(Meng) Advances in Atomic, Molecular, and Optical Physics Volume 47

Advances in
ATOMIC, MOLECULAR, AND OPTICAL PHYSICS
VOLUME 47
Editors
BENJAMIN BEDERSON
New York University
New York, New York
HERBERT WALTHER
Max-Plank-Institut für Quantenoptik
Garching bei München
Germany
Editorial Board
P. R. BERMAN
University of Michigan
Ann Arbor, Michigan
M. GAVRILA
F. O. M. Institut voor Atoom-en Molecuulfysica
Amsterdam, The Netherlands
M. INOKUTI
Argonne National Laboratory
Argonne, Illinois
W. D. PHILIPS
National Institute for Standards and Technology
Gaithersburg, Maryland
Founding Editor
SIR DAVID BATES
Supplements
1. Atoms in Intense Laser Fields, Mihai Gavrila, Ed.
2. Cavity Quantum Electrodynamics, Paul R. Berman, Ed.
3. Cross Section Data, Mitio Inokuti, Ed.
ADVANCES IN
ATOMIC,
MOLECULAR,
AND OPTICAL
PHYSICS
Edited by
Benjamin Bederson
DEPARTMENT OF PHYSICS
NEW YORK UNIVERSITY
NEW YORK, NEW YORK
Herbert Walther
UNIVERSITY OF MUNICH AND
MAX-PLANK INSTITUT FÜR QUANTENOPTIK
MUNICH, GERMANY
Volume 47
San Diego San Francisco New York

Boston London Sydney Tokyo
This book is printed on acid-free paper.
∞
Copyright
C 2001 by ACADEMIC PRESS
All Rights Reserved.

No part of this publication may be reproduced or transmitted in any form or by any
means, electronic or mechanical, including photocopy, recording, or any information
storage and retrieval system, without permission in writing from the Publisher.
The appearance of the code at the bottom of the first page of a chapter in this book
indicates the Publisher’s consent that copies of the chapter may be made for
personal or internal use of specific clients. This consent is given on the condition,
however, that the copier pay the stated per copy fee through the Copyright Clearance
Center, Inc. (222 Rosewood Drive, Danvers, Massachusetts 01923), for copying
beyond that permitted by Sections 107 or 108 of the U.S. Copyright Law. This consent
does not extend to other kinds of copying, such as copying for general distribution, for
advertising or promotional purposes, for creating new collective works, or for resale.
Copy fees for pre-2001 chapters are as shown on the title pages. If no fee code
appears on the title page, the copy fee is the same as for current chapters.
1049-250X/01 $35.00
Explicit permission from Academic Press is not required to reproduce a maximum of

two figures or tables from an Academic Press chapter in another scientific or research
publication provided that the material has not been credited to another source and that
full credit to the Academic Press chapter is given.
Academic Press
A Harcourt Science and Technology Company
525 B Street, Suite 1900, San Diego, California 92101-4495, USA
http://www.academicpress.com
Academic Press
Harcourt Place, 32 Jamestown Road, London NW1 7BY, UK
http://www.academicpress.com
International Standard Serial Number: 1049-250X
International Standard Book Number: 0-12-003847-1
PRINTED IN THE UNITED STATES OF AMERICA

01 02 03 04 05 06 SB 9 8 7 6 5 4 3 2 1
Contents
CONTRIBUTORS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Nonlinear Optics of de Broglie Waves

P. Meystre
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
II. s-Wave Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
III. A Microreview of Manybody Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
IV. Mean-Field Theory of Bose–Einstein Condensates . . . . . . . . . . . . . . . . . . 14
V. Four-Wave Mixing of de Broglie Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
VI. Mixing of Optical and Matter Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
VII. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
VIII. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Formation of Ultracold Molecules (T ≤ 200 μK) via

Photoassociation in a Gas of Laser-Cooled Atoms
Françoise Masnou-Seeuws and Pierre Pillet
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
II. Physical Concepts in Photoassociation and Formation of Molecules . . 57
III. Ultracold Molecule Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
IV. Theoretical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
V. Present Status of the Comparison between Experiment and Theory:
Formation Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
VI. New Schemes for Making Ultracold Molecules . . . . . . . . . . . . . . . . . . . . . . 116
VII. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
VIII. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
IX. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Molecular Emissions from the Atmospheres of Giant Planets

and Comets: Needs for Spectroscopic and Collision Data
Yukikazu Itikawa, Sang Joon Kim, Yong Ha Kim, and Y. C. Minh
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
II. Spectroscopy of the Giant Planets: Needs for Spectroscopic
and Collision Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
III. Spectroscopy of Comets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
IV. Spectral Databases and Improvements Needed . . . . . . . . . . . . . . . . . . . . . . 155
V. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
VI. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
VII. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
v
vi Contents
Studies of Electron-Excited Targets Using Recoil Momentum

Spectroscopy with Laser Probing of the Excited State
Andrew James Murray and Peter Hammond
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
II. Atomic Deflection Using Electron Impact . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
III. Doubly Excited States Studied via the Fluorescence Decay Product:
Recoiling Excited Atoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
IV. Stepwise Laser Probing of Deflected Metastable Targets . . . . . . . . . . . . . 187
V. Conclusions and Future Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
VI. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
VII. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Quantum Noise of Small Lasers

J. P. Woerdman, N. J. van Druten, and M. P. van Exter
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
II. Overview of Threshold Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
III. Good-Cavity versus Bad-Cavity Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
IV. Spontaneous Emission Factor β . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
V. Cavity QED and β . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
VI. Magnitude of β . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
VII. Relaxation Oscillation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
VIII. Disappearance of Fluctuation Threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
IX. Petermann Excess Quantum Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
X. Finite Number of Atoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
XI. Concluding Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
XII. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
XIII. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
SUBJECT INDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
CONTENTS OF VOLUMES IN THIS SERIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Contributors
Numbers in parentheses indicate the pages on which the authors’ contributors begin.
PETER HAMMOND (163), Department of Physics, University of Western Australia,

Nedlands, Perth WA6907, Australia
YUKIKAZU ITIKAWA (129), Institute of Space and Astronomical Science, 3-1-1
Yoshinodai, Sagamihara 229-8510, Japan
SANG JOON KIM (129), Department of Astronomy and Space Science, Kyung Hee
University, Suwon, 449-701, Korea
YONG HA KIM (129), Department of Astronomy and Space Science, Choong Nam
National University, Daejeon, 305-764, Korea
FRANÇOISE MASNOU-SEEUWS (53), Laboratoire Aimé Cotton, Bât. 505, Campus
d’Orsay, 91405 Orsay Cedex, France
P. MEYSTRE (1), Optical Sciences Center, The University of Arizona, Tucson,
AZ 85721
Y. C. MINH (129), Korea Astronomy Observatory, Hwaam, Yusong, Daejeon,
305-348, Korea
ANDREW JAMES MURRAY (163), Department of Physics and Astronomy, The
University of Manchester, Manchester M13 9PL, United Kingdom
PIERRE PILLET (53), Laboratoire Aimé Cotton, Bât. 505, Campus d’Orsay, 91405
Orsay Cedex, France
N. J. VAN DRUTEN (205), Department of Applied Physics, Delft University of
Technology, Lorentzweg 1, 2628 CJ Delft, The Netherlands
M. P. VAN EXTER (205), Huygens Laboratory, Leiden University, 2300 RA Leiden,
The Netherlands
J. P. WOERDMAN (205), Huygens Laboratory, Leiden University, 2300 RA Leiden,
The Netherlands
vii
This Page Intentionally Left Blank
Advances in
ATOMIC, MOLECULAR, AND OPTICAL PHYSICS
VOLUME 47
ADVANCES IN ATOMIC, MOLECULAR, AND OPTICAL PHYSICS, VOL. 47
NONLINEAR OPTICS
OF DE BROGLIE WAVES
P. MEYSTRE
Optical Sciences Center, The University of Arizona, Tucson, Arizona 85721
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
II. s-Wave Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
III. A Microreview of Manybody Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 8
IV. Mean-Field Theory of Bose–Einstein Condensates . . . . . . . . . . . . . . . . . 14
A. Zero Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
B. Finite Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
V. Four-Wave Mixing of de Broglie Waves . . . . . . . . . . . . . . . . . . . . . . . . 22
A. Mean-Field Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
B. Quantum Theory of Atomic Four-Wave Mixing . . . . . . . . . . . . . . . . . 27
VI. Mixing of Optical and Matter Waves . . . . . . . . . . . . . . . . . . . . . . . . . . 37
A. Parametric Amplification of Atomic and Optical Fields . . . . . . . . . . . . 38
B. Matter-Wave Superradiance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
C. Phase-Coherent Matter-Wave Amplification . . . . . . . . . . . . . . . . . . . 49
VII. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
VIII. References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
I. Introduction
One of the most profound revolutions brought about by quantum mechanics is that
it does away with the distinction between particles and waves. Planck and Einstein
first postulated the existence of light quanta to explain the blackbody radiation
spectrum and the photoelectric effect. They argued that one should associate to
the light quantum an energy E = hω and a momentum p = hk. Shortly thereafter,
de Broglie went one step further and proposed that electrons, and for that matter
all massive particles, should be thought of as waves of energy E = h 2 q 2 /2M and
wavelength = h/q, where q is the particle momentum [1].
One relatively recent application of these developments is atom optics. Its ex-
perimental foundations were laid as early as 1929, when O. Stern et al. demon-
strated the reflection and diffraction of atoms from metallic and crystalline surfaces
[2, 3]. A few years later, Frisch measured the deflection of atoms as a result of
absorption of light followed by spontaneous emission [4]. Atom optics is now an
important subfield of atomic, molecular and optical physics, and many passive
atom-optical elements, including mirrors, lenses, gratings, interferometers and
resonators, have been demonstrated.
1 Copyright C 2001 by Academic Press

All rights of reproduction in any form reserved.
ISBN 0-12-003847-1/ISSN 1049-250X/01 $35.00
2 P. Meystre
Just as the invention of the laser lead to a profound revolution in optics, and
enabled the development of nonlinear optics and quantum optics, a similar revo-
lution is taking place in atom optics now that experimentalists can more or less
routinely generate Bose–Einstein condensates. By extracting “atom laser” beams
from these condensates, it has been possible to carry out the first nonlinear atom
optics and quantum atom optics experiments. These are the developments reviewed
in this chapter.
The early experiments in atom optics considered low-density atomic samples,
where the atoms in the light fields evolve independently from each other. How-
ever, when the sample densities or atomic beam fluxes become large enough that
atom–atom interactions become important, the dynamics of a given atom become
dependent on the presence of other atoms. In traditional atomic physics, atom–
atom interactions are described as collisions. The point of view adopted in non-
linear atom optics [5–8] is different. One considers instead these interactions as a
mechanism that allows matter waves to interact and mix, very much like optical
waves in nonlinear optics [9–13].
The analogy between nonlinear optics and nonlinear atom optics is actually
quite profound: in conventional optics, effective nonlinear equations for the op-
tical fields follow from the elimination of the medium dynamics, while in atom
optics nonlinear matter-wave dynamics results from collisions, which are in turn
effective manybody interactions resulting from the (partial) elimination of the
electromagnetic field.
The goal of this chapter is to review the foundations and some of the recent de-
velopments of nonlinear atom optics, concentrating on the case of bosonic atoms.
Section II briefly discusses s-wave scattering, which provides an appropriate
description of the low-energy collisions of relevance in ultracold, low-density
samples of bosonic atoms. As such, this analysis determines the properties of
the nonlinearity most commonly used in the nonlinear manipulation of bosonic
Schrödinger fields. We then turn to a microreview of the formalism of manybody
theory, which allows one to treat atomic samples as fields, in close analogy to the op-
tical case. Section IV introduces the mean-field theory appropriate for the descrip-
tion of ultracold atomic samples, leading to the famous Gross–Pitaevskii equation
widely used in the dynamical description of Bose–Einstein condensates at zero tem-
perature. We also discuss ways to improve on this approach at finite temperatures.
With this formalism at hand, Sect. V discusses matter-wave four-wave mixing
in low-density atomic condensates both in a mean-field approach and in the frame-
work of a quantum optics-like description that analyzes the buildup of the side
modes from noise fully quantum mechanically.
Finally, the last section of this review turns to the nonlinear mixing between
optical and matter waves, giving in particular an analysis of “matter-wave super-
radiance” and of phase-coherent matter wave amplification.
To keep this review of a reasonable size, we have chosen to limit the discussion
to those aspects of nonlinear atom optics that have already received experimental
NONLINEAR OPTICS OF DE BROGLIE WAVES 3
confirmation. We have also ignored the fascinating aspects related to the generation
of matter-wave solitons, which is the topic of another review [14]. A considerably
more detailed presentation of this material can be found in a monograph, Atom
Optics, by P. Meystre, to be published by Springer-Verlag (2001).
II. s-Wave Scattering
Collisions play in atom optics a role similar to nonlinear susceptibilities in optics.

In low-density samples of ultracold bosonic atoms, the dominant source of colli-
sions is in the form of ground-state two-body interactions, which can be treated
relatively simply in the shapeless, or s-wave scattering approximation. The pur-
pose of this section is to review some key aspects of these collisions in the simplest
possible way. This leads to the concept of scattering length, which allows us to
describe them in terms of a single parameter. Since this problem is treated at great
length in many quantum mechanics texts, we merely outline the essential steps
here.
We proceed by describing the elastic scattering between two particles of masses
M1 and M2 in the partial waves formalism [15–19]. Introducing a coordinate system
in which the center of mass of the two-particle system is at rest, this problem
reduces to the scattering of a particle of reduced mass,
M1 M2
μ= (1)
M1 + M2
in a potential U(r), which we assume to be central in the following,
U (r) = U (r ). (2)
We consider a situation where the incident wave is a plane wave, while the outgoing
wave is asymptotically a spherical wave, that is,
eikr
ψ(r) ≃ eikz + f k (θ, φ) , (3)
r
where f k (θ, φ) is the scattering amplitude. In addition to the momentum k of the

incident wave, it depends in general on the spherical coordinates θ and φ defining
the direction in which the particle is scattered. The particle current,
−i h ∗
j= [ψ ∇ψ − (∇ψ ∗ )ψ], (4)
2μ
4 P. Meystre
is equal to
hk
ji = (5)
μ
for the incident particles and is approximately given by
hk
jo (r, θ, φ) ≃ | f k (θ, φ)|2 (6)
μr 2
for the outgoing wave.1 Consequently, the differential cross section is related to
the scattering amplitude by
dσ (θ, φ)
= | f k (θ, φ)|2 . (7)
d
In the case of a central potential, it is convenient to expand the wave function

for the relative atomic motion in terms of partial waves,
1
φk,l,m (r) = u k,l (r )Ylm (θ, φ), (8)
r
where u k,l (r ) are the solutions of the radial Schrödinger equation,
h2 d 2 l(l + 1)h 2 h2k2

− 2
+ 2
+ U (r ) u k,l (r ) = u k,l (r ), (9)
2μ dr 2μr 2μ
with u k,l (r = 0) = 0, and the Ylm (θ, φ) are spherical harmonics. For large r, one
can neglect both the effects of the central potential U (r ) and of the centrifugal
potential l(l + 1)h 2 /2μr 2 , so that
d2

2
+ k u k,l (r ) ≃ 0, (10)
dr 2
and

lπ
u k,l (r )|r →∞ ∝ sin kr − + δl . (11)
2
1
This approximate value takes into account the fact that while the r component of jo scales as
1/r 2 , its θ and φ components scale as 1/r 3 , since, as we recall, (∇)r = ∂/∂r, (∇)θ = (1/r )(∂/∂θ) and
(∇)φ = (1/r sin θ)(∂/∂φ). Hence they are negligible in the asymptotic region r → ∞.
The phase shift δl appearing in this expression is determined by the requirement

that u k,l → 0 as r → 0, the term lπ/2 being included explicitly to account for
the fact that in the free-particle case U (r ) = 0 the phase of u k,l (r ) equals lπ /2. In
other words, the phase δl is taken relative to the zero-potential solution. It is easily
shown that 2δl may be interpreted as the phase accumulated by the outgoing wave
relative to that of a free outgoing wave; see, e.g., [19].
We now expand the wave function (3) in terms of the partial waves (8), taking
advantage of the fact that for elastic collisions the final energy h 2 k 2 /2μ is the
same as the incident energy. We note in addition that for a central potential the
scattering wave function must be symmetric under rotations about the direction of
incidence of the incoming particle, the z axis. Hence, we need only include partial
waves of fixed momentum k and with m = 0, so that f k (θ, φ) → f k (θ) and
∞

ψ(r) = Cl φk,l,0 (r). (12)
l=0
The coefficients Cl can be determined from the requirement that the asymptotic
limit of this expansion is of the form (3). Noting that from Eqs. (8) and (11) we
have for r → ∞,
eikr e−i(lπ/2−δl ) − c.c. 0

φk,l,0 (r) ≃ Yl (θ), (13)
2ir
and using the expansion of the plane wave exp(ikz) in terms of spherical harmonics,
∞

eikz = i l 4π (2l + 1) jl (kr )Yl0 (θ)
l=0
∞
1 lπ
≃ i l 4π (2l + 1) sin kr − Yl0 (θ), (14)
l=0
kr 2
where we have used the asymptotic expansion of the spherical Bessel functions
jl (kr ) in the second line [19]. This requirement leads after some algebra to the
partial waves expansion of the scattering amplitude f k (θ) as
∞ ∞
1
f k (θ) = f k,l (θ) = 4π(2l + 1)eiδl sin δl Yl0 (θ). (15)
l=0
k l=0
We have seen in Eq. (7) that the differential scattering cross section is simply given
by | f k (θ )|2 . The total scattering cross section σ is obtained by integrating this
6 P. Meystre
result over the 4π solid angle. With the orthonormality relation

d Yl0∗ 0
′ (θ)Yl (θ) = δl,l ′ , (16)
this readily gives the familiar result

∞
4π
σ = (2l + 1) sin2 δl . (17)
k 2 l=0
For the ultracold applications of atom optics that we have in mind, we can
restrict the analysis to the case of particles of such low energy that
kr0 ≪ 1, (18)
where r0 is the range of the potential U (r ). As it turns out, the dominant partial
wave in the vicinity of the potential U (r ) is then the s-wave, ℓ = 0 [17].
We can gain an intuitive feeling for this important property by returning to the
expansion (14) in terms of spherical harmonics. More specifically, consider first a
free spherical wave,

(0) 2k 2
φk,l,m (r) = jl (kr )Ylm (θ, φ), (19)
π
where the jl are spherical Bessel functions. The associated probability for the
particle to be in the solid angle d about θ0 and φ0 and between r and r + dr is

(0)
2
Pl (θ0 ,φ0 ) ∝ r 2
φk,l,m (r)
dr d . (20)
For ultracold particles, kr → 0, the Bessel functions take the asymptotic form
(kr )l
jl (kr ) ≃ , (21)
(2l + 1)!!
and it can be shown that the function K 2r 2 jl2 (kr ) is negligible for

l(l + 1) > kr. (22)
(0)
The expansion (14) of a plane wave into free spherical waves φk,l,m (r) leads
therefore to the following physical interpretation: in the range r < r0 of U (r )
and for momenta k small enough that kr0 ≪ 1, the incident plane wave acts for all
practical purposes as an spherical wave, so that by conservation of angular momen-

tum, it must be scattered into another spherical wave. In other words, for particles
of sufficiently low energy, kr0 → 0, one needs consider only the partial wave
of angular momentum l = 0 in Eq. (17). This is the so-called s-wave scattering
approximation, which yields the cross section
4π
σ ≃ sin2 δ0 . (23)
k2
It can be shown [18] that in the low-energy regime, one has approximately
δl ∝ (kr0 )2l+1 , (24)
0
which
√ is much smaller than unity since kr0 ≪ 1. In that case, and with Y0 (θ, φ) =
1/ 4π Eq. (15) gives
δ0
f k,0 ≃ , (25)
k
and from the differential cross section (7), we find
σ = 4πa 2 , (26)
where we have introduced the scattering length,
δ0
a ≡ − f = −lim . (27)
k→0 k
Physically, the scattering length can be obtained by solving the radial wave equation
d2

− U (r ) u(r ) = 0, (28)
dr 2
the length a corresponding to the point where the asymptote of u(r ) intersects
the horizontal axis. In case of a hard-sphere potential of radius r0 , it corresponds
to that radius, a = r0 . In general, the scattering length may be either positive
or negative. Positive scattering lengths correspond to repulsive interactions, and
negative scattering lengths to attractive interactions.
To conclude this section, we note that it is often possible to replace the two-body
potential U (r ) by a pseudo-potential that gives the same scattering length as the
real potential and yielding the same s-wave scattering solutions. For unperturbed
8 P. Meystre
wave functions well-behaved at the origin, it takes the form
4πah 2
U (r ) = δ(r ). (29)
M
III. A Microreview of Manybody Theory

Very much as nonlinear optics is advantageously described in terms of Maxwell
fields, rather than in the particle (photon) picture, nonlinear atom optics is best
analyzed in terms of Schrödinger fields rather than single-atom dynamics. To
fully set the stage for our discussion, we therefore still need to introduce some key
elements of the formalism of second quantization. While certainly not necessary in
principle, this approach permits one to draw considerable intuition and inspiration
from traditional nonlinear optics. Additionally, the introduction of particle creation
and annihilation operators and of a Fock space for particles allows one to easily
account for systems where the total number of particles in a given mode is not
conserved. Finally, the inclusion of the quantum statistics of the particles proceeds
in a particularly simple way.
Second quantization finds its origin in the desire to treat the dynamics of N
identical particles satisfying either bosonic or fermionic statistics. One approach to
this problem consists in considering the N-particle wave function φ(r1 , . . . , r N , t),
evolving according to the N-particle Hamiltonian
N

HN = Hi , (30)
i=1
where Hi is the Hamiltonian of particle i. The N-particle wave function must be

either symmetric or antisymmetric under particle exchange, depending on the par-
ticles being bosons or fermions. Alternatively, one can describe this same problem
more conveniently using the field-theoretical approach of second quantization
[20–22].
It is far beyond the scope of this review to give a rigorous treatment of second
quantization starting from a canonical quantization approach. Rather, we postulate
that the Schrödinger wave function ψ(r) becomes an operator (r) ˆ which, for
bosonic particles, satisfies the commutation relations
ˆ
[(r), ˆ † (r′ )] = δ(r − r′ ),
(31)
ˆ
[(r), ˆ )] = 0,
(r ′
(32)
and for fermionic particles
ˆ
[(r), ˆ † (r′ )]+ = δ(r − r′ ),
(33)
ˆ
[(r), ˆ ′ )]+ = 0.
(r (34)
In this last expression, [. . .]+ is an anticommutator, [A, B]+ = AB + B A. As we

ˆ
shall see, (r) may be interpreted as an operator annihilating a particle at position
r and ˆ † (r) creates a particle at location r. The second-quantized Hamiltonian
corresponding to the single-particle Hamiltonian H = −h 2 ∇ 2 /2M + V (r) is
2
H= ˆ † (r) − h ∇ 2 + V (r) (r).
d 3r ˆ (35)
2M
Physically, one can understand this Hamiltonian in the following way: the field
ˆ
operator (r) picks a particle at location r, the Hamiltonian evolution H =
−h 2 ∇ 2 /2M is applied to this particle, and the operator ˆ † (r) puts it back into
place. Finally, the integral in Eq. (35) guarantees that all particles in the ensemble
are subject to this treatment.
From this Hamiltonian, it is easy to find the Heisenberg equation of motion for
ˆ
the field operator (r). For the case of bosons, we have
ˆ
d (r, t)
ih ˆ
= [(r, t), H]
dt
2
ˆ
= d 3r ′ (r, ˆ † (r′ , t) − h ∇ 2 + V (r′ ) (r
t), ˆ ′ , t)
2M
2
h ˆ
= − ∇ 2 + V (r) (r, t), (36)
2M
where we have used the relation [A, BC] = B[A, C] + [A, B]C and the commuta-
tion relations (33) and (34). It is easily seen that the same result holds for fermions,
despite the different commutation relations. Hence, the Heisenberg equation of mo-
ˆ
tion for the Schrödinger field operator (r, t) has the same form as the Schrödinger
equation for the wave function ψ(r, t) in usual quantum mechanics. We can think
ˆ
of the operator (r, t) as the quantized form of the single-particle Schrödinger
field ψ(r, t), in much the same way as we have quantized the electromagnetic
field, E(r, t) → Ê(r, t). Note, however, that while the quantization of the electro-
magnetic field leads to new physics, such is not the case for second quantization,
when applied to the nonrelativistic problems that we address here. It is merely a
convenient bookkeeping tool.
10 P. Meystre
In terms of the Schrödinger field creation operator ˆ † , the N-particle wave

function φ N (r1 , . . . r N , t) is
1

|φ N = √ ˆ † (r N ) . . .
d 3r1 . . . d 3r N φ N (r1 , . . . , r N , t) ˆ † (r1 )|0 , (37)
N!
where |0 is the vacuum state (absence of particles),
ˆ
(r)|0 = 0. (38)
Introducing the number operator

N̂ ≡ ˆ † (r)(r)
d 3r ˆ (39)
and using the commutator identity
N

[A, B N . . . B1 ] = B N . . . Bi+1 [A, Bi ]Bi−1 . . . B1 , (40)
i=1
it is easily shown that
N̂ |φ N = N |φ N . (41)
Similarly, one finds

ˆ N ˆ † (r N −1 ) . . .
ˆ † (r1 )|0 .
(r)|φ N = √ d 3r1 . . . d 3r N −1 φ N (r1 , . . . r N −1 , t)
N!
(42)
The equivalence of the formalism of second quantization with the conven-

tional quantum mechanical description of the N-body problem in terms of the
sum of N individual Hamiltonians can be demonstrated by applying the manybody
Hamiltonian (35) to the N-particle state |φ N , showing that
1

H|φ N = √ d 3 r d 3 r1 . . . d 3 r N ˆ
ˆ † (r)H (r)(r)
N!
× φ N (r1 , . . . r N )ˆ † (r N ) . . .
ˆ † (r1 )|0
N
1

= √ d 3 r1 . . . d 3 r N ˆ † (r N ) . . .
ˆ † (r1 )|0 Hi φ(r1 . . . r N ),
N ! i=1
(43)
where
h2 2
Hi ≡ − ∇ + V (ri ). (44)
2M i
This result shows that the multiparticle evolution governed by the Schrödinger
equation of motion
d
ih φ N (r1 , . . . r N , t) = HN φ N (r1 , . . . r N , t) (45)
dt
is equivalent to the evolution given by the Fock space Schrödinger equation
d |φ N
ih = H |φ N . (46)
dt
This equivalence is a direct consequence of the fact that the Schrödinger field
operator evolution is governed by the single-particle Hamiltonian H; see Eq. (36).
It is easily shown that the Schrödinger field operator satisfies a conservation of
probability law. Introducing the density
ˆ † (r, t)(r,
n̂(r, t) = ˆ t)
yields the continuity equation for the quantum field operators,
d n̂(r, t) ˆ † (r, t)
d ˆ
= ˆ
(r, t) + ˆ † (r, t) d (r, t)
dt dt dt
−i h ˆ † (r, t))(r,
ˆ ˆ † (r, t)∇ 2 (r,
ˆ
= [(∇ 2 t) − t)] = −∇ · ĵ, (47)
2M
where the flux

−i h
ˆ † (r, t)∇ (r,
ˆ ˆ † (r, t)](r,
ˆ

ĵ ≡ t) − [∇ t) , (48)
2M
compare with Eq. (4).

Very much as it is useful in optics to introduce a mode expansion of the electric
field operator E(r, t), we can introduce a mode expansion of the Schrödinger field
operator. We illustrate how this works by expanding (r, ˆ t) on the orthonormal
eigenfunctions ϕn (r) of the time-independent Schrödinger equation,
h2 2

− ∇ + V (r) ϕn (r) = E n ϕn (r)
2M
12 P. Meystre
as

ˆ
(r, t) = ϕn (r)ĉn (t), (49)
n
where the mode label n stands for the complete set of quantum numbers necessary
to characterize that mode. Inserting this expression and its hermitian conjugate
into the second-quantized Hamiltonian (35) gives

H= E n ĉn† ĉn , (50)
n
where we have made use of the orthonormality relation

d 3r ϕn∗ (r)ϕm (r) = δnm (51)
and have also assumed a discrete energy spectrum for simplicity. With Eq. (51) we
can express the operators ĉn from Eq. (49) as

ĉn (t) = ˆ
d 3r ϕn∗ (r)(r, t). (52)
For the bosonic commutation relations (31) and (32), this gives
†
[ĉn , ĉm ] = δnm ,
(53)
[ĉn , ĉm ] = 0,
which we recognize from the quantization of the electromagnetic field. For

fermionic particles one finds
†
[ĉn , ĉm ]+ = δnm ,
(54)
[ĉn , ĉm ]+ = 0.
The combination of the second-quantized Hamiltonian (50) and the bosonic com-
mutation relations (53) shows that in that case, we have reduced the problem
to that of a set of harmonic oscillators of energies E n . We can therefore inter-
†
pret ĉn as an annihilation operator, and ĉn as a creation operator for mode n,
with
√
ĉn |Nn = N n |Nn − 1 ,
† √
ĉn |Nn = Nn + 1|Nn + 1 , (55)
†
ĉn ĉn |Nn = Nn |Nn ,
where Nn is the number of particles in the mode. This shows that in second quan-
tization, the state of the system is simply expressed in terms of the number of
excitations of each single-particle mode. The total number of particles in the
system is given by

N̂ = N̂ n = ĉn† ĉn = ˆ † (r)(r),
d 3r ˆ (56)
n n
and is clearly a constant of motion for the Hamiltonian (50). Likewise, the indi-
vidual populations of all modes of the matter field are also constants of motion.
This latter property ceases to hold, however, as soon as interactions are permitted,
e.g., in the presence of a light field.
In case the matter–light coupling is via the electric dipole interaction, the
second-quantized interaction Hamiltonian is bilinear in the optical field creation
and annihilation operators, but quadratic in matter-field creation and annihilation
operators, a direct consequence of the conservation of the total number of particles
N̂ . An example of such an interaction Hamiltonian is
Va− f = hg(a ĉn† ĉm + h.c.),
whereby the absorption of a photon is accompanied by a particle being “annihi-

lated” from state m and “created” state n.
As is the case for the simple harmonic oscillator, nothing prevents us from
putting as many bosons in mode n as we wish. This, however, is no longer the case
for fermions: this is apparent from the anticommutation relation [ĉn , ĉm ]+ = 0,
† †
which for m = n yields ĉm ĉm = ĉm ĉm = 0. This indicates that the ground state of
a given mode is reached at the latest once a single particle has been removed. It is
furthermore impossible to populate a given mode with more than one particle, a
property further evidenced by the fact that the number operator N̂ n and its square
N̂ 2n are easily shown to be equal,
N̂ 2n = N̂ n .
Hence the population of a given mode must be either zero or one. In addition, one
14 P. Meystre
finds readily that
N̂ n ĉn† |0 = ĉn† |0 , (57)
†
where |0 is the vacuum state (absence of particle), indicating that ĉn |0 is an
eigenstate of mode n with value 1. This is nothing but Pauli’s exclusion principle,
expressed in the formalism of second quantization in terms of anticommutator
relations.
In the framework of second quantization, two-body collisions characterized by
the interaction energy V (ri , r j ) are described by the Hamiltonian
1

V= ˆ † (r)
d 3r d 3r ′ ˆ † (r′ )V (r, r′ )(r
ˆ ′ )(r),
ˆ (58)
2
ˆ ′ )(r)
the operator (r ˆ picking two particles at locations r and r′, to which the
two-body Hamiltonian V (r, r′ ) is applied before they are put back into place. The
application of V on the state vector |φ N gives
N N
1
V|φ N = √ × d 3r1 · · · d 3r N V (ri , r j )
N ! i=1 j>i
ˆ † (r N ) . . .
× φ N (r1 , . . . , r N ) ˆ † (r1 )|0 . (59)
One often encounters situations where the scalar Schrödinger field we have
considered so far needs to be extended to a multicomponent field, to account for
the internal states of the atoms involved. The generalization of the formalism of
second quantization to this case is straightforward.
IV. Mean-Field Theory of Bose–Einstein Condensates
A. ZERO TEMPERATURE
At zero temperature, a trapped Bose–Einstein condensate is characterized by the
property that all atoms occupy the same quantum state. In the absence of atom–
atom interaction, this is the ground state of the trap, but collisions change this
property somewhat, a point to which we return shortly. In this section, we derive
an effective nonlinear equation that describes the dynamics of such a condensate,
the so-called Gross–Pitaevskii equation [23–25]. We first use it to determine the
ground state of the system, and then extend the analysis to discuss the elementary
excitations of the condensate. This will give us a first peek at the profound analogy
between nonlinear optics and nonlinear atom optics. A considerably more de-
tailed theoretical discussion of Bose–Einstein condensation of low-density atomic
systems can be found in the review by Dalfovo et al. [26].
There are several ways to derive the Gross–Pitaevskii equation. For example, the
Hartree mean-field approach consists of factorizing the N-particle wave function
φ N (r1 , . . . r N ) of Eq. (37) as
rN

φ N (r1 , . . . r N ) = ϕ N (r), (60)
r=r1
where the effective single-particle states ϕ N (r) are assumed to be normalized. In

the so-called time-dependent Hartree approximation, the equations of motion for
these wave functions, also called Hartree wave functions, are determined from the
Hartree variational principle [27, 28],

δ ∂
= φ N |i h − H|φ N = 0. (61)
δϕ N∗ (r) ∂t
For a manybody Hamiltonian including a two-body interaction of the form (58),

this readily yields
∂ϕ N (r)

ih = H ϕ N (r) + (N − 1) d 3r ′ V (r, r′ )ϕ N∗ (r)ϕ N2 (r′ ) (62)
∂t
The (N − 1) factor appearing in this expression results from the fact that the two-
body Hamiltonian V involves two creation operators on the left of V ; see Eq. (58).
This leads to N (N − 1) equivalent terms, while the single-particle Hamiltonian,
which involves only one annihilation operator on the left of H, leads to N equivalent
terms.
Alternatively, one can describe the condensate in the context of another mean-
field theory whose basic ingredient consists in separating out the condensate con-
tribution from the Schrödinger field operator ˆ † (r) and treating it as a number,
the order parameter, or condensate wave function (r, t). Specifically, we assume
that
ˆ
(r, ˆ ′ (r, t),
t) = (r, t) + (63)
where
(r, t) = (r, t) . (64)

16 P. Meystre
In this approximation, it is assumed that the condensate possesses a well-defined

phase as a result of spontaneous symmetry breaking. While this is a conve-
nient and expeditious way to derive important results, we note that this assump-
tion is not needed in general. We also remark that in the thermodynamic limit
N → ∞ at constant density, this and the Hartree approach can be shown to be
equivalent.
ˆ ′ (r, t), which measures the departure of the condensate
Clearly, the operator
state from its mean-field value, must satisfy the Bose commutation relation
ˆ ′ (r, t),
[ ˆ ′† (r′ , t)] = δ(r − r′ ). (65)
The dynamics of the condensate wave function (r, t) can be obtained by taking
the expectation value of the Heisenberg equation of motion for the field operator
ˆ
(r, t). We have seen in Sect. II that in the case of s-wave scattering the two-body
potential can be approximated by the local interaction
V (r − r′ ) = gδ(r − r′ ), (66)
where
4π h 2 a
g= (67)
M
and a is the s-wave scattering length. In that case, we obtain the Gross-Pitaevskii
nonlinear Schrödinger equation [23–25],
h2 2

∂
ih (r, t) = − ∇ + Vtrap (r) + g|(r, t)|2 (r, t). (68)
∂t 2M
We observe that this derivation implicitly makes the factorization ansatz,
ˆ † (r, t)(r,
ˆ ˆ
t)(r, ˆ † (r, t) (r,
t) = ˆ ˆ
t) (r, t) . (69)
In addition to the familiar contributions of the kinetic energy and the trap
potential, the Gross–Pitaevskii equation contains an additional energy term,
g|(r, t)|2 (r, t), proportional to the local density |(r, t)|2 of the condensate.
This term is referred to as (twice) the mean-field energy of the condensate. It plays
a fundamental role in its dynamics. Most important in the context of this review,
the formal analogy between this term and the contribution of a local Kerr nonlin-
earity in optics suggests that the atom–atom interactions that are at its origin play
a role similar to that of a nonlinear medium for light. It is this observation which
opens up the way to nonlinear atom optical effects [5, 29, 30] and wave mixing
phenomena in Bose–Einstein condensates.
The fact that g|(r, t)|2 (r, t) is twice the mean-field energy results from
the observation that a Gross–Pitaevskii equation may also be obtained from the
Hartree-type variational principle,
∂ϕ(r, t) δE
ih = ∗ , (70)
∂t δϕ (r, t)
compare with Eq. (61), where E is the Gross–Pitaevskii energy functional,
h2

g
E[ϕ] = d 3r |∇ϕ|2 + Vtrap |ϕ|2 + |ϕ|4 , (71)
2M 2
and ϕ(r, t) is the Hartree condensate wave function. This allows us to identify the
mean-field energy as
g
E mf = |ϕ(r, t)|2 . (72)
2
The condensate ground state may be obtained by expressing the condensate

wave function as
(r, t) = (r)e−iμt/h , (73)
where μ is the chemical potential and the normalization of (r), taken to be real
for now, is

d 3r 2 (r) = N , (74)
N being the number of particles in the condensate. Substituting Eq. (73) into the
Gross–Pitaevskii equation (4) gives
h2 2

μ(r) = − ∇ + Vtrap (r) + g|(r, t)|2 (r). (75)
2M
With the normalization condition (74), this yields immediately the expression
1
μ= E kin + E pot + 2E mf . (76)
N
18 P. Meystre
In the limit of large condensates, the mean-field contribution dominates over the
kinetic energy term,2 except near the condensate boundary. In that case, one can
neglect that latter term, the so-called Thomas–Fermi approximation. Equation (75)
becomes then simply
μ(r) = [Vtrap (r) + g|(r, t)|2 ](r) (77)
yielding immediately the condensate density profile,
1
|(r)|2 = (μ − Vtrap ) (78)
g
for μ > Vtrap (r) and 0 otherwise. This discontinuity in the density profile is an
artifact of the Thomas-Fermi approximation, which is removed when taking the
quantum pressure properly into account.
So far, we have concentrated on the static properties of the condensate ground
state. The analysis of small perturbations away from that state can be performed
using a Bogoliubov quasi-particle approach [20, 31–33]. For trapped particles, it is
useful to work in coordinate space instead of the perhaps more familiar momentum
representation. We proceed by expressing the Schrödinger field as
†
ˆ
(r, t) = (r) + [u j (r)α̂ j (t) + v ∗j (r)α̂ j (t)], (79)
j>0
where (r) is the condensate wave function, corresponding to a macroscopic pop-

ulation No ≫ 1 of the ground-state mode j = 0, and the quantum numbers j label
elementary excitations above the condensate state. The quasi-particle operators
†
α̂ j (t) are required to satisfy the bosonic commutation relations [α̂ j , α̂l ] = δ jl . This
requirement, which leads to the normalization condition

d 3r [u i∗ (r)u j (r) − vi∗ (r)v j (r)] = δi j , (80)
is equivalent to seeking solutions of the condensate wave function of the form
(r, t) = e−iμt/h [(r) + u(r)e−iωt + v ∗ (r)eiωt ], (81)
where (r) is taken to be real for now. This ansatz, which is analogous to the side-
mode probe expansion familiar in nonlinear optics, will be useful in the discussion
of matter-wave mixing and squeezing of the following sections. In these cases,
2
The kinetic energy contribution is sometimes referred to as the “quantum pressure” term.
though, it will become necessary to remove the assumption that (r) is real, i.e.,
go past pure density arguments.
The functions u j (r) and v j (r) satisfy the coupled equations
hωu j (r) = [H0 − μ + 2g2 (r)]u j (r) + g2 (r)v j (r),

(82)
−hωv j (r) = [H0 − μ + 2g2 (r)]v j (r) + g2 (r)u j (r),
where H0 = −(h 2 /2M )∇ 2 ) + Vtrap (r). These equations must generally be solved
numerically for the trap potential at hand [34–40]. For spherical traps, the solutions
of Eqs. (82) are characterized by the quantum numbers n, l, and m, where n, is
the radial mode number, l is the angular momentum of the excitation, and m is its
z component. For axially symmetric traps, m is still a good quantum number.
B. FINITE TEMPERATURES
At finite temperatures, a nonvanishing portion of the atoms is in states other
than the condensate ground state, and one must account for this fact as well as
for so-called anomalous Bose correlations, a matter-wave analog of “squeezing”
contributions in quantum optics. We limit our review to a brief outline of some of
the general techniques used to approach this problem, without going into any detail
[41]. In addition to introducing standard approximations, this discussion paves the
way to the discussion of wave mixing of Sect. V, in particular, matter-wave phase
conjugation, which relies explicitly on the generation of anomalous correlations
in a Schrödinger field.
Our starting point is again the Heisenberg equation of motion (36) for the
Schrödinger field operator, including the s-wave two-body interaction, and the
ˆ
decomposition of (r, t) into a condensate and a noncondensate contribution; see
Eq. (63). In Sect. V we will use an expansion of the Schrödinger field on a restricted
set of matter-wave modes to handle the correlation functions of operator products
of the form ˆ †ˆˆ “exactly,” following standard quantum optics approaches.
Here, we follow a “condensed matter” approach, where the Heisenberg equations
of motion are truncated by an approximate factorization scheme. We first express
the product of the three field operators appearing in the Heisenberg equations of
motion as
ˆ †
ˆˆ = ||2 + 2||2
ˆ ′ + 2
ˆ ′† + ∗
ˆ ′
ˆ ′ + 2
ˆ ′†
ˆ +
ˆ ′†
ˆ ′
ˆ′ (83)
instead of performing the factorization ansatz (69). We then introduce the self-
consistent mean-field approximation,
ˆ ′†
ˆ ′
ˆ ′ ≃ 2
ˆ ′†
ˆ ′
ˆ ′ +
ˆ ′
ˆ ′
ˆ ′† . (84)
20 P. Meystre
The factor of 2 in this equation results from the sum of the direct (Hartree) and
exchange (Fock) contributions, which are identical in the case of a shapeless two-
body interaction.
As before, the equation of motion for the condensate wave function (r) is
obtained by taking the expectation value of the Heisenberg equation of motion
(83) to give
2 2
h ∇ ˆ † (r)(r)
ˆ (r)
ˆ
− + Vtrap (r) − μ (r) + g = 0, (85)
2M
which, together with the factorization scheme (84), yields

2 2
h ∇
− + Vtrap (r) − μ (r) + g[n c (r) + 2n ′ (r)](r) + gm ′ (r)∗ (r) = 0.
2M
(86)
In this expression, we have introduced [41] the condensate density,
n c (r) ≡ |(r)|2 , (87)
the noncondensate density,
ˆ ′† (r)
n ′ (r) ≡ ˆ ′ (r) , (88)
and the so-called anomalous density,
ˆ ′ (r)
m ′ (r) ≡ ˆ ′ (r) . (89)
The equation of motion for the condensate excitations, which are described by
ˆ ′ (r, t), is readily obtained by substracting Eq. (85) from the
the field operator
Heisenberg equation of motion (83). One finds
2 2
∂ ˆ′ h ∇ ˆ ′ (r, t) + g(
ˆ †
ˆˆ −
ˆ †
ˆ ).
ˆ
ih (r, t) = − + Vtrap (r) − μ
∂t 2M
(90)
Consistent with the mean-field approximation (84), we approximate the bilinear
operator products in this expression as
ˆ † (r)(r)
ˆ ˆ † (r)(r) ,
≃ ˆ
(91)
ˆ (r)
(r) ˆ ˆ (r) ,
≃ (r) ˆ
so that
ˆ † (r)(r)
ˆ (r)
ˆ ˆ † (r)(r)
− ˆ (r)
ˆ ˆ † (r)(r)
≃ 2 ˆ ˆ ′ (r) + (r)
ˆ (r)
ˆ ˆ ′† (r)
(92)
and
2 2
∂ ˆ′ h ∇ ˆ ′ (r, t)
i h (r, t) ≃ − + Vtrap (r) − μ
∂t 2M
ˆ ′ (r, t) + m(r)
+ g[2n(r) ˆ ′† (r, t)]. (93)
In this equation, we have introduced the total (condensate and noncondensate)

density n(r) via
ˆ † (r)(r)
n(r) ≡ ˆ ˆ ′† (r)
= |(r)|2 + ˆ ′ (r) ≡ n c (r) + n ′ (r) (94)
as well as the total anomalous density,
ˆ (r)
m(r) ≡ (r) ˆ ˆ ′ (r)
= (r)2 + ˆ ′ (r) ≡ |φ(r|2 + m ′ (r). (95)
It can be shown that Eqs. (86) and (93) can also be derived from the so-called
Hartree–Fock–Bogoliubov Hamiltonian [41, 42],
2 2
1

h ∇
HHFB = d 3r ∗ (r) − + + Vtrap (r) − μ + g |(r)|2 (r)
2M 2
2 2
+ d 3r ˆ ′† (r) − h ∇ + + Vtrap (r) − μ + 2gn(r) ˆ ′ (r)
2M

g ˆ ′† (r)
ˆ ′† (r) + H.c].
+ d 3r [m(r) (96)
2
This Hamiltonian can be diagonalized by the Bogoliubov transformation,
ˆ ′ (r) = u j (r)α̂ j − v ∗j (r)α̂ †j

(97)
ˆ ′† (r) = u ∗j (r)α̂ †j − v j (r)α̂ j

where the operators α̂ j satisfy Bose commutation relations. Under this transfor-
mation, the Hartree–Fock–Bogoliubov Hamiltonian (96) then takes the diagonal
22 P. Meystre
form
h2∇ 2

1

3 ∗ 2
HHFB = d r (r) − + + Vtrap (r) − μ + g|(r)| (r)
2M 2
†
− E j d 3r |v j (r|2 + E j α̂ j α̂ j , (98)
j j
where the functions u j (r) and v j (r) are given by the solutions of the coupled
Hartree–Fock–Bogoliubov eigenvalue equations
h2∇ 2

− + Vtrap (r) − μ + 2gn(r) u j (r) − gm(r)v j (r) = E j u j (r),
2M
2 2 (99)
h ∇
− + Vtrap (r) − μ + 2gn(r) v j (r) − gm ∗ (r)u j (r) = −E j v j (r).
2M
These same results can be obtained by inserting the ansatz,

