7 calculated with Multimode (VT) and the
reaction path Hamiltonian (RPH). Only harmonic terms are included in the normal coordinate force eld
Mode (D3h )
VT
RPH
0
1 (A1 )
792
798.79,798.92
0
635
640.01,640.12
2 (A1 )
3 (A002 )
936
938.95,939.05
00
4 (A2 )
562
565.98,566.16
0
5 (E )
1001;1001
1004.15,1004.33;1006.40,1006.59
0
6 ( E )
515;516
505.54, 505.76;519.31,519.44
0
162;163
162.17,162.54;195.75,197.40
7 (E )
00
8 (E )
498;499
500.78,500.86;502.61,502.77
27
324;325;325
321.24,324.27;366.99,368.98;384.57,388.70
943
961.23,962.64;992.68,993.23
1 + 7
2 + 7
797;798
802.45,803.82;835.51,836.10
3 + 7
1098;1098
1101.37,1102.78;1135.71,1136.29
4 + 7
724;725
729.95,730.75;764.76,766.07
5 + 7
1163;1164;1164;1164
1167.43,1168.74;1199.76,1200.33
6 + 7
678;678;678;679
667.63,669.13;681.47,683.33;703.07,703.79;718.18,718.92
660;661;661;661
662.77,664.57;666.77,667.26;694.16,694.66;703.95,705.04
8 + 7

Contents

195

196

Conclusions
Here we summarize the main results exposed in this work. The
chapters of this thesis originated several publications that are listed
in the Bibliography of the present section.
In chapter 1 [1] we have studied the four di erent types of Sturmian orbitals arising from the separation of variables in Schrodinger
equation for the hydrogen atom: polar, parabolic, spheroidal and
spheroelliptic. The de nition of spheroidal and spheroelliptic Sturmians includes a continuous parameter, which allow them to perform smooth transitions respectively from polar to parabolic Sturmians and from polar Sturmian oriented towards an axis to polar
Sturmian oriented towards a di erent axis. The symmetry properties of the alternative basis sets and the orthonormal transformations to each other have been investigated.
In chapter 2 [2] we have studied the corresponding problem for
Sturmians in momentum space (there are six types of them: spherical, Stark, Zeeman, spheroelliptic, elliptic cylindrical of type I and
II), exploiting in our investigation the method of trees and the
close relationship with hyperspherical harmonics. Then, we have
197

198
discussed the connections among basis sets in the con guration
and momentum space. Spherical, Stark and spheroelliptic Sturmians in momentum space are respectively the counterparts of polar,
parabolic and spheroelliptic Sturmians in con guration space. Elliptic cylindrical Sturmians of type I and II include a continuous
parameter that allows them smooth transitions between Stark and
Zeeman functions, and spherical and Zeeman functions respectively.
In chapter 3 we have applied one type of Sturmian orbitals in
con guration space (the polar ones) to the solution of Schrodinger
equation for atoms. The consequences arising from the use of Sturmians has been exposed: the kinetic energy operator disappears
from the secular equation; the representation of the nuclear attraction potential is diagonal; the eigenvalues are not the energies of the
p
2E ; since the orbitals are not orthogonal, matrix
system, but
elements are to be calculated in the framework of Slater-Condom
generalized rules; the e ect of external elds could be easily incorporated in our formalism. A calculation on Helium atom electronic
states has shown that the description of excited states is very good
allowing one to simulate easily the atomic spectrum. This is at
variance with conventional quantum chemistry methods. An application of the method in the case of external eld has been performed
in order to compute the rst order static polarizabilities of Lithium.
Chapter 4 deals with Sturmians and one-electron molecules [3].
Analytical formulas were obtained for nuclear attraction integrals
involving polar and parabolic Sturmians. A calculation of the po-

199
tential energy curve of H+2 has been performed in order to test the
convergence properties of Sturmian basis sets at di erent levels of
iteration of the secular equation.
In chapter 5 [4, 5] we have proposed an original method for
solving the molecular Schrodinger equation for molecules, inserting
Sturmian orbitals into a valence bond approach. We have not been
able to solve analytically multicenter repulsion integrals involving
Sturmians, so we had to expand the latters as linear combinations
of Gaussian functions and solve numerically the integrals. We calculated the potential energy curve for H2 , demonstrating that such
a method can be practically implemented.
Chapter 6 [6] is not closely related to the previous ones. We
studied the dynamics of PF5 , using a density functional potential
energy surface. The full surface has twenty D3h minima connected
by thirty C4v transition states. We used the theory of the molecular
symmetry group to understand the tunnelling splittings of the vibrational levels. We estimated a value for the splitting by studying
the motion between two equivalent minima. This splitting is very
small for the ground vibrational state. Therefore we then studied
the vibrations of PF5 in one of its minima, using normal coordinates
with anharmonic perturbation theory and variational theory. Our
predicted fundamental frequencies agree well with observation. Finally we used the theory of the Reaction Path Hamiltonian to study
the interaction of the vibrational modes with a large amplitude motion connecting two minima.

200

Bibliography
[1] V. Aquilanti, A. Caligiana and S. Cavalli,
Chem., 92, 99, (2003).

Int. J. Quantum

[2] V. Aquilanti, A. Caligiana, S. Cavalli and C. Coletti,


Quantum Chem., 92, 212, (2003).

Int. J.

[3] V. Aquilanti and A. Caligiana,


(2002).

Chem. Phys. Lett., 366, 157,

[4] V. Aquilanti and A. Caligiana,


submitted (2003).

J. Mol. Struct. (Theochem),

Sturmian Orbitals in Quantum Chemistry: An Introduction in "Fundamental world of

[5] V. Aquilanti and A. Caligiana,

quantum chemistry", Vol. I, 297-316, E. J. Brandas and E. S.


Kryachko (Eds.), 2003 Kluwer Academic Publishers.
[6] A. Caligiana, V. Aquilanti, R. Burcl, N.C. Handy, D.P. Tew,
Chem. Phys. Lett., 369, 335, (2003).