†
ˆ ′ (t, t) = u j (r)α̂ j e−i E j t − v ∗j (r)α̂ j ei E j t ,

(100)
j
into the equation of motion (93). This latter approach, which is more familiar to the
quantum optics community, leads to a simple interpretation of the quasi-particles
in terms of probe side modes of the Schrödinger field.
V. Four-Wave Mixing of de Broglie Waves
The preceding section showed that the departure of a Bose condensate from its
mean-field dynamics can be evaluated in a framework closely related to probe
spectroscopy and of the probe field analyses familiar from nonlinear and quantum
optics. The excitation of these side modes, or quasi-particles in the language of
manybody theory, results from the nonlinear dynamics associated with the cubic
nonlinearity of the system, that is, two-body collisions. In optics, cubic nonlinear-
ities are of course known to lead to four-wave mixing phenomena, hence it should
be obvious at this point that similar effects will also be possible with matter waves.
Four-wave mixing and the phase conjugation of atomic waves were first pro-
posed in [5, 43, 44]. The original theory, which predates the demonstration of
atomic Bose-Einstein condensation, used the near-resonant dipole–dipole inter-
action as a source of two-body collisions. In condensates, however, the interac-
tion with optical fields must typically be kept far off-resonant in order to avoid
spontaneous heating, and the dominant source of two-body interactions is there-

fore ground-state collisions. However, the mathematical form of the nonlinear
Schrödinger equation describing these two situations is identical, at least in the
shapeless approximation, so that the original predictions can be extended to the
situation of s-wave scattering without difficulty.
The first verification of matter-wave phase conjugation was realized by W. D.
Phillips and co-workers in a 23Na condensate [30] (see Fig. 1). The experimental
geometry involved atoms in one hyperfine state only, the four modes involved in
the mixing process being distinguished by their center-of-mass momentum. Mode
separation was achieved via Bragg scattering off an optical field. This experiment
constituted the first explicit experimental verification of a nonlinear atom optical
effect. Shortly thereafter, several other nonlinear atom optical effects were experi-
mentally demonstrated, including matter-wave superradiance [45], phase-coherent
matter-wave amplification [46, 47], and the generation of matter-wave solitons
[48, 49].
A full three-dimensional numerical solution of a closely related situation was
given [50], which found excellent agreement with the experimental work at
National Institute of Standards and Technology (NIST). We first review that analy-
sis, which relies on the Gross-Pitaevskii equation. The following subsection gives
an exact quantum mechanical calculation of four-wave mixing in a somewhat
different geometry.
A. MEAN-FIELD ANALYSIS
The NIST experiments proceed by first releasing a Sodium condensate centered
at r = 0 from its trap potential at time t = 0. It then evolves solely under the
influence of its mean-field energy until some time t1 . After this period of free
evolution, optical pulses are made to interact with the condensate, leading to Bragg
scattering. This produces three moving matter-wave wave packets of momenta
p1 ≃ 0, p2 , and p3 such that the momentum differences pi − p j is much larger
than the momentum spread of the initial condensate wave packet. The experimental
time it takes to create these side modes is very short compared to the time scale
over which the wave packets evolve, and can thus be taken to be instantaneous.
As a result, the three wave packets initially overlap and are copies of the initial
condensate. As they fly apart, they interact nonlinearly to produce a fourth wave
at p4 = p1 − p2 + p3 .
In these experiments, and in contrast to the situation considered in the following
section, the condensate has only a single spin component. We describe its dynamics
by the Gross–Pitaevskii equation (68),
h2 2

∂ 2
h (r, t) = − ∇ + Vtrap (r) + g|(r, t)| (r, t). (101)
∂t 2M
FIG. 1. Contour (a) and three-dimensional (b) gray-scale renditions of the experimental realization
of matter-wave four-wave mixing in a sodium Bose–Einstein condensate. (From Ref. [30]. Courtesy
S. Rolston.)
Immediately after the application of the Bragg pulses, its wave function is
3
1/2 ipi ·r/h

(r, t1 ) = (r) fi e , (102)
i=1

where f i = Ni /N is the fraction of atoms in wave packet i, with i f i = 1, and
(r) is the ground-state wave function of the condensate just before the Bragg
pulse.
Following the formation of the initial wave packets, the wave function (102)
evolves and the wave packets with different momenta start to separate. During this
separation, the mean-field energy acts as a cubic nonlinearity and generates a wave
packet of central momentum p4 = p1 − p2 + p3 . It can easily be shown that for
this momentum, both total energy E(t) and momentum p(t) are conserved during
the wave packet evolution, where
p2
E(t) = (t)| + g|(t)|2 |(t) (103)
2M
and
p(t) = −i h(t)|∇|(t) . (104)
At a subsequent time t, the condensate wave function may be expressed in the

form
4

(r, t) = ϕi (r, t)ei(ki ·r−ωi t) , (105)
i=1
where ki = pi /h and
hki2
ωi = . (106)
2M
In Eq. (105) we assumed that the envelopes ϕi (r, t) are slowly varying in space
and time on the scales of ωi−1 and ki−1 .
Substituting the form (105) into the Gross–Pitaevskii equation (101), collecting
in the familiar way of nonlinear optics terms multiplying the same phase factors
26 P. Meystre
[51], eliminating rapidly varying terms and keeping only the phase-matched terms
yields the set of four coupled partial differential equations for the envelopes ϕ(r, t),
h∇ 2

∂ hki i
+ ∇ −i ϕi (r, t) = − g δ(ki + k j − km − kn )
∂t M 2M h j,m,n
× δ(ωi + ω j − ωm − ωn )ϕ ∗j (r, t)ϕn (r, t),

(107)
where each of the indices can take any value between 1 and 4.
The momentum and energy conservation conditions implicit in the Kronecker
delta functions in these equations,
ki + k j − km − kn = 0 (108)
and
ωi + ω j − ωm − ωn = 0, (109)
are automatically satisfied if either i = j = m = n or i = j = m = n. The first of

these conditions corresponds to self-phase modulation and the second one to cross-
phase modulation. They are known from nonlinear optics to lead to effects such
as self-focusing and defocusing, but do not involve any real exchange of particles.
What this implies in the present context is that these contributions cannot lead
to the excitation of a new side mode from the three initial wave packets. Indeed,
particle exchange between the various wave packets can only occur when all four
indices are different, while at the same time conservation of momentum and energy
are satisfied.
The momentum conservation condition (108) implies that
ki + k j = km + kn = κ. (110)
It is always possible to construct a reference frame where κ = 0. In that frame

we have then ki = −k j and km = −kn , and the four-wave mixing process takes
the form of degenerate four-wave mixing [13, 51]. That is, all momenta are equal
in magnitude. We remark at this point that it is known from quantum optics that
degenerate four-wave mixing may be interpreted in terms of real-time holography.
The analogy between optics and atom optics therefore indicates that it is possi-
ble to achieve matter-wave holography via this four-wave mixing process [52].
The quantum optics analogy also leads to a direct interpretation of matter-wave
four-wave mixing in terms of density gratings: two of the wave packets, say those
of momenta ki and km , form a density grating from which the wave packet of
momentum k j is scattered, producing the conjugate, of momentum kn , of the mth
packet. We use a similar interpretation in Sect. VI to discuss the mixing of optical
and matter waves.
Alternatively, it is possible to think of atomic four-wave mixing in terms of a
scattering process in which one particle is annihilated in each wave packet belong-
ing to an initially populated pair, and simultaneously one particle is created in each
of two wave packets of the other pair, one of which is initially populated and the
other is initially in a vacuum. In optics terms, four-wave mixing removes one atom
each from the “pump” beams, and adds one each in the “probe” and “signal” beams.
The explicit form of the four-wave mixing equations of motion is readily ob-
tained from Eqs. (107) and read
h∇ 2

∂ hk1
+ ∇ −i ϕ1 (r, t)
∂t M 2M
ig ig
=− (|ϕ1 |2 + 2|ϕ2 |2 + 2|ϕ3 |2 + 2|ϕ4 |2 )ϕ1 − ϕ2 ϕ3∗ ϕ4 , (111)
h h
the other three equations being obtained by cyclic permutations. The first four terms
on the right-hand side are self-phase and cross-phase modulation contributions.
The factors of 2 in the cross-phase modulation contributions result from nonlinear
nonreciprocity, an effect also familiar from nonlinear optics. The last term is the
source term which either creates or annihilates particles in the wave packet being
propagated. These equations were solved numerically in [50, 53] and found to be
in excellent agreement with experiments. One typical numerical example in shown
in Fig. 2.
B. QUANTUM THEORY OF ATOMIC FOUR-WAVE MIXING

The preceding analysis uses a mean-field description to analyze the dynamics of
four-wave mixing in a multicomponent sodium condensate. We have seen earlier
that this approach makes strong implicit assumptions about the coherence proper-
ties of the system. While these are reasonable for large condensates, their validity
is much less clear at small particle numbers, and in particular when condensate
side modes build up from noise. This section addresses the question of the quan-
tum statistics of weakly excited side modes. Provided that only a few modes of
the system are actively involved in the dynamics, one can use as an alternative to
the standard manybody techniques an “essential states description” inspired from
quantum optics to obtain more accurate results. In general, such approaches still
28 P. Meystre
FIG. 2. (a)–(f ) Time sequence illustrating the generation of a fourth matter wave out of three initial
waves created out of a condensate by Bragg pulses. The three-dimensional gray-scale rendition at the
end of the process should be compared to Fig. 1b. These are results of theoretical simulations of the
NIST matter-wave four-wave mixing experiments. (From Refs. [50, 53]. Courtesy P. Julienne.)
need to make some approximations, for instance, an “undepleted mode” ansatz.

Occasionally, however, there is a situation that can be handled to all orders in the
fields. The example we present in the following is such a case.
We consider a Bose–Einstein condensate of 23Na atoms in the F = 1 hyperfine
ground state, with the three internal atomic states |m F = −1 , |m F = 0 , and
FIG. 2. (Continued )
|m F = 1 of degenerate energies, confined by a far-off-resonant optical dipole

trap. It can be shown [54, 55] that the three-component vector Schrödinger field,
ˆ −1 (r, t),
Ψ̂(r, t) = { ˆ 0 (r, t),
ˆ 1 (r, t)}, (112)
describing this system is governed by the second-quantized Hamiltonian,
p2

H= ˆ m† (r, t)
dr + U (r) ˆ m (r, t)
m 2M

h ˆ 1
ˆ 1
ˆ 1 †
ˆ1 + †
ˆ −1
ˆ −1 †
ˆ −1
ˆ −1 ]}†
+ dr{(c0 + c2 )[
2
ˆ 0 †ˆ 1
ˆ 0 ( †
ˆ −1
ˆ1 + †
ˆ −1 )] + c0
ˆ 0
ˆ 0
ˆ 0
ˆ0 † †
+ 2
ˆ 1†
+ 2(c0 − c2 ) ˆ 1 † ˆ
ˆ −1 −1
ˆ 1†
+ 2c2 (
† ˆ ˆ
ˆ −1 0 0 + H.c)}. (113)
In the language of nonlinear optics, the three terms quartic in one of the field
ˆ i†
operators only, i.e., of the form ˆ i†
ˆ i
ˆ i , are self-defocusing terms; the terms
involving two hyperfine states conserve the populations of the individual spin states
and merely lead to phase shifts; and the terms involving the central mode ˆ 0 and
30 P. Meystre
both side modes correspond to spin-exchange collisions. This “ four-wave mixing”

interaction, involving, e.g., the annihilation of a pair of atoms with m F = 0 and
the creation of two atoms in the states m F = ±1, leads to phase conjugation in
quantum optics and to matter-wave phase conjugation in the present case [56].
A four-wave mixing geometry can be realized when we consider a situation
where atoms in the m F = 0 state are placed in a linear superposition of two
counterpropagating center-of-mass modes of momenta ±hko . that is,
ˆ 0 (x) = √1 (eiko x a01 + e−ik0 x a02 ),

(114)
V
while the atoms of spin m F = ±1 are taken to be the running waves,
ˆ ±1 (x) = √1 e±iko x a±1 .

(115)
V
Here a01 and a02 are the annihilation operators of the two counterpropagating
†
m F = 0 modes, with [a0i , a0 j ] = δi j , i, j = 1 or 2, a1 , a−1 are the corresponding
operators for the running modes associated with m F = ±1, and V is the confine-
ment volume of the condensate. Inserting this mode expansion into the Hamiltonian
(113) and ignoring all non-phase-matched contributions yields the four-wave
mixing Hamiltonian
h 2 k02 hc0
H= N̂ + N̂ ( N̂ − 1)
2M 2
hc2 † † † † † †
+ (a a a1 a1 + a−1 a−1 a−1 a−1 − 2a1 a1 a−1 a−1
2 1 1
† † † †
+ 2(a1 a1 + a−1 a−1 )(a01 a01 + a02 a02 )
† † † †
+ 4a1 a−1 a01 a02 + 4a1 a−1 a01 a02 , (116)
where we have introduced the total number of atoms,
† † † †
N̂ = a1 a1 + a−1 a−1 + a01 a01 a02 a02 . (117)
An important distinction between optical and atomic nonlinear optics experi-

ments is related to the conservation of particle number: while the photon number
is not a conserved quantity, the number of atoms is, in the absence of losses. This
leads to important conservation laws, which we use in the following to describe
the dynamics of the system exactly. The most obvious conserved quantity of the
Hamiltonian (117) is the total number of atoms, as follows from the commutator,
[ N̂ , H] = 0 (118)
In addition, the population differences,
† †
D̂ 0 = a01 a01 − a02 a02 , (119)
† †
D̂ = a1 a1 − a2 a2 , (120)
are also conserved. These two conservation laws follow directly from the fact that
the Hamiltonian (116) describes the creation and annihilation of bosonic atoms in
pairs. For example, the annihilation of two atoms in the center-of-mass modes 1
and 2 of the m F = 0 hyperfine state, described by the operators pair a01 a02 , results
† †
in the creation of a pair of atoms in the m F = ±1 spin states via a1 a−1 .
Together with the conservation of the total number of atoms, these conservation
laws yield the two additional conserved quantities,
† †
N̂ 1 ≡ a1 a1 + a01 a01 (121)
and
† †
N̂ 2 ≡ a−1 a−1 + a02 a02 . (122)
These conservation laws make it is possible to define an angular momentum alge-

bra for this four-wave mixing system analogous to the Schwinger coupled boson
representation used in [57] for the description of two-mode condensates and in
[58] for the determination of the condensate ground state. However, the present
situation requires that these considerations be extended to a compound angular
momentum representation [59, 60].
We proceed by introducing the spinor operators a1 and a2 ,

a01 a2
a1 ≡ and a2 ≡ (123)
a1 a02
as well as the two angular momentum operators S1 and S2 ,

⎡ † † ⎤
a01 a1 + a01 a1
† 1⎢1 † † ⎥
S1 ≡ a1 σ a1 = ⎣ (a a − a01 a1 ) ⎦ (124)
2 i †01 1 †
a01 a01 − a1 a1
32 P. Meystre
and
⎡ † † † ⎤
a2 a02 + a02 a2
† 1⎢1 † †
S2 ≡ a2 σ a2 = ⎣ (a a − a a ) ⎦ , (125)
⎥
2 i †02 2 † 02 2
a2 a2 − a02 a02
where σ is the Pauli spin operator. The Casimir operators K j ≡ S2j associated with
the angular momenta S j are also constants of motion, since

N̂ j N̂ j
Kj = +1 . (126)
2 2
Expressed in terms of the total spin operator,
S ≡ S1 + S 2 (127)
with components S j = S1 j + S2 j , the Hamiltonian (116) becomes
h 2 k02 hc0
H= N̂ + N̂ ( N̂ − 1)
2M M
2
N̂ D̂ 0
+ 2hc2 S 2 − − . (128)
2 2
Together with the conserved Casimir operators, this form of the four-wave mixing
Hamiltonian suggests expressing the state of the system in terms of the complete
set of quantum numbers associated with the eigenstates of the operators S21 , S22 , S2 ,
and Sz . The Hamiltonian (128) can clearly be diagonalized in terms of these states,
yielding a complete set of eigenstates and eigenenergies.
To illustrate how this works, we consider a condensate consisting of N atoms
such that initially N1 = N2 = N /2. We further assume that there are m ≪ N /2
atoms in the hyperfine state m F = 1 and none in the state m F = 0. In the “natural”
basis of the operators {K 1 , S1z , K 2 , S2z }, the initial state of this system is described
by the state vector

N N N N
|φ(0) =
, − m; , −

. (129)
4 4 4 4
Expanding it in terms of a complete set of eigenvectors {|N /4, N /4, S, Sz } of the

Hamiltonian (128), we find
N

N
N N
|φ(0) = C − m, − ; S, −m
, ; S, −m ,

(130)
S
4 4 4 4
where

N N
N N
C(S1z , S2z ; S, −m) ≡ , ; S, −m
, S1z ; , S2z . (131)

4 4 4 4
are Clebsch–Gordan coefficients [61], which are nonzero only for Sz = S1z +
S2z = −m.
This conservation of Sz under the Hamiltonian (128) further implies that the
state vector of the system is given at time t by

N N
|φ(t) = α S (t)
, ; S, −m
S
4 4

≡ α S (t)|S, −m , (132)
S
where we have in the second equality made explicit use of the value of the con-
served quantities K 1 , K 2 , and Sz to simplify the notation via

N N

, ; S, −m → |S, −m . (133)

4 4
Note that this simplification is not general, but is appropriate for the initial condition
at hand.
The equations of motion for the probability amplitudes α S (t) follow from the
Schrödinger equation. For a condensate containing N atoms they read, in a frame
rotating at the frequency hk02 N /2M + c0 N (N − 1)/2,
i α̇ S (t) = c2 S(S + 1)α S (t), (134)
and can be integrated trivially with the initial condition

N N
α S (t = 0) = C − m, − ; S, −m .
4 4
We can apply these results to the study of the dynamics of population exchange
between the different modes of the condensate. For instance, the population of the
34 P. Meystre
m F = 1 hyperfine spin state can readily be determined from the expectation value
of S1z . With Eqs. (122) and the definition of S1z , we find
† N
a1 a1 = − S1z . (135)
4
From Eq. (132) we have

2

S1z = p
α S (t)C( p, −( p + m); S, −m
, (136)

p

S

where we inserted the identity operator
N
N N N
Î =
4 , p1 ; 4 , p2 4 , p1 ; 4 p2
(137)

p1, p2
and used the simplified notation (133) as well as the property that the Clebsch–
Gordan coefficients’ are nonzero only for Sz = S1z + S2z . Equation (136) can be
evaluated numerically.
The evolution of the population of the m F = 1 side mode is shown in Fig. 3
for N = 100 atoms in the system. Case (a) illustrates the buildup from noise,
†
m = 0, while in case (b) the initial mode population is a1 a1 = m = 5. In both
cases, the side-mode population exhibits an initial growth to the point where it
contains about one-third of the atoms in the first case and about half of the atoms
in the second. This is followed by a collapse to a quasi-steady-state population,
as well as a subsequent revival at 2c2 t1 = π. This dynamics then repeats itself
periodically, with revivals at at 2c2 tn = π n, independently of N.
FIG. 3. Evolution of the population of the m F = 1 side mode for N = 100 atoms in the system.
Case (a) illustrates the buildup from noise, m = 0, while in case (b) the initial mode population is
m = 5. (From Ref. 56.)
This remarkable independence is similar to the periodic revivals that occur

in two-photon Jaynes–Cummings [62]. During the periods of collapse, one has
S j z = −m/2, so that all modes are almost equally macroscopically populated
† † †
with a01 a01 = N /4 − m/2, a1 a1 = N /4 + m/2, a02 a02 = N /4 + m/2 and
†
a2 a2 = N /4 − m/2.
A particularly interesting prediction of this analysis is the possibility to obtain
quantum correlations between side modes. In optics, four-wave mixing provides a
method to study purely quantum mechanical effects such as squeezing and nonclas-
sical states of the radiation field, and also to prepare states of composite systems
exhibiting strong quantum mechanical entanglement [63]. These states are of im-
portance in tests of the foundations of physics as well as quantum information
processing such as quantum cryptography [64, 65] and quantum computing [66].
Macroscopic quantum states of massive particles present an interesting alterna-
tive to all-optical systems, hence it is worthwhile to determine to what extent
quantum entanglement between side modes can be achieved in Bose–Einstein
condensation.
One can quantify the amount of quantum entanglement between conden-
sate modes by determining the extent to which the Cauchy–Schwarz inequal-
ity is violated by the second-order cross-correlations functions between modes
[63]. In particular, for “classically looking” optical system with positive Glauber
P representations, the single-time two-mode second-order cross-correlation func-
tion is bound by the “classical” upper bound
(2) (2) (2) 1/2

G i, j (t) ≤ G i (t)G j (t) . (138)
In case the P representation is not positive or does not exist, in contrast, the upper
bound is higher, namely,
(2) (2) (1) (2) (1) 1/2

G i, j (t) ≤ G i (t) + G i (t) G j (t) + G i (t) . (139)
In these inequalities, we have introduced the single-time and single-mode first-

order correlation functions
(1) †
G j (t) ≡ φ(t)|a j a j |φ(t) (140)
as well as the single-time two-mode second-order correlation functions
(2) † †
G i j (t) ≡ φ(t)|ai ai a j a j |φ(t) (141)
36 P. Meystre
and the single-time, single-mode, second-order correlation functions
(2) † †
G j (t) ≡ φ(t)|a j a j a j a j |φ(t) . (142)
The single-time single-mode second-order cross-correlation between the side

modes m F = ±1 can be expressed in terms of the z component of the individ-
ual pseudo-spins as

(2) N N
G −1,1 (t) = − S1z − S2z
4 4
2
N N
= − (S1z + S2z ) + S1z S2z , (143)
4 4
where

2

[N /2]

S1z S2z = p(− p − m)
αs (t)C( p, −( p + m); S, m
(144)

p

s=m
and the sum can easily be evaluated numerically.

Figure 4 compares the time dependence of the normalized central-mode–side-
mode correlation function
(2)
(2) G 01,1 (t)
R01,1 (t) ≡ (145)
(2) (2)
G 01 (t)G 1 (t)
FIG. 4. Time evolution of the one-time normalized central mode-side mode correlation function
(2) (2)
R01.1 (t) (lower curve) and of the normalized side-mode–side-mode correlation function R−1.1 (t)
(upper curve) for N = 100 and (a) m = 0, (b) m = 5. The shaded region is classically forbidden.
(From Ref. 56.)
and the side-mode–side-mode correlation

(2)
(2) G −1,1 (t)
R−1,1 (t) ≡ (146)
(2) (2)
G 1 (t)G −1 (t)
for (a) the case m = 0 where the side mode builds up from quantum fluctuations
and (b) the case m = 5 of an injected signal.
These results illustrate how the correlations between the central mode and side
modes do satisfy the classical Cauchy–Schwarz inequality, while the side-mode–
side-mode cross-correlations violate them. The violation is particularly strong in
the case of buildup from noise, as should be intuitively expected. In that case the
hyperfine sidemodes m F = ±1 play symmetric roles, thus,
(2) (2) (1)
G −1,1 (t) = G 1 (t) + G 1 (t), (147)
(2)
i.e., R−1,1 = 1 Equation (139) then becomes an equality, leading to the maxi-
mum violation of the classical Cauchy-Schwarz inequality allowed by quantum
mechanics.
The difference in the behavior of the two-mode correlation functions between
side modes and those involving one central and one side mode can be intuitively
† †
understood from the form of the wave-mixing term a1 a−1 a01 a02 appearing in
the Hamiltonian (116). Indeed, the coupling between side modes, involving two
† †
annihilation operators, is reminiscent of the interaction a1 a2 in the Hamiltonian of
parametric amplification leading to squeezing and quantum entanglement between
two side modes. In contrast, the coupling between central and side modes involves
both an annihilation and a creation operator.
From these results, it appears that multicomponent condensates offer a fascinat-
ing potential method to create quantum entanglement at a truly macroscopic level,
a possibility made even more attractive by the fact that these systems suffer little
from dissipation, since they consist of ground-state atoms. The final section extends
these results to the nonlinear mixing of optical and matter wave fields, opening up
the additional possibility of optically controlling the quantum statistical properties
of Schrödinger fields.
VI. Mixing of Optical and Matter Waves
We have mentioned that conventional nonlinear optics relies for its description
on the elimination of the material dynamics, while in nonlinear atom optics one
eliminates (part of) the electromagnetic field, resulting in effective atom–atom
interactions. One should keep in mind, however, that these represent limiting cases.
38 P. Meystre
Outside these two regimes the atomic and optical fields are dynamically coupled,
and neither field is readily eliminated. This leads to intriguing new possibilities,
including the nonlinear mixing of optical and matter waves that we investigate in
this last section.
A. PARAMETRIC AMPLIFICATION OF ATOMIC AND OPTICAL FIELDS

We first discuss the parametric amplification of matter waves resulting from the
coupling between a Bose–Einstein condensate and both a strong laser field and a
weak single mode of a ring cavity. The strong pump laser is treated as a classical,
undepleted light field. It is assumed to be detuned far enough away from resonance
that spontaneous emission may be safely neglected. The cavity field, or ”probe”
field, is assumed to be weak relative to the pump, and is treated fully quantum
mechanically as a dynamical variable.
Assuming that the probe field begins in or near the vacuum state, and the atomic
field consists initially of a trapped condensate, the initial dynamics is dominated by
a single process: the absorption of a pump photon by a condensate atom followed
by the emission of a probe photon. This is a two-photon virtual transition in which
the excited atomic state population remains negligible. Due to atomic recoil, the
absorption/emission process transfers the atom from the condensate ground state
to a new state that is shifted in momentum space by the two-photon recoil. This
new state, which is a momentum side mode of the original condensate, can be
thought of as a second condensate component. It may or may not be in the same
internal ground state as the original condensate, depending on the polarizations of
the pump and probe photons. If the initial and final magnetic sublevels are different,
as might be the case in a spinor condensate, the transition would be termed a Raman
transition. This is to be contrasted with the case of a scalar condensate, in which
the initial and final sublevels are identical and the light-atom interaction may be
thought of as Rayleigh scattering or as two-photon Bragg scattering. The model
discussed in this section deals specifically with this Bragg-Rayleigh scheme, but
it can be extended to the Raman problem with only minor modifications.
The physics underlying the parametric amplification of optical and matter waves
can be understood intuitively in essentially classical terms: as the matter-wave side
mode is populated, it begins to interfere with the original condensate, resulting in
fringes [67]. These fringes are seen by the pump and probe fields as a spatial density
grating, which then enhances the photon scattering process. This interplay between
interference fringes and scattering can act as a positive feedback mechanism,
in which case the system is unstable and characterized by exponential growth.
Any small signal, including quantum noise, is sufficient to trigger the instability,
resulting in the generation of exponentially growing side-mode and probe fields.
Of course, this exponential growth is eventually reversed by high-intensity effects,
so that the long-time dynamics are characterized by large-amplitude nonlinear
oscillations. In many ways, this system is an extension into the ultracold regime
of the so-called collective atomic recoil laser (CARL) [68–70] We concentrate
here on the small-signal regime, characterized by exponentially growing fields.
As we shall see, one particularly interesting property of this system is that the
quantum state of the probe and side-mode fields depends strongly on the initial
conditions, so that, e.g., by injecting a small coherent light field into the probe,
one can create an entirely different quantum state than that generated from the
amplification of quantum vacuum fluctuations. The differences are manifested in
the quantum statistics of the individual field modes, as well as in nonclassical
correlations and entanglement between them.
Under conditions of far off-resonant excitation, the effective Hamiltonian
describing this system is [71]
h2q 2 †
†
H = d 3q ĉ (q)ĉ(q) + hδk b̂kλ b̂kλ
2M kλ
∗
g g ′ ′
kλ †
+ d 3 qh kλ b̂kλ b̂k′ λ′ ĉ† (q)ĉ(q + k − k′ ), (148)
′
kλk λ ′ δ
where the bosonic operator ĉ(q) annihilates a ground-state atom of momentum

hq, b̂kλ annihilates a photon of momentum hk and polarization λ, δ is the detuning
between the laser–atom frequency detuning, and gkλ is the atom-field coupling
constant. This Hamiltonian yields the Heisenberg equations of motion,
d hq 2 g ∗ gk′ λ′ †
kλ
ĉg (q) = −i ĉg (q) − i b̂kλ b̂k′ λ′ ĉg (q + k − k′ ) (149)
dt 2M kλk′ λ′
δ
and
∗
d gkλ gk′ λ′ †
b̂kλ = −iδk b̂kλ − i d 3q ĉg (q)b̂k′ λ′ ĉg (q + k − k′ ), (150)
dt k′ λ′

where δk = c|k| − δ.
The nonlinearity in the effective Hamiltonian (148) allows for cross-phase mod-
ulation and four-wave mixing phenomena between the atomic and optical fields.
The phase shift which the optical field acquires due to cross-phase modulation
with the atomic field is
|gkλ |2 N̂
φopt = t, (151)
δ
where N̂ is the operator for the total number of atoms, which may or may not be
40 P. Meystre
a constant of motion. Particularly in closed systems, where the number of atoms

is fixed, this nonlinear phase shift is nothing more than the dispersive shift to the
index of refraction found in linear optics. Nontrivial effects may occur in open
systems, e.g., if the number of atoms fluctuates in response to changes in the
optical field, but otherwise leads only to a negligible shift in the optical dispersion
relation. Similarly, the phase shift acquired by the atomic field due to cross-phase
modulation,
|gkλ |2 b̂† b̂kλ

kλ
φat = t, (152)
kλ
δ
is simply the spatially independent part of the Stark shift, which has no dynamical
effect on the atomic density.
In the examples we consider in the following, the Stark shift and the dispersion
of the optical field may be safely neglected. We focus instead on the effects of
four-wave mixing between atomic and optical fields, which leads to parametric
amplification and the generation of entanglement between atomic and optical fields.
In order to extract the basic physics underlying the four-wave mixing of optical
and matter waves, we ignore cavity losses and treat the pump laser field classically
in the following. The effective Hamiltonian (148) then reduces to
q2

H = hωrec b̂† b̂ + d 3 q 2 ĉ† (q)ĉ(q) + χ [ĉ† (q)ĉ(q − K)b̂ + H.c.] ,
K
(153)
where b̂ is the photon annihilation operator for the probe mode, whose frequency
is detuned from that of the pump by hωrec , and
K = k L − k. (154)
The effective atom-probe coupling constant is given by
g
χ= , (155)
2|δ|ωrec
where g is short hand for the coupling constant gkλ of the probe mode and is
the pump Rabi frequency.
We consider specifically the case where the atomic field initially consists of a
condensate well below the critical temperature Tc . Assuming that its initial mo-
mentum width is small compared to the recoil momentum hK, it is reasonable to
approximate it as a plane wave at q = 0. The effect of the optical fields is then to
couple the condensate to matter-wave side modes separated in momentum space

by integer multiples of hK. Making use for expediency of the symmetry breaking
approach of Sect. IV, we replace the condensate field operator ĉ(0) by a c-number
plus a perturbation,
√
ĉ(0) = N + ĉ0 , (156)
where N is number of atoms initially in the condensate. This is analogous to the

Bogoliubov approach, except that instead of computing a quasi-particle spectrum,
we proceed along the lines of the linear stability analyses familiar from quantum
optics. Specifically, we derive time-dependent equations
√ of motion for the per-
turbations ĉ0 about the condensate wave function N . For times short enough
that the fraction of atoms transferred into the side modes remains negligible, the
Hamiltonian (153) can be linearized, leading to
† †
√ †
Ĥ lin = hωrec [ĉ− ĉ− + ĉ+ ĉ+ + δ b̂† b̂ + χ N (b̂† ĉ− + b̂† ĉ+ + H.c.)], (157)
where ĉ± = ĉ(±K). This Hamiltonian yields a 3 × 3 set of linear equations for the
three coupled field modes (two matter waves and one optical mode),
⎛ ⎞ ⎛ √ √ ⎞⎛ ⎞
b̂ −δ
√ −χ N −χ N b̂
d ⎝ †⎠ † ⎠
ĉ− = i ⎝ χ √N 1 0 ⎠ ⎝ ĉ− , (158)
dτ
ĉ+ −χ N 0 −1 ĉ+
where τ = ωrec t. The time dependence of the three field modes is determined by
the eigenvalues of the 3 × 3 matrix on the r.h.s. of (158). These are found by solving
the characteristic equation
ω3 + ω2 − ω − + 2χ 2 N = 0, (159)
which has complex solutions provided that
(3 + 2 )3/2 − 3 + 9
χ2N > . (160)
27
When the condition (160) is satisfied, the system is unstable and the three fields
initially exhibit exponential growth. This growth will of course be arrested by
saturation effects as soon as the side-mode populations become significant.
†
The presence of the b̂† ĉ− term in Eq. (157) describes the creation of correlated
atom–photon pairs, and is analogous to the term found in the optical parametric
amplifier [63] which describes the generation of correlated photon pairs. One of
42 P. Meystre
the most interesting applications of the optical parametric amplifier is the gener-
ation of entangled quantum optical states. A similar entanglement occurs in the
present system, except that it is now between atomic and optical field modes. A
useful measure of quantum entanglement is the second-order equal-time intensity
correlation function, defined as
(2) N̂ i N̂ j − δi j N̂ i
gi j = . (161)
N̂ i N̂ j
These two-mode correlation functions (i = j) arise, e.g., if we consider a mea-

surement of the intensity difference between two modes, described by the operator
N̂ i j = N̂ i − N̂ j , (162)
whose variance is given by

(2)
(Ni j )2 = (Ni )2 − 2Ni N j gi j − 1 + (N j )2 .

(163)
(2)
For uncorrelated fields, gi j = 1, and we have the usual rule for the addition of
uncorrelated noise sources,
(Ni j )2 = (Ni )2 + (N j )2 . (164)
If, however, there are correlations between the fluctuations in the intensities of the
(2)
two modes, then we have gi j > 1, and the variance Ni j will be less than that
given by (164).
For classical fields (positive P function), the two-mode (i = j) correlations are
constrained by the Cauchy–Schwarz inequality,
(2) (2) 1/2 (2) 1/2

gi j ≤ gii gjj . (165)
Quantum mechanical fields, however, can violate this inequality and are instead
constrained by
1/2 1/2
(2) (2) 1 (2) 1
gi j ≤ gi j + gjj + , (166)
N̂ i N̂ j
which reduces to the classical result in the limit of large intensities. When triggered
from the vacuum, it is readily found that the single-mode correlation functions are
those of thermal fields g (2) (τ ) = 2. In this case the cross-correlation functions are
found to be [71]
1/2
1 1/2

(2) (2) 1
g12 = g23 = 2 + 2+ , (167)
N1 + N3 N2
and
(2)
g13 = 2, (168)
where 1 corresponds to the probe mode, 2 to the −K side mode, and 3 to the +K
mode. As N1 ≫ N3 we see that the intensity correlations between the probe and
−K side mode are very close to the maximum allowed by quantum mechanics,
whereas the other two-mode correlations are considerably more classical.
B. MATTER-WAVE SUPERRADIANCE
So far, we used an optical cavity to select the probe optical field mode. Recent
experiments by W. Ketterle and co-workers, however, have demonstrated that under
appropriate conditions mode selectivity can be achieved without the use of cavities
[45]. They observed highly directional scattering of pump photons along the long
axis of a cigar-shaped condensate when the pump polarization and wave vector
were both perpendicular to the condensate axis; see Fig. 5. The optical mode is
therefore selected by the condensate geometry, and this results in the buildup of
population in a condensate side mode with a single well-defined momentum.
Typical condensates are sufficiently localized in space that scattered photons
immediately exit the condensate, with a negligible cross section for multiple scat-
tering. Due to photon recoil, however, each such scattering event transfers an atom
from the condensate state into a state which is shifted in momentum space by the
two-photon recoil momentum, as discussed in the preceding section. The positive
feedback which leads to instability and parametric gain is then provided solely by
the recoiling atoms, which, due to their slow velocity, remain in the condensate
volume long enough to influence the direction of subsequent photon scattering.
To see quantitatively how this work, it is necessary to consider in more detail
the coupling coefficients gk,λ of the Hamiltonian (148). Their explicit form is

| 0 | c|k|d 2
g(k) = |k̂ × x̂|, (169)
2|| 2hǫ0 (2π)3
where d is the magnitude of the atomic dipole moment for the transition involved.
Because the geometry of the condensate is essential in understanding matter-
wave superradiance, it is no longer sufficient to consider a free-space description.
44 P. Meystre
FIG. 5. Superradiant matter-wave scattering. A cigar-shaped condensate is illuminated by an off-

resonant laser beam perpendicular to the long axis. (b)–(g) show absorption images after a 20-ms time
of flight of the atoms away from the trap, for various polarizations and durations of exposure to the
incident beam. For a polarization parallel to the long axis of the condensate, (b)–(d), normal Bragg
scattering is observed. For a polarization perpendicular to the long axis, the excitation of well-defined
condensate side modes is apparent. The pulse durations for (e)–(g) are 35, 75, and 100 μs, respectively.
(From Ref. [45].)
Instead, the atomic field is now taken to be a number state in which N atoms occupy
the ground state ϕ0 (r) of the trap containing the condensate. The effect of atomic
recoil during Bragg (or Rayleigh) scattering between the pump and the vacuum
mode k is therefore to transfer atoms into the state ϕ0 (r) exp[i(k L − k)r]. This
suggests expanding the atomic field operator onto quasi-modes according to

ˆ
(r, t) = r|q e−i(ωq +μ)t ĉq (t), (170)
q
where r|q = ϕ0 (r) exp(iq · r), and ωq = h|q|2 /2M. This is similar to the slowly
varying envelope approximation of optical physics, the envelope being given here
by ϕ0 (r).
A discrete quantization of the q values follows from the requirement that the
operators {ĉq } obey approximately the boson commutation relations
†
[ĉq , ĉq′ ] = q|q′ ≈ δq,q′ . (171)
Due to the finite size of the ground-state wave function ϕ0 (r), this means that q
and q′ must be well separated in k-space. Hence, the summation in Eq. (170) is
taken to include the condensate mode q = 0 as well as a grid of q values as closely
spaced as is consistent with orthogonality. Clearly, this expansion is not rigorously
orthogonal and complete, but it is sufficient to account for the quantum statistical
effects which occur above the critical phase-space density.
The quasi-mode populations experience losses as the recoiling atoms eventually
propagate out of the condensate volume. The lifetime of the quasi-mode q, however,
is on the order of Tq ≡ m L q /h|q|, where L q is the length of the condensate along
q. These losses tend to destroy the coherence between condensate and quasi-mode,
which accounts for the observation of a threshold for superradiance in the MIT
experiment: for insufficient laser power, the growth of matter-wave coherence
cannot overcome the losses. As this threshold is very small, we consider only the
situation far above threshold, in which case the losses may be neglected.
An important aspect of BEC superradiance is the generation of families of
higher-order side modes due to the scattering of pump photons by the first-order
side modes, as clearly evident in Fig. 5. For the present discussion, however, we
consider a simplified model containing only the primary Rayleigh scattering pro-
cess, whereby a condensate atom is transferred to a first-order side mode by scat-
tering a pump photon of momentum k L and frequency ω L . With this simplification,
the effective Hamiltonian (148) becomes for the trap situation under consideration:
t
3 †
d 3 k hg(k)ρq (k)eiωq ĉq† b̂† (k)ĉ0 + H.c. ,

Ĥ = d khω(k)b̂ (k)b̂(k) +
q=0
(172)
where

ρq (k) = d 3 r|ϕ0 (r)|2 exp [−i(k − k L + q) · r] (173)
is the Fourier transform of the ground-state density distribution centered at k =

k L − q.
From the Hamiltonian (172), it is straightforward to derive the equation of
motion for b̂(k), which upon formal integration yields

b̂(k, t) = b̂(k, 0)e−iω(k)t − i g(k)ρq (k)eiωqt
q=0
t
× dτ e−i[ω(k)+ωq ]τ ĉq† (t − τ )ĉ0 (t − τ ), (174)
0
where the first term gives the free electromagnetic field, i.e., vacuum fluctuations,
46 P. Meystre
and the second term is the radiation field due to Rayleigh scattering. A nonzero
†
expectation value of the coherence operator ĉq ĉ0 indicates the presence of in-
terference fringes, hence the radiated field increases as fringes build up. This is
the term that leads to the instability, where the memory of previous scattering
events, stored in the matter-wave interference fringes, enhances the present rate of
Rayleigh scattering.
Equation (174) is then substituted into the equation of motion for ĉq , which in
the Markov approximation yields

d Gq †
ĉq = −i d 3 kg(k)ρq (k)b̂† (k, 0)ei[ω(k)+ωq ]t ĉ0 + ĉ ĉq , (175)
dt 2 0
where

d 3 k|g(k)|2 |ρq (k)|2 δ ω(k) + ωq

G q = 2π (176)
is the single-atom gain. In deriving Eq. (175) we have used the orthogonality of the
states {|q } to make the approximation ρq∗ (k)ρq′ (k) ≈ |ρq (k)|2 δq,q′ , and neglected
the principal part which accompanies the δ function. If included, it would contribute
additional ground-state collisions due to the dipole–dipole interaction.
For a closed atomic system, the total number of atoms is conserved, hence
†

ĉ0 ĉ0 = N − ĉq† ĉq . (177)
q=0
†
For very short times we can therefore take ĉ0 ĉ0 ≈ N . In this case Eq. (175)
reduces to
d Gq
ĉq = N ĉq + fˆ†q (t), (178)
dt 2
where fˆq (t) is a noise operator whose correlation functions are given in the Markov
approximation by
†
fˆq (t) fˆq (t ′ ) = 0,
(179)
†
fˆq (t) fˆq (t ′ ) = G q N δ(t − t ′ ).
These noise operators allow the system to be triggered by quantum fluctuations,

and hence describe the “spontaneous” scattering which occurs in the absence of
any side-mode population.
Equation (178) can be solved exactly, giving

t
ĉq (t) = e(G q /2)N t ĉq (0) + dτ e(G q /2)N τ fˆ†q (t − τ ). (180)
o
From Eq. (180) it is possible to compute the probability Pq (n, t) of having n atoms
in mode q at time t, assuming zero population at t = 0. To accomplish this we first
compute the antinormally ordered characteristic function,
χq (η) = exp[−η∗ ĉq ] exp[ηĉq† ] , (181)
yielding
2
χq (η) = e−|η| [n̄ q (t)+1] , (182)
where
n̄ q (t) = exp(G q N t) − 1 (183)
is the mean population of mode q at time t. We can identify expression (182) as

corresponding to a chaotic field [63]. The number distribution for a chaotic field
is given by
1 −(n+1)

1
Pq (n, t) = 1+ , (184)
n̄ q (t) n̄ q (t)
which for n̄ q (t) ≫ 1 is well approximated by exp[−n/n̄ q (t)]/n̄ q (t).

To conclude the analysis, we still need to determine the geometric dependence
of the single-atom gain given by Eq. (176), and show that it is largest for radiation
along the long axis of the condensate. This is what allows the system to operate
without the benefit of an optical cavity, as already mentioned. We first note that
G q depends on g(k), which contains the dipole radiation pattern, as well as on
ρq (k) which depends on the geometry of the initial condensate. For a cigar-shaped
condensate, aligned along the ẑ axis, ρq (k) is a disk which lies parallel to the x̂–ŷ
plane in k-space. The dimensions of the disk in k-space are roughly the inverse of
the condensate dimensions in r-space. Thus, for a condensate whose dimensions
are large compared to an optical wavelength, the dimensions of ρq (k) are small
compared to k0 .
Since g(k) is slowly varying compared to ρq (k), it can be removed from the
integral in Eq. (176), and evaluated at the center of ρq (k). In addition, we neglect
the recoil shift ωq in the δ function, as it has negligible effect on the value of G q .
48 P. Meystre
The remaining integral then defines the solid angle q for the scattered radiation
associated with the qth mode according to
1

q = 2 d 3 k|ρq (k)|2 δ(|k| − k0 ). (185)
k0
This shows that only q values for which the center of ρq (k) lies at a distance k0
from the origin experience gain, a consequence of energy conservation. Thus, for
every active quasi-mode q there is a corresponding radiation direction k̂, such that
q = k0 (ŷ − k̂).
We can obtain a good estimate for q by taking |ρq (k)|2 to be an ellipsoid solid
with the inverse dimensions of the condensate. This gives
2 −1/2
4π 2 L 2
q = 2 2 cos θk̂,ẑ + sin θk̂,ẑ , (186)
k0 W W
where L is the length of the condensate along the ẑ axis, W is the radial diameter,
and θk̂,ẑ is the angle between the radiation direction and the long axis of the
condensate. Thus q is maximized for radiation along ẑ and −ẑ where it is given
by q = 4π/k02 W 2 . It is important to note that for the isotropic case L = W there
is no preferred direction, and a ring of radiation is instead observed.
Taking into account all of these considerations, the expression for the single-
atom gain becomes
sin2 θk̂,x̂
Gq = G , (187)
cos2 θk̂,ẑ + (L/W )2 sin2 θk̂,ẑ
where
3| |2 Ŵ
G= (188)
8||2 k02 W 2
is the maximum single-atom gain, Ŵ = k03 d 2 /3π hǫ0 being the single-atom spon-
taneous decay rate, and θk̂,x̂ is the angle between the radiation and polarization
directions. For the parameters of the MIT experiment we find G ≈ 4 × 10−4 · I,
where I is the laser intensity in mW/cm2, and G is given in hertz. A rough estimate
of the duration of a superradiant pulse for the case N = 106 and I = 100 mW/cm2
is t = ln(N )/G N ≈ 150μs, in excellent agreement with experimentally observed
time scales.
When modes with different values of G q compete, the competition is “unfair”
and the mode with the largest G q generally depletes all of the condensate atoms
before the populations of the other modes have a chance to grow. Modes with the
same G q , such as the quasi-modes corresponding to radiation along the ẑ and −ẑ
directions, instead compete “fairly,” and while the interplay is highly sensitive to
the random initial conditions, neither mode necessarily wins, i.e., there is no real
“winner-takes-all” effect, and often a “tie” will occur.
C. PHASE-COHERENT MATTER-WAVE AMPLIFICATION

Phase-coherent amplification goes one step further than the superradiance experi-
ment just discussed: in addition to increasing the number of atoms in a given mode
of the Schrödinger field, it also preserves the phase of the state being amplified.
The idea of phase-coherent matter-wave amplification is of course implicit in the
discussion of Sect. VI. B, but while these original theoretical proposals used optical
cavities to select the required optical field modes, the superradiance experiments
show that this is actually not necessary. It is this observation which is at the heart
of the phase-sensitive amplifier recently demonstrated.
In the absence of a seed signal, the parametric amplification of the side
mode builds up from density fluctuations, the atom-optical version of “vacuum
fluctuations.” To demonstrate the phase-preserving nature of the amplification,
it it necessary to amplify an injected signal of known properties. This was re-
cently achieved in experiments by Kozuma et al. [47] and by Inouye et al. [46].
In these experiments, the seed beam was generated by exposing the condensate
to a pulsed optical standing wave that transferred atoms into the appropriate side
mode by Bragg scattering. The phase of the amplified signal was then determined
FIG. 6. Phase coherent matter-wave amplifier A weak matter wave is split off a condensate by
application of a Bragg pulse. The probe is amplified by passing through the condensate “pumped” by a
strong laser beam. The coherence of the amplified matter wave is verified by observing its interference
with a reference matter wave produced by applying a second Bragg pulse to the condensate. (From
Ref. [46].)
50 P. Meystre
FIG. 7. Input–output characteristic of the matter-wave amplifier, showing the amplification of the
probe pulse and the depletion of the condensate used as an amplifying medium. (From Ref. [46].)
by atom interferometry, using as a reference a second matter wave also split off
the condensate by Bragg scattering; see Figs. 6 and 7.
Phase-coherent matter wave amplification can be seen as both the culmination
and the “coming of age” of atom optics. Particularly when used in combination
with atom lasers, coherent matter-wave amplifiers provide an exciting new tool in
the arsenal of atom optics, with potential applications in precision measurements,
the amplification of small signals, and atom lithography. In addition, very much
as optical parametric amplifiers are central to many aspects of modern quantum
optics, including in particular the generation of squeezed light and of quantum
entanglement in optical fields, it can be expected that similar developments will
now occur in atom optics.
VII. Acknowledgments
I have benefited from discussions with numerous collaborators and colleagues,

most notably E. V. Goldstein, G. Lenz, M. G. Moore, H. Pu, O. Zobay, and E. M.
Wright. This work is supported in part by the U.S. Office of Naval Research
under Contract 14-91-J1205, by the National Science Foundation under Grant
PHY98-01099, by the U.S. Army Research Office, and by the Joint Services Optics
Program.
VIII. References
1. de Broglie, L. (1924). Phil. Mag. 47, 446.
2. Stern, O. (1929). Naturwiss. 17, 391.
3. Knauer, F., and Stern, O. (1929). Z. Phys. 53, 779.
4. Frisch, R. (1933). Z. Phys. 86, 42.
5. Lenz, G., Meystre, P., and Wright, E. M. (1993). Phys. Rev. Lett. 71, 3271.
6. Lenz, G., Meystre, P., and Wright, E. M. (1984). Phys. Rev. A 50, 1681.
7. Zhang, W., and Walls, D. F. (1994). Phys. Rev. A 49, 3799.
8. Castin, Y., and Mølmer, K. (1995). Phys. Rev. A 51, R3426.
9. Franken, P. A., Hill, A. E., Peters, C. W., and Weinrich, G. (1961). Phys. Rev. Lett. 7, 118.
10. Armstrong, J. A., Bloembergen, N., Ducuing, J., and Pershan, P. S. (1962). Phys. Rev. 127, 1918.
11. Bloembergen, N. (1964). Nonlinear Optics. Benjamin (New York).
12. Shen, R. Y. (1964). “The Principles of Nonlinear Optics.” Wiley-Interscience, New York.
13. Boyd, R. W. (1992). “Nonlinear Optics.” Academic, San Diego.
14. Pötting, S., Meystre, P., and Wright, E. M. Nonlinear photonic crystals. In “Atomic Solitons in
Optical Lattices” (B. J. Eggleton and R. E. Slusher, Eds.), Springer-Verlag, in press.
15. Merzbacher, E. (1998). “Quantum Mechanics.” Wiley, New York.
16. Sakurai, J. J. (1995). “Modern Quantum Mechanics.” Addison-Wesley, Reading, MA.
17. Landau, L. D., and Lifshitz, E. M. (1977). “Quantum Mechanics.” Pergamon, New York.
18. Messiah, A. (1961). “Quantum Mechanics.” North Holland, Amsterdam.
19. Cohen-Tannoudji, C., Diu, B., and Laloë, F. (1997). “Quantum Mechanics.” Wiley, New York.
20. Bogoliubov, N. N. (1967). Lectures in Quantum Statistics. Gordon & Breach, New York.
21. Robertson, B. (1973). Am. J. Phys. 41, 678.
22. Ballentine, L. E. (1990). “Quantum Mechanics.” Prentice Hall, Englewood Cliffs.
23. Gross, E. P. (1961). Nuovo Cimento 20, 454.
24. Gross, E. P. (1963). J. Math. Phys. 4, 195.
25. Pitaevskii, L. P. (1961). Sov. Phys. JETP 13, 451.
26. Dalfovo, F., Giorgini, S., Pitaevskii, L. P., and Stringari, S. (1999). Rev. Mod. Phys. 71, 463.
27. Kerman, A. K., and Koonin, S. E. (1976). Ann. Physics (New York) 100, 332.
28. Negele, J. W. (1982). Rev. Mod. Phys. 54, 913.
29. Zhang, W., Walls, D. F., and Sanders, B. C. (1994). Phys. Rev. Lett. 72, 60.
30. Deng, L., Hagley, E. W., Wen, J., Trippenbach, M., Band, Y., Julienne, P. S., Simsarian, J. E.,
Helmerson, K., Rolston, S. L., and Phillips, W. D. (1999). Nature 398, 218.
31. Bogoliubov, N. N. (1947). J. Phys. (USSR) 11, 23.
32. Fetter, A. L. (1972). Ann. Phys. (New York) 70, 67.
33. Lifshitz, E. M., and Pitaevskii, L. P. (1980). “Statistical Physics,” Part 2. Pergamon, New York.
34. Edwards, M., Dodd, R. J., Clark, C. W., and Burnett, K. (1996). J. Res. Natl. Inst. Std. Technol.
101, 553.
35. Edwards, M., Ruprecht, P. A., Burnett, K., Dodd, R. J., and Clark, C. W. (1996). Phys. Rev. Lett.
77, 1671.
36. Singh, K. G., and Rokshar, D. S. (1996). Phys. Rev. Lett. 77, 1667.
37. Dalfovo, F., Giorgini, S., Guilleumas, M., Pitaevskii, L. P., and Stringari, S. (1997). Phys. Rev. A
56, 3840.
38. Esry, B. D. (1997). Phys. Rev. A 55, 1147.
39. Hutchinson, D. A., Zaremba, E., and Griffin, A. (1997). Phys. Rev. Lett. 78, 1842.
40. You, L., Hoston, W., and Lewenstein, M. (1997). Phys. Rev. A 55, R1581.
41. Griffin, A. (1996). Phys. Rev. B 53, 9341.
42. Fetter, A. L., and Walecka, J. D. (1971). “Quantum Theory of Many Particle Systems.” McGraw-
Hill, New York.
43. Goldstein, E. V., Plättner, K., and Meystre, P. (1995). Quantum Optics 7, 743.
44. Goldstein, E. V., Plättner, K., and Meystre, P. (1996). J. Res. Natl. Inst. Std. Technol. 101, 583.
45. Inouye, S., Chikkatur, A. P., Stamper-Kurn, D. M., Stenger, J., and Ketterle, W. (1999). Science
285, 571.
46. Inouye, S., Pfau, T., Gupta, S., Chikkatur, A. P., Görlitz, A., Pritchard, D. E., and Ketterle, W.
(1999). Nature 402, 641.
52 P. Meystre
47. Kozuma, M., Suzuki, Y., Torii, Y., Sigiura, T., Kuga, T., Hagley, E. W., and Deng, L. (1999). Science
286, 2309.
48. Burger, S., Bongs, K., Dettmer, S., Ertmer, W., Sengstock, K., Saopera, A., Shlyapnikov, G. V.,
and Lewenstein, M. (1999). Phys. Rev. Lett. 83, 5198.
49. Denschlag, J., Simsarian, J. E., Feder, D. L., Clark, C. W., Collins, L. A., Cubizolles, J., Deng, L.,
Hagley, E. W., Helmerson, K., Reinhardt, W. P., Rolston, S. L., Schneider, B. I., and Phillips, W. D.
(2000). Science 287, 97.
50. Trippenbach, M., Band, Y. B., and Julienne, P. S. (1998). Opt. Express 3, 530.
51. Meystre, P., and Sargent III, M. (1999). “Elements of Quantum Optics.” Springer-Verlag,
Heidelberg.
52. Zobay, O., Goldstein, E. V., and Meystre, P. (1999). Phys. Rev. A 60, 3999.
53. Trippenbach, M., Band, Y. B., and Julienne, P. S. (2000). Phys. Rev. A 62, 023608 (2000).
54. Ho, T. L. (1998). Phys. Rev. Lett. 81, 742.
55. Ohmi, T., and Machida, K. (1998). J. Phys. Soc. Japan 67, 1822.
56. Goldstein, E. V., and Meystre, P. (1999). Phys. Rev. A 59, 1509.
57. Milburn, G. J., Corney, J., Wright, E. M., and Walls, D. F. (1997). Phys. Rev. A 55, 4318.
58. Law, C. K., Pu, H., and Bigelow, N. P. (1998). Phys. Rev. Lett. 81, 5257.
59. Mattis, D. C. (1981). “The Theory of Magnetism I.” Springer-Verlag, New York.
60. Wang, L., and Eberly, J. H. (1993). Phys. Rev. A 47, 4248.
61. Abramowitz, M., and Stegun, I. A. “Handbook of Mathematical Functions.” Dover, New York.
62. Buck, B., and Sukumar, C. V. (1981). Phys. Lett. A 81, 132.
63. Walls, D. F., and Milburn, G. J. (1994). Quantum Optics. Springer-Verlag, Berlin.
64. Bennett, C. H. (1992). Science 257, 752.
65. Tittel, W., Ribordy, G., and Gisin, N. (1998). Phys. World 11, 41.
66. Ekert, A., and Jozsa, R. (1996). Rev. Mod. Phys. 68, 733.
67. Andrews, M. R., Townsend, C. G., Miesner, H. J., Durfee, D. S., Kurn, D. M., and Ketterle, W.
(1997). Science 275, 637.
68. Bonifacio, R., and De Salvo, L. (1994). Nuclear Instrum. Meth. Phys. Res. A 341, 360.
69. Bonifacio, R., De Salvo, L., Narducci, L. M., and d’Angelo, E. J. (1994). Phys. Rev. A 50, 1716.
70. Bonifacio, R. (1998). Optics Commun. 146, 236.
71. Moore, M. G., Zobay, O., and Meystre, P. (1999). Phys. Rev. A 60, 1491.
FORMATION OF ULTRACOLD
MOLECULES (T ≤ 200 μK ) VIA
PHOTOASSOCIATION IN A GAS
OF LASER-COOLED ATOMS
FRANÇOISE MASNOU-SEEUWS and PIERRE PILLET
Laboratoire Aimé Cotton, Bât. 505, Campus d’Orsay, 91405 Orsay Cedex, France
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
II. Physical Concepts in Photoassociation and Formation of Molecules. . . . . . 57
A. Long-Range Alkali Dimer Molecules: A Pair of Atoms . . . . . . . . . . . 59
B. The Le Roy–Bernstein Law for Bound Levels: Analogy with Rydberg
Law, Scaling Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
C. Theoretical Predictions for the Photoassociation Rates . . . . . . . . . . . . 70
D. The Reflection Principle for Calculation of Franck–Condon Factors . . . 73
E. The Nodal Structure of the Zero-Energy Scattering Wavefunction. . . . . 75
F. A Simple Theory of Photoassociation . . . . . . . . . . . . . . . . . . . . . . 77
G. How to Make a Molecule in the Ground State or Lower Triplet State . . . 78
III. Ultracold Molecule Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
A. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
B. Experimental Setup in the Orsay Experiment . . . . . . . . . . . . . . . . . . 80
1. Cold Atomic Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2. Photoassociation Laser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3. Detection of the Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
C. Photoassociation Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
1. Trap-Loss Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2. Ion Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
D. Mechanism for Cold Molecule Formation . . . . . . . . . . . . . . . . . . . . 91
E. Translational Temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
F. Photoassociation Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
G. Rotational and Vibrational Temperatures . . . . . . . . . . . . . . . . . . . . 96
H. Other Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
IV. Theoretical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
A. Dynamics: Determination of Vibrational Wavefunctions for
Photoassociated Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
1. Mapped Fourier Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
2. Numerov Approach and Other Methods . . . . . . . . . . . . . . . . . . . 107
3. Analytical Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
B. The Urgent Need for Accurate Molecular Potential Curves and
Electronic Dipole Transition Moments . . . . . . . . . . . . . . . . . . . . . . 108
V. Present Status of the Comparison between Experiment and Theory:
Formation Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

ISBN 0-12-003847-1/ISSN 1049-250X/01 $35.00
54 F. Masnou-Seeuws and P. Pillet
A. Photoassociation Rates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

B. Cold-Molecule Formation Rate . . . . . . . . . . . . . . . . . . . . . . . . . . 111
C. Tunneling Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
VI. New Schemes for Making Ultracold Molecules. . . . . . . . . . . . . . . . . . . 116
VII. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
VIII. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
IX. References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
I. Introduction
Interest in making cold (T < 1 K) and ultracold (T well below 1 mK) molecules
has long been recognized; among the possible applications, we may cite ultrahigh-
resolution spectroscopy measurements and metrology. Other applications can
be considered, for instance the search for an elementary dipole moment in a
paramagnetic molecule as a test of time-reversal symmetry (Kozlov and Ezhov,
1994; Hinds and Sangster, 1992; Sauer et al., 1994). Threshold laws for molecular
collisions at very low energies could also be investigated (Gray and Rice, 1985;
Forrey et al., 1989). More recently, the creation of a Bose–Einstein condensate of
molecules has appeared as a new challenge (Timmermans et al., 1999a,b, Wynar
et al., 2000, Koŝtrun et al., 2000) while Julienne et al. (1998) have proposed
creating a “molecular laser.”
In 2000, an extensive review written by Bahns et al. on the formation of cold
(T ≤ 1 K) molecules gave a detailed description of possible applications to high-
resolution spectroscopy, manipulation of molecules, cold collisions, production
of cold trimers and larger clusters, and finally Bose–Einstein condensation. Non-
optical cooling techniques were reviewed, including free-jet expansion and he-
lium buffer gas cooling, leading to subkelvin temperatures. To reach the ultracold
domain, i.e., submillikelvin temperatures, optical cooling techniques need to be
considered. However, direct laser cooling of molecules does not seem a promising
technique, and it appears more efficient first to cool the atoms and then make a
molecule out of cold atoms.
Indeed, laser cooling and trapping of atoms is now a well-mastered technique,
used in many applications such as atomic clocks. Temperatures below 10−6 K are
“easily” reached. Among the very few attempts to generalize optical cooling to
molecules, one may cite the deflection of a molecular beam achieved by Herman
et al. in 1979, and the demonstration, by Djeu and Whitney (1981), of laser cooling
by continuous anti-Stokes scattering, introduced long ago by Kastler (1950) as
“luminorefrigeration.” As discussed by Bahns et al. (1996), the main obstacle in
this research direction is that a molecule is essentially a multilevel system whereas
direct optical cooling requires a closed two-level system to recycle population.
Cooling of an atom takes place via a large number of optical pumping cycles in
which absorption and spontaneous emission transfer population from a lower level
to an upper one and back, as indicated in Fig. 1 for the case of Cs(6s) → Cs(6p3/2)
FORMATION OF ULTRACOLD MOLECULES 55
FIG. 1. (Left) Optical pumping scheme for cooling an atomic cesium sample, using the 6s →
6 p3/2 D2 resonance line. In order to cool the atom, many cycles of absorption–spontaneous emission
must be performed between the set of two (6s, f = 4) and (6p3/2, f = 5) hyperfine sublevels, varying
the red detuning of the laser. As the spontaneous emission may also be populating the (6s, f = 3) sub-
level, a repumping laser is used to put the atoms back in the upper state. (Right) The difficulty of
generalizing a direct cooling scheme to a molecule comes from the very large number of levels.
transition. The presence of neighboring levels, for instance, hyperfine sublevels,

is a nuisance, as absorption can populate another level, causing population loss by
spontaneous emission to the wrong lower level. When the number of neighboring
levels is limited, as in the chosen example, this inconvenience is corrected for by use
of repumping lasers. However, the number of repumping lasers has to stay within
reasonable limits, so the procedure cannot be easily generalized to a multilevel
system such as a molecule (Bahns et al., 1996, 2000).
An alternative is to take advantage of the efficiency of laser cooling techniques
for atoms: the idea is to cool free atoms in a first step, and then create ultracold
molecules by photoassociation of atomic pairs. This reaction in the cold regime
was first proposed by Thorsheim et al. (1987), and is indeed efficient to form cold
molecules in excited electronic states. Photoassociation has already been demon-
strated for all alkali atoms from Li to Cs: the first studies were for Na2 (Lett et al.,
1993), followed by Rb2 (Miller et al., 1993), Li2 (Abraham et al., 1995), K2 (Wang
et al., 1996), Cs2 (Fioretti et al., 1998), and NaK (Shaffer et al., 1999). Photo-
association has been further demonstrated for hydrogen (Mosk et al., 1999), he-
lium (Herschbach et al., 2000), and very recently for calcium (Zinner et al., 2000),
opening the way to photoassociation of other alkaline earth atoms. However, the
lifetime of a photoassociated molecule is short, and after a few nanoseconds it
decays by spontaneous emission, usually giving back a pair of atoms. The route
to long-lived ultracold molecules depends on the occurrence of particular sta-
bilization schemes in which spontaneous (or stimulated) emission also transfers
population toward bound levels of the ground singlet or triplet state. Theoreti-
cal models have been proposed by Band and Julienne (1995), and by Côté and
Dalgarno (1997, 1999). Other schemes have been found in experiments, and up
to now long-lived ultracold molecules have been observed for three alkali dimers
Cs2 (Fioretti et al., 1998, 1999; Comparat et al., 2000), K2 (Nikolov et al., 1999,
2000), and Rb2 (Gabbanini et al., 2000). Besides, Takekoshi et al. (1998, 1999)
have observed and trapped ultracold Cs2 molecules formed by a mechanism which
may not be photoassociation and could be three-body collisions. Very recently,
photoassociation experiments in a condensate have also been performed (Wynar
et al., 2000).
The field of cold molecules is experiencing very rapid evolution, both optical
and nonoptical cooling techniques being developed: the fact that it is possible to
write the present review chapter only one year after the previous one is a shin-
ing manifestation of this activity. We may say that spectacular progress has been
accomplished during the last three years in three main directions, two with non-
optical techniques and one with optical techniques.
At present (autumn 2000), the most efficient ways of cooling molecules to a
subkelvin temperatures below 1 K but still above 100 mK, with the perspective of
reaching a few millikelvin in a near future are as follow:
r The helium buffer gas cooling technique developed by the Harvard team
(Doyle et al., 1995; Friedrich et al., 1998, 1999; Weinstein et al., 1998),
together with magnetic trapping of molecules, as already described in the
review of Bahns et al. (2000). CaH molecules, produced via laser ablation of
solid CaH2, are cooled by cryogenic 3He down to a temperature of 400 mK,
and a typical number of 108 molecules are currently loaded in a magnetic
trap (deCarvalho et al., 1999).
r The new Stark decelerator technique introduced by Gerard Meijer and his
group in Nijmegen. They have demonstrated that a beam of neutral dipolar
molecules (CO, ND3) can be decelerated with a time-varying electric field
(Bethlem et al., 1999), bringing them to temperatures down to 350 mK. The
ND3 molecules are then trapped in a traveling potential well (Bethlem et al.,
2000a,b), with a density of ∼106 cm−3 molecules in a volume of ∼0.25 cm−3,
so that the total number of trapped molecules is ∼400,000.
For the formation of ultracold molecules, i.e., at temperatures well below
1 mK, the photoassociation scheme in a sample of laser-cooled ultracold atoms is
presently the best route, yielding a few millions of molecules at a translational tem-
perature in the microkelvin range, which very recently were successfully trapped.
The present review chapter will focus on this subject. The photoassociation re-
action being a crucial step in the process, we shall first review the concept of
photoassociation in Section II, emphasizing the physical ideas behind it. The next
step is the formation of ground-state ultracold molecules by spontaneous emission,
which requires a favorable branching ratio between bound–bound and bound–free
transitions. This is a key point of the experiments, which will be described in
Section III. Then the new theoretical methods developed to interpret such experi-
ments will be described in Section IV. Finally, we shall report on the present status
of the comparison between theory and experiment in Section V, new possible

schemes and the perspectives in the field being discussed in Section VI.
II. Physical Concepts in Photoassociation

and Formation of Molecules
In photoassociation experiments, first suggested by Thorsheim et al. (1987), a pair
of ultracold ground-state atoms absorbs a photon slightly (∼ a fraction of cm−1 to
1000 cm−1) red-detuned relative to the resonance line. Considering alkali atoms
A(ns), and a laser red-detuned either to the D1 or the D2 resonance line, the reaction
reads
A(ns) + A(ns) + E c + h ω0i − δ L → A∗z ∗ ( u,g (ns + np 2 P1/2,3/2 ); v, J ).

(1)
In Eq. (1), we have considered excitation to an electronic potential curve, u,g in

Hund’s case c representation, hereafter labeled i. This curve is correlated to one
of the two asymptotes (ns + np1/2 ) and (ns + np3/2 ), and 2π ω0 is the frequency
of the corresponding resonance line. hδ L is the energy detuning of the laser and
E c ≃ k B T the kinetic energy of the pair of atoms, where we have introduced
the Boltzman constant k B . In the present review, we shall not consider coupling
between various excited potential curves, and we shall not include effects linked
to the hyperfine structure of the atoms, therefore assuming a detuning larger than
the hyperfine splitting both in the ground state and in the excited state.
From the conservation of energy and angular momentum, taking as the energy
origin the asymptotic energy E(ns + ns) of the ground-state potential, we may
write (Pillet et al., 1997) the resonant condition for population of a bound level
i
with energy E v,J in an electronically excited potential curve i as
i
E c + h(ω0 − δ L ) = E Doppler + E recoil + E v, J. (2)
In Eq. (2), Ec is the kinetic energy associated with the relative motion of the
two atoms, E Doppler = V · hk L is the energy associated with the first-order non-
relativistic effect, and E recoil = (hk L )2 /8μ is the recoil energy. We have introduced
the reduced mass μ of the two atoms, the velocity V of the center of mass, and the
wavevector kL of the laser.
In the following, considering the energy Di of the asymptote of the potential
curve i, we shall often use the binding energy
iv,J = D i − E i (v, J ), (3)

iv = D i − E i (v, J = 0) = E vi . (4)
The resonance condition can be written
hδ L − iv = E c − E Doppler − E recoil . (5)
In this chapter, we shall consider temperatures of ∼20 to 200 μK, so that E c ≈

1.4 × 10−5 to 1.4 × 10−4 cm−1 , while in the case of cesium and for a red-detuning
of 1 cm−1 = 29.9792458 GHz, the two other quantities are E Doppler ≃ 100 kHz,
E recoil ≃ 1 kHz.
As we shall see below, since the typical accuracy of the photoassociation ex-
periments is at present of the order of 150 MHz or 5 ×10−3 cm−1, it is justified to
neglect the right-hand side of Eq. (5) and write
hδ L ≈ iv . (6)
The resonance condition is then an approximate equality between the binding

energy of the vibrational level and the energy shift of the laser relative to the
resonance line; this is why in photoassociation spectroscopy it is common to use
the word “detuning” for the binding energy iv of a level. This free–bound transi-
tion takes place under conditions somewhat unusual with respect to conventional
molecular spectroscopy:
r Due to the ultracold conditions, the width of the statistical distribution of the
kinetic energy Ec is markedly reduced. For a gas of atoms at thermal equilib-
rium with a temperature T, this width is of the order of k B T . It corresponds
to ∼200 cm−1 at room temperature (300 K), yielding the well-known appear-
ance of diffuse spectra, but only to ∼2 MHz or 7 × 10−5 cm−1, at 100 μK
yielding well-resolved spectral lines, similar to bound–bound transitions. Be-
sides, only a few partial waves have to be considered in the description of
the initial continuum state, so the rotational structure is most often limited
to J < 7–10, in contrast with conventional molecular spectra where a much
larger number of rotational levels are usually observed. The interest of pho-
toassociation as a new spectroscopic technique was reviewed by Stwalley
and Wang (1999).
r In the final state, the vibrational motion extends up to very large internuclear
distances, typically a few tens or even 100 a0. At small detunings, detailed
spectroscopic information has been obtained in the last decade on long-range
molecular states correlated to the (ns + np) dissociation limit of the alkali
dimers, as described in the reviews by Stwalley and Wang (1999) and by
Weiner et al. (1999).
In the present chapter, we shall not describe spectroscopy results but, rather,
concentrate on the determination of the number of molecules formed in a
given experiment. The first question is how many excited molecules are formed
per second by photoassociation. The absorption coefficient, and therefore the

reaction rate, is expected to depend on the electric dipole matrix element be-
tween the wavefunctions in the initial and final states. We shall analyze the final
state and the structure of the excited potential curves with asymptotic −C3i /R 3 be-
havior (see Section II.A). Close to the dissociation limit, vibrational levels in such
potentials can be described with the Le Roy–Bernstein analytical law, presented in
Section II.B, from which we shall deduce scaling laws for various matrix elements.
The theoretical predictions for the photoassociation rate at low laser intensity will
be summarized in Section II.C, and shown to be indeed proportional to the square
of the integral between the wavefunctions in the initial and final states. This inte-
gral will then be estimated by use of the reflection principle (Section II.D): within
a Franck–Condon picture, reaction (1) can be understood as a vertical transition
at a large distance RL = (C3i /hδ L )1/3 from the continuum of the ground state to a
vibrational level of the excited electronic state. In Section II.E, we shall discuss the
variation of the amplitude of the ground-state wavefunction at R = R L , and the
presence of nodes. The efficiency of the photoassociation process will be estimated
(Section II.F) as a function of physical parameters such as the detuning and the
temperature. Then, in Section II.G we shall introduce the problem of controlling
the decay of this short-lived excited molecule into a bound vibrational level of the
ground state rather than into a pair of atoms.
A. LONG-RANGE ALKALI DIMER MOLECULES: A PAIR OF ATOMS

The alkali atoms have been a benchmark for studies of thermal energy collisions,
line broadening, and laser cooling; for alkali dimers, besides the photoassocia-
tion spectroscopy experiments quoted above, there exist much data obtained by
conventional molecular spectroscopy experiments. Therefore the features of the
long-range potentials correlated to the first asymptote have been investigated by
many authors, starting with Dashevskaya et al. (1969), then Movre and Pichler
(1977) as well as Bussery and Aubert-Frécon (1985a, 1985b), Julienne and Vigué
(1991), Marinescu and Dalgarno (1995, 1996), Aubert-Frécon et al. (1998), all of
them quoted in the review of Stwalley and Wang (1999). A few ab initio calcula-
tions exist at shorter distances (Foucrault et al., 1992; Spies, 1989; Magnier et al.,
1993; Magnier and Millié, 1996).
One important feature is the existence of long-range wells: we shall focus on
their particular behavior for the heavy alkali dimers. The concept of a “pure long-
range molecule,” in which the motion of the two atoms is confined to the region
of large distances where only electrostatic interactions are active, any chemical
interaction being negligible, was introduced by Stwalley et al. in 1978.
As an example, we reproduce in Fig. 2 the potential curves for interaction
between two ground-state cesium atoms or, alternatively, a pair of Cs(6s) +
Cs (6 p1/2,3/2 ) atoms. The hyperfine structure being neglected, such figures were
FIG. 2. Potentials of the Cs2 molecule correlated to the (6s + 6s) and (6s + 6 p 2P1/2,3/2 ) asymp-
totes. Hund’s case c representation is used for the excited potentials, the 2g and 2u curves being not
represented.
obtained by matching the ab initio potential curves from Meyer’s group (Spies,
1989) with the long-range asymptotic expansion of Marinescu and Dalgarno
(1995). Two facts are striking:
r Due to the dipole–dipole R−3 asymptotic interaction, the excited potential
curves extend at much larger range than the ground-state curves with asymp-
totic R−6 behavior. This feature is general for homonuclear dimers, but would
not exist for heteronuclear dimers where the excited curves also display R−6
behavior at long range. Let us note for homonuclear dimers the exception of
i
the 0−g ( p1/2 ) curve for which the C 3 coefficient is zero, so that the asymptote
−6
varies as R .
r Two curves, of 0− g and 1u symmetry, both correlated to the (6s + 6 p3/2 )
asymptote, display a double-well structure, with ∼R −3 behavior in the entire
region of the outer well so that the external side of the barrier has a gentle
slope. Such wells exist for all homonuclear dimers, but we shall see that,
for the heavier ones (Rb2, Cs2), they are located at intermediate internuclear
distances and play a key role in the formation of ultracold molecules. All
symmetries have an exponential short-range repulsive potential for R ≤ 7a0 ,
with a steep slope.
What is the explanation for the existence of those long-range wells? In the
absence of fine-structure coupling, the long-range interaction between a ground-
state atom A (ns) and an atom in the first excited state A (np) is dominated by
the dipole–dipole interaction −C3 /R 3 . As only the excited atom has a permanent
dipole, this interaction corresponds to transfer of excitation between the two distant
atoms. The C3 coefficient is inversely proportional to the radiative lifetime of the
(np) excited atomic level and can be obtained from the square of the dipole moment
matrix element:
C3 = |ns|dz |np |2 . (7)
This definition is chosen by Julienne and Vigué (1991), and by Marinescu and
Dalgarno (1995). In the work of Aubert-Frécon et al. (1998), all coefficients are
related to a quantity M 2 = |ns|d|np |2 , which is 3 × C3. The C3 values computed
by Marinescu and Dalgarno (1995) are 5.503 for Li, 6.13 for Na, 8.665 for K, 9.202
for Rb, and 10.47 for Cs. In numerical examples, we note that for the heavy alkalis
C3 is of the order of 10 a.u. In such a model all exchange interaction is assumed
to be negligible, justifying the denomination “a pair of atoms.” Dashevskaya et al.
(1969) have classified the electronic states by their symmetry relative to exchange
of excitation between the two atoms as
ν = (−1) S w, (8)
where S = 0, 1 for singlet and triplet states, respectively, while w = 1, −1 for

g and u states, respectively. The long-range dipole–dipole interaction depends on
the orientation of the dipoles relative to each other and to the molecular axis, and
is found to be
C3
+2 for symmetric 1 g+ , 3 u+ states, (→ ←)
R3
C3
−2 3 for antisymmetric 1 u+ , 3 g+ states, (→ →)
R
C3
− 3 for symmetric 1 g , 3 u states, (↑ ↓)
R
C3
+ 3 for antisymmetric 1 u , 3 g states, (↑ ↑).
R
When the fine-structure splitting is introduced, in the Hund’s case c represen-

tation, the adiabatic g,u states for a given internuclear distance R at long and
intermediate range are obtained by diagonalization of the fine-structure effective

operator in the subspace of states with the same quantum numbers w and . (As
already stated, we restrict ourselves to distances such that the nuclear spins can be
considered as uncoupled.) Movre and Pichler (1977) have given the secular equa-
tions for all symmetries. This leads, close to the dissociation limit, to asymptotic
coefficients that will hereafter be noted C3i , and are linked by some angular factor
to the C3 coefficient defined above for interaction between an (ns) and an (np)
atom. This angular factor can readily be obtained by considering the asymptotic
expansion of a given i = g,u state on various Hund’s case a states for which
the long-range coefficient is C3 , −C3 , 2C3 , −2C3 according to the symmetry as
described in Eq. (8).
As an example, we shall consider the two 0− −
g ( p3/2 ) and 0g ( p1/2 ) states ob-
tained by diagonalization within the subspace of the (1) g and (1)3 g states.
3 +
For heavy alkali dimers due to the large value of the fine-structure splitting
(E at ( p3/2 )-E at ( p1/2 ) = 237.60 cm−1 for Rb, 554.11 cm−1 for Cs), it is more
convenient to use Hund’s case c representation at all internuclear distances; for
all dimers, the latter representation must be used at large distances to obtain the
correct dissociation limits (ns + np 2 p1/2 ) and (ns + np 2 p3/2 ). At any distance R,
the two 0− g potential curves are obtained (Marinescu and Dalgarno, 1996; Aubert-
Frécon et al., 1998; Fioretti et al., 1999) by diagonalization of the potential matrix
√
E at ( p3/2 ) + 32 V (R) + 13 V (R) 2
" #
3
(V (R) − V (R))
V(R) = √ ,
2
3
(V (R) − V (R)) E at ( p1/2 ) + 13 V (R) + 32 V (R)
(9)
where V (R) and V(R) are the potential curves of the (1) 3 g+ and (1) 3 g
electronic states, displayed in Fig. 3 for Cs2 and both correlated to the (ns + np)
dissociation limit. In Hund’s case c representation, the correct dissociation limits
are obtained at energies E at ( p3/2 ) and E at ( p1/2 ). The coupling between the two
atomic fine-structure levels at finite distance, due to the anisotropy of the mole-
cular potentials, is proportional to the splitting [V (R) − V (R)]. The restriction
to the subspace of states correlated to the (ns + np) asymptote is no longer valid
at small distances.
In the long-range region, where the dipole–dipole interaction dominates, the
V (R) and V(R) curves display their asymptotic −2(C3 /R 3 ) and (C3 /R 3 ) behavior
(see above) and the potential matrix can be approximated by
" √ #
C3 2C3
E at ( p3/2 ) − R3
− R3
V(R) ≈ √ . (10)
2C3
− R3
E at ( p1/2 )
FIG. 3. V (R) and V (R) potential curves (see text) for the Hund’s case a 3 g+ and 3 g states of
the Cs2 molecule. The open circles correspond to ab initio calculations by Spies (1989). Arrows
indicate the distance where the matching with the asymptotic expansion is taking place, choosing
C3i , C6i , C8i coefficients from Marinescu and Dalgarno (1995). In the upper curve, direct matching
being not possible, an interpolation was performed in the region between the two arrows.
The two 0− g asymptotic curves are obtained by diagonalization of this matrix, and
can be considered as R −3 curves with slowly R-varying C3 coefficients. At very
large distances, the upper curve 0− with C3i = C3 ,
g ( p3/2 ) is asymptotically attractive√
while the lower curve is flat (C3 = 0). As the nondiagonal term 2(C3 /R 3 ) is
i
increasing when the atoms get closer,√ there is an avoided crossing at a distance Rmin
3
such that E at ( p3/2 ) − E at ( p1/2 ) ≈ ( 2C3 /Rmin ). This explains why the external
well is located at shorter and shorter distances when the fine-structure splitting
increases from the light alkali atoms (Li, Na) to the heavier ones (Rb, Cs). This
also explains why the repulsive part of this external well, located on the left of this
avoided crossing, is displaying typical R−3 behavior: for R ≤ Rmin , the repulsive
branch of the external well may be approximated by
√ C3
Uup (0−
g ( p3/2 )) ≈ E( p3/2 ) + ( 2 − 1) 3 . (11)
R
For Rb2 and Cs2 molecules, higher-order R−6 and R−8 multipole terms, as well
as asymptotic exchange terms, contribute to the repulsive branch (Aubert-Frécon
et al., 1998; Fioretti et al., 1999), yielding corrections of a few cm−1; nevertheless,
the physics of the long-range external wells is dominated by the dipole–dipole
interaction between a pair of separated atoms. This is why we may say that we
form “a pair of atoms” rather than a usual molecule; the same idea is expressed in
the concept of a “pure long-range molecule” by Stwalley et al. (1978).
This concept is even more adapted for the long-range well in the 1u ( p3/2 ) curve,
which is located at a distance so large that the hyperfine-structure coupling has to
be explicitly considered. The first observation of a 1u pure long-range molecule
was done by Wang et al. (1998) for K2 (De = 0.5394 cm−1 ) and the Cs2 state was
analyzed recently by Comparat et al. (2000). (see Section III.C)
In contrast, for all attractive curves differing from 0−g ( p3/2 ) and 1u ( p3/2 ) the
vibrational motion extends from large to small internuclear distances, with reflec-
tion on the steep inner barrier. As we shall see later, the time spent in the inner
region is comparatively very short.
B. THE LE ROY–BERNSTEIN LAW FOR BOUND LEVELS: ANALOGY

WITH RYDBERG LAW, SCALING LAWS
In this section, we shall neglect any rotation effect, assuming J = 0. Let us then
consider a vibrational level with energy E vi close to dissociation limit D i in a
typical i = g,u excited potential curve U i (R) with asymptotic behavior
C3i
U i (R) ≈ D i − . (12)
R3
In the present section we shall also neglect the dynamical coupling with other
channels i, so that the i label can safely be dropped to lighten the notation. A level
will be characterized by its binding energy iv = Di − E vi hereafter denoted v .
However, we shall keep the notation C3i for the long-range coefficient, as it differs
from C3 by an angular factor depending on the symmetry of the curve i.
The vibrational motion takes place both in the long-range region, where the mo-
tion in the asymptotic potential is very slow, and in the short-range region where,
due to the attractive chemical potential, the motion is very quick. Besides, as the
binding energy is negligible compared to the short-range potential, the local de
Broglie wavelength at small internuclear distances is fairly independent of v , so
that this rapid motion is similar for a wide range of levels. As an example, we dis-
play in Fig. 4 two vibrational wavefunctions, for the potential Cs2 (1g , 6s + 6 p3/2 ),
corresponding to binding energies v = 1 and 10 cm−1. Such functions have
been computed according to the numerical method described in Section IV.A. It
is clear that the probability density is concentrated at the outer turning point Rv ,
where the potential can safely be approximated by its asymptotic behavior given in
Eq. (12). This corresponds to the physical idea of two atoms standing still most of
the time at distance R = Rv . A look at the behavior of the two wavefunctions at
small distances, displayed in Fig. 5, confirms that their nodal structure is similar,
the only dependence on the binding energy appearing in their amplitude. This can
FIG. 4. Vibrational wavefunctions in the 1g (6s + 6 p3/2 ) potential of the Cs2 molecule, for two
values of the binding energy, v = 1 and 10 cm−1.
be justified as stated above by the negligible value of v compared to the short-

range potential, or alternatively, by the short time spent by the system in the inner
region, leading through the uncertainty relation to a weak energy dependence.
The physical interpretation is thus very similar to the description of the motion
of a Rydberg electron in the potential of an atomic ion, where the effect of the
FIG. 5. Same as Fig. 4, concentrating on the short-range behavior of the wavefunctions for the
two levels with binding energy v = 1 and 10 cm−1. The nodal structure of the two wavefunctions
5/12
is similar, the energy dependence being concentrated on the amplitude, which scales as v , and
therefore increases by ∼2.6 when the detuning is multiplied by 10.
short-range non-Coulomb core is characterized by a quantum defect η in the

Rydberg law,
Ryd
E(n, l) = I − , (13)
[n − η(l)]2
linking the energy of a level to the principal quantum number n. In Eq. (13),
I is the energy of the ionization threshold, and Ryd is the Rydberg constant.
The quantum defect is weakly energy-dependent, but depends on the azimuthal
quantum number l.
Considering motion in a potential with power-law asymptotic behavior R−n,
with n ≥ 3, such as in Eq. (12), a generalization of Eq. (13) was derived in 1970
by Le Roy and Bernstein. This law was found to be such an important tool in the
understanding of photoassociation spectroscopy that it seems worthwhile to recall
the main steps of its demonstration (Le Roy and Bernstein, 1970; Stwalley, 1973).
The scaling laws are discussed in the review of Vigué (1982), and the link with
quantum defect theory analyzed by Ostrovsky et al. (2001).
Assuming the semiclassical picture to be valid, the vibrational wave function for
a bound level with energy E v in the single potential U(R) may be written (Landau
and Lifshitz, 1977) as
R
1 1
v (R) = Nv sin k(R ′ ) d R ′ + π , (14)
k(R) v
Rin 4
with the Bohr–Sommerfeld quantization condition on the phase,

Rv
′ ′ 1
βBS (E v ) ≡ k(R ) d R = v + π. (15)
v
Rin 2
v
In Eqs. (14) and (15), Rin and Rv are, respectively, the inner and outer turning points
of the vibrational motion. For a diatomic system with reduced mass μ, we have
defined the local wavenumber kv (R) and local de Broglie wavelength λv (R) as
1 2π
kv (R) = 2μ[E v − U (R)] = . (16)
h λv (R)
Therefore Eq. (14) is not valid close to a classical turning point, where
v
kv (Rin ) = kv (Rv ) = 0. We shall give below a more general expression that is also
valid close to a turning point. The normalization factor is linked to the classical
period of motion (Landau and Lifshitz 1977) through
Rv
d R′

hTvib
2[Nv ]−2 = ′
= , (17)
v
Rin kv (R ) 2μ
where Tvib is defined as the time necessary to go from the outer turning point to
the inner one and back. The physical explanation given by Lefebvre-Brion and
Field (1986) is that the probability amplitude in Eq. (14) is linked to the classical
probability density, which is inversely proportional to the velocity times the period.
From differentiation of the Bohr–Sommerfeld condition (15) over the quantum
number, the same integral gives the inverse of spacing between the levels,
Rv
d R′

π
dv
h 2
= . (18)
μ d E E=Ev

v
Rin k(R ′ )
From this an important relation between the normalization factor and the level
spacing,
π h2 2

d E v
= N , (19)
dv E=Ev

2μ v
can be derived, which we shall use later to determine a scaling law.

For highly excited vibrational levels, the integral in Eqs. (15) and (18) is de-
termined mostly by the outer domain of distances R, where the potential can be
approximated by the general asymptotic formula U (R) = D − Cn /R n ; this al-
v
lowed Le Roy and Bernstein (1970) to evaluate it in the limit Rin /Rv → 0 with an
analytical formula,
Rv
d R′ 2n π h 2

′
= (D − E v )−(n+2)/2n , (20)
v
Rin k(R ) n − 2 2μHn

2π (n − 2) Ŵ(1 + 1/n) −1/n
Hn = h Cn . (21)
μ 2Ŵ(1/2 + 1/n)
By integration of Eq. (20), the Le Roy–Bernstein law is obtained, generalizing the

Rydberg law to any potential in R −n , n = 2,
E v = D − [Hn (v D − v)]2n/(n−2) , (22)
where the noninteger number v D is the accumulated phase at the dissociation

threshold divided by π . We may define an effective quantum number,
veff = v D − v, (23)
such that the levels are now numbered starting from the dissociation limit. In the
spirit of the quantum defect theory (Seaton, 1983; Friedrich, 1998) it can be shown,
as illustrated in Fig. 5, that for levels close to the dissociation limit, the wavefunc-
tions exhibit the same nodal structure at short distances, the energy dependence
being concentrated in the normalization factor. Therefore any change in the short-
range potential results into a modification of the phase in Eq. (15) equivalent for
all such levels: whereas the total phase πv D is modified, the difference π (v D − v)
remains a constant.
In the case of photoassociation into an excited curve with long-range R −3
behavior as described by Eq. (12), the law giving the dissociation energy reads,
using Ŵ(4/3)/ Ŵ(5/6) = 0.790903 = γ3 and
$
π i −1/3 −1/3
H3 = 0.790903h C3 = A3 C3i , (24)
2μ
and considering the cesium dimer, with C3i ≈ 10 a.u. and atomic mass M A = 133
as a reference,
133 3 10 2

v = H36 (v D − v)6 = veff
6
× 1.17 10−12 cm−1 , (25)
MA C3i
where M A is the atomic mass so that μ = 1822.8885(M A /2). Close to the dis-
sociation limit there is a quasi-continuum of vibrational levels: for a detuning of
1 cm−1, the numbering of a Cs2 level from the dissociation limit is already ∼100!
This is why it is more convenient to analyze the present vibrational series by con-
sidering an energy scale rather than a numbering of levels. For a given dimer, the
vibrational progression in Eq. (25) is a signature of the excited electronic state i
populated by photoassociation, through the C3i coefficient. The vibrational spacing
rapidly increases as a function of detuning, according to

d E
dv
=− = +6(H3 )6 (veff )5 = −6H3 (v )5/6 . (26)
dv E=Ev

dv
We can give an estimate of the level spacing taking the cesium dimer as reference,

1/3
dv
−2 133 10
= −6.14 × 10 (v )5/6 cm−1 . (27)
dv
E=Ev M A C3i
In the case of Cs2, choosing C3i = 10.47 from Marinescu and Dalgarno (1995),
the level spacing is increasing from ∼0.06 to ∼0.4 cm−1 when the detuning varies
from 1 to 10 cm−1. The classical outer turning point,
1/3
C3i

Rv = , (28)
v
then decreases from 131.9 to 61.5 a0, verifying a (v )−1/3 or (veff )−2 scaling law,
which, we shall see later, clearly manifested in the photoassociation spectra. The
scaling law verified by the normalization factor in Eq. (19), may be deduced from
the vibrational spacing,

2 −2 M A dv
−1
(Nv ) ([a0 ] ) = 0.3516 × (cm )

. (29)
133 dv E=E v
We can verify in Figs. 4 and 5 that the inner part of the wavefunctions is in-
deed scaled by a factor ∼2.6 when the detuning is varying from 1 to 10 cm−1,
corresponding to a change of (Nv )2 by a factor 105/6 ≈ 6.8. As this (Nv )2 normal-
ization factor scales any matrix element of a physical quantity corresponding to
short-range coupling, the (veff )5 or (v )5/6 scaling law is then equivalent to the
well-known ν −3 scaling law for Rydberg states (Seaton, 1983; Friedrich, 1998).
We should mention that in the case of an asymptotic R −5 potential, Gouédard
et al. (1989) gave for the first time experimental verification of the corresponding
scaling law near a dissociation limit of the Cl2 molecule.
The formation of ground-state molecules in low enough vibrational levels can
be considered as a short-range process, as it involves the overlap with short-
range vibrational wavefunctions in the ground state. Therefore, the corresponding
5/6
probability is expected to scale as v as a function of the binding energy, and
5/6
hence as δ L as a function of detuning. Thus, large detunings are expected to be
more favorable. In contrast, we shall see later that, the photoassociation process is
definitely more favorable at small detunings.
It is also interesting, within a time-dependent picture, to estimate the classical
vibrational period in the (ns + np) excited potential curve of a dimer A2,
1/3 $
C3i [a.u.]

MA
Tvib [ps] ≈ (v [cm−1 ])−5/6 × 534 ps. (30)
10 133
Assuming C3i values around 10 a.u., and for a 1-cm−1 detuning, the vibrational
period has a magnitude of about 500 ps. At even smaller detunings, this classi-
cal vibration period becomes of the order of magnitude of the radiation lifetime
(a few nanoseconds), so that a model neglecting spontaneous emission is no longer
valid, and time-dependent models must be implemented.
To be through, we should give a correct expression, thorough for the semi-
classical wavefunction in the vicinity of a turning point, where the approximation
given in Eq. (14) is no longer valid. We shall follow the derivation proposed by
Vigué (1982), Lefebvre-Brion and Field (1986), and Child (1991) and write the
uniform semiclassical (USC) wavefunction (Miller, 1968), valid for all values of
R, as
1/4
N Z (R)
vUSC (R) = √ Ai[−Z (R)], (31)
π k(R)2
where
% R we have introduced an Airy function of the phase Z (R) =
[ 32 R v k(R ′ ) d R ′ ]2/3 . The sine behavior as a function of the phase is reached
in
for Z > 1.5, where the uniform semiclassical function becomes equivalent to the
sine function
√ in Eq. (14). The factor multiplying the Airy function in Eq. (31) is
then N π(h[Rv ]2 )1/3 [6μC3i ]−1/6 , where Rv is the outer turning point linked to
the detuning through Eq. (28).
The semiclassical arguments developed in the present section, the choice for
the norm, and therefore the derivation of the scaling laws, are no longer valid
close to the dissociation limit, when the wavefunction has an important extension
in the nonclassical region and behaves as an evanescent wave. Corrections to the
semiclassical model have been proposed by many authors, and will be discussed in
Section IV.A.3. Such corrections concern very few levels close to the dissociation
limit. Bendes numerical calculations are available in all cases.
Scaling laws offer an invaluable tool in the interpretation and fitting of experi-
ments: the Le Roy–Bernstein formula has been widely used in order to identify
the excited electronic states or to fit C3i coefficients to the observed spectrum [see
the review of Stwalley and Wang (1999), and references therein]. Examples will
be given in Section III.C.
Generalization of the present derivation to two coupled channels has been pro-
posed recently by Kokoouline et al. (2000a) and Ostrovsky et al. (2001). The
energies of the levels for two coupled vibrational series can be plotted as Lu Fano
plots and fitted by three parameters, which are two generalized quantum defects
and a reduced coupling parameter.
C. THEORETICAL PREDICTIONS FOR THE PHOTOASSOCIATION RATES

The problem is to estimate how many excited molecules in a given ro-vibrational
level will be created in a photoassociation experiment. Many different groups have
been working on the theory of photoassociation, using both “time-independent”
formalism, with stationary wavefunctions (Napolitano et al., 1994; Pillet et al.,
1997; Côté and Dalgarno, 1998; Mackie and Javanainen, 1998), and “time-
dependent” formalism, with wavepacket propagation (Mackholm et al., 1994;
Vardi et al., 1997; Boesten et al., 1999; Vala et al., 2001; Vatasescu et al., 2001).
After some controversy, there is presently agreement between various approaches
for “time-independent” calculations. Up to now, “time-dependent” calculations
have focused on the presentation of new schemes rather than quantitative predic-
tion of the photoassociation rates, and will be described in Section VI. A precise
comparison between the predictions of “time-dependent” and “time- independent”
approaches is still lacking.
The quantum formulation of the transition between a bound level and a contin-
uum level of a molecule due to coupling by the electromagnetic field between two
electronic states was considered in the early days of quantum mechanics (Condon,
1928; Winans and Stueckelberg, 1928; James and Coolidge, 1939) using various
approximations. Doyle (1968) has given a detailed derivation of the absorption co-
efficient in a sample of hydrogen atoms: the rate is then defined as the probability
of absorption of a photon for an incident beam of one photon crossing unit area
per second and per unit frequency interval.
Napolitano et al. (1994) as well as Côté and Dalgarno (1998) use this definition
to compute the absorption rate coefficient for the inelastic process yielding a bound
molecule out of two ground-state 2S alkali atoms via absorption of a photon with
energy hω L = h(ω0 − δ L ) according to the reaction (1)
∞
2 πh 2
K (T, ω L , v) = (n at ) (2J + 1)|S(E, J ; i, v, J ; ω L )| (32)
k J =0
where n at is the atomic density and S(E, J ; i, v, J ; ω L ) is the free–bound transition

amplitude between the initial continuum level at energy E and a bound level v, J in
the excited potential curve i. Here, . . . . . . . . . is an average over the distribution
of initial velocities. If a Maxwellian distribution at temperature T is assumed,
expression (32) yields for the absorption coefficient
∞
n2 ∞

K (T, ω L , v) = (2J + 1) at × d E e−E/kT |S(E, J ; i, v, J ; ω L )|2 ,
J =0
h QT 0
(33)
where the translational partition function Q T = (2π μk B T / h 2 )3/2 has been intro-
duced. In most papers, the numerical results are given for the quantity K (T, ω L , v)
divided by the incident photon flux and by the square of the atomic density.
Various approximations have been discussed to simplify the calculations. Al-
though the S-matrix element has to be computed by solving coupled equations,
Napolitano et al. (1994) show that their close-coupling calculations for sodium, in
a temperature range from 0.01 to 10 mK, can be approximated by a very simple
resonant scattering expression involving the width of the bound level defined as a
function of the spontaneous emission rate and the stimulated emission rate back
to the ground state. Such expression is similar to a Breit–Wigner formula, and
Gardner et al. (1995) have generalized it taking account of the directional depen-
dence. Photoassociation can then be viewed as an optically induced Feshbach reso-
nance, and the richness of this point of view has been recently demonstrated by the
experiments of Fatemi et al. (2000), who measured such resonances in a sodium
sample. At low laser intensities, in the framework of a perturbative treatment, the
widths factors in the Breit–Wigner formula involve the electronic transition mo-
ment between the unperturbed initial and final states. In the present work we shall
not discuss the shape of the photoassociation lines, assuming that the resonance
condition(s) can be expressed by a δ function. By use of a Fermi-golden-rule type
of approximation, Côté and Dalgarno (1998) find
I
& '
2
S(E, J ; i, v, J ; ω L ) ∼ 8π 3
ψ E,J (R)|D(R)|φv,J
i
(R)
δ(E − hδ L + v ),
c
(34)
where we have introduced the coupling matrix element, proportional to the square
root of the laser intensity and to the molecular transition dipole D(R), between the
energy normalized ground-state scattering wavefunction ψ E,J (R) and the unper-
i
turbed vibrational function ψv,J (R). With an R-centroid approximation (Lefebvre-
Brion and Field, 1986), the electronic transition moment can be further approxi-
mated by
& i
' &
i '
ψ E,J (R)|D(R)|φv,J (R) ≈ D(Rv ) ψ E,J (R)
φv,J (R) , (35)
i.e., the product of D(Rv ), value of the dipole moment at the outer classical turning
point of the vibrational motion, and the overlap integral, the square root of the
Franck–Condon factor. Therefore, for a laser tuned at resonance, the photoas-
sociation rate is controlled by the Franck–Condon factor between the radial wave
functions in the initial continuum level and the final bound level.
The approach developed by Pillet et al. (1997) uses an atomic physics point
of view and a density-matrix description of a collection of N atoms, interacting
by two-body interaction, in the presence of laser light at time t. The density ma-
trix at time t = 0 is computed for an assembly of atoms when the laser is off,
assuming a Maxwell–Boltzmann distribution at temperature T. The coupling ma-
trix element is described by a δ function approximation in the spirit of Eq. (34),
considering that the rotational quantum number may vary between the initial and
final states. The time evolution equation of the density matrix, in interaction rep-
resentation, is solved in a perturbative approach. For a low intensity I of the pho-
toassociation laser tuned at resonance, the photoassociation rate RPA (in s−1),
defined as the number of photoassociated molecules, formed in an individual level
(v, J), divided by the total number Nat = n at V of atoms in a trap of volume V , is
found as
3/2
3 h
′
RPA = A(J, J , ǫ P A ) n at λ3th
2π 2

−Er '
2
K 2
ψ E,J (R)
φv,J

&
i
× exp ′ (R)

, (36)
kB T
where A(J, J ′ ), ǫPA )is an angular factor depending on J, J ′ and on the laser
polarization. λth = h 1/(3μk B T ) is the thermal de Broglie wavelength, and 2K
is the atomic Rabi frequency. The latter is a function of the laser intensity I,
Ŵ I
K2 = , (37)
8 I0
where we have defined a saturation intensity I0 = π hcŴ/(3λ3PA ) from the natural

width of the atomic np level, Ŵ/2π . In the case of cesium, Ŵ/2π = 5.22 MHz,
so the saturation intensity has a value of 1.1 mW/cm2 at the PA laser wavelength
λPA = 2πc/ω0 . The rate RPA in Eq. (36), divided by n at and by the incident photon
flux φ = I /hωPA yields a rate κ per photon and per unit density (its dimension is a
length to the fifth power), which is similar to the expressions given by most other
authors (Napolitano et al., 1994; Julienne, 1996; Côté and Dalgarno, 1998). The
quasi-continuum approach developed by Javanainen and Mackie (1998), Mackie
and Javanainen (1999) avoids the use of delta functions and yields a similar result.
D. THE REFLECTION PRINCIPLE FOR CALCULATION

OF FRANCK–CONDON FACTORS
The estimation of the photoassociation probability thus relies on knowledge of the
Franck–Condon factor,
'
2
F(E ∼

&
i
= k B T, J ; i, v, J ′ ) =
ψ E,J (R)
φv,J ′ (R)

. (38)
In the following we shall consider J = J ′ and drop the angular factor A(J, J ′ , ǫPA ).
Optical transitions between a bound and a continuum level of a molecule were
studied in the early days of quantum mechanics and viewed as vertical transi-
tions, at a given internuclear distance, between two electronic states (Franck,
1925; Condon, 1926, 1928). In his review on the Franck–Condon principle in
bound–free transitions, Tellinghuisen (1985) gives an extensive discussion of the
various approaches to the problem, through classical, semiclassical, or quantum
methods.
In a semiclassical approach, the overlap integral can be estimated within a
stationary phase approximation. Such a procedure, derived by Jablonski (1945),
has been adapted to the photoassociation problem by Julienne (1996), and further
discussed by Wang and Stwalley (1998) and Boisseau et al. (2000b). The main
idea of this derivation is to write the product of two oscillating functions as the
sum of two functions, and to neglect the high-frequency component, in order
to keep only the low-frequency term. The approach is valid provided the two
wavefunctions oscillate at frequencies that are not too different, so that the sum of
the frequencies is indeed much larger than their difference. When the continuum
function is oscillating much more slowly, i.e., at very low temperatures, Pillet et al.
i
(1997) consider for the vibrational function in the excited potential φv,J (R) an Airy
function behavior in the vicinity of the outer classical turning point. We shall see
later that the same formula is obtained by the two methods. The Franck–Condon
factor obtained with the stationary-phase method is
i
d E v,J 1
F(E, J ; i, v, J ) ≈ i
i
|ψ E,J (Rv,J )|2 , (39)
dv D Rv,J
i i
proportional to the value ψ E,J (Rv,J ) of the continuum wavefunction at R = Rv,J ,
in accord with the image of a vertical transition where the vibrational motion is
stopped. We shall use this formula for J = 0, and drop the J index hereafter.
The level spacing is introduced because of the normalization factor in the
vibrational wavefunction [see Eq. (19)], while, as in the Landau–Zener formula,
the quantity D(Rvi ) is the difference between the slopes of the two potential curves
at R = Rvi ,
d
D Rvi = [U i (R) − Vg (R)]| R=Rvi ,

(40)
dR
optimal conditions corresponding to a minimum of this quantity. The Condon

point Rvi has been defined in Eq. (28). Vg (R) is the ground-state potential, which
g
in the asymptotic region behaves as ∼−C6 /R 6 and therefore exhibits a much
i
weaker slope than the excited potential U (R). We may write
i 4
1 R 1 1/3
≈ v i ≈ C3i (v )−4/3 . (41)
D Rvi 3C3 3
A look at Eq. (39) shows that, as far as the choice of the final bound level is
concerned, the efficiency of the photoassociation reaction involves the competition
of two effects:
r In order to optimize the difference of slopes, which scales as (v )−4/3 , or
(Rvi )4 , the photoassociation process should take place at small detunings,
corresponding to a Condon point Rvi located at large distances, where the
two curves have similar slopes.
r In contrast, the normalization factor for the vibrational level v, related to the
level spacing d E vi /dv, scales as (v )5/6 , or (Rvi )−5/2 , so that if we keep the
constraint of populating individual vibrational levels, this factor decreases

when photoassociation is taking place at large distances Rvi .
The competition of these two effects gives a (v )−1/2 dependence on the detuning.
Of course, the Franck–Condon factor also depends on the behavior of the ground-
state wavefunction, which, as discussed in the next section, favors excitation at
small detunings, or vertical transition at a large distance Rvi .
E. THE NODAL STRUCTURE OF THE ZERO-ENERGY

SCATTERING WAVEFUNCTION
From Eq. (39), we see that the Franck-Condon factor reflects the nodal structure of
the ground-state continuum wavefunction, in the sense that the s-wave contribution
to the photoassociation signal will be zero when the Condon distance Rvi is such
that ψ E (Rvi ) = 0. We display, in Figs. 6 and 7, an example of the wavefunctions
ψ E (R) describing s-wave scattering of two ground-state cesium atoms, interacting
through the ground X1 g+ or lower-triplet a 3 u+ potential, for a collision energy of
FIG. 6. Energy-normalized wavefunction E (R) for scattering of two ground-state cesium atoms
in the X1 g+ potential at a temperature of 140 μK. The computed wavefunction corresponds to a
ground-state scattering length a S = −33a0 . The two arrows indicate the distances Rvi = 131.9 a0 and
Rvi = 61.5 a0 corresponding to photoassociation at detuning v = 1 and 10 cm−1, respectively, for an
excited potential varying asymptotically as −C3i /R 3 , with C3i = 10.47 a.u.
FIG. 7. Energy-normalized wavefunction E (R) for scattering of two ground-state cesium atoms
in the a 3 u+ potential at a temperature of 140 μK. The computed wavefunction corresponds to a
negative scattering length aT = −530a0. The two arrows indicate the distances Rvi = 131.9a0 and
Rvi = 61.5a0 corresponding to photoassociation at detuning v = 1 and 10 cm−1, respectively, for an
excited potential varying asymptotically as −C3i /R 3 , with C3i = 10.47 a.u.
140 μK. The scattering lengths were chosen a S = −33a0 from Dion et al. (2001)
and aT = 530a0 from Drag et al. (2000b). The main feature is the existence of two
very different regions:
r At long range, when the ground-state potential is negligible, the wavefunction
has sine behavior, with anamplitude inversely proportional to the square root
of the wavenumber k = (2μE/h 2 ) (or to E 1/4 ) and a phase depending on
the scattering length a S,T ,
1/4
2μ
E (R) = sin[k(R − a)]. (42)
π h2 E
2
It is expected that the photoassociation probability is enhanced at low col-

lision energies as E −1/2 . If the scattering length is positive, a minimum
in the photoassociation signal is observed for a Condon point located at
distance a.
r At very low temperature, the Wigner threshold law behavior is reached, and
in the asymptotic region E (R) is approximated by linear behavior as a
function of (R − a) (Côté et al., 1995; Julienne, 1996; Wang and Stwalley,

1998),
$
2μ R
0 (R) ≈ 2
sin[δ0 (k)] − 1 + O(k 2 ) (43)
πh k a
$
2μk a
≈ 2
R 1− , (44)
πh R
where we have introduced the zero-energy s-wave phaseshift δ (k).

0
r At shorter distance, due to the presence of the ground-state potential and
to quantum reflection effects, the amplitude of the wavefunction E (R) is
dramatically decreased, leading to a loss of more than one order of magni-
tude in the photoassociation efficiency, and to the existence of minima in
the photoassociation signal reflecting the nodal structure in the ground-state
wavefunction. Such minima, predicted by Côté and Dalgarno, were observed
for the first time by Hulet’s group (Côté et al., 1995) and then in many
photoassociation experiments (Julienne, 1996), leading to accurate determi-
nation of the scattering length (Drag et al., 2000b).
The two conclusions that we may draw are the following:

r First, the bottleneck in the photoassociation probability lies in the amplitude
of the ground-state wavefunction, which depends on the scattering length
and increases at low collision energies. It is more efficient to implement pho-
toassociation in the long-distance region where the continuum wavefunction
has reached asymptotic behavior. For instance, in the example of cesium
photoassociation chosen in the present review, we may see in Figs. 6 and 7
that by increasing the detuning from 1 to 10 cm−1, Rvi is decreasing from
131 gao to 61.5 ao and the amplitude of E (Rvi ) decreases by one order
of magnitude, and hence the photoassociation probability decreases by two
orders of magnitude.
r Second, accurate calculations of the overlap integrals should be performed,
the analytical formulas being helpful only to estimate the orders of magni-
tude in the low-detuning region, for temperatures low enough that the linear
approximation of the wavefunction in Eq. (44) is valid. We shall describe
accurate numerical methods in Section IV.A.
F. A SIMPLE THEORY OF PHOTOASSOCIATION

The perturbative approach adapted to the weak laser-field regime, developed by
Pillet et al. (1997), may be used for the calculation of photoassociation rates as
long as the saturation regime is not reached. We will assume E = k B T in the

following. In a region where the ground-state potential is negligible, and at very
low energies, the wavefunction E (R) can be approximated by a linear variation,
so that the Franck–Condon factor may be written from Eqs. (39) and (12) as
a 2

4γ3 2/3 −7/6
F(E, J = 0; i, v, J = 0) ∼ √ 2 C3i v μE 1/2 1 − , (45)
πh Rv
which is equivalent to the formula given by Wang and Stwalley (1998), in a paper
where they compare homonuclear and heteronuclear systems. From Eqs. (19), (39)
−7/6
and (41) one may predict for the overlap integral a dependence in v , since
the linear increase of the ground-state wavefunction as a function of distance R
−2/3
is introducing an extra v factor. The E1/2 energy dependence is linked to the
linear approximation of the wavefunction, and would change to a E −1/2 depen-
dence if the asymptotic sine behavior of Eq. (42) was considered for the initial
continuum wavefunction. Using Airy function representation for the excited-state
wave function, as in the work of Pillet et al. (1997), the same formula is obtained
for the estimation of the Franck–Condon overlap by analytical integration of the
product of an Airy function and a sine function. Care must be taken to use the
normalization of the vibrational wavefunction defined in Eq. (31), which is a fac-
tor 1.29 larger than the norm used by Pillet et al. (1997) from linearization of the
potential. Numerical tests of the validity of the estimation in Eq. (39) are presented
by Pillet et al. (1997) and by Boisseau et al. (2000b). In the latter case, the va-
lidity range of the reflection approximation is extensively discussed, and overlap
integrals for p and d waves are given as well as for s waves. The conclusion is
that the reflection approximation is valid in a large domain of physical situations,
justifying the present discussion.
The conclusion of the paragraph is that photoassociation is much more efficient
at large internuclear distances or at small detunings, creating excited molecules
in loosely bound vibrational levels, which most of the time look like a pair of
two atoms at large distance Rvi . This molecule, being in an excited electronic
state, decays by spontaneous emission, the lifetime being of the order of a few
nanoseconds.
G. HOW TO MAKE A MOLECULE IN THE GROUND STATE

OR LOWER TRIPLET STATE
The next problem is how to use those short-lived photoassociated molecules to cre-
ate long-lived molecules in a bound level v0 of the ground state or lower triplet state.
The spontaneous emission process is most likely to take place through vertical tran-
sition at a large distance Rvi . As the ground potential curves, with R−6 asymptotic
behavior, display a much narrower well (see Fig. 2), the vibrational motion does
not extend that far, so spontaneous decay populates mainly continuum levels, and
eventually a few of the uppermost bound levels. Once a photoassociated molecule
is formed, it usually decays back into a pair of free ground-state atoms,
A2 [ u,g (ns + np1/2,3/2 ); v, J ], → A(ns) + A(ns) + E c′ + h ω0i − δ2 .

(46)
The gain in relative kinetic energy [E c′ − E c = h(δ2 − δ L ) > 0] generally allows

both atoms to escape from the trap, so most photoassociation experiments have
indeed been analyzed through trap-loss measurements.
In order to make ground-state molecules, several schemes have been proposed
by theoretical goups. Côté and Dalgarno (1997, 1999) discuss how spontaneous
decay of the photoassociated molecule populates the uppermost vibrational levels
of the ground state.
A2 [ u,g (ns + np1/2,3/2 ); v, J ], → A2 [1,3 g,u

+
(ns + ns; v0 , J )] + h(ω0 + δ3 ).
(47)
Their study for one excited potential curve of Li2 confirms that the branching ratio
between reactions (47) and (46) is very low, leading to a small population transfer
into the last least-bound level of the ground X 1 g+ state, which is hardly a bound
molecule. A similar conclusion has been obtained recently in the Cs2 case by Dion
et al. (2001). Therefore, Côté and Dalgarno (1997, 1999) propose implementing
a Franck–Condon pumping scheme in order to transfer the population to lower
vibrational levels. They consider sequences with absorption of laser light to a
lower and lower vibrational level v1 < v of the A1 u+ or 13 g+ excited potential
curves, followed by spontaneous emission. Detailed calculations of the absorption
and spontaneous emission probabilities show that a large part of the population is
lost at each step by spontaneous emission to the continuum [reaction (46)], and
the authors conclude that the scheme is not efficient unless induced emission is
considered instead of reaction (47).
Another scheme was proposed by Band and Julienne (1995), who suggested a
two-photon sequence in which the photoassociation step is followed by absorption
to a Rydberg state with good Franck–Condon overlap with low vibrational levels
in the ground state. This scheme was implemented for K2 in the experiments of
Nikolov et al. (1999), and we shall see discuss its efficiency later.
Long-lived ultracold molecules were first observed in the cesium photoassocia-
tion experiment of Fioretti et al. (1998), owing to a formation scheme that was not
originally considered by the theoreticians, and is clearly linked to the particular
features of long-range wells in heavy alkali dimers. Indeed, we shall show that an
efficient mechanism has been found in situations where the vibrational motion of
the photoassociated molecule is gradually slowed down at intermediate distances,
leaving the possibility for vertical transition into a bound level of the ground
(or lower-triplet)-state potential curve.
III. Ultracold Molecule Experiments
A. INTRODUCTION
In this section, we present the experimental results for the formation of ultra-
cold molecules via photoassociation of laser-cooled atoms in a magnetooptical
trap device (MOT). We give a detailed description of the experimental scheme in
Orsay that led to the first observation of ultracold molecules in cesium (Fioretti
et al., 1998). Compared to the photoassociation spectroscopy experiments, the
cold molecule experiments differ by the direct detection of the formed molecules,
based on a photoionization process coupled with a time-of-flight mass spectrom-
eter. Most of the time, the states studied in the photoassociation experiments do
not give formation of cold molecules. This is due to the very weak branching ratio
in favor of bound–bound transitions from the excited photoassociated molecules,
which generally dissociate after spontaneous emission of one photon. We analyze
carefully the mechanisms of formation of cold molecules in the case of cesium
atoms, which are particularly efficient, making the cesium dimer a quite ideal
case. We expose the precise measurements for the translational temperature in the
molecular sample and for the rate of cold-molecule formation. We present finally
the recent demonstration of the stimulated Raman photoassociation process to
prepare state-selected cold molecules, so opening the way to the preparation of
molecules, not only translationally cold but also vibrationally and rotationally cold.
We discuss the other experimental schemes (Nikolov et al., 1999, 2000; Gabbanini
et al., 2000). We briefly present the possible developments for trapping the formed
cold molecules and analyze the experiment using a CO2 laser quasi-electrostatic
trap by Takekoshi et al. (1998). Finally, we mention very briefly the formation of
cold molecules in a condensate (Wynar et al., 2000), a subject which exceeds the
frame of this review.
B. EXPERIMENTAL SETUP IN THE ORSAY EXPERIMENT
1. Cold Atomic Source

A scheme of the setup is shown in Fig. 8. Photoassociation (PA) experiments with
cold alkali atoms are performed with the usual tools for providing cold atomic
samples: a magnetooptical trap (MOT) (Raab et al., 1987; Monroe et al., 1990),
FIG. 8. (a) Scheme of the experimental

√ setup. (b) Detection time sequence. The Cs+ and Cs+
2 ion
pulses arrival times are in a ratio 2. (c) Overlap between the PA laser beam and the cold atomic
sample.
dark SPOT (dark spontaneous force optical trap) (Ketterle et al., 1993; Townsend
et al., 1996), or a dipole trap such as FORT (far off-resonance trap) (Miller et al.,
1993.) Very recently, PA experiments have also been developed by using a rubidium
condensate (Wynar et al., 2000). In the cesium experiments, the cold atom source
is provided by the use of either a Cs vapor-loaded MOT or a dark SPOT. Details of
the experimental setup have been published in several previous articles (Comparat
et al., 1999). Using a dark SPOT instead of a MOT implies a different initial state
of the pair of atoms in the PA collisional process: the atoms are in the 6s1/2 , f = 3
hyperfine level, instead of being in f = 4.
In a vapor-loaded MOT, the cold Cs atoms are produced at the intersection of
three pairs of mutually orthogonal, counterpropagating σ + – σ − laser beams of
intensity 1–2 mW/cm2 and of diameter 6 mm, at the zero magnetic field point
of a pair of anti-Helmholtz coils with a magnetic field gradient of 15 gauss/cm.

The residual pressure is 2 ×10−9 torr. The cooling and trapping laser beams are
split from a slave diode laser (SDL 5422-H1, 150 mW, single mode, λ ≈ 852 nm)
injection locked to a master diode laser. The master laser (SDL 5412-H1, 100 mW)
is stabilized by optical feedback from an extended, grating-ended cavity. Locking
the master laser frequency to a saturated absorption line of a cesium vapor ensures
its long-term stabilization. The trapping laser frequency is tuned about 13 MHz
(≃2.5 natural linewidths) on the red of the frequency ν 4→5 of the 6s1/2 , f =
4 → 6 p3/2 , f ′ = 5 atomic transition (see Fig. 1, left). A repumping laser beam
(SDL 5712-H1, 100 mW, λ ≈ 852 nm) of frequency ν 3→4 resonant with the
61/2 , f = 3 → 6 p3/2 , f ′ = 4 transition is superimposed with two of the beams of
the cooling laser, preventing atoms from being optically pumped in the untrapped
f = 3 hyperfine level of the ground state. The total number of cold atoms inside
the trap is determined by using a calibrated photodiode to record the fluorescence
signal due to the trapping laser. We use the expression for the detected power of
fluorescence in the solid angle , given by
Ŵ C12 ωtot
2
/2
P = Nat hνtrap . (48)
4π 2 δ 2 + Ŵ 2 /4 + C22 ωtot
2
/2
In Eq. (48), ωtot is the total Rabi frequency due to the six trapping laser beams at
the frequency ν trap, Ŵ is the inverse of the characteristic time of spontaneous emis-
sion. C1 and C2 correspond to effective coefficients, C12 ≃ C22 ≃ 0.73, measured
experimentally by Townsend et al. (1995). These coefficients do not correspond to
those expected (≃0.4) by assuming an equirepartition on all the Zeeman sublevels
of the atomic ground state f = 4. They indicate a larger population in the magnetic
sublevels with maximum absolute quantum numbers.
Depending on the experimental conditions (laser detuning and intensity), the
number of atoms lies in the range between 2 and 5 × 107 and the temperature T
of the cold atomic sample ranges between 20 and 200 μK (Fioretti et al., 1999).
For a detuning of 14 MHz, an intensity of ∼1 mW/cm2 for each laser beam, and
a temperature of 140 ± 20 μK, the number of atoms is Nat of 4.7 × 107 estimated
within a factor of 2. The atomic sample is approximately spherical. We assume
that the density is proportional to a gaussian distribution exp(−r 2 /2σ R2 ), with
σ R = 300 ± 50 μm in the three space dimensions. The mean density is of the
order of 4 × 1010 atoms/cm3 (∼1011 atoms/cm3 for the peak density) and is known
within a factor of 3. In these experimental conditions, the characteristic loading
time for the trap has been measured in the range 600–800 ms, depending on day-
to-day fluctuations of the experimental conditions. In the MOT and dark SPOT
devices, atoms can be further cooled below the Doppler limit, down to T ≤ 30
μK, by detuning, during 7 ms, the trapping laser to nine natural linewidths and
simultaneously reducing the beam’s intensity by a factor of 2 through a Pockels
cell. The MOT is shifted into a dark SPOT configuration by modifying the re-
pumping arrangement to transfer most of the atoms (>90%) in the “dark” state
f = 3 in the center of the trap. The repumping beam is now partly screened with
a 3-mm-diameter black spot of insulating tape, stuck onto a microscope slide. The
spot is imaged at the center of the trapping region using a lens of focal length
l = 400 mm, placed at a distance 2l apart. Just after the lens, a beam splitter splits
the repumping laser beam into two beams. They are superimposed with two cool-
ing laser beams. The images of the spot for the two laser beams correspond to the
center of the trapping zone. The difficulties in applying the dark SPOT technique to
heavy atoms have already been noticed (Townsend et al., 1996). To obtain a large
fraction of the atoms in a “dark” state, we illuminate the cold atomic sample with
a depumping laser beam of about 10 mW cm−2, tuned about 25 MHz (∼5 natural
linewidths) on the blue of the frequency ν 4→3 of the 6s1/2 , f = 4 → 6 p3/2 , f ′ = 3
atomic transition. For that purpose, the diode laser is locked on the level cross-
ing 6s1/2 , f = 4 → 6p3/2 , f ′ = 3 and 6s1/2 , f = 4 → 6p3/2 , f ′ = 5 and the laser
beam passes through an acoustooptic modulator which shifts its frequency by
200 MHz. We have verified by photoabsorption that more than 90% of the atoms
in the center of the trapping zone are in the “dark” state f = 3.
2. Photoassociation Laser
The trapped, cold Cs atoms are illuminated with a continuous-wave (CW) laser
with wavelength λPA = 2πcωPA
to produce the photoassociation reaction; from Eqs. (1)
and (6), neglecting the kinetic energy terms in Eq. (5), we consider two kinds of
transitions, according to the hyperfine level in the initial state,
2Cs (6s1/2 , f = 4) + hωPA → Cs2 [ u,g (6s1/2 + 6 p1/2,3/2 ; v, J )] (49)
for the MOT experiments, and
2Cs (6s1/2 , f = 3) + hωPA → Cs2 [ u,g (6s1/2 + 6 p1/2,3/2 ; v, J )] (50)
for the dark SPOT experiments. The rovibrational level of a long-range molecu-
lar state u,g —labeled here in Hund’s case c notation—is thus populated. PA is
achieved by continuously illuminating the cold Cs atoms with the beam (λPA
852 or 894 nm) of a Ti:sapphire laser (“Coherent 899” ring laser) pumped by an
argon-ion laser, red-detuned from the 6s( f = 4) → 6 p3/2 ( f ′ = 5) atomic transi-
tion (11732.183 cm−1 (Avila et al., 1986), or from the 6s( f = 4) → 6 p1/2 ( f ′ = 3)
atomic transition (11178.07 cm−1). The frequency scale is calibrated using a Fabry–
Perot interferometer, and the absorption lines of iodine (Gerstenkorn et al., 1982).
The maximum absolute uncertainty is estimated to be ±150 MHz, mainly due
to the uncertainties on the position of the iodine lines. Thanks to the Perot–Fabry
interferometer, which has a free spectral range of 750 MHz, the local uncertainty
is reduced to about ±10 MHz. The maximum available power of the laser beam
is 650 mW. We approximate the intensity by a Gaussian profile, exp(−2ρ 2 /w02 ),
with w0 = 300 ± 50 μm, leading to a maximum intensity of 450 W cm−2 for the
PA laser.
3. Detection of the Molecules

The Orsay experiment uses two kinds of detection schemes for the photoassociation
process and the formation of molecules in the ground or lower triplet state.
r As in many photoassociation experiments (see Stwalley and Wang, 1999, and
references therein), trap-loss analysis is performed recording the fluorescence
yield from the trap, collected by a photodiode. The variations of the signal
are interpreted as decrease of the number of atoms in the trap due both
to dissociation of PA molecules into a pair of atoms that escape the trap
[Eq. (46)] and possibly to the formation of cold ground-state molecules.
r Second, Cs+ 2 ions are detected through a time-of-flight mass spectrometer,
after photoionization of the translationally cold Cs2 molecules. The photons
are provided by a pulsed dye laser (λ2 ∼ 716 nm, 7-ns duration) pumped by
the second harmonic of a Nd-YAG laser; the ionization is a REMPI process,
using as intermediate step the vibrational levels of an electronic molecular
state correlated to the (6s + 5d3/2,5/2 dissociation limits (Fioretti et al., 1998)
(see Fig. 9). There is a window of a few hundred wavenumbers around the
wavelength λI ≈ 716 nm and another window is observed for λ′I ≈ 554 nm,
corresponding to the REMPI process via vibrational levels correlated to
6s + 7s. The time-of-flight scheme discriminates atomic and molecular ions.
The sensitivity of ion detection for probing the formation of molecules is a
key point of these experiments. The upper vibrational levels of the ground
state cannot be detected in the present scheme (Dion et al., 2001), due to the
selection rule forbidding (6s → 5d) atomic transitions. Estimations show
that presently the detection efficiency is only about 5%.
C. PHOTOASSOCIATION SPECTRA
The fluorescence and the Cs+ 2 ion spectra are recorded as a function of the PA
laser frequency. A summary of the Orsay data is shown in Fig. 10 with typical
spectra obtained by using a MOT atomic sample. The origin of the energy scale is
fixed at the 6s1/2 , f = 4 → 6 p3/2 , f ′ = 5 atomic transition, which corresponds to
an energy of 11732.183 cm−1 above the 6s1/2 , f = 4 + 6s1/2 , f = 4 asymptote.
FIG. 9. The detection scheme in the Orsay experiment. It is a two-photon resonant process (REMPI)
via vibrational levels in the (2) 3 g potential curve correlated to the Cs (6s) + Cs (5d) dissociation
limit. Ultracold molecules can be detected both in the ground X 1 g+ and metastable a 3u+ states.
For detunings smaller than 0.1 cm−1, the MOT is destroyed by the PA laser. The
fluorescence and the Cs+
2 ion spectra are quite different.
r First, we clearly observe resonance lines up to a PA laser detuning of 80 cm−1

while the trap-loss spectrum stops beyond 40 cm−1 and of only 40 cm−1 for
the trap-loss.
r Second, the density of resonance lines in the trap-loss spectrum is much more
important.
1. Trap-Loss Spectrum
In the fluorescence spectrum, recorded as a function of the detuning of the
photoassociation laser, we observe dips interpreted as vibrational progressions.
The presence of a dip corresponds either to a photoassociation reaction followed
by dissociation,
Cs2 u,g 6s + 6 p 2 P3/2 ; v, J , → Cs(6s) + Cs(6s) + E c′ + h ω0i − δ2 ,

(51)
where it is assumed that E c′ is large enough so that the two atoms live the trap,
or to a photoassociation reaction followed by spontaneous emission into a bound
FIG. 10. Cs+ 2 ion signal (lower signal) and trap fluorescence yield (upper signal) versus detuning of
the PA. In the inset, details of the rotational structure of the 0−
g v = 10 level is shown. The dashed line
indicates the correspondance of vibrational levels of the 0− g state on both spectra. Notice the different
scales for the axis. The spectra are the results of a large number of scans.
level of the ground state,
Cs2 [ u,g (ns + np1/2,3/2 ); v, J ] → Cs2 [1,3 g,u

+
(6s + 6s; v0 , J )] + h(ω0 + δ3 ).
(52)
The observed vibrational progressions can be assigned to the 1g , 0+ u , and

0−g (6s 1/2 + 6 p 3/2 ) states. About 80 lines for each vibrational progression are well
resolved in the range 2–45 cm−1. The 1u state does not seem to be present in the
fluorescence spectrum. The reason is that the dissociation of the photoassociated
1u molecules after spontaneous emission do not produce “hot” enough atoms to
escape the trap: in other words, E c′ is smaller than the trap depth. In the energy
range of about 5 cm−1, the observed linewidths are mostly due to the hyperfine
structure and in qualitative agreement with the computed ones of about 6, 1.5, 0.6,
and 6 GHz for the 1g , 0+ −
u , 0g , and 1u, respectively (Fioretti et al., 1998).
To verify the identification of the lines, a fit can be performed with the semi-
classical Le Roy–Bernstein law already discussed in Section II.B,
1/6 −1/3
D i − E vi = A3 (v D − v) C3i

, (53)
where the dissociation limit D i , the constant A3, and quantum number v D have
been defined in Eqs. (24) and (25). The relation between C3i and C3 = 6s|dz |6 p 2
is given by Julienne and Vigué (1991):
√
5 7+2
C31 = C3 (0+
u) = C3 C32 = C3 (1g ) = C3 (54)
3 3
The fit allows us to identify the excited electronic states, and to obtain val-
ues for their dissociation energies and for the constant C3 (see the reviews
by Weiner et al., 1999; Stwalley and Wang, 1999). In the cesium experi-
ments not reviewed in the quoted papers the following values are obtained:
D fit (1g ) = −3 ± 1 GHz, v fitD (1g ) = 212.9 ± 0.5 (this refers to numerotation v = 0
for δ1 = −44.141 cm−1 ) and C3fit (1g ) = 15.72 ± 0.05 atomic units for the 1g state
fit
and D fit (0+ +
u ) = −7.8 ± 1.1 GHz, v D (0u ) = 214.0 ± 1 (this refers to numerota-
tion v = 0 for δ1 = −44.336 cm , and C3fit (0+
−1
u ) = 16.1 ± 0.4 atomic units for
the 0+ u state. C 3 coefficients are in 4% and 9% agreement with the calculated values
[C3 (1g ) = 16.22 a.u., C3 (0+ u ) = 17.46 a.u.], Marinescu and Dalgarno (1995). But
the atomic lifetime τ deduced from the fitted C3 coefficients,
3
hǫ0 λ
τ= (55)
2e2 6s|dz |6 p 2 2π
is in 1% and 5% agreement with the experimental one, τ6 p3/2 = 30.5 ± 0.1 ns

(Volz and Schomoranzer, 1996). Notice that the fits are performed in a range
of detuning (larger than 4 cm−1) where the hyperfine structure is not expected to
perturb noticeably the −C3 /R 3 behavior. The uncertainty of the fitting coefficients
illustrates their variation according to the number of lines included in the fit. Some
more accurate fit, including the higher multipolar expansion in the asymptotic
part of the potential, should give more accurate prediction. Other effects, such as
predissociation of 0+ u (Kokoouline et al., 2000b) should also be considered. As
mentioned by Fioretti et al. (1998), the Le Roy–Bernstein approach does not work
very well for fitting the asymptotic potential of 0− g , due to its particular double-
well character, with smooth variation of the potential in the bump zone between the
inner and outer wells. Similar conclusions were obtained by Fioretti et al. (2001)
for the equivalent double-well in the Rb2 case.
2. Ion Spectrum
The ion signal is more selective than the trap-loss one, only the vibrational pro-
+
gressions of the 0−
g and 1u states being present in the Cs2 spectrum. The detection
−
procedure can be optimized, for either 0g or 1u, by adjusting the wavelength of
the pulsed laser used for the photoionization. The detection is here sensitive up to
a detuning range of 80 cm−1 for the PA laser.
+
Cs2 0−g Pure Long-Range State. The Cs2 ion spectrum exhibits 133 well-
resolved structures assigned as the vibrational progression in the outer well of
the 0−
g ( p3/2) state, starting at v = 0. The rotational structure, shown for v = 10
in the inset of Fig. 10, is resolved up to J = 8 for most of the vibrational
levels below v = 74.1 The energies of the spectral lines have been fitted with a
Rydberg–Klein–Rees (RKR) and near-dissociation expansion (NDE) approach,
giving, for the outer well, an effective potential curve with a 77.94 ± 0.01 cm−1
depth and an equilibrium distance Re = 23.36 ± 0.10a0 (Fioretti et al., 1999).
This provides good knowledge for the vibrational wavefunctions and for the inner
and outer turning points of the classical vibrational motion up to v = 74. The 0− g
long-range well filled potential of various alkali dimers has been characterized
by photoassociation spectroscopy: the experimental values for the spectroscopic
constants have been determined for Na2 (Ratliff et al., 1994), where hyperfine
structure has to be included, for K2 (Wang et al., 1996), who obtain a well depth
and position De = 6.49 cm−1 , Re = 14.63a0 , Rb2 (Cline et al., 1994), where
1
The large number of observed rotational levels is due to a cooperative effect between the cooling
laser and the PA laser (Fioretti et al., 1999). In fact, if the cooling is switched off during the PA phase,
at a temperature T ≤ 30 μK, only s-waves have to be considered in the experiment with excitation of
only J = 0 and J = 2 rotational levels. At T ≃ 200 μK s-, p-, and d-waves are essentially present and
the excitation of rotational state up to J = 4 is possible.
De = 28.295 cm−1 . As discussed in Section II.A, the well depth De = 77.94 cm−1
found by Fioretti et al. (1999) is much larger for Cs2.
Two giant structures at detunings δ1 = 2.14 cm−1 and δ2 = 6.15 cm−1 will be
described in Section V.C. We remark that the variation of level spacing follows
approximately the 5/6 scaling law as a function of detuning: in the vicinity of a
1-cm−1 detuning, the level spacing is ∼0.6 cm−1, whereas close to hδ L = 10 cm−1
the level spacing increases up to ∼3.6 cm−1, in qualitative agreement with the
6.8 factor predicted by the scaling law. More refined verification should take into
account the derivative of the level energy, and the variation of the C3i coefficient
as a function of distance Rvi . The modulation of the line intensities, due to the
nodal structure of the ground-state wavefunction, described in Section II.E, is
clearly visible. It has been noticed (see, for instance, Côté et al., 1995; Tiesinga
et al., 1996) that this modulation can be used for the determination of collision
parameters such as scattering lengths. Up to six of the eight visible nodes could
be used by Drag et al. (2000b) to fit the ground triplet scattering length aT.
The slow variation of the maximum intensity as a function of detuning can be
qualitatively explained by a compensation beween the −7/6 variation of the pho-
toassociation rate and the 5/6 variation of the bound–bound transition probability
provided we may prove that the molecular ion signal is a signature of the presence
of molecules in the a 3 u+ lower triplet state.
Cs2 1u Pure Long-Range State. The large structures in the Cs+ 2 ion spectrum in
the range (3–6 cm−1) are assigned to the 1u state. The Cs+ 2 ion spectra, recorded as
a function of the PA laser frequency for the MOT and the dark SPOT devices, are
shown in Fig. 11. The MOT spectrum exhibits both the 0− g and the 1u vibrational
progressions, while the 0− g progression is dominated by the 1u one in the dark
SPOT spectrum. The observed intensities corresponding to the same vibrational
levels are different for the MOT and the dark SPOT experiments, which is not
surprising because the two spectra correspond to two different initial states of the
collisional process [see Eqs. (50) and (51)], with two free atoms in either the f = 4
or f = 3 state. For the 1u state, the intensities corresponding to the dark SPOT
are on average 10 times larger. One notices also that the vibrational level, v = 0,
appears in the dark SPOT spectrum but not in the MOT spectrum. Conversely,
higher vibrational levels, up to v = 12, are observed in the MOT spectrum.
Figure 12 shows the well-resolved structures of the vibrational levels v = 0,
1, and 2 of the 1u state. The spectrum does not exhibit any simple progression,
neither hyperfine nor rotational. To interpret these complex structures, Comparat
et al. (2000) have performed systematic asymptotic calculations including both
hyperfine structure and molecular rotation, for all the electronic states involved in
1u photoassociation. They had not to introduce exchange energy in the calculations
since this quantity remains small at the position (about 25a0) of the inner wall of
FIG. 11. Cs+2 ion signal versus detuning of the photoassociation laser in the dark SPOT device
(a) and in the MOT (b). For the dark SPOT spectrum, the frequencies are shifted by 18.384 GHz.
FIG. 12. Zoom of Cs+ 2 ion signal versus detuning of the PA laser in the dark SPOT device, for the
1u vibrational levels v = 0 (a), v = 1 (b), and v = 2 (c).
FIG. 13. Structure of the v = 0 line of the 1u vibrational progression, compared in mirror with the
calculated intensities. The frequency of the first observable line is arbitrarily taken as zero.
the 1u potentials. The complete calculation of the positions and the intensities of
the lines, in which the only adjustable parameters are the initial populations of the
different partial waves (from s to g, here), agree very well with the experiment
(see Fig. 13). More than the 0− g state, the 1u state can really be labeled as a pure
long-range state, corresponding to a pair of atoms, the cohesion of which is given
by the electrostatic long-range multipole interaction (Stwalley et al., 1978).
D. MECHANISM FOR COLD MOLECULE FORMATION

Up to now, although we have identified selective population of the 0− g ( p3/2 ) and
lu ( p3/2) excited states in the ion spectrum, we have not yet proved the presence of
ground-state ultracold molecules. Indeed, the molecular ions could be produced
directly by photoionization of an excited state. To rule out this explanation, we
analyze the Cs+ 2 ion signal by considering the following temporal sequence: first
the PA laser beam is applied during 15 ms, then the ionizing laser pulse (7-ns
width) is delayed compared to the switching on of the PA laser. We observe that
the Cs+ 2 ion signal decreases with a characteristic time of the order of 10 ms
(see Fig. 14). This time is five orders of magnitude larger than the radiative lifetime
of any singly excited molecular state with electric-dipole allowed transition to the
ground state. Indeed, this characteristic time is of the order of the time during
FIG. 14. Time evolution of the Cs+ 2 ion signal when the photoassociation laser is turned on and off.
The exponential decrease corresponds to a characteristic time Td ≈ 10 ms, much larger than the radiative
lifetime of the photoassociated molecules (Tsp ≈ 30 ns). This result demonstrates that the photoion-
ization scheme is indeed starting from long-lived ultracold molecules in the ground singlet or triplet
state. The decrease of the signal is due to the fact that those molecules are falling down from the trap.
which molecules can move significantly out of the trap because of gravity. This
result clearly indicates that ions are not produced by direct photoionization of PA
excited molecules, but indeed by photoionization of the ground-state molecules.
Why, in contrast with earlier discussion, has reaction (52) a favorable branching
ratio in the case of two particular symmetries? This can be explained by the very
particular shape of the outer well for the two relevant potential curves, discussed in
Section II.A. Due to the smooth ∼R−3 behavior of the inner branch of the well, the
vibrational motion is gradually slowed, and there is a non-negligible probability
of presence at the inner turning point, as is manifested in Fig. 15, to be compared
with Fig. 4 representing an “ordinary” vibrational wavefunction. The efficiency
of the mechanism for the formation of cold molecules comes from the existence
of a Condon point at intermediate distance, as shown schematically in Fig. 16. In
all cases photoassociation occurs at long-range distance. The dissociation reaction
(51) can always take place (i), but spontaneous emission may also occur at a short
enough interatomic distance [case reaction (ii) or (iii)], forming cold ground-state
(ii) or lower triplet-state (iii) molecules. This is due to the particular vibrational
motion in the 0− g or 1u external wells, where a long time delay is spent at both
the outer and inner turning points. In contrast, in the case of the 1g or 0+ u state,
the reflection on the inner repulsive wall is very rapid, so that the vibration of the
excited molecule keeps the two atoms at large interatomic distance most of the time.
Therefore, the formation of translationally cold molecules is attributed to the
particular shape of some of the external potential wells in the heavy-alkali dimers,
which offer at the same time an efficient photoassociation rate and a reasonable
branching ratio for spontaneous emission toward the ground state. A confirmation
of this mechanism has been demonstrated in rubidium by Gabbanini et al. (2000).
FIG. 15. Vibrational wavefunction in the outer well of the 0−

g (6s + 6 p3/2 ) potential of Cs2, for a
binding energy v = 6 cm−1 .
E. TRANSLATIONAL TEMPERATURE
These cold ground-state Cs2 molecules are not trapped by the MOT and can be
detected below the trap zone (Fioretti et al., 1998; Comparat et al., 1999). The
spatial analysis of the ballistic expansion of the falling molecular cloud yields
FIG. 16. Mechanism for the formation of ultracold molecules, from the relevant potential curves of
Cs2. The photoassociation process occurs at large distance. Line (i) represents spontaneous emission
toward continuum states, with dissociation of the molecule; lines (ii) and (iii) represent spontaneous
emission toward bound states, with formation of stable cold molecules.
FIG. 17. Temperature measurement through time of flight: recordings (i–iv) correspond to the
temporal analyzis of Cs+
2 ion signal at the MOT position (i), and at 0.95 mm (ii), 1.90 mm (iii), and
2.85 mm (iv) below the MOT position.
a measure of the temperature of the molecular cloud. The ground-state cold Cs2
molecules are photoionized into Cs+ 2 ions, using here the dye laser focused on spot
of 300 μm diameter. Two kinds of measurements of the molecular temperature
have been performed. The first one consists of photoassociating the cold atomic
sample during a time of 3 ms. The fall of the molecular could is then temporally
analyzed at different distances below the atomic trap. Figure 17 shows a time-of-
flight analysis at different positions below the cold-atom trap. The theoretical fit
of the experimental data gives access to the temperature, and also allows one to
determine precisely the height of the fall. The other measurement of the tempera-
ture of the molecules relies on the analysis of the ballistic expansion for different
heights, using the dye laser beam focussed on a spot with the same 300 μm
diameter. Figure 18 depicts the spatial analysis of the falling molecular cloud. A
model proposed by A. Lambrecht et al. (1996), taking into account the formation
and the fall of the cold molecules, allows one to derive the molecular temperature
from the data. A temperature as low as 20+15
−5 μK has been determined. The atomic
temperature, measured similarly by photoionizing the cold Cs atoms into Cs+ ions,
lies in the range 20–30 μK, showing no difference with the molecular temperature.
F. PHOTOASSOCIATION RATE
The measurement of the rate of formation of cold molecules could be obtained
directly by using the Cs+
2 ion signal. Unfortunately, this signal is difficult to analyze,
because the efficiency of the photoionization process is badly calibrated, probably
not better than 10%. Taking into account the ion recollection and the detector
efficiency, the global efficiency of the detection does not exceed a few percent.
FIG. 18. Temperature measurement through ballistic expansion: recordings correspond to the spatial
analyzis of Cs+
2 ion signal at the MOT position (i), and at 1.90 mm (ii) and 3.80 mm (iii) below the
MOT position.
This first estimation leads to a cold molecule formation rate of 0.1 per atom and
per second in the conditions of a MOT. For a precise determination, it is preferable
first to measure the PA rate by using the trap-loss signal and then to deduce the
formation rate by using the calculated branching ratio between bound–bound and
bound–free transitions of the photoassociated molecules.
Drag et al. (2000a) have so performed accurate trap-loss measurements for the
two states 0+ −
u and 0g (6s + 6 p3/2 ). To determine the experimental PA rates, they
have considered the dynamic trap equation (Hoffmann et al., 1994) expressing the
balance between the loading rate and the various loss rates, for a total number
of trapped atoms Nat and a density nat,

d Nat
= − γ Nat − (β + βPA ) n 2at (r ) d 3r, (56)
dt vol
where γ is the loss rate due to background gas collisions, β is the loss rate due
to binary collisions among the trapped atoms, and βPA is the loss rate resulting
from PA among the trapped atoms.2 In the steady-state regime, the total number
of trapped atoms is measured with or without the PA laser. In the former case,
the resonance line intensity for the considered rovibrational level is recorded.
Considering the Gaussian distribution of the atomic density and assuming that it
is not too much affected by PA, determine the ratio
NPA γ + βn at
≃ . (57)
Nat γ + (β + βPA )n at
2
The coefficient βPA is weighted by a factor corresponding to the partial overlap between the cold
atomic volume and the laser beam (see Fig. 8c).
Knowing the quantity γ + βn at , one may deduce the parameter βPA . The lat-
ter quantity corresponds to the inverse of the characteristic loading time of the
MOT, which has been measured as τ ≃ 735 ms, with the experimental con-
ditions of the MOT n at = 1011 cm−3, T = 140 μK. They obtain in this way
−2 −1
n at βPA (0+
u , 200 W cm ) = 3.5 ± 0.5 s , for the level corresponding to a res-
−1
onance line located at 14.35 cm below the dissociation limit and for a PA laser
intensity of 200 W cm−2. They have measured the rate n at βPA (0− −2
g , 55 W cm ) =
2.45 ± 0.6 s−1 for the v = 77, J = 2 level and for a PA laser intensity of 55 W
cm−2 . More generally, in the Orsay experiments, for all the states observed in the
energy range 5–10 cm−1 below both dissociation limits 6s + 6 p3/2 and 6s + 6 p1/2 ,
PA rates measured through trap losses have values ranging roughly between 1 and
5 s−1 (Drag et al., 2000a). The determination of the formation rate of cold molecules
is discussed below in 5 V.B.
G. ROTATIONAL AND VIBRATIONAL TEMPERATURES

The molecules formed through PA are indeed translationally cold. But after sponta-
neous emission the cold molecular sample is found in a statistical mixture of a few
rovibrational levels, determined by the Franck–Condon factors. To obtain ground-
state molecules that are cold in all degrees of freedom (translation, vibration, and
rotation) is still a challenge. Putting all the molecules in the same rovibrational
level is a necessary step to get a molecular sample useful for further experiments.
Stimulated Raman PA, where the emission on a given bound–bound transition is
stimulated (see the scheme of the relevant levels in Fig. 19), offers such a possibil-
ity (Bohn and Julienne, 1996; Vardi et al., 1997). Stimulated Raman PA has been
used in a Rb condensate, where molecules should be formed, but up to now no
direct evidence of their presence could be produced (Wynar et al., 2000). It has
been demonstrated for the cesium atom, by using a MOT device. (Laburthe-Tolra
et al., 2001)
The principle of the experiment consists, for a pair of colliding cold Cs atoms
prepared in the hyperfine level f = 3, in having a Raman two-photon transition to
form directly a ground-state X 1 g+ molecule in a well-defined final rovibrational
level v, J . In fact, as the ground-state vibrational levels reached are located energy-
wise in the vicinity of the dissociation limit, the gerade/ungerade symmetry is
broken due to hyperfine coupling, and the electronic ground state corresponds
to a superposition of singlet and triplet characters (Comparat et al., 2000). The
intermediate level, labeled 1 in Fig. 19, is a hyperfine-rotational component of the
vibrational level v = 0 or 1 of the 1u state, which will be specified later. The first
laser, labeled L1, is detuned by a value on either the red side or the blue side of
the resonance of the PA transition,
2Cs (6s, f = 3) + hν1 → Cs∗2 (1) (58)

FIG. 19. Relevant energy levels of the stimulated Raman PA transition, from a continuum of states
α of the two colliding atoms to a final bound level (2) of the ground state molecule. and δ correspond
to the detunings of lasers L1 and L2 compared to the PA transition α → 1 and to the stimulated Raman
PA transition α → 2. The laser L2 can also produce one-photon PA.
so that no PA due to the laser L1 can be observed. We scan the frequency of the
laser L2 to make resonant (δ = 0) the Raman transition toward the final ground-state
level, labeled 2. One has
Cs2∗ (1) + hν2 → Cs2 (X1 g+ (6s, f ′ + 6s, f ′′ ); v, J ) (59)
Stimulated Raman PA is achieved by applying at t = 0, on the atomic cloud

continuously illuminated by the laser beam L2, a pulsed laser beam L1 (τ = 2.5 ms;
this time is chosen to avoid a too significant decreasing of the atomic density). The
laser beam L1 is provided by the Ti:sapphire laser with an available intensity in the
MOT zone of 450 W cm−2. A Pockels cell is used for switching. The laser beam
L2 is provided by a DBR diode laser (SDL 5712-H1, 100 mW) with a maximum
available intensity of 50 W cm−2.
Figure 20a shows the PA Cs+ 2 ion spectrum obtained by applying only the
laser L2 with an intensity of 30 W cm−2, and by scanning its frequency around
the resonance corresponding to one-photon excitation (process shown by the
dot-dashed arrow in Fig. 19) of the vibrational level v = 1 of the 1u (6s + 6 p3/2 )
state. One has
2Cs(6s; f = 3) + hν2 (60)

→ Cs2∗ [1u (6s + 6 p3/2 ); v = 1]. (61)
FIG. 20. (a) PA spectrum of the 1u v = 1 level, by scanning laser L2. (b) Same as (a) but in the
presence of laser L1. (c) Stimulated Raman PA resonance. The inset analyzes the Raman character of
resonance (c).
The maximum of detected ions is here of the order of 100 per shot. The dashed
arrow shows on this spectrum a very well resolved and isolated line. This resonance
corresponds to the excitation of the hyperfine level, labeled 1, with a total angular
momentum F = 7, and with a wave function close to
|(6s + 6 p3/2 ) 1u , It = 7, M I = 7; F = 7, M F . (62)
It is the total nuclear spin, and MI and are respectively the projections of It
and of the total electronic angular momentum, Jt = 1, on the molecular axis. The
spectrum in Fig. 20b was obtained under the same conditions as previously, but
by applying at the same time the Ti:sapphire laser (L1). The frequency of laser
L1 (see the dotted line of Fig. 20) is now detuned by −20 MHz compared to the
resonance 1. The Cs+ 2 ion signal is 1.5% of the ion signal with laser L1 tuned
to resonance 1, which means less than 10 ions per shot. This number is com-
parable to the Cs+ 2 ion background due to the presence of cold molecules in the
MOT even without applying any PA laser (Fioretti et al., 1998; Gabbanini et al.,
2000; Takekoshi et al., 1998). Nevertheless, the presence of the laser L1 perturbs
the L2 spectrum. This perturbation is not fully understood, but it probably cor-
responds to further excitation of the 1u molecules toward highly excited states.
Another characteristic of spectrum (b) compared to (a) is the appearance of new
broad structures as shown by the full-line arrow. Reducing the intensity of the
laser L2 down to only 2 W cm−2 produces no more direct PA signal due to L2, but
maintains the existence of the extra structures which appear now as very narrow
resonances (see Fig. 20c) corresponding to the formation of cold Cs2 molecules
in a well-defined level of the ground state through stimulated Raman PA. The
Raman character of the resonance has been tested by red- and then blue-detuning
by quantities ≃ ±40-MHz laser L1 from resonance 1. The inset in Fig. 20
shows the shift of the resonance by a quantity ∼80 MHz. It has been verified
experimentally that the position of the line is fixed by the frequency difference of
the two lasers. The stimulated Raman PA signal of Fig. 20c corresponds to about
50–100 detected ions, meaning a number of cold molecules of about 1000, within
a factor of 3. The difficulty of a precise estimation of the cold-molecule num-
ber comes from the rough estimation for photoionization efficiency (Drag et al.,
2000a; Dion et al., 2001).
H. OTHER EXPERIMENTS
The case of the cesium atom is very demonstrative for the use of PA for the for-
mation of cold molecules. We have already demonstrated the possibility to ob-
tain cold molecules in the ground state or the lowest triplet state at temperature
in the range 20–200 μK with a formation rate typically varying between 0.05
and 0.2 molecule per atom and per second for the experimental conditions in
Orsay (atomic temperature 20–200 μK, density 4 × 1010 atoms/cm3). For 50 mil-
lion trapped atoms, 2.5 to 10 million ultracold molecules are formed per second.
We can also form state-selected molecules with similar characteristics by using
the stimulated Raman photoassociation. Is it possible to extend the case of the
cesium atom and dimer to other species? The answer is not obvious, because the
efficiency of the mechanisms is based on the particular double-well shape in the
0−g and 1u (6s + 6 p3/2 ) potentials curves, creating a Condon point at intermediate
distance. Among other alkali atoms, only the rubidium present a similar char-
acteristic for the 0−g (5s + 5 p3/2 ) state. The formation of cold Rb2 molecules in
the lowest triplet state has been demonstrated by Gabbanini et al. (2000). The
formation of translationally cold K2 molecules in their singlet ground state has
been demonstated via photoassociation of the A1 u+ (4s + 4 p3/2 ) state (Nikolov
et al., 1999). Here the molecular configuration does not correspond to a long-range
outer well. The branching ratio for bound–bound transitions is not as important
as for the cesium and rubidium experiments, and the rate for the formation of
cold molecules is only 1000 per second for the considered experimental condi-
tions. This number is nevertheless more important as compared to most of PA
experiments, and is due to an accidentally non-negligible Franck–Condon fac-
tor for a given bound–bound transition. More complex schemes of formation of
cold molecules have been proposed, with PA of highly excited molecular states
(Band and Julienne, 1995; Bohn and Julienne, 1996). An efficient production of
K2 molecules has indeed been achieved by using a two-step excitation scheme.
In this case, highly excited molecules radiate toward various vibrational levels of
the X ground state, with a rate comparable to the one achieved for cesium atom
(Nikolov et al., 2000). The supplementary difficulty is the use of a second pho-
toassociation laser but the method is a general one.
The conclusion of this section is that the formation of cold molecules via PA
can certainly be extended to any species which can be previously laser-cooled.
The demonstration of novel mechanisms for the formation of ultracold molecules
make this route very promising. However, a last difficulty in making these cold
molecular samples useful for applications is to be able to trap them. The trapping
of cold molecules with a CO2 laser quasi-electrostatic trap has been demonstrated
(Takekoshi et al., 1998,1999). These trapped cold molecules are not formed via PA,
but are those naturally present in the MOT trap. This result should be extended to
molecules formed directly by photoassociation in such a trap. Unpublished results
obtained in Orsay have demonstrated the possibility of trapping cold molecules
in the lowest triplet state by using a quadrupolar magnetic trap. Trap devices
for molecules should show rapid developments in the next years. They should
allow one to accumulate up to 104–106 ultracold molecules in a 10- to 100-μK
temperature range, eventually in a well-defined rovibrational level of the ground
state (or the lowest triplet state) if stimulated Raman PA is used. The PA experiments
using a condensate should also show interesting developments (Wynar et al., 2000).
We did not develop in this review this kind of very exciting results, which should
lead to the possibility of observing a molecular condensate or a molecule laser.
The coherence of the Raman process is here a key point making the conception of
these experiments a little different.
IV. Theoretical Methods

The theoretical treatment of cold-atom collisions has stimulated many new devel-
opments, which can be found, for instance, in the review of Weiner et al. (1999). In
the present review we shall focus on progress that was specially motivated by the
research on ultracold molecules, insisting upon the necessary accuracy in estima-
tion of Franck–Condon overlap as described in Section II.D. The specific technical
problems are:
r The dynamics of ultracold molecules is characterized by the large extension

of the vibrational motion in regions where the potential is very weak. Accurate
description of the nuclear motion requires methods capable of describing
both the short-range region, where chemical potentials are acting, and the
long-range region.
r In addition, the dynamics is very sensitive to the accuracy in the electronic
potential curves, and the requirement is well beyond the present accuracy of
ab initio calculations. Methods to fit the potentials to accurate experimental

data should be revisited for the present problem.
A. DYNAMICS: DETERMINATION OF VIBRATIONAL WAVEFUNCTIONS

FOR PHOTOASSOCIATED MOLECULES
We have seen above that the estimation of the photoassociation rate is controlled
by the matrix element of the electronic dipole moment between the continuum
wavefunction in the ground state and the bound vibrational wavefunction in the
excited state [see Eqs. (34) and (35)]. Numerical accuracy in the estimation of this
quantity is a key for a good comparison between theory and experiment. Although
well-established techniques have long been developed in molecular physics for
that purpose, the problem of cold molecules presents some specific difficulties.
We give in Fig. 8 an example of a typical vibrational wavefunction for a level close
to the dissociation limit of the Cs21g( p3/2). The local de Broglie wavelength λ(R),
in atomic units, is with h = 1, is
2π
λ(R) = , (63)
2μ[V∞ − V (R)]
and depends on the value of the potential V (R), which varies by several orders
of magnitude from the short-range region to the asymptotic region. It is clearly
visible in Fig. 8 that the local de Broglie wavelength is varying from a few tens of
a0 in the outer region to a value less than 1a0 in the inner region, i.e., by more than
one order of magnitude from the short-range region to the long-range region. The
numerical or analytical methods should be adapted to this situation, and the use
of a radial coordinate adapted to the local de Broglie wavelength is the key for all
developments.
1. Mapped Fourier Method

Grid numerical methods using fast Fourier transform have proved to be very effi-
cient for quantum molecular dynamics (Kosloff, 1988, 1996). Wavefunctions and
operators are represented on grids in both the momentum and coordinate spaces.
It is convenient to use constant grid steps, which is linked to the minimum value
of the local de Broglie wavelength. For calculations of the vibrational levels of
a diatomic molecule, a Fourier grid method, originally proposed by Marston and
Balint-Kurti (1989) and by Colbert and Miller (1992), was further developed by
Monnerville and Robbe (1994), Dulieu and Julienne (1995), and Dulieu et al.
(1997), and proved to be very successful. The wavefunction may be represented
by its values at a set of N equally spaced radial distances between the two distances
R1 and R1 + N × s = R1 + L,
N /2

ψ(Ri ) = ak gk (Ri )(i, j = 1, N ), (64)
k=(−N /2+1)
L
Ri = R1 + (i − 1)s , s = (65)
N
where N is assumed to be an even number, provided we may define a set of delta

functions at the grid points through the relation
N /2

gk (Ri )gk (R j ) = δ(Ri − R j ) (i, j = 1, N ). (66)
k=(−N /2+1)
(67)
This can be achieved by a discrete Fourier expansion on a set of N ingoing and

outgoing plane waves, with wavelength varying from 2s to L, choosing

1 2kπ R
gk (R) = √ exp −i , (68)
N L

N N
kǫ − + 1, . (69)
2 2
The coefficients ak are then readily calculated as
N
1 i2kπ R j
ak = √ ϕ(R j ) exp , (70)
N k=1 L
and can be interpreted as the amplitude ak( p) of the wavefunction on equally

spaced grid points in the momentum space, pk = 2kπ/L. The two representations
are connected by Fourier transform, the basis set in momentum representation,
conjugated to the coordinate representation basis set defined in Eq. (68), being
1
f j ( p) = √ exp(−i p R j ). (71)
N
The maximum value of momentum that can be represented in such a basis is
π π
pN = N = , (72)
L s
where s is the constant grid step defined in Eq. (65). It is then straightforward to
represent all operators by N × N matrices either in coordinate or in momentum
representation. The radial Schrödinger equation,
[T + V(R)]ψ(R) = Eψ(R), (73)
involves two operators. The matrix V(R) is diagonal in the coordinate representa-
tion, while the kinetic energy operator T( p) is diagonal in the momentum repre-
sentation and can be easily evaluated. After Fourier transform, it is written in the
coordinate representation as
π2 N2 + 2
Tii = , (74)
μL 2 6
π2 1
Ti j = (−1)i− j 2 2
. (75)
μL sin [(i − j)π/N ]
The eigenvalues are obtained by diagonalization of the N × N matrix T + V(R).

The computation effort involved in the diagonalization procedure scales
as N3. The choice of the constant step, and therefore of N, is linked to the maxi-
mum momentum in the physical problem or to the minimum value of the local de
Broglie wavelength,
λ(Re ) π
s ≤ = , (76)
2 2μ(V∞ − VRe )
obtained at the equilibrium distance Re where the potential is minimum. In the ex-
ample of the v = 332 excited wavefunction in the 1g potential presented in Fig. 8,
the calculations would require at least diagonalization of a 8200 × 8200 matrix.
However, the grid step is unnecessarily small in the long-range region. Fattal et al.
(1996) and Kosloff (1996) have proposed a mapping procedure in order to optimize
the use of phase space in grid methods, with application to Coulomb potentials. This
mapping procedure has been generalized to the present problem by Kokoouline
et al. (1999), who have introduced a variable grid step adapted to the local
de Broglie wavelength,
π λ(R)
s(R) = = , (77)
2μ[V∞ − V (R)] 2
or more generally, depending on an enveloping potential Venv (R) located below

the physical potential for all internuclear distances, for instance, the analytical
asymptotic potential −C3i /R 3 or −C6i /R 6 . The grid step is now
π
senv (R) = . (78)
2μ[V∞ − Venv (R)]
The number of grid points is markedly reduced when the step size is defined by
Eq. (77) and the same accuracy can be obtained by diagonalization of a 564 × 564
matrix. This matrix is obtained by defining a grid in a new coordinate x such that the
variable step in R corresponds to a constant step in x. We therefore use a Jacobian
J (x) such that
d R = J (x) d x, (79)
λ(R)
J (x) = . (80)
λ(Re )
The radial Schrödinger equation
d2 J′ d 5 (J ′ )2 1 J ′′

1 1
− + + V + − + φ(x) = Eφ(x),
2μJ 2 d x 2 μJ 3 d x 2μ 4 J4 2 J3
(81)
may be written in a symmetrical form by considering a new wavefunction φ(x),
φ(x) = J +1/2 (x)ψ(x), (82)
and by introducing a new potential, v̄(x) such that
7 (J ′ )2 1 J ′′

1
V (x) = V (x) + − . (83)
2μ 4 J4 2 J3
The calculation of the kinetic energy operator is then straightforward,

( )
π2 1 1 1
Ti, j = (−1)i− j 2 2 2
+ 2 , (84)
2μL sin [(i − j)π/N ] Ji Jj
if i = j, while
π2 N2 + 2 1
Ti,i = . (85)
μL 2 6 Ji2
Therefore calculations are performed in the working grid using the coordinate x
with a constant grid step. The vibrational energies are then obtained by diagonal-
ization of the Hamiltonian in the new representation. The wavefunctions have to
be transformed making use of Eqs. (83) and (81) to obtain results in the physical
grid R. If the potential V (R) is given as a set of numerical values, the procedure
is straightforward. However, as the derivation involves first and second derivatives
of the Jacobian J (x), care must be taken that such derivatives are well defined. In
many applications, the potentials are obtained by matching short-range potentials
computed by ab initio method to long-range asymptotic expansion, and interpola-
tion procedures have to be carried out carefully. An alternative procedure consists
of considering an enveloping analytical potential for the definition of the new
variable x. Choosing, for instance, a potential Venv (R) = −Cn /R n with n = 2, the
change of variable in Eq. (81) becomes analytical, and for a grid starting at distance
Rin we may define an adaptative coordinate through the formula
√ ( )
2 2μCn 1 1
x= − (n−2)/2 . (86)
π (n − 2) R (n−2)/2 R
in
The interpretation of this change of variable will be given on page 108.

As discussed by Kokoouline et al. (1999), this procedure is less efficient than
the preceding one, making use of the real potential, as far as optimization of the
phase space is considered, but the calculations now use analytic formulas, which
may save computing time. Finally, we should note that the wavefunction defined
at grid points can be interpolated without further loss of accuracy by use of the
Fourier expansion,
N
π
ψ(q) = ψ(q j ) sinc (q − q j ) , (87)
j=i
q
where q j is a working uniform grid (with or without mapping), q is any intermedi-

ate point, q is a grid step, and we have defined the function sinc(z) = [sin(z)/z].
In Fig. 21 we display the wave function, computed with sinc interpolation at a
large number (Ninterp = 10,000) of q values, for the (v = 332) vibrational level of
Cs2(1g). Comparison with linear interpolation clearly illustrates the good quality
of the second interpolation. Of course, we have checked the accuracy of the inter-
polated wavefunctions by comparing to standard methods. The wavefunction in
the adaptative coordinate x is displayed in Fig. 22, illustrating the regularity of the
oscillations.
The grid points can be used for the Fourier expansion of the wavefunction,
yielding useful properties in numerical quadrature. It is possible to use MFGR
representation with the same grid for the ground- and excited-state wavefunctions
FIG. 21. (Upper) The v = 332 vibrational wavefunction for the 1g (6s + 6 p3/2 ) potential of Cs2,
with binding energy v = 0.0419 cm−1 , Computed at the grid points defined with the mapping of
the Eq. (82). Here, the grid extends up to 500a0, and contains N = 564 points. (Lower) Interpolation
procedure. The wave function ψ(R), R = R j , is interpolated as Eq. (89) in text.
so that excellent accuracy can be obtained for an overlap integral of the two func-
tions; this seems one of the main advantages of a grid method. It can be easily
generalized to double-well potentials, as will be discussed later in the example
of tunneling effect, and to calculations with several channels, including predisso-
ciation problems where coupling between bound and continuum levels has to be
considered (Kokoouline et al., 2000b). In the method described above, all vibra-
tional levels are computed up to a maximum energy. It may be convenient to limit
the calculations to a given energy range by use of filtering techniques: in particular,
in the vicinity of a dissociation limit, there is a large density of levels. A mapping
procedure is introduced by Tiesinga et al. (1998) in a filtering method making use
of Green-function operators to resolve the hyperfine structure for the levels close
to the Na(3s ) + Na(3p ) asymptote.
The choice of the number of grid points is presently determined by optimization
of the use of the classical phase space (Fattal et al., 1996; Kokoouline et al., 1999).
FIG. 22. Interpolated wavefunction in the x variable, adapted to the local de Broglie wavelength.
The graph shows only the part corresponding to large R. Comparing with Fig. 21, one can appreciate
the regularity of the oscillations in the mapped wavefunction.
For low-energy scattering, or for levels close to the dissociation limit, the extension
of the wavefunction in the classically forbidden region increases, and this effect
has been up to now empirically addressed by an increase of the number N of points.
A proper treatment of the classically forbidden region, with a priori calculation
of the number of points necessary to represent the classically forbidden region,
still has to be implemented; it should benefit from the expertise developed by the
groups working on analytical representation of the wavefunction (Boisseau et al.,
2000b).
2. Numerov Approach and Other Methods

Many groups working in the field of cold collisions and long-range molecules
have used the well-known Numerov–Cooley algorithm (Cooley, 1971, and ref-
erences therein), both for bound and continuum wavefunctions, often using the
Blatt criterion (Blatt, 1967) to choose the step length by minimizing local trunca-
tion error. For bound and predissociated levels, the Cooley–Cashion–Zare routine
SCHR has been regularly modified and improved by the group at the University
of Waterloo (Canada), and the present version (Le Roy, 2000) includes variable
steps. We should also mention introduction of a variable step in Numerov integra-
tion by Côté and Jamieson (1995), the step size being divided by a factor of 2 each
time the local wavenumber becomes smaller than a given parameter. Comparison
with the grid methods described above still has to be carefully done, in order to
check accuracy and efficiency. The present situation is such that potentials data
being usually given for a limited number of internuclear distances, with a large
mesh size, results depend on the method chosen for interpolation. An accurate
comparison should consider a potential in an analytical form, or implement the
same interpolation procedure.
In the case of a two-channel problem, with both bound and continuum solutions,
a recent paper by Rawitscher et al. (1999) compares various approaches to compute
scattering wavefunctions and extract the scattering length and effective range. For
two coupled potentials of Lennard–Jones type, with an exponential coupling term,
equivalent results are obtained by
r A spectral-type integral equation based on Chebyshev expansions,

r A noniterative eigenchannel variant of the R-matrix method (Aymar et al.,
1996),
r A Gordon method (Gordon, 1969) as modified by Mies (1973).
The third one is easier to implement but necessitates a much larger number of
points. In contrast, an improved Numerov method is shown to yield much poorer
accuracy. The authors conclude with the necessity of performing similar compar-
ison under more complex conditions.
3. Analytical Treatment
Analytical solutions of the radial Schrödinger equation have been proposed by
many authors, mainly in the framework of the JWKB approximation. At zero
scattering energy, Gribakin and Flanbaum (1993), in a paper which is considered
one of the main references in cold collisions, obtained an analytical expression of
the scattering length by transforming the radial equation into a Bessel equation.
We should note that the change of variable is
√
2 2μCn 1
x= (n−2)/2
, (88)
(n − 2) h R
similar to the change of variable introduced in Eq. (86) at the limit R0 → ∞, Rin
with h = 1. Future work might benefit from this analogy. For the description of
bound levels, many authors have worked within the framework of the semiclassical
approximation defined above. Close to the dissociation limit, where the part of the
wavefunction that extends into the nonclassical region becomes important, correc-
tions to the first-order JWKB approximation have been implemented. Friedrich and
Trost (1996a,b), Trost and Friedrich (1997); Trost et al. (1998) introduce a Maslov
index modifying the /4 reflexion phase at the outer turning point, and changing
the Bohr–Sommerfeld quantization condition in Eq. (15). Another approach, by
Boisseau et al. (1998, 2000b), considers second-order JWKB terms and numerical
corrections to the quantization condition. The two derivations give the same re-
sults. Gao (2000) has discussed in detail the breakdown of Bohr’s correspondence
principle. As discussed above, numerical calculations can also be implemented in
the region where the first-order semiclassical treatment fails.
B. THE URGENT NEED FOR ACCURATE MOLECULAR POTENTIAL CURVES

AND ELECTRONIC DIPOLE TRANSITION MOMENTS
One reason why the comparison between theory and experiment happened to be
relatively easy in the case of K2, Rb2 and Cs2 dimers is that accurate potentials
were available at both short and long internuclear distances. We should cite, for
short distances, the pseudo-potential ab initio calculations of Meyer’s group (Spies,
1989) for Cs2, of Foucrault et al. (1992) for Rb2 and Cs2, of Magnier and Millié
(1996) for K2 and NaK, and of Magnier et al. (1993) for Na2. In addition, a lot
of spectroscopic information is available, very accurate potentials being fitted to
experimental spectra: let us cite, among many other examples, the recent fit of
ground-state potential for the Rb2 molecule by Seto et al. (2000), where 99.8%
of the potential well can be deduced from a fit to 12,148 transition frequencies
in a high-resolution A–X emission spectrum. For implementation of the detection
scheme using absorption to an intermediate state, knowledge of excited potential
curves and transition dipole moments is essential. In that case, the information
collected in the experiments of Pichler’s group, who performed absorption exper-

iments with dense alkali vapors in a heat pipe (Pichler et al., 1983; Beuc et al.,
1982; Veza et al., 1998), has also proved very helpful. In contrast, recent results on
photoassociation of calcium (Zinner et al., 2000) have not been fully interpreted
due to the lack of theoretical information.
The matching between short-range ab initio calculations and long-range expan-
sion (Marinescu and Dalgarno, 1995, 1996; Derevianko et al., 1999), and refer-
ences herein. Next must be performed (see Fig. 3). Discussion on the accuracy of
long-range coefficients is beyond the scope of the present chapter, and can be found
in various papers (Stivalley and Wang, 1999). The computed values either agree
with experiment or must be modified by a few percentage. This can be verified
on the values given in §C1 for the C3 coefficients. Finally, it should be wise to
remember from the two previous sections that when potentials are computed on a
set of internuclear distances, the mesh size should be sufficiently small to avoid
loss of accuracy due to the interpolation procedure.
V. Present Status of the Comparison between Experiment

and Theory: Formation Rates
As examples of comparison between theory and experiment, we shall discuss
the determination of photoassociation rates and absolute number of ultracold Cs2
molecules obtained in Orsay, as well as the demonstration of a tunneling effect
from analysis of the spectra.
A. PHOTOASSOCIATION RATES
Using for Cs2 the set of potentials obtained by matching the ab initio calculations of
Meyer’s group (Spies, 1989) to asymptotic curves using the long-range coefficients
of Marinescu and Dalgarno (1995), the photoassociation rates were computed with
the numerical methods developed above and compared to experiment (Drag et al.,
2000a).
In Fig. 23 we display the quantities RPA in s−1 (number of photoassociated
molecules formed per second divided by the total number of atoms) defined in Eq.
(36) and κPA = RPA /n at /(I /hωPA ), defined per photon and per unit density, for
photoassociation into a vibrational level of the double-well 0− g (6s + 6 p3/2 ) curve,
and of the long-range attractive 0−g (6s + 6 p1/2 ) curve. The detuning δ L of the laser
is relative to the D2 and D1 resonance lines, respectively, and is varied on the entire
range covered by the 0− g (6s + 6 p3/2 ) external well. The photoassociation rate
displays minima at detunings such that Rvi corresponds to the nodes in the ground-
state wavefunction, as discussed in Section II.E. The intensity of the maxima
decreases as a function of the detuning of the laser: for photoassociation into a
FIG. 23. Computed cesium photoassociation rates, RPA (in s−1) and κPA (in cm−5), for molecu-
lar states of 0−
g symmetry correlated to 6s + 6 p3/2 (closed circles) and to 6s + 6 p1/2 (open circles).
Experimental parameters are chosen: T = 140 μK, n at = 1011 cm−3 , I = 55 W/cm2 . We notice a mod-
ulation in the rates due to the nodal structure of the initial continuum state wavefunction (see text). The
dot-dashed lines indicate the PA rate given with the approximate formula (55) in the work of Pillet et
al. (1997), approximating the continuum wavefunction by its asymptotic sine behavior.
level of the 0− g ( p3/2 ) state, the decrease verifies the (δ L )

−7/6
behavior predicted
by the simple model described in Section II.F. We should notice the importance
of the decrease, the rate varying by two orders of magnitude when the detuning
increases from 1 cm−1 to 50 cm−1.
−6
In contrast, as the 0− g ( p1/2 ) potential curve displays R asymptotic behavior
i
(due to cancellation of the C3 coefficient), the scaling law is different (Wang and
Stwalley, 1998) and the decrease is slower. We notice that, despite the C6 long-range
behavior, the corresponding PA rates are not negligible, being roughly one order
of magnitude smaller than in the previous case. This result is of some interest for
future studies of PA in heteronuclear species, which also present R−6 long-range
behavior (Wang and Stwalley, 1998).
An interesting result of the calculations is the importance of the correct de-
termination of the correct ground-state wavefunction. In Fig. 23, we also display
rates computed with a E (R) continuum wavefunction approximated by a free-
wave sine function: the corresponding curve is overestimated by more than one
order of magnitude, and of course does not represent the minima at the nodes of
the real E (R) function. This confirms that the bottleneck in the photoassociation
process is the amplitude of the initial wavefunction, demonstrating the advantage
of performing experiments at small detunings.
Precise comparison between experiment and theory has been presented by
Drag et al. (2000a) for photoassociation into the vibrational levels of the 0− g and
−1
0+u (6s + 6 p 3/2 ) curves, at detunings 6.92 and 14.35 cm , respectively, below the
6s( f = 4) + 6 p3/2 ( f ′ = 5) limit, for which atomic loss rates have been measured.
The main source of uncertainty lies in the experimental parameters (atomic density
nat, temperature T, laser intensity I ) which have to be introduced in the calculations.
The measured atom loss rates are determined under experimental conditions with
nat = 1011 cm−3, T = 140 ± 20 μK, I = 55 W/cm2 for 0− g and I = 200 W/cm
2
+ 2
for 0u . The calculations use T = 140 μK and I = 55 W/cm in both cases.
For the 0− +
g and 0u photoassociation rates, the theoretical values are, respectively,
of 6.6 and 5.8 molecules s−1, in good agreement with the experimental values
of 2.4 and 3.5 molecules s−1 (assuming simply that the atom loss rate is twice
the PA rate). Both calculated and experimental rates lie within the margin of 1–5
excited molecules formed per second and per atom, for a total number of atoms
Nat = 5 × 107, estimated within a factor of 2. This means that all atoms in a MOT
could be transformed into excited molecules in a time less than 1 s.
B. COLD-MOLECULE FORMATION RATE

The next step is the estimation of the percentage of the photoassociated molecules
that may be stabilized into long-lived molecules. The rate equation for the number
Nexc(v, J ) (or simply Nexc) of molecules photoassociated in a rovibrational state
(v, J) by the laser can be written as
d
Nexc = RPA Nat − Ŵexc Nexc , (89)
dt
where Ŵexc is the spontaneous emission probability of level (v, J ) and Nat is the
number of atoms in the trap at time t. Provided that RPA ≪ Ŵexc ≈ Ŵ, the number
of atoms Nat is assumed to be constant, as the trap loading time is longer (∼700 ms)
than all other characteristic times in the problem. In such conditions, the system
−1
reaches within a time Ŵexc a stationary state with a constant number N̄ exc of excited
molecules,
RPA
N̄ exc = Nat (0). (90)
Ŵexc
The number of long-lived ultracold molecules Ncold (t) produced at time t by spon-
taneous emission can be estimated as
(t) bound
Ncold = Ŵexc N̄ exc t, (91)
bound
where Ŵexc is the spontaneous emission probability of the excited level v toward
all levels v ′′ of the molecular ground and lower triplet states. One can write,
FIG. 24. (a) Probability of populating vibrational levels of the ultracold molecules in the a 3 u+ state,
after spontaneous decay from the photoassociated levels lying in the external well of the 0− g (6s + 6 p3/2 )
state (black circles) or in the 0−
g (6s + 6 p1/2 ) curve. (b) Number of ultracold molecules formed in the
a 3 u+ state after photoassociation and spontaneous emission.
according to Eq. (90),
t bound RPA
Ncold = Ŵexc Nat (0)t. (92)
Ŵexc
Neglecting the R dependence of the transition dipole moment, the ratio between
bound–bound and total spontaneous emission probabilities in Eq. (93) can be
approximated by
bound
Ŵexc
= |v0 |v |2 = S(v → g), (93)
Ŵexc v0
where the sum runs over all bound vibrational levels v0 of the molecular ground
or (metastable triplet) state g. Selection rules for the PA rate are the same as for
spontaneous emission. We present in Fig. 24 the S(v → g) sums computed for the
3 +
transitions 0−
g (6s + 6 p3/2 ; v, J = 0) → a u over detunings covering the entire
−
energy range of the 0g external well. In the calculations of Drag et al. (2000a), the
vibrational levels v0 in the ground triplet state have been considered up to a binding
energy of 0.4 cm−1, the upper ones being not detected in the experiment. As ex-
pected, the variation of S as a function of detuning is opposite to the variation of the
5/6
PA rate, due to the v dependence of the normalization factor Nv2 [see Eq. (29)].
The decay occurs by vertical transition at intermediate distance, and the probabil-
ity increases at large detuning due to the increase of the normalization factor. The
lowest levels of the 0− g (6s + 6 p3/2 ) long-range well are shown to decay almost com-
pletely into bound levels. For comparison we have also reported the bound–bound
branching ratios for the 0− g (6s + 6 p1/2 ; v, J = 0) state, where, due to a different
scaling law, the dependence on detuning is much weaker. At all detunings, the weak-
3 +
ness of S[0− g (6s + 6 p1/2 ; v, J = 0) → a u ] population transfer, due to the lack
of a Condon point at intermediate distance, confirms the difficulty of identifying
the vibrational levels of the 0− g (6s + 6 p1/2 ) state in the ion spectra, since the prob-
ability of forming metastable a 3 u+ molecules is small.
When the photoassociation rate is multiplied by the probability of decay into a
bound level, we obtain the curves given in Fig. 24b, where the number of ultracold
molecules formed per second in the trap is given as a function of detuning. The
minima correspond to the zeros in the photoassociation rate at nodes of the ground-
state wavefunction. The maxima display a slow decrease as a function of detuning,
−7/6
due to the balance between the v decrease of the photoassociation rate and
5/6
the v increase of the probability of spontaneous decay into a bound level.
This result explains an interesting feature in the Cs+ 2 photoassociation spec-
trum for the 0− g (Fioretti et al., 1999) double-well long-range state (see Fig. 10),
where, apart from the intensity minima already discussed in previous works (Côté
et al., 1995; Fioretti et al., 1998; Drag et al., 2000b), the average number of
detected molecular ions is roughly constant over the entire energy range investi-
gated (∼80 cm−1). Assuming comparable ionization efficiency for all vibrational
levels v0 , this would indicate that the formation rate of ultracold molecules for
this particular double-well state in Cs2 is nearly independent of the choice of v
in the photoassociation process. This has been predicted in Section II from the
balance between the two scaling laws: a level v close to the dissociation limit of
the excited state will be populated efficiently (due to favorable Franck–Condon
factors at large distances), but will poorly emit toward bound levels v0 of the
lowest electronic states, due to the low probability amplitude in the vibrational
i
wavefunction φv,J (R) at intermediate distance. The situation is exactly oppo-
site for the lowest levels v of the excited state: the very large branching ratio
for decay to bound levels in the 3 u+ curve is associated with a very small PA
rate, such that the calculations predict no formation of cold molecules, as is
effectively observed (Fioretti et al., 1999). The conclusion is that, in order to
get a large rate for ultracold-molecule formation, one needs a good compromise
between the PA rate and the branching ratio for the bound–bound spontaneous
emission.
Except at very low detunings, the calculations yield an average theoretical rate
oscillating beween 0.1 and 0 triplet ultracold molecules per second and per atom.
This number is in reasonable agreement with the rates obtained directly from
trap-loss measurements, where a number of 0.05 to 0.2 molecule per second and
per atoms seems a good estimate (Drag, 2000a). It is also compatible with the
values deduced from the measured number of ions: a global rate for the formation
of cold molecules is between 2 × 105 and 106 molecules per second, in a trap
containing Nat ≈ 50 million atoms.
Concerning other alkali atoms, only rubidium offers a similar outer well in its
0−g (5s + 5 p3/2 ) state. The number of cold Rb2 molecules is found to be smaller
than for Cs2 as the outer well is located at a greater distance, reducing by a factor of
3 the bound–bound branching ratio (Fioretti et al., 2001). The number of cold K2
molecules obtained by Nikolov et al. (1999) by one-step photoassociation is only
about 1000 molecules per second, due to the absence of a similar external well.
In contrast, the two-step scheme of Nikolov et al. (2000) produces 105 molecules
per second and per vibrational level, or 106 molecules per second when summing
over all vibrational levels, which is the same formation rate as in the cesium case.
The time delay to transform half the atoms into molecules is therefore estimated
to be less than 1 minute in all the present experiments, showing that the process is
indeed very efficient.
An even better efficiency might be obtained if the photoassociation process was
simultaneously populating a large number of levels, using, for instance, broad-
band pulsed laser excitation: at low detunings, the small value of the amplitude of
the vibrational wavefunction of each individual level in the excited state would be
compensated by the number of levels, increasing the flux toward the ground state.
C. TUNNELING EFFECT
Another scheme for formation of ultracold molecules in low vibrational levels
of the a 3 u+ Cs2 state was identified in the photoassociation spectrum of Fioretti
et al. (1999), manifested by detection of a large number of molecular ions. In
the experimental ion spectrum displayed in Fig. 10, we can see two “giant” lines,
corresponding to binding energies 1 = 2.14 cm−1 , 2 = 6.153 cm−1 relative to
the 6s + 6 p3/2 dissociation limit. Details of the observed spectrum are given in
Fig. 25. The “giant” structures have a well-resolved rotational structure in con-
trast to the neighboring lines identified as vibrational levels of the outer well of
the 0−g (6s + 6 p3/2 ) potential curve. Vatasescu et al. (2000) have analyzed the data
showing that the rotational constants Bv1 = 137 MHz and Bv2 = 189 MHz associ-
ated to vibrational levels of the two structures are typical of motion in the inner
well. This can be deduced from the definition of the rotational constant as Bv =
h 2 /2μR 2 with μ = 121135.83 a.u., giving an estimate of the mean value of the
internuclear distance R̄ = h/(2μBv )−1/2 . For G1 and G2 structures, respectively,
FIG. 25. Details of the photoassociation spectra in the region of the (a) G1 and (b) G2 structures,
with their J labeling. Such structures are embedded in the vibrational progression of regular levels
(a) v = 102–105, (b) v = 79–81 with unresolved rotational structure. The rotational constants of the
giant structures G1 and G2 are obtained from the slopes of the lower curves giving the variation of the
binding energy as a function of J (J + 1).
this yields R̄ 1 ≈ 14a0 and R̄ 2 ≈ 12a0 , whereas the hump in the double-
well potential is located at a distance Rb ≈ 15a0 . The spectroscopy results could
be reproduced by theoretical calculations, and we show in Fig. 26 an example
of a vibrational wavefunction with tunneling. In the inner region, coupling with
vibrational motion in the potential curve 0− g (6s + 5d) has to be considered, so
that the function displays several Condon points favorable to spontaneous decay
toward low-lying bound levels of the a 3u+ Cs2 state. The large intensity of the ion
signal at the detunings corresponding to 1 and 2 binding energies could be an
FIG. 26. Tunneling effect: vibrational wavefunction extending both in the outer well of the 0− g (6s +
6 p 2 p3/2 ) potential curve of Cs2 and in the inner well. In the inner region, coupling with the 0−
g (6s + 5d)
vibrational series has to be considered.
indication of a good efficiency for this scheme, but better analysis of the detection
efficiency as a function of the vibrational number v 0 of the final molecule has to
be performed.
VI. New Schemes for Making Ultracold Molecules

Due to the complexity of potential curves in a given molecule, it seems that many
new schemes can be proposed owing to collaboration between theory and exper-
iment. If we stay within the present scheme of photoassociation with continuous
laser light, it seems wise to perform photoassociation at large internuclear dis-
tances and to look for situations where the vibrational motion is gradually stopped
at intermediate distance, providing opportunity to populate bound levels in the
ground-state potential curve by spontaneous emission. Besides potential barriers
with a gentle slope, which are not easily encountered, a more general mechanism is
offered by resonant coupling. In many situations, the excited-state potential curve
i is dynamically coupled to another one, i′ . Looking for resonances beweeen the
two vibrational series, one can transfer population from a level v i of the first series
to a level v i ′ of the other one. Dion et al. (2001) have recently shown that a large
number of ground-state Cs2 molecules are indeed observed in experiments where
a loosely bound level v i of the 0+ u (6s + 6 p1/2 ) curve is populated by photoassoci-
ation, and the population transferred to a level v i ′ of the narrower 0+u (6s + 6 p3/2 )
curve, which is has good Franck–Condon overlap with bound levels of the ground
state. This scheme is summarized in Fig. 27. Two-step photoassociation, with
transfer of population to bound levels of the lowest Na2 1u Rydberg excited
curve, has been discussed by Band and Julienne (1995), while Almazor et al.
(1999) proposed to populate a double-well curve in the Na2 (6) 1 g+ excited state.
Another possibility consists of considering photoassociation with pulsed lasers,
and trying to control the formation of molecules. This has been discussed by
several authors (Vardi et al., 1997; Vala et al., 2001), but no absolute rate has been
given up to now in the condition of the experiments. Vala et al. (2001) propose to
enhance the photoassociation rate by use of chirped pulses. In the work of Vardi
et al., two nanosecond pulses are used, one for photoassociation, another one for
population transfer by induced emission to a precise bound level v0 in the ground
excited curve. The estimation of 6 × 10−6 molecules formed per pulse and per atom
indicates that this process is much less efficient that the present schemes using CW
lasers. Half the atoms could be converted to molecules in ∼25 mn, whereas the CW
experiments predict times of the order of a few seconds. However, all long-lived
molecules would be produced in the same vibrational level v0 , which is not the
case in the present experimental schemes with CW lasers. The main advantage
of pulsed excitation is that a large number of levels can be photoassociated at the
same time, increasing at low detunings the flux to bound levels of the ground state
as discussed in Section V.B. We have shown in Section II.B that at low binding
energies all photoassociated vibrational levels have a similar wave function at short
and intermediate internuclear distance, and therefore similar spontaneous emission
probabilities toward levels of the ground state.
VII. Conclusion
While direct laser cooling of molecules is not at present a promising technique,
this chapter shows that an efficient way of making ultracold molecules is to start
from a sample laser-cooled atoms, and to implement a two-step process: first, a
photoassociation reaction creates, out of two atoms, a bound molecule in an ex-
cited electronic state, and second, this short-lived molecule decays by spontaneous
emission to bound levels of the ground electronic state (or lower triplet state).
The photoassociation process, first suggested by Thorsheim et al. (1987), uses
laser light slightly red-detuned with respect to an atomic resonance line, and can
be understood as a vertical (Franck–Condon) transition at large distance Rvi be-
tween a continuum level (two colliding atoms) and a bound level v in an ex-
cited electronic curve i of the molecule. The photoassociated molecule can be
viewed as a pair of distant atoms, weakly bound by dipole–dipole interaction: for
loosely bound vibrational levels, the two atoms are most of the time moving in the
asymptotic region of weak interaction, the outer classical turning point Rvi being a
decreasing function of the detuning. The dynamics is similar to that of a Rydberg

electron, the binding energy of a level being linked to its vibrational number by the
Le Roy–Berstein formula generalizing the Rydberg formula. Scaling laws are
found, showing that the photoassociation process is favored at very small detun-
ings, corresponding to vertical transitions at large distance. In such conditions, the
photoassociated molecule usually decays back into a pair of ground-state atoms.
Indeed, the ground-state potential curve, with asymptotic R−6 behavior, is much
narrower than the excited curve, and bound levels can be efficiently populated by
spontaneous emission only when the nuclei are at short or intermediate distance. In
most cases, the vibrational motion of the photoassociated molecule is such that the
amount of time spent in the short-range region is too small, making the branching
ratio between stabilization and dissociation negligible.
However, experiments by Fioretti et al. (1998, 1999), using photoassociation
in a cesium sample and detection of molecular ions by a REMPI scheme, have
demonstrated for some symmetries (0− g , 1u ) of the photoassociated molecule, an
efficient mechanism for population of bound levels in the ground (or lower-triplet)-
state potential curve. This can be explained by the double-well structure of the 0− g
and 1u curves, where a barrier at intermediate distance is gently slowing down
the vibrational motion, increasing the time spent at distances where spontaneous
emission can efficiently populate a bound level of a ground (or lower triplet) state
of the molecule. A similar mechanism exists for rubidium, and was demonstrated
by Gabbanini et al. (2000). At present, in a cesium trap containing 50 million atoms,
at a density nat = 1011 cm−3 and temperature T = 140 μK, using continuous pho-
toassociation lasers with intensity ≤1 kW/cm2, it is possible to create from 105 to
106 molecules per second, that is, to transform 10% of the atoms into molecules in
a time of a few seconds. Similar results were obtained by Nikolov et al. (2000) for
K2 using another mechanism where the photoassociated molecule is excited by a
second laser to a Rydberg state which decays by spontaneous emission to bound
levels of the ground electronic state. Presently, we should note that only a few
molecules (∼5%) thus formed are detected. The sensitivity of the REMPI detec-
tion should be improved, and the ionization step optimized, in collaboration with
molecular spectroscopy groups. Once most molecules are detected, the formation
of ultracold molecules will appear as a very strong effect.
The interpretation of experiments and design of new schemes relies on close
collaboration between theory and experiment. New theoretical methods have been
developed to treat ultracold molecules, and their comparison with more traditional
methods has to be implemented. There is good agreement between measured pho-
toassociation rates and theoretical calculations performed in the framework of a
perturbative treatment. An important result is the rapid decrease of the photo-
association rate as a function of detuning, interpreted by scaling laws, and by
the behavior of the initial continuum-state wavefunction, which has a reduced
probability amplitude at small internuclear distances. The photoassociation pro-
cess is thus more efficient at low detuning. In contrast, the branching ratio for
stabilization into bound levels by the spontaneous emission process is decreasing

at low detuning, due to a smaller amplitude of the vibrational wavefunctions of
the photoassociated molecule in the intermediate distance region where to bound
levels is possible. Present experiments have to choose a compromise between the
efficiency of the photoassociation step and the efficiency of the stabilization step.
Some new schemes that rely on tunneling effect, resonant coupling, or two-step
excitation were also shown to be efficient, and must be further explored.
The photoassociation rate is presently estimated in the framework of a per-
turbative treatment, and comparison with more sophisticated theoretical studies
using time-dependent formalism still has to be implemented. Photoassociation us-
ing pulsed lasers has been little explored as yet experimentally (Boesten et al.,
1996, 1998) but seems a promising route. The conclusions of the present chap-
ter concern photoassociation rates for population of individual vibrational levels.
Different conclusions would be drawn if broad-band excitation was used in the
photoassociation process, as the population of a large number of levels, with sim-
ilar wavefunctions in the short- and intermediate-distance region, would increase
the efficiency of the stabilization process. The discussion of pulsed laser experi-
ments connects the subject of cold-molecule formation to the general problem of
coherent control.
Current results could be extended to all homonuclear systems provided a dense
enough sample can be obtained. Formation of heteronuclear molecules, although
presenting smaller photoassociation rates, would have the advantage of making
the trapping process easy with a dipole trap. Indeed, at present the molecules stay
for roughly 10 s inside the trap, and then fall down. The next important step is to
trap them. A CO2 laser, used in the experiments of Takekoshi et al. (1998, 1999)
seems promissing. Magnetic trapping has been used for triplet molecules, and first
demonstration has been achieved recently in Orsay. The rotational temperature of
the ultracold molecules is low, but they are formed in a large number of vibra-
tional levels, so a Raman scheme to lower the vibrational temperature should be
implemented. Coherent control techniques (Bartana et al., 1993, 1997) also offer
interesting possibilities.
The optimization of the rate of molecule formation should be searched for in
an increase of the atomic density rather than an increase of the laser intensity,
which may lead to saturation or reexcitation. We are reaching there the field of
atomic condensates, which is beyond the scope of the present review but offers
very interesting opportunities: the choice between two routes toward molecular
condensates, either creating molecules within an atomic condensate (Wynar et al.,
2000) or bringing a dense ensemble of ultracold molecules to conditions where
Bose condensation can take place, is an interesting perspective. The molecule laser
would be an important application.
Other applications lie in threshold laws for collisions (atom–molecule, Rydberg
atom–molecule, molecule–molecule collisions), very little explored at present
(Forrey et al., 1998, 1999a,b), on the formation of more complex species (trimers
small clusters). Ultracold molecular sources would be ideal for molecular inter-
ferometry experiments (Bordé et al., 1994; Chapman et al., 1995; Lisdaf et al.,
2000). In short, a new research field is open.
VIII. Acknowledgments
This review relies on work done within the team Molécules et atomes froids in
Laboratoire Aimé Cotton, so we are very indebted to Ph.D. students Cyril Drag,
Bruno Laburthe Tolra, Salah Boussen, and Nathalie Hoang for experiments, Marie-
Laure Almazor, Viatcheslav Kokoouline, Philippe Pellegrini, Benoit T’Jampens,
and Mihaela Vatasescu for theory, to former or present postdocs Andrea Fioretti,
Wilson de Souza Melo, Ricardo Gutteres, and Claude Dion, and to our young or
senior colleagues Daniel Comparat, Olivier Dulieu, Samuel Guibal, Claude Amiot,
and Anne Crubellier. Special thanks are due to Olivier Dulieu and Claude Dion
for their help in preparing the manuscript and figures. Stimulating collaboration
is gratefully acknowledged with R. Kosloff in Jerusalem, (owing to the Arc en
Ciel-Keshet cooperation program and to invitation of F. M-S by the Fritz Haber
Institute where part of the paper has been written), with E. Tiemann, University of
Hannover, (within the Procope Franco-German cooperation program), and with the
group of W. Stwalley in Storrs, University of Connecticut, (within CNRS-NSF co-
operation). We thank T. F. Gallagher (Virginia), G. Pichler (Zagreb), E. Luc-Koenig
(Orsay), V. N. Ostrovsky (St. Petersburg) for collaboration. Finally, we are grateful
to M. Allegrini, R. Côté, A. Dalgarno, J. Dalibard, C. Gabbanini, Ph. Gould, P.
Lett, J. Pinard, L. Pruvost, C. Salomon, and C. Williams for useful discussions.
IX. References
Abraham, E. R. I., Ritchie, N. W. M., McAlexander, W. I., and Hulet, R. G. (1995). Photoassocia-
tive spectroscopy of long-range states of ultracold 6Li2 and 7Li2. J. Chem. Phys. 103, 7773–
7778.
Almazor, M.-L., Dulieu, O., Elbs, M., Tiemann, E., and Masnou-Seeuws, F. (1999). How to get access
to long range states of highly excited molecules. Eur. Phys. J. D 5, 237–242.
Aubert-Frécon, M., Hadinger, G., Magnier, S., and Rousseau, S. (1998). Analytical formulas for long
(+/−)
range energies of the 16 g,u states of alkali dimers dissociating into M(ns) + M(np). J. Mol.
Spectrosc. 188, 182–189.
Avila, G., Gain, P., deClercq, E., and Cerez, P. (1986). New absolute wavenumber measurement of the
D2 line of caesium. Metrologia 22, 11–14.
Aymar, M., Greene, C. H., and Luc-Koenig, E. (1996). Multichannel Rydberg spectroscopy of complex
atoms. Rev. Mod. Phys. 68, 1015–1124.
Bahns, J. T., Gould, P. L., and Stwalley, W. (1997). Laser cooling of molecules: A sequential scheme
for rotation, translation and vibration. J. Chem. Phys. 104, 9689.
Bahns, J. T., Stwalley, W. C., and Gould, P. L. (2000). Formation of cold (T ≤ 1 K) molecules. Adv.
Atomic Mol. Opt. Phys. 42, 171–224.
Band, Y. B., and Julienne, P. S. (1995). Ultracold-molecule production by laser-cooled atom photoas-
sociation. Phys. Rev. A 51, R4317–R4320.
Bartana, A., Kosloff, R., and Tannor, D. J. (1993). Laser cooling of molecular internal degrees of
freedom by a series of shaped pulses. J. Chem. Phys. 99, 196–210.
Bartana, A., Kosloff, R., and Tannor, D. J. (1997). Laser cooling of internal degrees of freedom II.
J. Chem. Phys. 106, 1435–1448.
Bethlem, H. L., Berden, G., and Meijer, G. (1999). Decelerating neutral dipolar molecules. Phys. Rev.
Lett. 83, 1558–1561.
Bethlem, H. L., Berden, G., Crompvoets, F. M. H., Jongma, R. T., van Roij, A. J. A., and Meijer, G.
(2000a). Electrostatic trapping of ammonia molecules. Nature 406, 491–494.
Bethlem, H. L., Berden, G., van Roij, A. J. A., Crompvoets, F. M. H., and Meijer, G. (2000b). Trapping
neutral molecules in a traveling potential well. Phys. Rev. Lett. 84, 5744–5747.
Beuc, R., Milosevic, S., Movre, M., Pichler, G., and Veza, D. (1982). Satellite bands in the far blue
wing of the potassium first resonance doublet. Fizika (Yu) 14, 345–349.
Blatt, J. M. (1967). J. Comput. Phys. 1, 382.
Boesten, H., Tsai, C. C., Verhaar, B. J., and Heinzen, D. J. (1996). Observation of a shape resonance in
cold-atom scattering by pulsed photoassociation. Phys. Rev. Lett. 77, 5194.
Boesten, H., Tsai, C. C., Heinzen, D. J., moonen, A. J., and Verhaar, B. J. (1999). Time-independent
and time-dependent photoassociation of rubidium. J. Phys. B 32, 287–308.
Bohn, J. L., and Julienne, P. S. (1996). Semianalytic treatment of two-color photoassociation spec-
troscopy and control of cold atoms. Phys. Rev. A 54, R4637–R4640.
Boisseau, C., Audouard, E., and Vigué, J. (1998). Quantization of the highest levels in a molecular
potential. Europhys. Lett. 41, 349–354.
Boisseau, C., Audouard, E., Vigué, J., and Flanbaum, V. V. (2000a). Analytical corrections to the
JWKB quantization conditions for the highest levels in a molecular potential. Eur. Phys. J. D 12,
199–209.
Boisseau, C., Audouard, E., Vigué, J., and Julienne, P. S. (2000b). Reflection approximation in pho-
toassociation spectroscopy. Phys. Rev. A 62, 052705.
Bordé, C. J., Courtier, N., du Burck, L., Goncharov, A. N., and Gorlicki, M. (1994). Molecular inter-
ferometry experiments. Phys. Lett. A188, 187–190.
Bussery, B., and Aubert-Frécon, M. (1985a). Multipolar long-range electrostatic, dispersion and in-
duction energy terms for the interaction between two identical alkali atoms Li, Na, K, Rb and Cs
in various electronic states. J. Chem. Phys. 82, 3224–3234.
Bussery, B., and Aubert-Frécon, M. (1985b). Potential energy curves and vibration-rotation energies
for the two purely long-range bound states 1u and 0− g of the alkali dimers M2 dissociating to
M(ns2S1/2) + M(np2P3/2) with M = Li, Na, K, Rb, and Cs. J. Mol. Spectrosc. 113, 21–27.
Chapman, M. S., Ekstrom, C. R., Hammond, T. D., Rubinstein, R. A., Schiedmayer, J., Wehinger, S.,
and Pritchard, D. E. (1995). Optics and interferometry with Na2 molecules. Phys. Rev. Lett. 74,
4783–4786.
Child, M. S. (1991). “Semiclassical Mechanics with Molecular Applications.” Clarendon Press, Oxford.
Cline, R. A., Miller, J. D., and Heinzen, D. J. (1994). Study of Rb2 long-range states by high resolution
photoassociation spectroscopy. Phys. Rev. Lett. 73, 632–635.
Colbert, D. T., and Miller, W. H. (1992). A novel discrete variable representation for quantum mechani-
cal reactive scattering via the S-matrix Kohn method. J. Chem. Phys. 96, 1982–1991.
Comparat, D., Drag, C., Fioretti, A., Dulieu, O., and Pillet, P. (1999). Photoassociative spectroscopy
and formation of cold molecules in cold cesium vapor: Trap-loss spectrum versus ion spectrum.
J. Mol. Spectrosc. 195, 229–235.
Comparat, D., Drag, C., Laburthe Tolra, B., Pillet, P., Crubellier, A., Dulieu, O., and Masnou-Seeuws,
F. (2000). Formation of cold Cs2 ground state molecules through photoassociation in the 1− u long
range state. Eur. Phys. J. D 11, 59–71.
Condon, E. U. (1926). A theory of intensity distribution in band systems. Phys. Rev. 28, 1182–1201.
Condon, E. (1928). Nuclear motions associated with electron transitions in diatomic molecules. Phys.
Rev. 32, 858–872.
Cooley, J. W. (1961). An improved eigenvalue corrector formula for solving the Schrödinger equation
for central fields. Math. Comput. 15, 363.
Côté, R., and Dalgarno, A. (1997). Mechanism for the production of vibrationally excited ultracold
molecules of 7Li2. Chem. Phys. Lett. 279, 50–54.
Côté, R., and Dalgarno, A. (1998). Photoassociation intensities and radiative trap loss in lithium. Phys.
Rev. A 58, 498–508.
Côté, R., and Dalgarno, A. (1999). Mechanism for the production of 6Li2 and 7Li2 utracold molecules.
Côté, R., and Jamieson, M. J. (1995). comparison of phaseshift calculations by asymptotic fit and
quadrature in ultra-low temperature scattering. J. Comput. Phys. 118, 388–391.
Côté, R., Dalgarno, A., Sun, Y., and Hulet, R. G. (1995). Photoabsorption by ultra-cold atoms and the
scattering length. Phys. Rev. Lett. 74, 3581–3583.
Dashevskaya, E., Voronin, A. I., and Nikitin, E. E. (1969). Theory of excitation transfer between
identical alkali atoms. I. Identical partners. Can. J. Phys. 47, 1238–1248.
deCarvalho, R., Doyle, J. M., Friedrich, B., Guillet, T., Kim, J., Patterson, D., and Weinstein, J. D.
(1999). Buffer-gas loaded magnetic traps for atoms and molecules: A primer. Eur. Phys. J. D 7,
289–309.
Derevianko, A., W. R, J., Safronova, M. S., and Babb, J. F. (1999). High-precision calculations of
dispersion coefficients, static dipole polarizabilities and atom-wall interaction constants for alkali-
metal atoms. Phys. Rev. Lett. 3589–3592.
Dion, C. M., Drag, C., Dulieu, O., Laburthe Tolra, B., Masnou-Seeuws, F., and Pillet, P. (2001). Resonant
coupling in the formation of ultracold ground state molecules via photoassociation. Phys. Rev.
Lett. 86, 2252–2254.
Djeu, N., and Whitney, W. T. (1981). Laser cooling by spontaneous anti-Stokes scattering. Phys. Rev.
Lett. 46, 236.
Doyle, J. M., Friedrich, B., Kim, J., and Patterson, D. (1995). Buffer-gas loading of atoms and molecules
into a magnetic trap. Phys. Rev. A 52, R2515–R2518.
Doyle, R. (1968). The continuous energy spectrum of the hydrogen quasi-molecule. J. Quant. Spectrosc.
Radiat. Transfer 8, 1555–1569.
Drag, C. (2000). Photoassociation d’atomes de césium froids. Formation et caractérisation d’un
nuage froid de molécules diatomiques de césium. Ph.D. thesis, Université Paris-Sud, Centre
d’Orsay.
Drag, C., Laburthe Tolra, B., Dulieu, O., Comparat, D., Vatasescu, M., Boussen, S., Guibal, S., Crubel-
lier, A., and Pillet, P. (2000a). Experimental versus theoretical rates for photoassociation and
formation of cold molecules. IEE Quantum Electron. 36, 1378–1388.
Drag, C., Tolra, B. L., T’Jampens, B., Allegrini, M., Crubellier, A., and Pillet, P. (2000b). Photoas-
sociative spectroscopy as a self-sufficient tool for the determination of the Cs triplet scattering
length. Phys. Rev. Lett. 85, 1408–1411.
Dulieu, O., and Julienne, P. S. (1995). Coupled channel bound states calculations for alkali dimers
using the Fourier grid method. J. Chem. Phys. 103, 60–66.
Dulieu, O., Kosloff, R., Masnou-Seeuws, F., and Pichler, G. (1997). Intermediate long range molecules:
Bound and quasibound states for alkali dimers. J. Chem. Phys. 107, 10633–10642.
Fatemi, F. K., Jones, K. M., and Lett, P. D. (2000). Observation of optically induced Feschbach reso-
nances in collisions of cold atoms. Phys. Rev. Lett. 85, 4462–4465.
Fattal, E., Baer, R., and Kosloff, R. (1996). Phase space approach for optimizing grid representations:
The mapped Fourier method. Phys. Rev. E 53, 1217–1227.
Fioretti, A., Comparat, D., Crubellier, A., Dulieu, O., Masnou-Seeuws, F., and Pillet, P. (1998).
Formation of cold Cs2 molecules through photoassociation. Phys. Rev. Lett. 80, 4402–
4405.
Fioretti, A., Comparat, D., Drag, C. Amiot, C., Dulieu, O., Masnou-Seeuws, F., and Pillet, P. (1999).
Photoassociative spectroscopy of the Cs2 0− g long range state. Eur. Phys. J. D 5, 389–403.
Fioretti, A., Amiot, C., Dulieu, O., Mazzoni, M., and Gabbanini, C. (2001). Cold rubidium molecule
formation through photoassociation: A spectroscopic study of the 0− g long range state of Rb2. Eur.
Phys. J. D. (2001) (in press).
Forrey, R. C., Balakrishnan, N., Karchenko, V., and Dalgarno, A. (1998). Feshbach resonances in
ultracold atom-diatom scattering. Phys. Rev. A 58, R2645–R2647.
Forrey, R. C., Balakrishnan, N., Dalgarno, A., Haggerty, M. R., and Heller, E. J. (1999a). Quasiresonant
energy transfer in ultracold atom-diatom collisions. Phys. Rev. Lett. 82, 2657–2660.
Forrey, R. C., Karchenko, V., Balakrishnan, N., and Dalgarno, A. (1999b). Vibrational relaxation of
trapped molecules. Phys. Rev. A 59, 2146–2152.
Foucrault, M., Millié, P., and Daudey, J. P. (1992). Non-perturbative method for core-valence correlation
in pseudopotential calculations: Application to the Rb2 and Cs2 molecules. J. Chem. Phys. 96,
1257–1264.
Franck, J. (1925). Trans. Faraday Soc. 21, 536.
Friedrich, H. (1998). “Theoretical Atomic Physics.” Springer-Verlag, Berlin.
Friedrich, H., and Trost, J. (1996a). Phase loss in WKB waves due to reflection by a potential. Phys.
Rev. Lett. 76, 4469–4473.
Friedrich, H., and Trost, J. (1996b). Non integral Maslov indexes. Phys. Rev. A 54, 1136–1145.
Friedrich, B., deCarvalho, R., Kim, J., Patterson, D., Weinstein, J. D., and Doyle, J. M. (1998). Towards
magnetic trapping of molecules. J. Chem. Soc. Faraday Trans. 94, 1783–1791.
Friedrich, B., Weinstein, J. D., deCarvalho, R., and Doyle, J. M. (1999). Zeeman spectroscopy of CaH
molecules in a magnetic trap. J. Chem. Phys. 110, 2376–2383.
Gabbanini, C., Fioretti, A., Lucchesini, A., Gozzini, S., and Mazzoni, M. (2000). Observation of
translationally cold ground state rubidium molecules. Phys. Rev. Lett. 84, 2814–2817.
Gao, B. (2000). Breakdown of Bohr’s correspondence principle. Phys. Rev. Lett. 83, 4225–4228.
Gardner, J., Cline, R. A., Miller, J. D., Heinzen, D. J., Boesten, H. J. M., and Verhaar, B. J. (1995).
Collisions of doubly spin-polarized, ultracold 85Rb atoms. Phys. Rev. Lett. 74, 3764–3767.
Gerstenkorn, S., Vergès, J., and Chevillard, J. (1982). Atlas du spectre d’absorption de la molécule
d’iode. Tech. Rep., Laboratoire Aimé Cotton, Orsay, France.
Gordon, R. G. (1969). New methods for constructing wavefunctions for bound states and scattering.
J. Chem. Phys. 51, 14–25.
Gouédard, G., Girard, B., Billy, N., and Vigué, J. (1989). ωl scaling law near a molecular dissociation
limit: Theory and experimental tests in Cl2 B state. Chem. Phys. 132, 385–390.
Gray, S. K., and Rice, S. A. (1985). A scattering resonance description of very low energy collision
induced vibrational relaxation. J. Chem. Phys. 83, 2818–2828.
Gribakin, G. F., and Flanbaum, V. V. (1993). Calculation of the scattering length using the semiclassical
approximation. Phys. Rev. A 48, 546–553.
Herman, A., Leutwyler, S., Woste, L., and Schumacher, E. (1979). Molecular spectroscopy by pho-
todeflection of Na2 in a supersonic beam. Chem. Phys. Lett. 62, 44–47.
Herschbach, N., Tol, P. J. J., Wassen, W., Hogervorst, W., Woestenenk, G., Thomsen, J. W., der Straten,
P. V., and Niehaus, A. (2000). Photoassociation spectroscopy of cold He(23s) atoms. Phys. Rev.
Lett. 84, 1874–1877.
Hinds, E. A., and Sangster, K. (1992). Testing time-reversal symmetry with molecules. AIP Conf. Proc.
270, 77–83.
Hoffmann, D., Feng, P., and Walker, T. (1994). Measurements of Rb trap-loss collision spectra. J. Opt.
Soc. Am. B 11, 712–720.
Jablonski, A. (1945). General theory of pressure broadening in spectral lines. Phys. Rev. A 68, 78–93.
James, H. M., and Coolidge, A. S. (1939). Continuous spectra of H2 and D2. Phys. Rev. 55, 84.
Javanainen, J., and Mackie, M. (1998). Probability of photoassociation from a quasi-continuum ap-
proach. Phys. Rev. A 58, R789–R792.
Julienne, P. S. (1996). Cold binary collisions in a light field. J. Res. Natl. Inst. Std. Technol. 101,
487–503.
Julienne, P. S., and Vigué, J. (1991). Cold collisions of ground- and excited-state alkali metal atoms.
Phys. Rev. A 44, 4464–4485.
Julienne, P., Burnett, K., Band, Y., and Stwalley, W. (1998). Stimulated Raman molecule production
in Bose-Einstein condensation. Phys. Rev. A 58, R797–R800.
Kastler, A. (1950). Quelques suggestions concernant la production optique et la détection optique
d’une inégalité de population des niveaux de quantification spatiale des atomes. application à
l’expérience de Stern et Gerlach et à la résonance magnétique. J. Phys. France 11, 255.
Ketterle, W., Davies, K. B., Joffe, M. A., Martins, A., and Pritchard, D. E. (1993). High density of cold
atoms in a dark spontaneous-force optical trap. Phys. Rev. Lett. 70, 2253–2256.
Kokoouline, V., Dulieu, O., Kosloff, R., and Masnou-Seeuws, F. (1999). Mapped Fourier methods for
long range molecules: Application to perturbations in the Rb2(0+ u ) spectrum. J. Chem. Phys. 110,
9865–9876.
Kokoouline, V., Dulieu, O., Kosloff, R., and Masnou-Seeuws, F. (2000a). Theoretical treatment of
channel mixing in excited Rb2 and Cs2 ultra-cold molecules: Determination of predissociation
lifetimes with coordinate mapping. Phys. Rev. A 62, 032716.
Kokoouline, V., Dulieu, O., and Masnou-Seeuws, F. (2000b). Theoretical treatment of channel mixing in
excited Rb2 and Cs2 ultra-cold molecules. Perturbations in 0+ u photoassociation and fluorescence
spectra. Phys. Rev. A 62, 022504.
Kosloff, R. (1988). Time dependent methods in molecular dynamics. J. Phys. Chem. 92, 2087–2100.
Kosloff, R. (1996). Quantum molecular dynamics on grids. In “Dynamics of Molecules and Chemical
Reactions” (R. H. Wyatt and J. Z. H. Zhang, Eds.), p. 185. Marcel Dekker, New York.
Koŝtrun, M., Mackie, M., Côté, R., and Javanainen, J. (2000). Theory of coherent photoassociation of
a Bose-Einstein condensate. Phys. Rev. A62, 063616.
Kozlov, M. G., and Ezhov, V. F. (1994). Enhancement of the electric dipole moment of the electron in
the bf molecule. Phys. Rev. A 49, 4502–4507.
Labiorthe-Tobra, B., Drag, C., and Pillet, P. (2001). Observation of cold state-selected cesium molecules
formed by stimulated Raman photoassociation, PRA in press.
Lambrecht, A., Giacobino, E., and Reynaud, S. (1996). Atomic number fluctuations in a falling cold
atomic cloud. Quantum Semiclass. 8, 457.
Landau, L. D., and Lifshitz, E. M. (1977). “Quantum Mechanics (Non-Relativistic Theory)”. Pergamon,
Oxford.
Le Roy, B. (1980). Applications of Bohr quantization in diatomic molecule spectroscopy in “Semi-
classical methods in molecular scattering and spectroscopy” M. S. Child editor, 1980 Reidel Publ.
Dordredt, Holland.
Le Roy, R. J. (2000). A computer program for solving the radial Schrödinger equation for bound and
quasi-bound levels. Univ. Waterloo Chem. Phys. Res. Rep. CP 642 R 1.
Le Roy, R. J., and Bernstein, R. B. (1970). Dissociation energy and long-range potential of diatomic
molecules from vibrational spacings of higher levels. J. Chem. Phys. 52, 3869–3879.
Lefebvre-Brion, H., and Field, R. W. (1986). “Perturbations in the Spectra of Diatomic Molecules.”
Academic Press, New York.
Lett, P. D., Helmerson, K., Philips, W. D., Ratliff, L. P., Rolston, S. L., and Wagshul, M. E.
(1993). Spectroscopy of Na2 by photoassociation of laser-cooled Na. Phys. Rev. Lett. 71, 2200–
2203.
Lisdat, C., Franck, M., Knöckel, H., Almazor, M.-L., and Tiemann, E. (2000). Realization of a Ramsey-
Bordé matter wave interferometer on the K2 molecule. Eur. Phys. J. D 12, 235–240.
Mackholm, M., Giusti-Suzor, A., and Mies, F. H. (1994). Photoassociation of atoms in ultracold
collisions probed by wavepacket dynamics. Phys. Rev. A 50, 5025–5036.
Mackie, M., and Javanainen, J. (1999). Quasi-continuum modeling of photoassociation. Phys. Rev. A
60, 3174–3187.
Magnier, S., and Millié, P. (1996). Potential curves for the ground and numerous excited electronic
states of the K2 and NaK molecule. Phys. Rev. A 54, 204–218.
Magnier, S., Millié, P., Dulieu, O., and Masnou-Seeuws, F. (1993). Potential curves for the ground
and excited states of the Na2 molecule up to the (3s + 5p) dissociation limit. J. Chem. Phys. 98,
7113–7125.
Marinescu, M., and Dalgarno, A. (1995). Dispersion forces and long-range electronic transition dipole
moments of alkali-metal dimer excited states. Phys. Rev. A 52, 311–320.
Marinescu, M., and Dalgarno, A. (1996). Analytical interaction potentials of the long range alkali-metal
dimers. Z. Phys. D 36, 239–248.
Marston, C. C., and Balint-Kurti, G. (1989). The Fourier grid Hamiltonian method for bound state
eigenvalues and eigenfunctions. J. Chem. Phys. 91, 3571–3576.
Mies, F. H. (1973). Molecular theory of atomic collisions: Calculated croos sections for H+ + F(2P).
Phys. Rev. A 7, 957–967.
Miller, J. D., Cline, R. A., and Heinzen, D. J. (1993). Photoassociation spectrum of ultracold Rb atoms.
Phys. Rev. Lett. 71, 2204–2207.
Miller, W. H. (1968). Uniform semiclassical approximations for elastic scattering and eigenvalue
problems. J. Chem. Phys. 48, 464–467.
Monnerville, M., and Robbe, J. M. (1994). Optical potential coupled to discrete variable representation
for calculations of quasibound states: Application to the CO(B 1 + − D′ 1 +) predissociating
interaction. J. Chem. Phys. 101, 7580–7581.
Monroe, C., Swann, W., Robinson, H., and Wieman, C. (1990). Very cold trapped atoms in a vapor
cell. Phys. Rev. Lett. 65, 1571–1574.
Mosk, A. P., Reynolds, M. W., Hijmans, T. W., and Walraven, J. T. M. (1999). Photoassociation of
spin-polarized hydrogen. Phys. Rev. Lett. 82, 307–310.
Movre, M., and Pichler, G. (1977). Resonance interaction and self-broadening of alkali resonances
lines I: Adiabatic potential curves. J. Phys. B: At. Mol. Opt. Phys. 10, 2631–2638.
Napolitano, R., Weiner, J., Williams, C. J., and Julienne, P. S. (1994). Line shapes of high resolution
photoassociation spectra of optically cooled atoms. Phys. Rev. Lett. 73, 1352–1355.
Nikolov, A. N., Eyler, E. E., Wang, X., Li, J., Wang, H., Stwalley, W. C., and Gould, P. (1999).
Observation of translationally ultracold ground state potassium molecules. Phys. Rev. Lett. 82,
703–706.
Nikolov, A. N., Ensher, J. R., Eyler, E. E., Wang, H., Stwalley, W. C., and Gould, P. (2000). Efficient
production of ground-state potassium molecules at sub-mK temperatures by two-step photoas-
sociation. Phys. Rev. Lett. 84, 246–249.
Ostrovsky, V. N., Kokoouline, V., Luc-Koenig, E., and Masnou-Seeuws, F. (2001). J. Phys. B: At. Mol.
Opt. Phys. 34, L27–L38.
Pichler, G., Milosevic, S., Veza, D., and Beuc, R. (1983). Diffuse bands in the visible absorption spectra
of dense alkali vapors. J. Phys. B: Atomic Mol. Opt. Phys. 16, 4619–4631.
Pillet, P., Crubellier, A., Bleton, A., Dulieu, O., Nosbaum, P., Mourachko, I., and Masnou-Seeuws, F.
(1997). Photoassociation in a gas of cold alkali atoms. I: Perturbative quantum approach. J. Phys.
B 30, 2801–2820.
Raab, E. L., Prentiss, M., Cable, A., Chu, S., and Pritchard, D. E. (1987). Trapping of neutral sodium
atoms with radiation pressure. Phys. Rev. Lett. 59, 2631.
Ratliff, L. P., Wagshul, M. E., Lett, P. D., Rolston, S. L., and Phillips, W. D. (1994). Photoassociative
ionization of the 1g , 0+ −
u and 0g states of Na2. J. Chem. Phys. 101, 2638–2641.
Rawitscher, G. H., Esry, B. D., Tiesinga, E., J. P. Burke, J., and Koltracht, I. (1999). Comparison of
numerical methods for the calculation of cold atom collisions. J. Chem. Phys. 111, 10418–10426.
Sauer, B. E., Wang, J., and Hinds, E. A. (1994). YbF—A new way to measure the electron dipole
moment. Bull. Am. Phys. Soc. Ser. H. 39, 1060.
Seaton, M. J. (1983). Quantum defect theory. Rep. Prog. Phys. 46, 167–257.
Seto, J., Vergès, L. J., and Amiot, C. (1995). Direct potential fit analysis of the X1 g+ state of Rb2:
Nothing else will do. J. Chem. Phys. 113, 3067–3076.
Shaffer, J. P., Chalupezak, W., and Bigelow, N. P. (1999). Photoassociative ionization of heteronuclear
molecules in a novel two-species magneto-optical trap. Phys. Rev. Lett. 82, 1124–1127.
Spies, N. (1989). Ph. D. thesis. Fachbereich Chemie, Universitat Kaiserslautern.
Stwalley, W. C. (1973). Expectation values of the kinetic and potential energy of a diatomic molecule.
J. Chem. Phys. 58, 3867–3870.
Stwalley, W., and Wang, H. (1999). Photoassociation of ultracold atoms: A new spectroscopic technique.
Stwalley, W., Uang, Y., and Pichler, G. (1978). Pure long-range molecules. Phys. Rev. Lett. 41, 1164–
1167.
Takekoshi, T., Patterson, B. M., and Knize, R. J. (1999a). Observation of cold ground state cesium
molecules produced in a magneto-optical trap. Phys. Rev. A 59, R5–R7.
Takekoshi, T., Patterson, B. M., and Knize, R. J. (1999b). Observation of optically trapped cold cesium
molecules. Phys. Rev. Lett. 81, 5105–5108.
Tellinghuisen, J. (1985). The Franck–Condon principle in bound-free transitions. In “Photodissociation
and Photoionization”. (K. P. Lawley, Ed.), pp. 299–369. John Wiley and Sons Ltd. Advances in
Chemical Physics 60, 299–369.
Thorsheim, H. R., Weiner, J., and Julienne, P. S. (1987). Laser-induced photoassociation of ultracold
sodium atoms. Phys. Rev. Lett. 58, 2420.
Tiesinga, E., Williams, C. J., and Julienne, P. S. (1996). A spectroscopic determination of scattering
lengths for sodium atom collisions. J. Res. Natl. Inst. Std. Technol. 101, 505–520.
Tiesinga, E., Williams, C. J., and Julienne, P. S. (1998). Photoassociative spectroscopy of highly excited
vibrational levels of alkali dimers: Green’s function approach for eigenvalues solvers. Phys. Rev.
A 57, 4257–4267.
Timmermans, E., Tommasini, P., Côté, R., Hussein, M., and Kerman, A. (1999a). Rarefied liquid
properties of hybrid atomic-molecular Bose-Einstein condensates. Phys. Rev. Lett. 88, 2691–94.
199–230.
Timmermans, E., Tommasini, P., Hussein, M., and Kerman, A. (1999b). Feshbach resonances in atomic
Bose-Einstein condensates. Phys. Rep. 315, 199–230.
Townsend, C., Edwards, N., Cooper, C., Zetie, K., Foot, C., Steane, A., Szriftgiser, P., Perrin, H.,
and Dalibard, J. (1995). Phase-space density in the magneto-optical trap. Phys. Rev. A 52, 1423–
1440.
Townsend, C. G., Edwards, N. H., Zetie, K. P., Cooper, C. J., and Foot, C. J. (1996). High density
trapping of cesium atoms in a dark magneto-optical trap. Phys. Rev. A 53, 1702–1714.
Trost, J., and Friedrich, H. (1997). WKB and exact wave functions for inverse power-law potentials.
Phys. Lett. A228, 127–133.
Trost, J., Eltschka, C., and Friedrich, H. (1998). Quantization in molecular potentials. J. Phys. B 31,
361–374.
Vala, J., Dulieu, O., Masnou-Seeuws, F., Pillet, P., and Kosloff, R. (2001). Coherent control of cold
molecule formation through photoassociation using chirped pulsed laser field. Phys. Rev. A 63,
013412.
Vardi, A., Abrashkevich, D., Frishman, E., and Shapiro, M. (1997). Theory of radiative recombi-
nation with strong laser pulses and the formation of ultracold molecules via stimulated photo-
recombination of cold atoms. J. Chem. Phys. 107, 6166–6174.
Vatasescu, M., Dulieu, O., Amiot, C., Comparat, D., Drag, C., Kokoouline, V., Masnou-Seeuws, F., and
Pillet, P. (2000). Multichannel tunneling in the Cs2 0− g photoassociation spectrum. Phys. Rev. A
61, 044701-1-4.
Vatasescu, M., Dulieu, O., Kosloff, R., and Masnou-Seeuws, F. (2001). Toward optimal control of
photoassociation of cold atoms and photodissociation of long range molecules: Characteristic
times for wavepacket propagation. Phys. Rev. A 63, 033407-1-13.
Veza, D., Beuc, R., Milosevic, S., and Pichler, G. (1998). Cusp satellite bands in the spectrum of Cs2
molecule. Eur. Phys. J. D 2, 45–52.
Vigué, J. (1982). Semiclassical approximation applied to the vibration of diatomic molecules. Ann.
Phys. Fr 3, 155–192.
Volz, U., and Schomoranzer, H. (1996). Precision lifetime measurements on alkali atoms and on helium
by beam gas laser spectroscopy. Physica Scripta 65, 48–56.
Wang, H., Gould, P. L., and Stwalley, W. (1996). Photoassociative spectroscopy of ultracold 39K atoms
in a high density vapor-cell magneto-optical trap. Phys. Rev. A 53, R1216–R1219.
Wang, H., Gould, P. L., and Stwalley, W. (1997). Long-range interaction of the 39K(4s)+39K(4p)
asymptote by photoassociativespectroscopy. I. the 0− g pure long-range state and the long-range
potential constants. J. Chem. Phys. 106, 7899–7912.
Wang, H., and Stwalley, W. (1998). Ultracold photoassociative spectroscopy of heteronuclear alkali-
metal diatomic molecules. J. Chem. Phys. 108, 5767–71.
Wang, X., Wang, H., Gould, P. L., Stwalley, W., Tiesinga, E., and Julienne, P. S. (1998). First observation
of the pure long range 1u state of an alkali dimer by photoassociative spectroscopy. Phys. Rev. A
57, 4600–4603.
Weiner, J., Bagnato, V. S., Zilio, S. C., and Julienne, P. S. (1999). Experiments and theory in cold and
ultracold collisions. Rev. Mod. Phys. 71, 1–86.
Weinstein, J. D., deCarvalho, R., Guillet, T., Friedrich, B., and Doyle, J. M. (1998). Magnetic trapping
of calcium monohydride molecules at millikelvin temperatures. Nature 395, 148.
Winans, J., and Stueckelberg, E. C. G. (1928). The origin of the continuous spectrum of the hydrogen
molecule. Proc. Natl. Acad. Sci. USA 14, 867–871.
Wynar, R., Freeland, R. S., Han, D. J., Ryu, C., and Heinzen, D. J. (2000). Molecules in a Bose-Einstein
condensate. Science 287, 1016.
Zinner, G., Binnewies, T., Riehle, F., and Tiemann, E. (2000). Photoassociation of cold Ca atoms. Phys.
Rev. Lett. 85, 2292–2295.
MOLECULAR EMISSIONS FROM

THE ATMOSPHERES OF GIANT
PLANETS AND COMETS: NEEDS
FOR SPECTROSCOPIC AND
COLLISION DATA
YUKIKAZU ITIKAWA
Institute of Space and Astronomical Science, 3-1-1 Yoshinodai, Sagamihara 229-8510,
Japan
SANG JOON KIM

Department of Astronomy and Space Science, Kyung Hee University, Suwon, 449-701,
Korea
YONG HA KIM
Department of Astronomy and Space Science, Choong Nam National University, Daejeon,
305-764, Korea
Y. C. MINH
Korea Astronomy Observatory, Hwaam, Yusong, Daejeon, 305-348, Korea
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 130

II. Spectroscopy of the Giant Planets:
Needs for Spectroscopic and Collision Data . . . . . . . . . . . . . . . . . . . . . 131
A. Ultraviolet and Visible Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
B. Infrared Spectra. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
C. Molecular Abundances Inferred From Spectroscopic Observations . . . . 135
D. Aurorae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
E. Titan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
III. Spectroscopy of Comets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
A. Ultraviolet and Visible Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
B. Infrared Spectra. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
C. Radio Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
D. Isotopic Abundances. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
E. Collisional Processes for Cometary Spectra . . . . . . . . . . . . . . . . . . . 154
IV. Spectral Databases and Improvements Needed. . . . . . . . . . . . . . . . . . . . 155
V. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
VI. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
VII. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

ISBN 0-12-003847-1/ISSN 1049-250X/01 $35.00
130 Y. Itikawa et al.
I. Introduction
The atmospheres of the giant planets and comets are believed to be least modi-
fied from their original compositions compared with the atmospheres of terrestrial
planets and the tenuous atmospheres of icy satellites in the outer solar system.
Planetary astronomers have been studying possible connections between the con-
stituents of these atmospheres and those of local interstellar clouds. Current theo-
ries on the solar system formation are mainly based on the elemental abundances
derived from the spectra of these planetary atmospheres. Comparisons between the
chemical compositions in these atmospheres and those in local interstellar clouds
should be the starting point to elucidate the chemical evolution of constituents
from interstellar nebulae through the proto-solar nebula to the current planetary
atmospheres.
Spectroscopy is the most powerful means for astronomers to unveil the true
nature of these heavenly bodies. Analyzing the spectra of the planetary objects, as-
tronomers have been able to identify molecules, atoms, radicals, ions, dimers, and
haze particles in the planetary atmospheres, and derive abundances of these con-
stituents. The spectra have also been utilized to derive temperatures and pressures
and to infer detailed radiative and collision processes occurring in the atmospheres
of the planetary objects. Spectroscopic observations of the giant planets and comets
from Earth-based and space-borne observatories during the last three decades have
produced a long list of molecules, radicals, and ions. Analyses of these planetary
spectra have often been hampered by the lack of relevant laboratory spectroscopic
data for the constituents of the atmospheres. The crucial data for the analyses in-
clude line positions, quantum-number designations of the upper and lower states,
and line intensities or transition probabilities. In particular, the identification of
the planetary molecules is only possible with the help of laboratory data on the
line positions. There are still hundreds of unidentified lines in the visible region
of cometary spectra. Many of these lines are suspected to arise from the high-
quantum-number transitions of known species. Consequently, proper laboratory
studies of these molecular bands are urgently needed to identify molecules, and
to analyze the spectra of the atmospheres of comets and the giant planets. Recent
high-resolution spectroscopy in the visible range provides opportunity to separate
isotopic molecular lines from adjacent normal molecular lines. These spectro-
scopic observations of the atmospheric molecules leave a great task not only to
planetary scientists, but to experimentalists and theorists in the field of atomic and
molecular physics.
Historically, laboratory spectroscopic studies for the visible region were inten-
sively carried out several decades ago, and most of the visible spectroscopic data
are only stored in the form of hard copies in the literature. Spectroscopists’ in-
terests have moved to other spectral ranges nowadays, leaving behind insufficient
databases for the identification of the cometary and planetary lines. In the infrared
MOLECULAR EMISSIONS FROM THE ATMOSPHERES 131
region of cometary and planetary spectra, there are also unidentified lines, although
infrared databases have been expanding during the last two decades. Many of the
unidentified infrared lines are speculated to be overtone and combination bands
of molecules already identified, but the relevant spectroscopic information is not
sufficient.
Most of the molecular emissions from the giant planets and comets are caused
by the fluorescence of solar radiation. To analyze the spectra quantitatively, ra-
diative transfer calculations are performed. The fluorescence spectra depend on
the rotational/vibrational states of the fluorescing molecules. Those molecular
states are populated through collisions with atoms, molecules, ions, and electrons.
Even some chemical reactions contribute to the balance of the molecular states.
Thus the analysis of planetary spectra also requires the collision data (i.e., excita-
tion/deexcitation cross sections). It should be noted that those collision processes
are directly responsible for auroral emission and infrared thermal emission or radio
waves from the planetary atmospheres.
In the present chapter, an overview is presented for the emission spectra of the
atmospheres of the giant planets and comets. Emphasis is placed on the recent high-
resolution observations with space probes and ground-based telescopes, including
the detailed studies of the recent bright comets, Hale–Bopp and Hyakutake. One
of the main objectives of the spectral observations is to determine the elemen-
tal abundances of the atmospheres, especially isotopic abundances, which are of
great importance in deducing the origin and evolution of the planetary system.
As mentioned above, spectroscopic and collision data play an essential role in the
analysis of planetary spectra. Many of the necessary data, however, are still lacking
or insufficient. The present chapter specifically reports those data needs.
II. Spectroscopy of the Giant Planets:

Needs for Spectroscopic and Collision Data
In this section we present spectra of the atmospheres of the giant planets: Jupiter,
Saturn, Uranus, and Neptune. Sections II.A and II.B summarize observational
results of ultraviolet, visible, and infrared spectra of the planets, and radiative
transfer processes in the atmospheres. Among the information obtained from those
spectra, the most important quantity is the molecular abundance, including isotopic
abundance, which is discussed in Section II.C. Section II.D describes the aurorae
of the giant planets. Auroral emissions are induced by the precipitations of high-
energy electrons, protons, or ions on to the high-altitude atmospheric layers from
the magnetosphere. In Section II.E we present spectra of Titan, the biggest satellite
of Saturn and the only satellite which has a thick atmosphere perhaps retaining
signatures of primordial atmospheric constituents due to least modification by faint
sunlights.
A. ULTRAVIOLET AND VISIBLE SPECTRA

In the early 1930s, R. Wildt performed a series of spectroscopic observations
of Jupiter, and identified CH4 and NH3 lines in the visible range (Wildt, 1932).
Almost two decades later, the detection of the most abundant molecule, H2, in
Uranus was achieved by Herzberg (1952), and later in Jupiter by Kiess et al.
(1960) via observations of several lines in the 3 0 band of H2. Voyager 1 and 2
ultraviolet spectrometers observed H2 and H emissions from Jupiter’s polar regions
(Broadfoot et al., 1979). In the ultraviolet region, space-borne investigations were
mostly concentrated on the auroral spectra of polar regions of the Jovian planets,
because the auroral emission is intense compared to airglow emission in general.
Continuum and absorption features of the giant planets in the ultraviolet region
have been studied with the International Ultraviolet Explorer (IUE) since 1978
(e.g., Hunt and Moore, 1984). These observations as well as recent observations
with the Hubble Space Telescope (HST) and the Galileo spacecraft are summarized
in Sect. II.D (Aurorae) in detail. The giant planets passively reflect sunlight in the
visible range, not much showing molecular signatures, but in the infrared their
cool atmospheres emit radiation revealing rich information on their constituents
and thermal characteristics.
B. INFRARED SPECTRA
In the late 1960s and early 1970s, the detections of rotational–vibrational bands
of molecules in infrared became possible because of rapid technological devel-
opment in infrared detector systems, especially Fourier transform multiplexing
spectrometers, of which light-gathering ability was higher than other conventional
spectrometers at that time. Gillett et al. (1969) found that the 8.0–11.5 μm range of
Jupiter is depressed substantially below the computed H2 pressure-induced absorp-
tion continuum. By detecting two Q branches of the v2 NH3 band, they confirmed
that the NH3 is an important opacity source near 10 μm. Ridgway (1974a,b) de-
tected C2H2, C2H6, and PH3 in 10-μm high-resolution spectra of Jupiter obtained
with Fourier transform spectrometer (FTS) on the 4-m telescope at Kitt Peak. The
5-μm atmospheric window also provided an opportunity to detect lines of CO, H2O,
and GeH4. Subsequently, Tokunaga et al. (1981) reported a positive detection of
HCN on Jupiter near 13.5 μm.
During the encounter with Jupiter in 1979, the Infrared Interferometer Spec-
trometer and Radiometer (IRIS), a Fourier transform spectrometer on the Voyager
spacecraft, confirmed hydrocarbon molecules detected in the 1970s from Earth-
based observations. Voyager IRIS data taken from the north polar auroral region of
Jupiter showed several diminutive structures, which were later identified as C2H4,
C3H4, C6H6, and possibly a radical, CH3 (Kim et al., 1985). Recently, Saturn
and Neptune spectra obtained by the Infrared Space Observatory (ISO) show an
emission of CH3, a photochemically produced radical, near 16 μm (Lellouch,

2000). The rotational structures of the v2, v3, and v4 bands of CH3 have been an-
alyzed by chemists (e.g., Spirko and Bunker, 1982; Yamada et al., 1981; Hirota
and Yamada, 1982; references therein). However, CH3 line intensities or Einstein
A coefficients, which are valuable for the derivation of a column abundance of
CH3, are needed for future analyses of the rotational structure of the band. ISO
also detected C3H4 and C4H2 on Saturn; C6H6 on Jupiter and Saturn; CO2 on
Jupiter, Saturn, and Neptune; and H2O on Jupiter, Saturn, Uranus, and Neptune
(e.g., see a summary of the ISO results in Bézard et al., 1999, and references
therein).
Noll et al. (1989, 1990) reported detections of AsH3 in the atmospheres of
Jupiter and Saturn near 5 μm. The possibility of the existence of H2S in Saturn,
Uranus, and Neptune has been proposed by several microwave observations
(Grossman, 1990; de Pater et al., 1991), although the spectral lines of the molecule
were never been unambiguously identified. Eventually, the mass spectrometer on
the Galileo probe detected H2S with a mole fraction of 1 × 10−5 (Niemann et al.,
1996). The mixing ratios of the major constituents of Jupiter’s atmosphere, H2 and
He, have also been determined by the Galileo probe mass spectrometer along with
mixing ratios or upper limits for less abundant molecules including CH4, H2O,
NH3, C2H6, C3H8, C2H4, and H2S, and the inert gases Ne, Ar, Kr, and Xe.
Planetary infrared airglows are due to either fluorescence or thermal emission
of molecules and ions in the upper atmospheres. The existence of global H3+
emission was suspected when the H3+ auroral images of Jupiter were taken at
the Infrared Telescope Facility (IRTF) simultaneously with the spectra of the au-
rora at the Canada-France-Hawaii Telescope (CFHT) (Kim et al., 1991). Ballester
et al. (1994) obtained latitudinal variation of temperatures derived from spatially
resolved H3+ spectra of Jupiter. The global H3+ emissions are found to be de-
creased at the impact site of the comet SL9 fragments on Jupiter, compared with
undisturbed regions at the same latitude (Kim et al., 1996). During the spectral
observations of Jupiter, Trafton et al. (1989) noticed unidentified strong airglow
emission lines near 2.1 μm in addition to the global airglow emissions of H2
quadrupole lines and H3+ lines on November 23, 1988. The unidentified emis-
sions were widespread, but it seems that the emissions might be highly transient
because the emissions were never detected again.
The short-wavelength spectrometer on the ISO successfully detected 3-μm line
emissions from the v3 fundamental and the ν 3+v4−v4 band of CH4 on the disks of
Jupiter and Saturn (Drossart et al., 1999). The similar 3-μm line emissions of CH4
were also detected on Titan from a ground-based observation (Kim et al., 2000; see
the section on Titan). The 3-μm CH4 fluorescence emissions from giant planets and
Titan have not been detected previously from the ground because of strong telluric
absorption, except during an extraordinary event, the collision between comet SL9
and Jupiter in July 1994. From the impact sites of SL9 on Jupiter, a part of the
FIG. 1. Strong emission of the ν 3 P branch of CH4 at the A impact site of comet SL9 on Jupiter
(solid line) and a corresponding model spectrum of the P branch (broken line). (From Kim et al., 1996.)
P branch of the CH4 v3 band was detected in emission as shown in Fig. 1 (Kim et al.,
1996; Dinelli et al., 1997). In the future, we expect to detect additional emissions
from other bands of CH4, and from bands of other hydrocarbons, although these
band intensities are expected to be weaker than the CH4 v3 and v3+v4−v4 bands.
We need a database for the Einstein A coefficients of these molecular lines in order
to analyze the hydrocarbon airglows.
For the radiative transfer calculations of thermal infrared radiation in the plan-
etary tropospheres and stratospheres, local thermodynamic equilibrium (LTE) has
been assumed for approximation (e.g., Knacke et al.,1982; Kim et al.,1985). The
LTE approximation greatly simplifies the numerical calculation for the otherwise
complicated radiative transfer processes of thermal infrared radiation in the plan-
etary atmospheres. Above a certain altitude, where atmospheric density becomes
tenuous, the validity of the LTE breaks down, because the collisional deexcita-
tion rates of molecular vibrational states become less than spontaneous vibrational
transition rates (e.g., Appleby, 1990). However, even above this altitude, ground
rotational states are usually in LTE, since the collisional deexcitation rates of the
ground rotational states are often greater than the spontaneous transition rates
(e.g., Kim et al., 2000). Vibrational states are mostly populated by pumping from
the ground states via solar infrared radiation, not by collisions with other atmo-
spheric species. Since the ground states are the dominant reservoir of the pop-
ulation, excited rovibrational states often follow the rotational population of the
ground states. Therefore, the rotational temperatures derived from emission spec-
tra of infrared airglows and infrared aurorae often represent the ground rotational
temperatures.
C. MOLECULAR ABUNDANCES INFERRED FROM

SPECTROSCOPIC OBSERVATIONS
Voyager IRIS spectra show very wide smooth continuum absorption in the range
15–40 μm, which is mainly caused by collision-induced absorptions (CIA) of
H2–H2 and H2–He absorptions (Fig. 2). This absorption continuum was utilized
to derive He abundance indirectly with the help of temperature–pressure profiles
derived from Voyager radio occultation experiment (e.g., Gautier et al., 1981).
The He abundances for Uranus and Neptune are 15% and 18% by volume,
FIG. 2. A typical Voyager IRIS spectrum taken from an equatorial region of Jupiter. The smooth
continuum between 15 and 40 μm is caused by pressure-induced H2–H2 and H2–He absorptions. This
continuum has been used to derive a He mixing ratio in Jupiter (e.g., Gautier et al., 1981).
respectively, which are consistent with the solar composition of ∼16%. The de-
rived values for Jupiter and Saturn are 13.6% and 3%, respectively, suggesting
that He has been depleted from the upper atmospheres of both planets. The
slight depletion in the atmosphere of Jupiter was confirmed by the in-situ ob-
servations of the Galileo probe (e.g., Niemann et al., 1998). However, a recent
reanalysis of the Voyager measurements by Conrath and Gautier (2000) shows
that the He abundance of Saturn is between 11% and 16%. This new devel-
opment creates a dispute about the current interior models of the giant planets
(e.g., Gautier and Owen, 1989), which are based on the significant He depletion
in Saturn. There have been early attempts to construct temperature-dependent
CIA models (e.g, Trafton, 1967; references therein); and subsequent efforts to
refine the models (e.g., Borysow and Tang, 1993; Zheng and Borysow, 1995;
references therein), which have been compared with laboratory spectra and ap-
plied to the atmospheres of the giant planets. Recent detections of cool brown
dwarfs and Jupiter-size planets orbiting nearby stars demand CIA models for
broad ranges of temperatures and pressures (e.g., Borysow et al., 1997; Trafton,
1998).
Gautier and Owen (1983, 1989) reviewed carbon abundances derived from both
ground and Voyager observations. The C/H ratios compared with the solar value
are 2.3 times for Jupiter, 2.1 times for Saturn, ∼20 times for Uranus, and ∼25
times for Neptune. The carbon and other heavy-element enrichments in Jupiter
were recently confirmed by the measurement of the Galileo probe mass spectrom-
eter. The enrichments seem to increase in the sequence Jupiter, Saturn, Uranus, and
Neptune, and appear to be roughly correlated with the ratio of core mass to total
planet mass (Gautier and Owen, 1989). Gautier and Owen suggested that it may
arise either from efficient upward convection of gases produced from the cores
during the formation of these planets or from the dissolution of infalling plan-
etesimals in the early atmospheres. Recently, based on the Galileo results, Owen
et al. (1999) showed that Ar, Kr, and Xe in Jupiter’s atmosphere are enriched to
the same extent as the other heavy elements.
The detected C3H4 is methyl acetylene. There are two other forms of C3H4:
allene, which has a different order of bonding; and cyclopropene, which has a
simple ring of three carbon atoms surrounded by four hydrogen atoms. Allene
and cyclopropene have not been detected in the giant planets. Benzene (C6H6), a
molecule with a ring of six carbon atoms, was observed in the giant planets, as
mentioned above. On the other hand, cyclopropane (C3H6), another ring molecule,
has not been observed, although propane, C3H8, was detected in the atmospheres of
the giant planets and Titan. It is interesting to investigate why the atmospheres of the
giant planets and Titan do not have observable amounts of allene, cyclopropene,
and cyclopropane. Therefore, good compilation of spectroscopic parameters is
needed for the identification or the derivations of the upper limit abundances of
these molecules.
Isotopic molecules such as CH3D, HD, and 13CH4 were also detected on Jupiter.
Later, weak lines of 15NH3 were observed nearby strong 14NH3 lines around
10 μm; and H218O was detected through the 5-μm atmospheric window. The
derived 15N and 18O abundances seem to be similar to the solar values, as the
fractionations of heavy species, such as carbon, nitrogen, and sulfur, are less sen-
sitive to chemical reactions in planetary atmospheres, and this is also the case for
cometary isotopic abundances for heavy elements, as discussed in the comet sec-
tion. Gautier and Owen (1983) summarized pre-Galileo D/H values of the giant
planets obtained from Voyager and ground-based observations.
Recently, the mass spectrometer of the Galileo probe accurately measured iso-
tope ratios for D/H, and 3He/4He (Mahaffy et al., 1998), and the D/H ratio is
consistent with Voyager and ground-based data, and recent spectroscopic results
from ISO (e.g., Bézard et al., 1999). The D/H ratio is comparable or a little less
than the value in the local interstellar medium, 1.6 ± 0.12 × 10−5 (Mahaffy et al.,
1998). The D/H ratio of Saturn is similar to that of Jupiter, but the ratios of
Uranus and Neptune are several times higher than those of Jupiter and Saturn (e.g.,
de Bergh et al., 1990). In fact, the values of Uranus and Neptune are equal to or
somewhat less than the D/H ratios of Titan, ∼1.5 × 10−4 (e.g., de Bergh et al.,
1990), P/Halley, ∼3.1 × 10−4 (e.g., Eberhardt et al., 1995), and Earth oceans,
∼1.6 × 10−4. The measured D/H ratios in the atmospheres of the giant planets
suggest that the giant planets may be made from accretions of icy planetesimals,
and the capture of gases from the surrounding proto-solar nebula. The 3He/4He
ratio obtained by the Galileo probe is equal to or a little higher than that found
in meteoritic gases, 1.5 ± 0.3 × 10−4. Niemann et al. (1998) present isotopic ra-
tios measured from the Galileo probe for 13C/12C, 20Ne/22Ne, 38Ar/36Ar, and for
isotopes of Kr and Xe. The derived isotope ratios and heavy-element abundances
are the center of disputes regarding the origin, internal structure, and evolution
of the giant planets (e.g., Owen et al.,1999; Atreya et al., 1999; Manuel et al.,
1998).
Thus far, only CH3D, HD, 13CH4, 15NH3, and H218O have been used to remotely
derive D, 13C, 15N, and 18O abundances from ground spectroscopic observations.
Although the Galileo probe has been successful in the derivation of various iso-
topic abundances, we still need isotopic abundances locked in various molecules,
such as PH3, H2S, etc., in order to investigate the degree of isotopic fractionations
in the atmospheres of the giant planets. The determination of the isotopic fraction-
ation among the molecules will help us in better understanding the chemistry and
convection of the atmospheres. For this purpose we need spectroscopic parame-
ters for the bands of isotopic counterparts of the above molecules, which are not
completely known in the literature.
Weak absorption features in the far-infrared spectra of the Voyager IRIS spectra
of Jupiter and Saturn were attributed to H2–H2 dimers by McKellar (1984) and
Frommhold et al. (1984), after the features were first recognized by Hanel et al.
FIG. 3. Spectra of Jupiter at 2.10–2.13 μm compared with a laboratory spectrum of the H2–H2
dimer (bottom). The positions of five S and Q branch lines of H2–H2 and of the S1(1) line of the H2
are indicated. Solar lines are also indicated. (From Kim et al., 1995.)
(1979). Subsequently, Fox and Kim (1984) and Borysow and Frommhold (1986)
attributed features in the far-infrared IRIS spectra of Titan to the H2–N2 dimer. The
2–2.5 μm atmospheric window has been utilized to detect the 2.1-μm features of
H2–H2 dimers from Jupiter (Fig. 3; Kim et al., 1995), Saturn, and Neptune (Fig. 4;
Trafton et al., 1997). The dimer detections through the 2-μm atmospheric window
also suggest the possibility of detecting inert gases such as argon and neon utilizing
sharp spectral structures of H2–Ar and H2–Ne (McKellar, 1995). Since inert gases
do not have strong lines in the visible and infrared ranges, it has been difficult to
derive their abundances in the atmospheres of the giant planets except the in situ
measurements by the Galileo probe (Niemann et al., 1998).
Although extensive laboratory measurements of various dimers have been con-
ducted (e.g., McKellar et al., 1999; McKellar, 1995, 1996; references therein),
theoretical models have rarely been constructed. Even for a simple dimer, H2–H2,
theoretical models have been constructed only for the far-IR S0(0) and S0(1) tran-
sitions by Schaefer and McKellar (1990). The 2-μm v = 1–0 structures, which
can be observed directly with ground-based IR telescopes (e.g., Kim et al., 1995),
have not been modeled. Many spectral lines of various dimers, which have been ob-
served in laboratories by McKellar and his colleagues during the last two decades,
have not been properly designated quantum mechanically.
FIG. 4. H2–H2 dimer features in near-infrared spectra of Neptune and Saturn (Trafton et al., 1997).
A laboratory absorption spectrum of H2 is shown below Saturn’s spectrum; it includes both the H2 S1(1)
quadrupole line and the associated H2–H2 dimer spectrum, as indicated. A synthetic solar spectrum is
also shown. A schematic spectrum for CH4 is shown near the top for comparison, and the propagated
error spectra for the Neptune and Saturn spectra are shown at the bottom.
D. AURORAE
Auroral particle (electrons, protons, or other ions) precipitation produces electro-
magnetic waves from X ray through visible light to radio range. Jupiter’s strong
radio emission has been detected from ground-based observations since the 1950s.
Voyager 1 and 2 ultraviolet spectrometers unequivocally identified H2 Lyman and
Werner bands and H Ly-α emissions from Jupiter’s polar regions (Broadfoot et al.,
1979). Shortly after the Voyager encounter with Jupiter, the IUE confirmed that the
auroral UV emissions (Clarke et al., 1980). Visible auroral images were first taken
by the imaging experiment on Voyager 1 (Smith et al., 1979), and then soft X-ray
emissions were detected by the Einstein Observatory (Metzger et al., 1983). The
spectroscopic observations of the Jovian UV aurora with the HST were first made
by Trafton et al. (1994). Recently, Gladstone et al. (1998) extensively observed
the Jovian X-ray emissions with ROSAT. High-quality X-ray auroral images of
the Jovian planets are expected with the Chandra X-ray telescope launched in
1999.
Voyager/IRIS spectra taken from the northern auroral region of Jupiter showed
several weak emission structures, which were later identified as C2H4, C3H4, and
C6H6 by Kim et al. (1985). Trafton et al. (1987) observed 2-μm H2 quadrupole
lines in emission over the polar regions of Jupiter. The auroral H2 quadrupole line
emissions were theoretically predicted by Kim and Maguire (1986). Subsequently,
Trafton et al. (1989) detected unidentified strong emission lines at 2.1 μm in
addition to the H2 quadrupole lines near Jupiter’s northern limb. These unidentified
lines were eventually identified as the 2v2 band of H3+ (Fig. 5) by Drossart et al.
(1989) using infrared spectra obtained by the FTS on the CFHT. Maillard et al.
(1990), Oka and Geballe (1990), and Miller et al. (1990) detected the fundamental
band (v2) of H3+ at 3.5 μm in the wide areas of the both polar regions. Subsequently,
Kim et al. (1991) and Baron et al. (1991) successfully obtained auroral images,
which revealed detailed auroral structures. During the impacts of the fragments of
comet SL9 on Jupiter in 1994, Kim et al. (1996) obtained 3.5-μm H3+ emission
FIG. 5. H3+ and H2 emission lines over the polar haze continuum of the auroral regions of Jupiter
(Drossart et al., 1989). All the lines are attributed to the 2v2 band lines of H3+ except a line at 2.1218 μm,
which is due to the S1(1) line of H2.
spectra of the auroral regions, and they found that the temperatures of the southern
H3+ aurora were normal, suggesting no significant influence of the impacts on
the auroral activities. A comprehensive review of the multispectral nature of the
Jupiter’s aurorae was given by Kim et al. (1998), and here we will not present
detailed descriptions about Jupiter’s aurorae.
Since Jupiter’s magnetosphere contains large amount of sulfur and oxygen ions
from Io, it has been expected that these ions may precipitate on to the Jupiter’s
auroral regions. Recently, Trafton et al. (1998) searched for heavy-ion precipitation
with HST, and found no evidence of heavy-ion features in the auroral spectra. They
detected an unidentified feature near 1254 Å at the 4-σ level, which might have
been the suspected SII line at 1256 Å in the low-resolution IUE spectra of Waite
et al. (1988).
The Galileo orbiter has been revolving around Jupiter and its satellites since its
arrival in December 1995. During the passages near Jupiter, the Galileo orbiter has
observed Jupiter’s aurorae with the solid-state imager (SSI), the extreme ultraviolet
spectrometer (EUVS), and the ultraviolet spectrometer (UVS). The SSI has ob-
served prominent visible aurora on the night side of Jupiter (Ingersoll et al., 1998).
The SSI has taken images of the visible aurora with red, violet, green, methane,
and clear filters. Ajello et al. (1998) analyzed Galileo EUV and FUV spectra taken
simultaneously at both the north and south polar regions. The Galileo UVS has ob-
served Jupiter’s aurorae in the near-ultraviolet (NUV) and mid-ultraviolet (MUV)
ranges for the first time. Pryor et al. (1998) analyzed the MUV auroral spectra
using, James et al.’s (1998) laboratory spectra.
Prior to the Voyager 1 and 2 encounters with Saturn, IUE and Pioneer ob-
servations suggested the presence of aurorae on Saturn, but both observations
were inconclusive. Analyzing the spectra of polar regions observed by the Voyager
UVS, Broadfoot et al. (1981a,b) reported clearly defined aurorae. The Voyager
UVS detected an aurora of Uranus near the magnetic pole (Broadfoot et al., 1986).
Broadfoot et al. (1989) tentatively identified faint aurora from the dark side of
Neptune based on Voyager 2 UVS data.
Since the initial detections of the UV aurorae on Saturn, Uranus, and Neptune
by Voyager 1 and 2, there have been a few recent developments in the investigations
of these aurorae—Geballe et al. (1992) first reported the detection of H3+ emission
from Saturn. A dark auroral oval on Saturn was discovered in a 2200-Å band in
HST/FOC UV images (Ben Jaffel et al., 1995). Trafton et al. (1993) reported
detection of H3+ emission from Uranus. They found that the emission intensity
was greater than that of Saturn. Trafton et al. (1995) detected auroral H2 emission
at 2 μm from Uranus. Subsequently, Lam et al. (1997) obtained 2-μm H2 emission
images of Uranus.
In the planetary auroral regions, precipitating particles from the magnetosphere
cause electronic and vibrational excitations of atmospheric species. In order to
calculate excitation rates of the atmospheric species, we need to know excitation

cross sections of all those electronic and vibrational states of the atmospheric
species upon collisions with the precipitating particles. The energies of precipi-
tating particles decrease due to multiple collisions with the atmospheric species,
and at the same time secondary electrons and ions are produced from the colli-
sions. The excitation rates for the various states of atmospheric molecules and
atoms have been calculated by sophisticated computer programs (e.g., Bisikalo
et al., 1996). For the calculations, cross-section data are needed for all the colli-
sion processes expected between precipitating particles (e.g., electrons, protons,
or ions) and planetary molecules. Cross sections for the electronic excitation
or ionization for molecules in the giant planets, perhaps except for H2, H, He,
and CH4 (e.g., Kim, 1988), are not well known. Particularly important molecules
are C2H2, C2H4, C2H6, C3H8, C3H4, and C6H6. Eventually, the primary and sec-
ondary particles slow down and stop near the bottom of the planetary ionospheres.
The low-energy electrons efficiently excite the rovibrational states of molecules.
Electron-impact vibrational or rotational cross sections of H2 and CH4 have been
measured previously in laboratories. However, those for other molecules, C2H2,
C2H4, C2H6, C3H8, C3H4, and C6H6, by electron, proton, and ion collisions are
scarcely known.
Pure rotational excitations of molecules by electrons are not very important
in the auroral regions of the giant planets, because of high atmospheric densities
in the auroral ionospheres. In the ionospheres of the giant planets, the ground
rotational states of CH4 and H2 are in local thermodynamical equilibrium (LTE),
because rotational deexcitation rates by collisions with H2 and CH4 are signifi-
cantly greater than rotational Einstein A coefficients, which are almost zero for a
spherical-top molecule such as CH4 and for homonuclear diatomic molecules such
as H2 and N2. However, the vibrational states of these molecules are in non-LTE,
because the vibrational deexcitation rates by neutral collisions are approximately
equal to or less then vibrational Einstein A coefficients. Since the ground states
are the dominant reservoir of the CH4 and H2 population, the rovibrational states
should approximately follow the rotational population of the ground states, and the
rotational temperature derived from the observed emission intensities of the rovi-
brational states should approximately represent the ground rotational temperature,
which in turn represents the local kinetic temperature of the ionosphere.
The excited atmospheric species lose their energies via radiation and/or col-
lisions with other species. The last stage of the auroral particle precipitations is
the thermalization of the auroral atmosphere, emitting infrared radiation through
rovibrational bands of molecules. In order to calculate infrared emissions from
molecules and ions, we need cross sections for vibrational–vibrational (v–v)
and vibrational–translational (v–t) energy transfers between vibrationally excited
species and other surrounding molecules/ions. For v–v and v–t energy transfer rates
of H2 with H2, He, and H collisions, and of CH4 with H2, He, N2, O2, and CH4
collisions are partially available. However, the v–v and v–t rates of C2H2, C2H4,
C2H6, C3H8, C3H4, and C6H6 by collisions with H2, He, H, CH4, etc., are scarcely
available in literature, especially at low temperatures in the planetary atmospheres.
Some crude theoretical models are often used to estimate the energy transfer rates,
but their results contain large ambiguities. Comprehensive quantum mechanical
formulas, which can be used readily by scientists in other fields, are needed.
E. TITAN
The existence of Titan’s atmosphere was first recognized spectroscopically by
Kuiper (1944), who discovered gaseous CH4 (methane). Almost three decades later,
a tentative detection of H2 in the visible range, and a probable detection of C2H6
(ethane) at 12.2 μm were reported. A model atmosphere with a large temperature
inversion was proposed (e.g., Danielson et al., 1973). Additional hydrocarbon
bands of C2H2, C2H4, and a deuterated methane, CH3D, were positively detected
in emission in the range 7.8–13.3 μm along with the C2H6 band (Gillett, 1975).
Unlike ground-based telescopes, which barely resolve the Titan disk, the IRIS
on the Voyager 1 and 2 spacecraft obtained spatially resolved infrared spectra in
the range 5–50 μm (e.g., Hanel et al., 1981). The Voyager IRIS team identified
emissions bands of C3H4 (methyl acetylene), C3H8 (propane), and HCN (hydro-
gen cyanide) in the spectra of equatorial and temporal regions of Titan. In the
spectra of north polar regions, where dark haze layers were clearly seen in visible
images, additional molecules, C4H2 (diacetylene), C2N2 (cyanogen), and HC3N
(cyanoacetylene), were identified by the IRIS team. The detections of the several
nitrogen-containing molecules confirmed the N2-dominant atmosphere, which was
first recognized by the results from extreme UV observations of N2 emissions
(Broadfoot et al., 1981b), and from radio occultation experiment (Tyler et al.,
1981) on Voyager 1. The N2-dominant atmosphere was remarkably similar to an
atmospheric model of Titan proposed by Hunten (1978). Further detailed analyses
of the IRIS spectra revealed a CO2 emission band, a dominant CH3D feature at
8.6 μm over a C3H8 band feature there, and a C4N2 ice band. A ground infrared
observation also yielded a CO band in absorption at 1.6 μm (Lutz et al., 1983).
Recently, Kim et al. (2000) detected 3-μm emission lines of the v3 and
v3+v4−v4 bands of CH4 from Titan’s high-altitude atmosphere using the CGS4 on
the UKIRT (Fig. 6). Analyzing the spectrum, they found that the CH4 emissions
form mainly in the 530–630 km altitude range, where temperatures are 137–145 K,
consistent with a low-temperature mesosphere as proposed by Yelle (1991). It is
surprising that the emission intensity of the weak combination band is comparable
to that of the strong v3 fundamental band, and that this is also the case for Jupiter
and Saturn (Drossart et al., 1999).
III. Spectroscopy of Comets

This section reviews the emission spectra of comets. The spectra in the ultravi-
olet/visible, infrared, and radio ranges are described in Sections III.A, III.B, and
III.C, respectively. Section III.D summarizes isotopic abundances derived from
the observed spectra of cometary comas. Most of the cometary emissions are
arisen as a fluorescence of the solar radiation. Collisions among the atmospheric
constituents (i.e., atoms, molecules, radicals, ions, and electrons), however, have
sometimes significant effects on the fluorescence spectra. These collisional pro-
cesses are discussed in Section III.E.
Since Swings et al. (1941) began spectroscopic observations of comets in the
visible range, parent molecules in cometary nuclei have been considered to be pris-
tine, keeping information about elemental abundances in the presolar nebula. The
visible range, however, hardly shows parent molecules, but mostly shows radical
emissions. The abundances of parent molecules in the nuclei have been indirectly
inferred from the radical abundances observed by using laboratory and/or theoreti-
cal data on chemical reactions between molecules, radicals, and ions. The inferred
abundances of parent molecules from various spectral ranges have been com-
pared with those in interstellar media (e.g., Bockelée-Morvan et al., 2000; Irvine
et al., 2000). Chemistry models of various interstellar media have been significantly
improved during the last decade by theorists who have been trying to explain the el-
emental abundances observed in interstellar ices (e.g., Fegley et al., 1997; Aikawa
et al., 1999; and references therein). The comparisons between abundances of
cometary molecules and those of interstellar ices (e.g., Bockelée-Morvan et al.,
2000; Irvine et al., 2000) show that there is an intimate similarity suggesting the
interstellar origin of the cometary material.
Comets produce abundant atoms, molecules, radicals, and ions. Some radicals
in cometary comas could not be identified immediately because corresponding
laboratory spectra were not available. Since free radicals are extremely reactive, it
has not been easy to obtain their spectra in laboratories [e.g., see the introduction
section of Herzberg (1971)]. Indeed, the identification of the radical bands has been
difficult, often took many years, and sometimes was not correct. For example, the
4050 Å band was first discovered in 1881 by Huggins (1882), but it was not
correctly identified for 70 years. The band even confused a Nobel Prize laureate,
FIG. 6. (Top) Spectrum of Titan obtained on 1999 September 14 (UT) (Kim et al., 2000). Strong
telluric absorption by the Q-branch lines of the CH4 v3 band blocks Titan’s Q-branch emission. The
S/N ratios at the peaks of typical v3 and the v3+v4−v4 lines are estimated to be 2–4 and 8, respectively.
(Bottom) A model spectrum, including the v3 and v3+v4−v4 bands with a rotational temperature of
140 K. The long vertical lines indicate the positions of the v3 band lines, some of which are blended
with the lines of the v3+v4−v4 band. The short vertical lines and other weak features are all identified
as those of the v3+v4−v4 band. Only the v3 lines are labeled.
G. Herzberg, who attributed the 4050 group to a band of CH2 (e.g., Herzberg, 1942).
It was ultimately identified as the bands of C3 by Clusius and Douglas (1954).
A. ULTRAVIOLET AND VISIBLE SPECTRA

Observations of comets in the ultraviolet were begun by using sounding rockets
and space-borne observatories (e.g., Feldman et al., 1974). The detected strong
H I Ly-α envelope and OH emission at 3085 Å confirmed the icy conglomerate
model, “dirty snowball model,” for cometary nuclei proposed by Whipple (1950,
1951). Since 1978, the IUE has been active in the observations of comets, most
notably comet P/Halley, during its operation period of nearly 18 years. Strong
atomic emissions of H, C, S, and O; radical emissions of CS, OH, and a weak C2
band; and ionic emissions of CO2+ have been usually shown in UV spectra. Before
the bright comets Hyakutake and Hale–Bopp, the only comet which exhibited S2
emission in the UV region was comet IRAS–Araki–Alcock 1983VII (A’Hearn
et al., 1983).
In the visible range (3000–9000 Å), radicals, such as OH, NH, CH, CN, C2,
C3, NH2 (e.g., Arpigny et al., 1991; Brown et al., 1996); ions, such as CO+, OH+,
CH+, H2O+, N2+, CO2+ (e.g., Wyckoff et al., 1996); and only one stable molecule,
S2 (Fig. 7; Kim et al., 1990), have been observed. During the last five decades,
the spectral resolving power (λ/δλ) of spectrometers in the visible range has been
increased from an order of ∼1000 (Swings et al.,1941; Dossin et al.,1961) to
∼100,000 (Brown et al., 1996) nowadays. The high-resolution spectroscopy in
the visible range provides a wealth of emission lines from the radicals, making it
possible to investigate detailed fluorescence and collision excitation processes of
the radicals in cometary comas (e.g., Schleicher and A’Hearn, 1982; Kim et al.,
1989); and to separate isotopic molecular lines from adjacent normal molecular
lines (e.g., Kleine et al., 1994 ,1995). On the other hand, Brown et al. (1996) listed
hundreds of unidentified lines between 3800 and 9900 Å. Wyckoff et al. (1996)
also reported unidentified lines in the plasma tail spectra of P/Halley.
Schleicher and A’Hearn (1982, 1988) constructed a fluorescence model of OH
bands that appeared in the IUE high-dispersion spectra of comets in the near-UV
range, which can also be observed from the ground. Kim et al. (1989) constructed
an A–X (0–0) band model of NH, which occurs in the range 3345–3375 Å of
P/Halley spectra. They found that the NH spectrum is due to pure fluorescence
with negligible influence by collisions. Gredel et al. (1989) updated the C2 band
model of A’Hearn (1978) and compared it with the high-resolution spectra of
P/Halley, and the fit was satisfactory. However, the authors pointed out that the ac-
curate determinations of the transition probabilities of the electronic bands should
improve their model significantly.
In cometary comas, molecular number densities are usually less than those
of planetary ionospheres or thermospheres, so collision deexcitation rates are in-
significant in fluorescence processes. Therefore, molecules in cometary comas are
FIG. 7. The solid line is a ground-based spectrum of comet IRAS–Araki–Alcock 1983VII between
3050 and 4350 Å, and the dashed line is a fluorescent equilibrium model described by Kim et al. (1990).
The identified B–X bands of S2 in the observed spectrum are marked by ∗ .
mostly in non-LTE except for the regions very close to the surfaces of massive
outgassing comets, such as comet Hale–Bopp. Note that in pure fluorescent equilib-
rium, ground-state populations are constantly redistributed mainly by fluorescence
transitions from upper states, and thus the ground rotational state populations can-
not represent the local kinetic temperatures.
Among the detected neutral radicals in the range 3000–9000 Å, line-by-line
spectral structures of NH2 and C3 bands have not been modeled in detail. NH2 is
an asymmetric top molecule, and therefore its band structure is very complicated
compared with other linear and symmetric molecules. While line-by-line quantum-
number designations of the NH2 A–X bands, have been carried out by Dressler and
Ramsay (1959), Johns et al. (1976), and in unpublished works by spectroscopists
at the Herzberg Institute of Astrophysics, a complete line list of the A–X bands
including high J lines has not been available thus far. Furthermore, individual line
transition probabilities of the bands have not been completely determined either,
although vibrational transition moments and lifetimes have been discussed by
Jungen et al. (1980), and Einstein A coefficients for certain rotational lines of the
(0,9,0)–(0,0,0) and (0,8,0)–(0,0,0) transitions have been calculated by Kawakita
et al. (2000). The lack of accurate transition probabilities prevents the development
of a proper fluorescence model for the NH2 bands.
The situation for the C3 bands is even worse. Although Gausset et al. (1965)
listed quantum-number designations for strong lines, the list lacks weak line des-
ignations, especially for the spectral range 3880–3950 Å, where the strong B–X
band of CH occurs (Kim et al., 1997a) and the strong lines of the A–X band
of CO2+ (Kim, 1999) also occur (Fig. 8). Furthermore, the line list of Gausset
et al. does not provide line intensities and therefore transition probabilities, which
are important parameters used for the derivations of C3 abundances in comets. In
the 3880–3950 Å range of cometary spectra, there are many unidentified lines,
some of which are expected to be due to weak lines of the C3 bands. In order
to properly model the B–X band of CH and the A–X band lines of CO2+, it is
important to clear out those unidentified lines in this spectral range. Since the NH2
and C3 bands are the most prominent emission structures, usually appearing in
the wide spectral range of cometary visible spectra, models of these bands are
urgently needed to properly analyze high-resolution visible spectra of the recent
bright comets.
FIG. 8. Comparison of the best-fit model (solid line) and a tail spectrum (dashed line) of comet
P/GZ observed in 1985 with a spectral resolution of 20 Å (Kim, 1999). The NH band is blended with a
CO2+ band. The negative intensities occur because atmospheric transmission rapidly decreases toward
shorter wavelengths, and the negative intensities in the GZ spectrum indicate the degree of uncertainties
in the comet spectrum.
Among the detected ions, CO+, OH+, CH+, H2O+, N2+, CO2+, only the bands
of two ions, CO+ and CO2+, have been modeled. Magnani and A’Hearn (1986)
constructed a fluorescent equilibrium model of the visible bands of CO+, including
Swings and Greenstein effects, which depend on heliocentric radial velocity. Kim
(1999) constructed a fluorescence equilibrium model of the A–X and B–X band
systems of CO2+, and compared the model with the spectra of comets Austin
(1989c1) and P/Giacobini–Zinner (GZ) (Fig. 8). The model, however, includes
only vibrational–vibrational transitions, ignoring rotational transitions. In order to
construct the line-by-line model of the CO2+ bands, accurate line positions and
line transition probabilities are needed. Wyckoff et al. (1996) reported unidentified
lines, which are fairly strong and presumably due to ionic lines, in the spectra of
plasma tails of several comets. Bands of other ionic molecules, such as OH+, CH+,
H2O+, and N2+, are awaiting modeling for use in the analyses of plasma tail spectra
and in the identification of those unidentified lines.
B. INFRARED SPECTRA
Before P/Halley, the low-resolution spectroscopy of comets in the infrared region
yielded only continuum emissions of dust and/or ice through the 3-, 10-, and 20-μm
atmospheric windows (e.g., Tokunaga et al., 1984). Recently, spectroscopic obser-
vations of comet Hale–Bopp made by the ISO produced a spectacular panorama
of the infrared spectrum of dust particles covering the range 7–45 μm (Crovisier
et al., 1997). The spectral structure of the dust spectrum has been found to be
remarkably similar to that of Mg-rich olivine-like forsterite (Crovisier, 1998).
The FTS on the NASA/Kuiper Airborne Observatory (KAO) successfully re-
solved individual lines of the 2.65-μm band of H2O in the coma of P/Halley
(Mumma et al., 1986). A low-resolution infrared spectrometer (IKS) on the Vega
spacecraft also detected molecular emissions from the bands of H2CO, CO2, and
possibly CO, in addition to the H2O band in the 2.5–5 μm region of P/Halley
(Moroz et al., 1987). After P/Halley, Hoban et al. (1993) detected emission lines
of the 3.52-μm band of CH3OH in comet P/Swift–Tuttle.
The recent bright comets Hale–Bopp and Hyakutake with high-resolution in-
frared spectroscopy provided opportunities to detect several parent molecules un-
ambiguously. In particular, symmetric molecules, such as CH4, C2H6, and C2H2
having no permanent electric dipole moment could not be detected by radio obser-
vations. Mumma et al. (1996) positively detected infrared lines of C2H6, CH4, and
CO in the coma of Hyakutake. They found an abundant C2H6/CH4 ratio, which is
consistent with the production of C2H6 via hydrogenation of C2H2 on grain sur-
faces, or UV photolysis of ice mixtures in interstellar clouds (Mumma et al., 1996;
Hasegawa and Herbst, 1993), and not consistent with a thermochemically equili-
brated region in the proto-solar nebula (e.g., Prinn and Fegley, 1989). Almost at the
same time as Mumma et al. (1996), Brooke et al. (1996) observed Hyakutake and
FIG. 9. The plots show CSHELL/IRTF spectra (solid line) of comet Hale–Bopp in the regions having
gaseous emissions from H2O, CH4, C2H6, C2H2, CO, and HCN (Weaver et al., 1997). The spectra have
not been corrected for atmospheric transmittance; gases in the terrestrial atmosphere produce the many
observed absorption features. The gaseous emissions are superimposed on a strong continuum emitted
by the cometary grains, which have been modeled (dashed line).
detected the 3-μm band lines of C2H2. The same molecular species, H2O, CH4,
C2H6, C2H2, and CO, in addition to HCN were detected in Hale–Bopp (Fig. 9;
Weaver et al., 1997). OCS was first detected in Hyakutake in radio (Woodney
et al., 1997), and subsequently in Hale–Bopp in infrared at 4.85 μm (Dello Russo
et al., 1998). The ratio of the OCS production rates relative to water production
rates are 3 × 10−3 for Hyakutake (Woodney et al., 1997), and 3–5 × 10−3 for
Hale–Bopp (Dello Russo et al., 1998).
Recently, Weaver et al. (1999) observed comet 21P/GZ and found that C2H6
is depleted at least by a factor of ∼10 compared to its relative abundances in
Hale–Bopp and Hyakutake. This depletion is similar to that observed for C2 and
C3 from optical observations of GZ (A’Hearn et al., 1995), suggesting that the
formation of volatile carbon-chain molecules was inhibited in GZ. Almost at the

same time, Mumma et al. (2000) also observed comet 21P/GZ, and found that
C2H6 is substantially less abundant compared with Hale-Bopp and Hyakutake.
Recent Keck observations (Mumma et al., 1999) of comet C/Lee provide high
signal-to-noise-ratio spectra in the range 1–5.5 μm showing many unidentified
lines, which are awaiting identification.
C. RADIO SPECTRA
Until the positive detections of millimeter lines of H2CO, H2S, and CH3OH in
several comets reported by Bockelée-Morvan et al. (1991) and Crovisier et al.
(1991) in the early 1990s, radio observations of comets were not very successful.
The only parent molecule unambiguously detected in the microwave range before
1990 was HCN in comets P/Halley and Kohouteck 1973 XII (e.g., Schloerb et al.,
1987). The radio spectra of the recent bright comets Hyakutake and Hale–Bopp
yielded an avalanche of new parent molecules: NH3, HDO, HNC, CH3CN, CS2,
and possibly OCS and HNCO from Hyakutake; and HCOOH, HCOOCH3, HC3N,
NH2CHO, SO, SO2, H2CS, and positively OCS and HNCO from Hale–Bopp
(Fig. 10). (For summaries and reviews of these detections, see Lis et al., 1997;
Crovisier and Bockelée-Morvan, 1999; Bockelée-Morvan et al., 2000).
One of the surprises from these observations was the detection of SO and
SO2, as very stringent upper limits on their abundances in comets were deter-
mined previously from UV observations (Kim and A’Hearn, 1991). Recently, Kim
et al. (1997b), using updated information available in the literature, found that Kim
and A’Hearn (1991) overestimated the solar excitation rate of the UV transition
of SO by ∼2 orders of magnitude using an erroneous radiative lifetime available
at that time in the literature. For SO2, the nondetection in the UV range is still
in question. Since the UV band of SO2 has not been sufficiently analyzed in the
literature, detailed laboratory and theoretical analyses of the band are needed to
resolve this issue.
The derived abundances of parent molecules in Hyakutake and Hale–Bopp
are found to be approximately consistent with those inferred in interstellar ices,
hot cores of molecular clouds, and bipolar flows around protostars (e.g., Bockelée-
Morvan et al., 2000; Irvine et al., 2000). The derived abundances from radio obser-
vations along with results from infrared observations suggest that cometary mate-
rial was mainly formed by similar nonequilibrium processes as produce interstellar
ices: grain-surface reactions, condensation of products of ion–molecule reactions,
and UV processing. The subsequent process during the planetary formation in
the proto-solar nebula did not seem to significantly influence the compositions of
the cometary material (e.g., Bockelée-Morvan et al., 2000). These new findings
become a great challenge to theorists—the nonequilibrium chemical models of
Willacy et al. (1998) and Finocchi et al. (1997) produce molecular abundances
FIG. 10. Millimeter spectra of Hale–Bopp containing emission lines of SO (observed at Caltech
Submillimeter Observatory, CSO), SO2 (Institut de Radioastronomie Millimétrique, IRAM, Plateau-
de-Bure interferometer, PdB), OCS (CSO), HC3N (CSO), HNCO (CSO), NH2CHO (IRAM 30-m),
HCOOH (IRAM PdB), and HCOOCH3 (IRAM 30-m) (Bockelée-Morvan et al., 2000). The velocity
frame is with respect to the comet nucleus velocity. The dashed line superimposed on the observed
spectrum of HCOOCH3 is a synthetic profile calculated using HC3N line at 227.419 GHz observed
at the same time. It takes into account that the HCOOCH3 line at ∼225.562 GHz is a blend of eight
transitions whose positions are shown.
in the inner part of the proto-solar nebula, which are significantly different from
those of comets.
D. ISOTOPIC ABUNDANCES
Kleine et al. (1995) derived a 12C/13C ratio of 95 ± 12 analyzing high-resolution
CN spectra of P/Halley between 3860 and 3890 Å, and the result agrees with the
solar ratio of ∼90. The agreement suggests that comets should be coevolved from
the same molecular cloud as other members of the solar system. The recent bright
comets Hale-Bopp and Hyakutake yielded high-signal-to-noise-ratio spectra of
radicals in the visible region (Cochran et al.,1997; Meier et al.,1998a). Meier
et al. (1998a) analyzed high-resolution spectra of the NH (3340–3380 Å) and
CH (4280– 4340 Å) bands of Hyakutake, and derived upper limits of ND/NH
and CD/CH (Fig. 11) ratios, which are 6 × 10−3 and 3 × 10−2, respectively. Millar
et al. (1989) investigated gas-phase chemistry including ion–molecule interactions
in dense interstellar clouds in order to study deuterium fractionations of various
molecules. According to their model calculations, the ND/NH and CD/CH ratios
FIG. 11. Measured (top) and modeled (bottom) A–X (0–0) bands of CH and CD of comet Hyakutake
(Meier et al., 1998a). Solid lines represent CH features and open lines belong to CD lines. Note that
the modeled spectrum has a different y scale than measurements.
should be in the ranges of 1 × 10−2 to 2 × 10−4 and 1 × 10−2 to 4 × 10−3, re-

spectively, and these ratios are significantly greater than the HD/H2 ratio, which
is ∼ 3 × 10−5.
Isotopic molecules, such as HDO and H13CN in Hyakutake, and DCN, H13CN,
HC15N, and C34S in Hale–Bopp, were also detected in the radio range (e.g.,
Bockelée-Morvan 1998; Meier et al., 1998b, 1998c; Crovisier and Bockelée-
Morvan, 1999). The derived D/H ratios in H2O from the H2O and HDO radio
lines of Hyakutake and Hale–Bopp as well as the D/H ratio of P/Halley measured
from a mass spectrometer on the Giotto spacecraft are all ∼3 × 10−4. This value
is approximately consistent with observed D/H ratios in the hot cores of molecular
clouds, but lower than those measured in carbonaceous meteorites and interstel-
lar ices. The lower D/H ratios in comets than in interstellar ices suggest that the
diffusion process in the warm inner part of the proto-solar nebula partially diluted
the originally higher ratios in the interstellar ices (Bockelée-Morvan et al., 1998,
2000; Irvine et al., 2000). The D/H ratios from the three comets are clearly ∼2
times higher than terrestrial standard mean ocean water (SMOW).
The D/H ratio in HCN derived from HCN and DCN lines is 2.3 × 10−3 (Meier
et al.,1998b), and the ratio is in the range reported in the hot cores of molecular
clouds (Hatchell et al.,1998). The derived ratio is consistent with results from
interstellar ion–molecule chemistry model (Millar et al.,1989). The hot cores,
however, might be too young to reach the steady state, and the observed ra-
tio might have been diluted from an originally higher value of interstellar ices
(Hatchell et al.,1998). Therefore, additional observations, laboratory, and theo-
retical works are needed to clarify the connection between the cometary mate-
rial and the interstellar matter. The isotopic ratios of carbon, nitrogen, and sulfur
appear to be cosmic, because the fractionations of heavy species, such as car-
bon, nitrogen, and sulfur, are less sensitive to chemical reactions in cometary
nuclei.
E. COLLISIONAL PROCESSES FOR COMETARY SPECTRA

As shown from the above OH and NH research in the visible spectroscopy, the
rotational structures of the radicals may be influenced by collisions with neutrals
and electrons in the cometary environment. The difficulty in applying the colli-
sion influences to the spectral structures is that experimental results on the cross
sections of the radicals are very limited in the literature. In particular, rotational
excitation rates of cometary molecules via collisions with cometary molecules
are poorly known. Only several cases, such as H2O <− (excited by) H2, CO <−
H2O, CO <− N2, N2 <− H2O, etc., have been investigated in detail by S. Green
(e.g., Phillips et al., 1996; Green, 1993, 1995; references therein), whose untimely
death prevented further progresses in this important area. S. Green constructed
cross sections theoretically. The theoretical results were tested by experimental
measurements of pressure-broadening line widths, although the experimental line
widths are not sufficient to generate the excitation rates inversely. Since the most
abundant neutral species in comas are H2O, OH, and CO, the collision cross sec-
tions of cometary radicals with these neutrals at coma temperatures (10–500 K)
are needed. In order to calculate collision frequency, neutral densities of H2O,
OH, and CO are needed. The OH density, which should be similar to the H2O
density, has been measured in the coma of comet P/Halley by instruments on
Vega and Giotto (e.g., Fig. 4 of Kim et al., 1989). The rotational excitations by
molecule–molecule collisions are important only in inner coma regions, where
neutral densities are sufficiently high to excite or deexcite the cometary molecules
compared with other excitation processes, such as electron collisions and radiation
excitations.
Electron densities in the outer coma region (1 × 104 to 9 × 105 km from the
nucleus) of P/Halley as a function of cometocentric distance were measured by
Vega 1 and 2 (Pederson et al., 1987). Kinetic energies (which can be translated into
kinetic temperatures) of electrons for P/Halley were also derived to be 0.5–1.0 eV
in the outer coma region (Pederson et al., 1987). An ion mass spectrometer on
Giotto measured ion densities, which may be used as electron densities assuming
an overall charge neutrality in the inner coma region (103–104 km) of P/Halley.
In the outer coma regions, where neutral densities decrease exponentially, colli-
sions with electrons become significant (e.g., Fig. 4 of Kim et al., 1989). Infor-
mation on electron impact cross sections for rotational excitations of cometary
molecules is very limited. In particular, virtually no experimental data are avail-
able for these processes so far. A few approximate theoretical methods have been
proposed for the calculation of the rotational cross sections, but their reliability is
uncertain.
The rotational excitations and deexcitations of cometary molecules by colli-
sions with electrons, ions, and neutrals equally influence the rotational structures
of the ultraviolet and visible (electronic), infrared (rovibrational), and radio (pure
rotational) bands. It is also noted that vibrational or rovibrational excitations of
molecules by electrons, ions, or neutrals are not very important in cometary co-
mas compared with excitations by solar infrared radiation, but these vibrational
transitions are important processes in planetary auroral regions, as mentioned in
Section II.D. For a massive comet, such as Hale–Bopp, the vibrational deactivation
rates of cometary molecules by collisions with water molecules near the nucleus
is significant compared with vibrational Einstein A coefficients of the cometary
molecules, and, therefore, a LTE state can be formed near the nucleus.
IV. Spectral Databases and Improvements Needed

Analysis of observed spectra relies largely on available spectral databases. These
databases, however, are not sufficient for the analysis of the planetary or cometary
spectra. In this section we propose possible improvements of the spectral databases.
Since the early 1980s, spectroscopists at the Air Force Geophysical Laboratory
(AFGL), the Goddard Space Flight Center (GSFC), the Jet Propulsion Labora-
tory (JPL), and the Centre National de la Recherche Scientifique (CNRS) have
constructed/compiled spectroscopic databases (the HITRAN, MLA, ATMOS, and
GEISA databases, respectively) in standardized formats mainly for infrared lines
of the molecules mostly in the Earth’s atmosphere (e.g., Rothman et al., 1987;
Bjoraker et al., 1986; Brown and Toth, 1985; and Husson, et al., 1986, respec-
tively). These databases accommodated available spectroscopic line parameters
published in the literature, and added new spectroscopic information generated
by the spectroscopists in their institutions. Thus, the majority of the line informa-
tion in these databases overlaps each other. In the submillimeter, millimeter, and
centimeter ranges, Poynter and Pickett (1985) have been compiling spectroscopic
parameters for more than 100 molecules and atoms.
Although the above spectroscopic databases have been continually updated by
the authors over the last 20 years, planetary scientists frequently find that needed
spectroscopic parameters for certain molecules are not included in the databases.
For example, line positions, quantum-state designations for the lines, line inten-
sities, line widths, and lower-state energies, which are basic information to be
used for the analyses of planetary spectra, are very poorly known for the over-
tone and combination bands of hydrocarbons, such as C2H2, C2H4, C2H6, C3H8,
C3H4, C4H2, and C6H6; the nitrogen-containing molecules NH3 and HCN; and
molecules containing heavy elements, such as H2S, AsH3, GeH4, and PH3. Even
for H2O, CO2, CO, and CH4, which are relatively well studied compared with the
above molecules, the spectroscopic parameters for the overtone and combination
bands are not completely studied. The intensities of far-infrared bands are severely
affected by hot bands due to their relatively low energy states. The intensities of
far-infrared fundamental bands are often substantially lower than the total hot band
intensities even at room temperature. The hot band influence is significant even in
outer planets’ cold environment (60–150 K). Some of the fundamental bands of
C2H2, C3H4, C4H2, C6H6, CO2, and H2O occur at wavelengths longer than 10 μm,
where hot band intensities become significant. The spectroscopic parameters for
these hot bands are available only in part, with the exception of CO2 and H2O.
As we discussed in Section II.B, we expect to detect additional airglow emissions
from the bands of CH4 and other hydrocarbons in planetary atmospheres. The most
extensive compilation of these CH4 band transitions can be found in a recent work
of Wenger and Champion (1998), who listed absorption line intensities of the CH4
bands. For the analysis of the emission lines, we need Einstein A coefficients for
individual line transitions. Although Einstein A coefficients for individual lines
can be calculated from the absorption line intensities using Eq. (2–17) of Penner
(1959), it is not usually straightforward to obtain the Einstein A coefficients. For
speedy analyses of planetary airglow emissions, we need a column of Einstein A
coefficients for planetary molecules in the databases, such as HITRAN, ATMOS,
GEISA, etc.
V. Conclusions
The successful spectroscopic observations of atmospheric molecules in the atmo-
spheres of the giant planets, Titan, and comets, and the derived elemental abun-
dances in these atmospheres during the last three decades leave a fundamental
task for planetary scientists—how to explain the origin and chemical evolution
of the solar system bodies based on the derived elemental abundances. The cur-
rent elemental abundances of the giant planets support a planetary formation sce-
nario that the cores and atmospheres of the giant planets were made from accre-
tions of icy planetesimals and from the capture of gases from the surrounding
proto-solar nebula. The current elemental abundances derived from comets
Hyakutake and Hale–Bopp are approximately consistent with molecular abun-
dances derived from interstellar ices, hot cores of molecular clouds, and a subse-
quent planetary formation process. There have also been active spectroscopic and
image observations of aurorae and airglows on the giant planets. However, detailed
auroral and airglow processes are not clearly determined compared with those of
Earth’s aurorae and airglows, partially due to lack of information on the collision
cross sections of auroral molecules. Hundreds of lines in cometary and planetary
spectra are unidentified. Proper spectroscopic studies of these molecular bands are
urgently needed to identify molecules, and to properly analyze the spectra of the
atmospheres of comets and the giant planets. Additional laboratory and theoretical
works are essential to improve the current solar system formation scenario and to
better understand auroral and airglow processes. Success of planned future mis-
sions and ground-based observations as well as further theoretical and laboratory
studies of molecules will provide significant progress in understanding the origin
and evolution of the solar system, and chemical processes of molecules in the
atmospheres of the giant planets, Titan, and comets.
VI. Acknowledgments
The present review chapter is a product of the cooperative research of the Core Uni-
versity Program on Energy Science and Engineering between the Seoul National
University and the Kyoto University. SJK, YHK, and YCM acknowledge finan-
cial support provided by a grant (1999-1-113-001-5) from the Interdisciplinary
Research Program of the Korean Science and Engineering Foundation. We would
like to thank D. Bockelée-Morvan for providing Fig. 10.
VII. References
A’Hearn, M. F. (1978). Astrophys. J. 219, 768.

A’Hearn, M. F., Feldman, P. D., and Schleicher, D. G. (1983). Astrophys. J. 274, L99.
A’Hearn, M. F., Millis, R. L., Schleicher, D. G., Osip, D. J., and Birch, P. V. (1995). Icarus 118, 223.
Aikawa, Y., Umebayashi, T., Nakano, T., and Miyama, S. M. (1999). Astrophys. J. 519, 705.
Ajello, J., et al. (1998). J. Geophys. Res., 103, 20149.
Appleby, J. F. (1990). Icarus 85, 355.
Arpigny, C., Dossin, F., Woszczyk, A., Donn, B., Rahe, J., and Wyckoff, S. (1991). An abstract presented
at the International Conference on Asteroids, Comets, Meteors 1991, Flagstaff, AZ, June 24–28,
1991.
Atreya, S. K., Wong, M. H., Owen, T. C., Mahaffy, P. R., Niemann, H. B., de Pater, I., Drossart, P., and
Encrenaz, Th. (1999). Planet. Space Sci. 47, 1243.
Ballester, G. E., Miller, S., Tennyson, J., Trafton, L. M., and Geballe, T. R. (1994). Icarus 107, 189.
Baron, R., Joseph, R. D., Owen, T., Tennyson, J., Miller, S., and Ballester, G. E. (1991). Nature 353,
539.
Ben Jaffel, L., Leers, V., and Sandel, B. R. (1995). Science 269, 951.
Bézard, B., Encrenaz, T., Lellouch, E., and Feuchtgruber, H. (1999). Science 283, 800.
Bisikalo, D. V., Shematovich, V. I., Gerard, J.-C., Gladstone, G. R., and Waite, Jr., J. H. (1996).
J. Geophys. Res. 101, 21157.
Bjoraker, G. L., Larson, H. P., and Kunde, V. G. (1986). Icarus 66, 579.
Bockelée- Morvan, D., Colom, P., Crovisier, J., Despois, D., and Paubert, G. (1991). Nature 350, 318.
Bockelée- Morvan, D., et al. (1998). Icarus 133, 147.
Bockelée- Morvan, D., et al. (2000). Astron. Astrophys. 353, 1101.
Borysow, A., and Frommhold, L. (1986). Astrophys. J. 303, 495.
Borysow, A., and Tang, C. (1993). Icarus 105, 175.
Borysow, A., Jørgensen, U. G., and Zheng, C. (1997). Astron. Astrophys. 324, 185.
Broadfoot, A. L., et al. (1979). Science 204, 979.
Broadfoot, A. L., et al. (1981a). J. Geophys. Res. 86, 8259.
Broadfoot, A., et al. (1981b). Science 212, 206.
Brooke, T. Y., Tokunaga, A. T., Weaver, H. A., Crovisier, J., Bockelee-Morvan, D., and Crisp, D. (1996).
Nature 383, 606.
Brown, L. R., and Toth, R. A. (1985). J. Opt. Soc. Am. B 2, 842.
Brown, M. E., Bouchez, A. H., Spinrad, A. H., and Johns-Krull, C. M. (1996). Astrophys. J. 112, 1197.
Clarke, J. T., Moos, H. W., Atreya, S. K., and Lane, A. L. (1980). Astrophys. J. 241, L179.
Clusius, K., and Douglas, A. E. (1954). Can. J. Phys. 32, 319.
Cochran, A. L., Barker, E. S., Cochran, W. D., and Lambert, D. L. (1997). Bull. Am. Astron. Soc. 29,
1050.
Conrath, B. J., and Gautier, D. (2000). Icarus 144, 124.
Crovisier, J. (1998). Bull. Am. Astron. Soc. 30, 1059.
Crovisier, J., and Bockelée-Morvan, D. (1999). “Composition and Origin of Cometary Materials” p. 1.
Kluwer Academic, Dordrecht,, The Netherlands.
Crovisier, J., Despois, D., Bockelée-Morvan, D., Colom, P., and Paubert, G. (1991). Icarus 93, 246.
Crovisier, J., Leech, K., Bockelée-Morvan, D., Brooke, T. Y., Hanner, M. S., Altieri, B., Keller, H. U.,
and Lellouch, E. (1997). Science 275, 1904.
Danielson, R. E., Caldwell, J. J., and Larach, D. R. (1973). Icarus 20, 437.
de Bergh, C., Lutz, B. L., Owen, T., and Maillard, J. P. (1990). Astrophys. J. 355, 661.
de Pater, I., Romani, P. N., and Atreya, S. K. (1991). Icarus 91, 220.
Dello Russo, N., Disanti, M. A., Mumma, M. J., Magee-Sauer, K., and Rettig, T. W. (1998). Icarus
135, 377.
Dinelli, B. M., et al. (1997). Icarus 126, 107.
Dossin, F., Fehrenbach, Ch., Haser, L., and Swings, P. (1961). Ann. Astrophys. 24, 519.
Dressler, K., and Ramsay, D. A. (1959). Phil. Trans. R. Soc. A 251, 553.
Drossart, P., et al. (1989). Nature 340, 539.
Drossart, P., Fouchet, Th., Crovisier, J., Lellouch, E., Encrenaz, Th., Feuchtgruber, H., and Champion,
J.-P. (1999). Proc. Conf. “The Universe as Seen by ISO,” Paris, France, 20–23 Oct. 1998 (ESA
SP-427, March 1999).
Eberhardt, P., Reber, M., Krankowsky, D., and Hodges, R. R. (1995). Astron. Astrophys. 302, 301.
Fegley, B., Jr. (1997). In “Analysis of Returned Comet Nucleus Samples,” (S. Chang, Ed.), p. 73.
NASA CP 10152.
Feldman, P. D., Takacs, P. Z., Fastie, W. G., and Donn, B. (1974). Science 185, 705.
Finocchi, F., Gail, H.-P., and Duschl, W. J. (1997). Astron. Astrophys. 325, 1264.
Fox, K., and Kim, S. (1984). Bull. Am. Astron. Soc. 16, 706.
Frommhold, L., Samuelson, R., and Birnbaum, G. (1984). Astrophys. J. 283, L79.
Gausset, L., Herzberg, G., Lagerqvist, A., and Rosen, B. (1965). Astrophys. J. 142, 45.
Gautier, D., and Owen, T. (1983). Nature 304, 691.
Gautier, D., and Owen, T. (1989). In “Origin and Evolution of Planetary and Satellite Atmospheres,”
S. K. Atreya et al., Eds.), p. 487. Tucson: University of Arizona Press.
Gautier, D., et al. (1981). J. Geophys. Res. 86, 8713.
Geballe, T. R., Jagod, M. F., and Oka, T. (1992). IAU Circ. No. 5566.
Gillett, F. C. (1975). Astrophys. J. 201, L41.
Gillett, F. C., Low, F. J., and Stein, W. A. (1969). Astrophys. J. 157, 925.
Gladstone, G. R., Waite, J. H., Jr., and Lewis, W. S. (1998). J. Geophys. Res., 103, 20083.
Gredel, R., van Dishoeck, E. F., and Black, J. H. (1989). Astrophys. J. 338, 1047.
Green, S. (1993). Astrophys. J. 412, 436.
Green, S. (1995). Astrophys. J. Suppl. 100, 213.
Grossman, A. W. (1990). Ph.D. thesis, California Institute of Technology.
Hanel, R., et al. (1979). Science 204, 972.
Hanel, R., et al. (1981). Science 212, 192.
Hasegawa, T. I., and Herbst, E. (1993). Mon. Not. R. Astron. Soc. 263, 589.
Hatchell, J., Millar, T. J., and Rodgers, S. D. (1998). Astron. Astrophys. 332, 695.
Herzberg, G. (1942). Astrophys. J. 96, 314.
Herzberg, G. (1952). Astrophys. J. 115, 337.
Herzberg, G. (1971). “The Spectra and Structures of Simple Free Radicals: An Introduction to Molec-
ular Spectroscopy.” Cornell University Press, Ithaca,, New York.
Hirota, E., and Yamada, C. (1982). J. Mol. Spectrosc. 96, 175.
Hoban, S., Reuter, D. C., DiSanti, M. A., Mumma, M. J., and Elston, R. (1993). Icarus 105, 548.
Huggins, W. (1882). Proc. R. Soc. (Lond.) 33, 1.
Hunt, G. E., and Moore, V. (1984). J. Quant. Spectrosc. Radiat. Transf. 32, 439.
Hunten, D. M. (1978). NASA Conf. Publ. 2068, 127.
Husson, N., Chedin, A., and Scott, N. A. (1986). Ann. Geophys. 4A, 185.
Ingersoll, A. P., et al. (1998). Icarus 135, 251.
Irvine, W. M., Schloerb, F. P., Crovisier, J., Fegley, B., Jr., and Mumma, M. J. (2000). In “Protostars and
Planets IV” (V. Mannings, A. Boss, and S. Russel, Eds.), p. 1159. University of Arizona Press,
Tucson.
James, G. K., Ajello, J. M., and Pryor, W. R. (1998). J. Geophys. Res. 103, 20113.
Johns, J. W. C., Ramsay, D. A., and Ross, S. C. (1976). Can. J. Phys. 54, 1804.
Jungen, Ch., Hallin, K-E. J., and Merer, A. J. (1980). Mol. Phys. 40, 25.
Kawakita, H., Ayani, K., and Kawabata, T. (2000). P. Astrom. Soc. Japan 52, 925.
Kiess, C. C., Corliss, C. H., and Kiess, H. K. (1960). Astrophys. J. 132, 221.
Kim, S. J., Caldwell, J., Rivolo, A. R., Wagener, R., and Orton, G. S. (1985). Icarus 64, 233.
Kim, S. J., (1988). Icarus 75, 399.
Kim, S. J. (1999). Earth, Planets, Space 51, 139.

Kim., S. J., and A’Hearn, M. F. (1991). Icarus 90, 79.
Kim, S., and Maguire, W. (1986). NASA Conf. Pub. 2441, 95.
Kim, S. J., A’Hearn, M. F., and Cochran, W. D. (1989). Icarus 77, 98.
Kim, S. J., A’Hearn, M. F., and Larson, S. M. (1990). Icarus 87, 440.
Kim, S. J., Drossart, P., Caldwell, J., Maillard, J-P, Herbst, T., and Shure, M. (1991). Nature 353, 536.
Kim, S. J., Trafton, L. M., Geballe, T. R., and Slanina, Z. (1995). Icarus 113, 217.
Kim, S. J., Orton, G. S., Dumas, C., and Kim, Y. H. (1996). Icarus 120, 326.
Kim, S. J., Brown, M., and Spinrad, H. (1997a). J. Geomag. Geoelect. 49, 1165.
Kim, S. J., Bockelée-Morvan, D., Crovisier, J., and Biver, N. (1997b). Earth Moon Planets 78, 65.
Kim, S. J., Lee, D. H., and Kim, Y. H. (1998). Rept. Prog. Phys. 61, 525.
Kim, S. J., Geballe, T. R., and Noll, K. (2000). Icarus 147, 588.
Kleine, M., Wyckoff, S., Wehinger, P. A., and Peterson, B. A. (1994). Astrophys. J. 436, 885.
Kleine, M., Wyckoff, S., Wehinger, P. A., and Peterson, B. A. (1995). Astrophys. J. 439, 1021.
Knacke, R. F., Kim, S. J., Ridgway, S. T., and Tokunaga, A. T. (1982). Astrophys. J. 262, 388.
Kuiper, G. P. (1944). Astrophys. J. 100, 378.
Lam, H. A., Miller, S., Joseph, R. D., Geballe, T. R., Trafton, L. M., Tennyson, J., and Ballester, G. E.
(1997). Astrophys. J. 474, L73.
Lellouch, E. (2000). In “Astrochemistry: From Molecular Clouds to Planetary Systems” (Y. C. Minh
and E. F. van Dishoeck, Eds.), p. 491. IAU, 2000, International Astronomical Union. Sheridan
Books, Inc., Chelsea, Michigan.
Lis, D. C., et al. (1997). Icarus 130, 355.
Lutz, B. L., de Bergh, C., and Owen, T. (1983). Science 220, 1374.
Magnani, L. and A’Hearn, M. F. (1986). Astrophys. J. 302, 477.
Mahaffy, P. R., Donahue, T. M., Atreya, S. K., Owen, T. C., and Niemann, H. B. (1998). Space Sci.
Rev. 84, 251.
Maillard, J-P., Drossart, P., Watson, J. K. G., Kim, S. J., and Caldwell, J. (1990). Astrophys. J. 363, L37.
Manuel, O., Windler, K., Nolte, A., Johannes, L., Zirbel, J., and Ragland, D. (1998). J. Radioanal.
Nuclear Chem. 238, 119.
McKellar, A. R. W. (1984). Can. J. Phys. 62, 760.
McKellar, A. R. W. (1995). In “Collision and Interaction Induced Spectroscopy” (G. C. Tabisz and
M. N. Neuman, Eds.), p. 467. Kluwer, Dordrect, The Netherlands.
McKellar, A. R. W. (1996). J. Chem. Phys. 105, 2628.
McKellar, A. R. W., Roth, D. A., Pak, I., and Winnewisser, G. (1999). J. Chem. Phys. 110, 9989.
Meier, R., Wellnitz, D., Kim, S. J., and A’Hearn, M. F. (1998a). Icarus 136, 268.
Meier, R., Owen, T. C., Jewitt, D. C., Matthews, H. E., Senay, M., Biver, N., Bockelée-Morvan, D.,
Crovisier, J., and Gautier, D. (1998b). Science 279, 1707.
Meier, R., Owen, T. C., Matthews, H. E., Jewitt, D. C., Bockelée-Morvan, D., Biver, N., Crovisier, J.,
and Gautier, D. (1998c). Science 279, 842.
Metzger, A. E., Gilman, D. A., Luthey, J. L., Hurley, K. C., Schnopper, H. W., Seward, F. D., and
Sullivan, J. D. (1983). J. Geophys. Res. 88, 7731.
Millar, T. J., Bennett, A., and Herbst, E. (1989). Astrophys. J. 340, 906.
Miller, S. V., Joseph, R. D., and Tennyson, J. (1990). Astrophys. J. 360, L55.
Moroz, V. I., et al. (1987). Astron. Astrophys. 187, 513.
Mumma, M. J., Weaver, H. A., Larson, H. P., Williams, M., and Davis, D. S. (1986). Science 232, 1523.
Mumma, M. J., DiSanti, M. A., Dello Russo, N., Fomenkova, M., Magee-Sauer, K., Kaminski, C. D.,
and Xie, D. X. (1996). Science 272, 1310.
Mumma, M. J., Dello Russo, N., DiSanti, M. A., Gilbert, A. M., Graham, J. R., McLean, I. S., Becklin,
E. E., Figer, D. F., Larkin, J. E., Levenson, N. A., Teplitz, H. I., and Wilcox, M. K. (1999). Bull.
Am. Astron. Soc. 31, 1124.
Mumma, M. J., DiSanti, M. A., Dello Russo, N., Magee-Sauer, K., and Rettig, T. W. (2000). Astrophys.
J. 531, L155.
Niemann, H. B., et al. (1996). Science 272, 846.
Niemann, H. B., et al. (1998). Adv. Space Res. 21, 1455.
Noll, K. S., Geballe, T. R., and Knacke, R. F. (1989). Astrophys. J. 338, L71.
Noll, K. S., Larson, H. P., and Geballe, T. R. (1990). Icarus 83, 494.
Oka, T., and Geballe, T. R. (1990). Astrophys. J. 351, L53.
Owen, T., Mahaffy, P., Niemann, H. B., Atreya, S., Donahue, T., Bar-Nun, A., and de Pater, I. (1999).
Nature 402, 269.
Pedersen, A., Grard, R., Trotignon, J. G., Beghin, C., Mikhailov, Y., and Mogilevsky, M. (1987). Astron.
Astrophys. 187, 297.
Penner, S. S. (1959). “Quantitative Molecular Spectroscopy and Gas Emissivities.” Addison-Wesley,
London.
Phillips, T. R., Maluendes, S., and Green, S. (1996). Astrophys. J. Suppl. 107, 467.
Poynter, R. L., and Pickett, H. M. (1985). “Submillimeter, Millimeter, and Microwave Spectral Line
Catalogue.” JPL Pub. 80–23, Rev. 2.
Prinn, R. G., and Fegley, B., Jr. (1989). In “Origin and Evolution of Planetary and Satellite Atmo-
spheres.” (S. K. Atreya et al., Eds.), p. 78. University of Arizona Press, Tucson.
Pryor, W. R. (1998). J. Geophys. Res. 103, 20149.
Ridgway, S. T. (1974a). Astrophys. J. 187, L41.
Ridgway, S. T. (1974b). Bull. Am. Astron. Soc. 6, 376.
Rothman, L. S., et al. (1987). Appl. Opt. 26, 4058.
Schaefer, J., and McKellar, A. R. W. (1990). Z. Phys. D 15, 51.
Schleicher, D. G., and A’Hearn, M. F. (1982). Astrophys. J. 258, 864.
Schleicher, D. G., and A’Hearn, M. F. (1988). Astrophys. J. 331, 1058.
Schloerb, F. P., Kinzel, W. M., Swade, D. A., and Irvine, W. M. (1987). Astron. Astrophys. 187, 475.
Smith, B. A., et al. (1979). Science 204, 951.
Spirko, V., and Bunker, P. R. (1982). J. Mol. Spectrosc. 95, 381.
Swings, P., Elvey, C. T., and Babcock, H. W. (1941). Astrophys. J. 94, 320.
Tokunaga, A. T., Beck, S. C., Geballe, T. R., Lacy, J. H., and Serabyn, E. (1981). Icarus 48, 283.
Tokunaga, A. T., Hanner, M. S., Veeder, G. J., and A’Hearn, M. F. (1984). Astron. J. 89, 162.
Trafton, L. M. (1967). Astrophys. J. 147, 765.
Trafton, L. (1998). In “Planetary, Cometary, Interstellar Atmospheres” (A. A. Vigasin and Z. Slanina,
Eds.), p. 177. World Scientific, Singapore.
Trafton, L., Lester, D., Carr, J., and Harvey, P. (1987). International Workshop on Time Variable
Phenomena in the Jovian System, Flagstaff, AZ.
Trafton, L., Lester, D. F., and Thompson, K. L. (1989). Astrophys. J. 343, L73.
Trafton, L. M., Geballe, T. R., Miller, S., Tennyson, J., and Ballester, G. E. (1993). Astrophys. J. 405,
761.
Trafton, L. M., Gérard, J. C., Munhoven, G., and Waite, J. H., Jr. (1994). Astrophys. J. 421, 816.
Trafton, L. M., Miller, S., Geballe, T., Ballester, G., and Tennyson, J. (1995). Bull. Am. Astron. Soc.
27, 34.
Trafton, L. M., Kim, S. J., Geballe, T. R., and Miller, S. (1997). Icarus 130, 544.
Trafton, L. M., Dols, V., Gérard, J.-C., Waite, J. H., Jr., Gladstone, G. R., and Munhoven, G. (1998).
Astrophys. J. 507, 955.
Tyler, G. L., Eshleman, V. R., Anderson, J. D., Levy, G. S., Lindal, G. F., Wood, G. E., and Croft, T. A.
(1981). Science 212, 201.
Waite, Jr., J. H., Clarke, J. T., Cravens, T. E., and Hammond, C. M. (1988). J. Geophys. Res. 93, 7244.
Weaver, H. A., Brooke, T. Y., Chin, G., Kim, S. J., Bockelée-Morvan, D., and Davies, J. (1997). Earth,
Moon, Planets 78, 71.
Weaver, H., et al. (1999). Icarus 142, 482.

Wenger, C. H., and Champion, J. P. (1998). J. Quant. Spectrosc. Radiat. Transfer 59, 471.
Whipple, F. L. (1950). Astrophys. J. 111, 375.
Whipple, F. L. (1951). Astrophys. J. 113, 464.
Wildt, R. (1932). Veröff. Univ. Sernwarte Göttingen 2(22), 171.
Willacy, K., Klahr, H. H., Millar, T. J., and Henning, Th. (1998). Astron. Astrophys. 338, 995.
Woodney, L. M., McMullin, J., and A’Hearn, M. F. (1997). Planet. Space Sci. 45, 717.
Wyckoff, S., Fox, R., Wehinger, P., Heyd, R., and Ferro, A. (1996). Bull. Am. Astron. Soc. 28, 1093.
Yamada, C., Hirota, E., and Kawaguchi, K. (1981). J. Chem. Phys. 75, 5256.
Yelle, R. V. (1991). Astrophys. J. 383, 380.
Zheng, C., and Borysow, A. (1995). Icarus 113, 84.
STUDIES OF ELECTRON-EXCITED
TARGETS USING RECOIL
MOMENTUM SPECTROSCOPY
WITH LASER PROBING OF
THE EXCITED STATE
ANDREW JAMES MURRAY
Department of Physics and Astronomy, The University of Manchester, Manchester M13
9PL, United Kingdom
PETER HAMMOND
Department of Physics, University of Western Australia, Nedlands, Perth WA6907,
Australia
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
II. Atomic Deflection Using Electron Impact. . . . . . . . . . . . . . . . . . . . . . . 167
A. Principles of the Deflection Technique . . . . . . . . . . . . . . . . . . . . . . 167
B. Experimental Recoil Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . 173
C. Differential Cross-Section Measurements . . . . . . . . . . . . . . . . . . . . 177
III. Doubly Excited States Studied via the Fluorescence Decay Product:
Recoiling Excited Atoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..... 182
A. Principles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..... 182
B. Experimental Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . ..... 184
C. Observations of Momentum-Analyzed Doubly-Excited-State
Fluorescence Decay Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
IV. Stepwise Laser Probing of Deflected Metastable Targets . . . . . . . . . . . . . 187
A. Principles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
B. Experimental Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
C. Results from Laser Probing of Metastable Targets . . . . . . . . . . . . . . . 192
V. Conclusions and Future Experiments. . . . . . . . . . . . . . . . . . . . . . . . . . 199
VI. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
VII. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
I. Introduction
Historically, experiments studying electron scattering from gas-phase targets have
overwhelmingly concentrated on detection of the scattered electron. This ex-
perimental bias is due to the relative ease of producing an electron beam in a

ISBN 0-12-003847-1/ISSN 1049-250X/01 $35.00
164 Andrew James Murray and Peter Hammond
well-defined state prior to the interaction, where the energy and momentum of the
electron are accurately controlled [1, 2]. Further, analysis of the scattered electron
following elastic scattering [3, 4], inelastic scattering [5], or ionization [6, 7] of
the target can be achieved to high accuracy using electrostatic detectors designed
to measure the electron momentum.
It is less usual to detect the target momentum following the interaction, although
by conservation of momentum and energy the same information can in principle
be extracted. As in conventional electron-scattering experiments, it is necessary
to define to high accuracy the initial and final momenta of the target undergoing
the interaction to obtain meaningful results. In most gas-phase experiments, an
effusive beam is employed from either a hypodermic needle or from an oven, and
so the distribution of initial target velocities and directions is not well defined. It
is therefore difficult to accurately determine collisional parameters from the target
momenta in these studies.
Although experiments which measure the target recoil are difficult, they have
been performed for a number of years, since there can be advantages in using
such techniques. Bederson [8] pioneered these studies by employing a target of
well-defined initial momentum and measuring the deflection (recoil) following
electron interaction. Since these initial experiments, Bederson and co-workers
have exploited this technique to measure cross sections for elastic, inelastic, and
superelastic scattering from atoms and for molecules [9–12]. These experiments
measure cross sections in the low-energy regime at forward scattering angles,
where it is difficult to employ conventional techniques because of to the difficulty
of observing electrons scattered in the forward direction.
Bederson and co-workers employ a crossed-beam geometry where the electron
beam and target beam interact at right angles as shown in Fig. 1. The targets,
deflected through a recoil angle α by electrons scattering through an angle θe,
FIG. 1. Recoil of a target in a crossed-beam interaction. The incident target of well-defined momen-
tum interacts with an electron which scatters through an angle θe . Momentum conservation requires
that the target is deflected through an angle α.
STUDIES OF ELECTRON-EXCITED TARGETS 165
are detected simultaneously with the direct, unscattered target beam using the
same detector. It is therefore possible to measure absolute cross sections. These
measurements are constrained to forward electron-scattering angles which are
inferred from the deflected target angles.
Exploitation of a well-defined initial target momentum has also been employed
in a new type of ionization experiment recently invented, known as cold target recoil
ion momentum spectroscopy (COLTRIMS) [13–17]. This technique can produce
large amounts of information, and has been used in electron-impact, ion-impact,
and photoionization studies. A target beam of well-defined initial momentum is
produced, and the momenta of the recoiling ion and electron(s) following the
collision is measured. COLTRIMS experiments can sample the entire momentum
space, and so cross sections that are differential in both energy and angle are
determined. Since the emitted electrons are detected over a 4π solid angle, absolute
measurements can be performed.
One further experiment that exploits the initial target momentum and which
is relevant here is the inelastic electron-scattering process which leaves the target
in a metastable state. In this case, the interaction produces a deflection and the
target is left with internal energy. Momentum and energy are again conserved, and
so it is possible to determine the electron-scattering angle which correlates with
the deflected target. The cross section, differential in energy and angle, can then
be calculated. One advantage of this technique is that the internal energy can be
employed for detection of the neutral metastable target. When a metastable target
interacts with a surface, the internal energy is transferred to the surface and an
electron may be emitted. This electron can be detected with high efficiency using
an electron detector such as a channel electron multiplier (CEM) [18], channel
plate (CP) [19], or discrete dynode electron detector [20]. In many experiments
the internal energy of the target is sufficiently high to directly emit an electron
from the front surface of the detector (around 8 eV of internal energy is required
for most commercial devices [21]). In this case, the detector can be used directly
to measure the neutral metastable atoms. Metastable targets with lower internal
energy can also be detected in collision with heated low-work-function materials,
with the electron emitted from the surface being subsequently detected [22].
A number of experiments have exploited the internal energy of metastable
targets. Brunt et al. [23] and Buckman et al. [24] used a high-resolution electron
source to study short-lived negative ion states that decay by autodetachment to
produce metastable helium. These experiments used an effusive helium beam from
a hypodermic needle, and calculated an average deflection angle from the peak
of the target Doppler profile. The metastable atom detector, consisting of a CEM
with a cone-shaped entrance aperture, was placed at this angle and the total yield
of excited atoms was measured as a function of electron impact energy as shown
in Fig. 2. The cross section was therefore measured differential in energy, with no
angular information being ascertained. A rich variation of metastable yield was
FIG. 2. Threshold metastable excitation of helium using a high-energy-resolution electron beam.

(From Ref. [24].)
observed as the incident energy varied from threshold to ionization at 24.6 eV,
arising from the decay of temporary negative ion states by autodetachment into
the metastable states, together with cascade contributions from higher states. This
technique has been used to study a variety of different atomic and molecular targets
[25, 26, and references therein].
Metastable atom recoil angular measurements were made by Shpenik and co-
workers [27], who studied electron-impact excitation of metastable levels of all
noble gases. In these experiments the incident target beam momentum was well de-
fined using a supersonic expansion. A cross section differential in energy was there-
fore determined. These authors compared their work with calculations using a close
coupling method, which included effects arising from negative ion states decaying
by autodetachment and cascade contributions from higher-lying states [28–30].
Target recoil experiments which also utilized time-of-flight measurements were
first reported by Pearl et al. [31] and were later extended by Defrance et al. [32].
In molecules, Oshima et al. [33] studied N2 excited to the E 3 g+ metastable state
and H2 excited to the c3u metastable state. In these experiments the time-of-flight
technique was used to measure the momenta of deflected targets, thereby producing
a cross section differential in both energy and angle. It is these experiments that
most closely resemble the work discussed in Sect. II of this chapter.
Finally, it is appropriate to mention new techniques which produce a well-
defined target momentum and which are being used for both scattering and atom
optics experiments. These techniques use resonant laser cooling and manipula-
tion of the target beam momentum, exploiting the interaction of a single-mode
laser beam with the target [34]. In these experiments, the momenta of many
resonant laser photons is transferred to the target in such a way that the target
slows down, and the velocity profile is compressed to a narrow distribution. The
velocity can also be reduced in a direction orthogonal to the target beam direction,
producing a high-brightness beam [35–37]. The interaction is controlled using
the polarization, power, and detuning of the laser beam. It is possible to produce
a slow, intense beam of atoms which can be used for atom optics experiments
[38], atom interferometers [39], atom trapping [40], atom cooling [41], Bose–
Einstein condensation experiments [42], and for electron-scattering experiments.
These laser cooling and compression techniques promise to open up new avenues
which exploit the momentum of the target beam. Since the deflection angle fol-
lowing electron impact is inversely proportional to the target momentum, reducing
and controlling this momentum will allow very high-resolution experiments to be
conducted.
In this report, new experiments are described which exploit the metastable
target recoil technique and which also use laser probing of the electron-impact-
excited target to access information unobtainable by other means. To describe
these experiments it is necessary to consider the metastable target recoil arising
from electron impact, and the appropriate equations of motion are introduced in
Sect. II. The experimental apparatus designed and built at Manchester is described,
and results from these studies are presented.
The target deflection technique can also be used to study doubly excited states
of atoms created by electron impact. These experiments are discussed in Sect. III,
and although they are still in their infancy, they are expected to open new avenues
for the study of these very delicate atomic states.
The technique of laser probing the recoiling electron-impact-excited targets is
then discussed in Sect. IV, together with the concept of stepwise resonant laser
excitation and field ionization adopted in these studies. These techniques are used
to probe atomic and molecular targets, and results obtained from these experiments
are presented.
Finally, the direction in which future experiments may proceed is considered
in Sect. V. It is expected that these experiments will use more sophisticated laser
probing techniques, and will exploit the techniques of laser cooling and control of
the initial target momenta. By combining these methods a rich and diverse field of
exploration promises to be opened up in the near future.
II. Atomic Deflection Using Electron Impact
A. PRINCIPLES OF THE DEFLECTION TECHNIQUE

When an electron interacts with a target atom or molecule, a number of differ-
ent processes can occur. The electron may scatter elastically from the target, or
the interaction may lead to excitation or ionization. If the target is a molecule,
fragmentation may also occur. During these processes the total energy and mo-
mentum of the system must be conserved. Hence, if the momenta of the ini-
tial products prior to the reaction are known, the momenta of the final products
are also known. In the work discussed here, the target is excited to a metastable
state, and so internal energy is stored in the target following the reaction. This
energy remains stored until the experiment is concluded, and so it is not nec-
essary to consider effects due to emission of radiation or due to collisional
processes.
Since both energy and momentum are conserved, the recoil deflection of the
metastable target following electron excitation can be determined. It is usual to
derive the equations of motion in the center-of-mass coordinate frame [27, 32],
but it is also possible to derive these in the laboratory frame. This latter approach
is adopted here. In the laboratory frame a coordinate system is defined with the
x axis in the direction of the incident target beam and the z axis in the direction
of the incident electron beam as in Fig. 3. Pei , Pef are the momenta of the incident
electron prior to and following the reaction, Pai , Paf are the momenta of the target
prior to and following the reaction, ma is the mass of the target, me is the mass
of the electron, Eexc is the excitation energy of the state formed in the collision,
(θe, φ e) are the scattered electron angles (in spherical geometry) and (θa, φ a) are
the target recoil angles. Conservation of energy then requires
Pei2 P2 Pe2f Pa2f

E incident = + ai = + + E exc = E final , (1)
2m e 2m a 2m e 2m a
FIG. 3. The general coordinate system in the laboratory frame of reference. The target has an initial
momentum Pai in the x direction and is deflected through to spherical angles (θa, φa ) with a final
momentum Pa f . The electron has initial momentum Pei in the z direction and is scattered to a spherical
angle (θe, φe ) with momentum Pe f .
and conservation of linear momentum requires
Pai = Pe f sin θe cos φe + Pa f sin θa cos φa (x direction)

0 = Pe f sin θe sin φe + Pa f sin θa sin φa (y direction) (2)
Pei = Pe f cos θe + Pa f cos θa (z direction)
These four equations must be solved simultaneously to deduce the reaction of

the system to the collision. Apart from the condition where the incident electron
energy is very close to the excitation energy of the state Eexc, it can be assumed
that the kinetic energy of the target is much smaller than the kinetic energy of
the electron. It is only in the energy region where the electron kinetic energy is
similar to the incident target kinetic energy that deviations between a complete
dynamical model and the model used here become apparent. This assumption
allows simplification of these equations so that the final target momentum can be
written as

Pa f = Pei2 + Pai2 + Pe2f − 2ζPe f sin(θe + δ),
Pei

where ζ = Pai2 cos2 φe + Pei2 δ = tan−1 ,
Pai cos φe

Pe2f sin2 θe + Pai2 − 2Pai Pe f sin θe cos φe
tan θa = , (3)
Pe f cos θe − Pei
Pe f sin θe sin φe
tan φa = .
Pe f sin θe cos φe − Pai
If a geometry is chosen where the incident and deflected targets are confined to a
plane with φa = φe = 0◦ , Eq. (3) simplifies to

Pa f = Pei2 + Pai2 + Pe2f − 2ξ Pe f sin(θe + ),
Pei

where ξ = Pai2 + Pei2 = tan−1 , (4)
Pai
Pai − Pe f sin θe
tan θa = φa = 0.
Pei − Pe f cos θe
In this planar scattering case, the recoil angle α = (90◦ − θ a) increases as the
initial target momentum decreases. Vector momentum diagrams representing
planar scattering are shown in Fig. 4. For forward electron scattering (Fig. 4a),
FIG. 4. (a) Forward, (b) backward, and (c) general electron-scattering conditions in planar geometry.
The extremes of forward and backward scattering of the metastable target are shown, together with the
more general case. Two solutions can exist along a given path of the outgoing target such that energy
and momentum are conserved.
momentum transfer to the target results in small recoil angles. For backward elec-
tron scattering (Fig. 4b), large recoil angles result from the collision. However,
in general (Fig. 4c), for any given recoil angle α, two possible values of the out-
going target momentum Paf can exist, each corresponding to a different electron
scattering angle θe.
To illustrate the dual nature of the reaction, and the relationship between the
electron-scattering angle θe and the target recoil angle α, Fig. 5 shows the results
of modeling the collision for an incident electron of energy 40 eV with a helium
target of initial velocity 1960 m/s (this is typical for a supersonically expanded
beam as used at Manchester). The 23S1 metastable helium atom with excitation
energy Eexc = 19.82 eV is assumed to be excited in the reaction, and so the final
energy of the scattered electron is 20.18 eV. The momentum of the target prior to
the collision is therefore known, as is the momentum of the incident and scattered
electron. The final target momentum Paf can then be determined as a function of
electron-scattering angle from θe = 0◦ through to θe = 180◦ using Eq. (4).
In Fig. 5a, the atom deflection angle α is plotted as a function of θ e. Under
these conditions, the atoms recoil through angles from around 4.2◦ for forward
electron scattering through to around 24.8◦ for backward scattering. Figure 5b
illustrates the dual nature of the reaction, where the deflection angle is plotted
as a function of the final target momentum following the collision. At any given
deflection angle between the recoil angle limits, two distinct values of metastable
atom momentum are possible. By experimentally selecting the deflected targets
at a given recoil angle between these limits, the metastable atoms are expected to
travel with two distinct velocities to the detector. By locating the detector a set
FIG. 5. (a) Deflection of helium target traveling at 1960 m/s excited to the 23 S metastable state, as a
function of electron scattering angle for E inc = 40 eV. The atom recoil angles θa and α are shown.
(b) Momentum of helium target traveling at 1960 m/s excited to the 23 S metastable state, for E inc =
40 eV, showing the dual nature of the reaction.
FIG. 6. Atomic deflection angle as a function of incident electron energy for helium traveling at
1960 m/s.
radius from the interaction region, these velocities transform into different arrival
times for different groups of atoms. Time-of-flight measurements therefore yield
information about the scattering process which produces these different groups.
Figure 6 shows the minimum and maximum deflection angles expected for ex-
citation to specified states of helium as a function of incident electron energy. Near
threshold for excitation of the target state, a reduced angular range is expected.
As the incident energy increases, the range of target deflection angles increases.
Excitation to the metastable states is illustrated, together with excitation to highly
excited Rydberg states which can have lifetimes allowing them to reach the de-
tector before decaying. It is clear that if the target incident momentum is not well
defined, differentiation of atoms excited to specific metastable states is experimen-
tally difficult at incident energies much above threshold. If this condition is not
satisfied, cross sections obtained directly from deflected target yield can only give
information on a mixture of both states. Cascade contributions due to electron
excitation of higher lying states feeding into the metastable states must also be
considered when comparing theory to experiment.
One advantage of measuring the deflected target yield compared to conven-
tional electron-scattering measurements is that a complete set of scattering angles
is accessible, from forward scattering (as measured by Bederson and co-workers)
through to backscattering where θe = 180◦ . In conventional experiments, when
forward-scattering measurements are attempted, the electron detector is often sat-
urated by electrons emitted directly from the electron gun. The backscatter re-
gion is also usually inaccessible, due to the electron source and detector collid-
ing as θe = 180◦ is approached. These regions have therefore remained largely
unexplored using conventional techniques until recently, when a new type of ex-
periment was invented at Manchester. This technique allows both forward and
backscatter regions to be investigated while detecting the electrons, and since this
produces results which can be compared with those obtained from atom recoil
experiments, it is relevant to consider this technique here.
The new method is called the magnetic angle changing technique and was
invented by Frank Read and co-workers in Manchester [43–46]. In these experi-
ments an electron spectrometer is modified to include a pair of current-carrying
coils which straddle the interaction region [43]. These coils produce a magnetic
field localized to the interaction region so that the magnetic field is negligible at
the electron source and the electron detector. By contrast, within the interaction
region the electrons traverse the magnetic field lines and experience a force which
curves their trajectories. The magnetic force deviates the electrons onto velocity-
dependent paths, and so electrons scattered from the target arrive at the detector
at a different angle than that through which they were initially scattered.
It is possible to adjust the magnetic field and detector angle so as to measure
the yield of electrons scattered through all possible angles. Further, since the elas-
tically scattered and direct beam of electrons have different energies from those
inelastically scattered, the inelastically scattered electrons arrive at the detector at
a different angle. Both small- and large-angle inelastic scattering cross sections
can therefore be measured. In addition, by using high-energy-resolution detectors,
the experiments yield cross sections free of cascade contributions. By also using
a high-energy-resolution incident electron beam, different states can be distin-
guished within the resolution of both the detector and source. This resolution is
around 20 meV for the most sophisticated spectrometers operating with a gas-phase
target.
B. EXPERIMENTAL RECOIL TECHNIQUES

The apparatus developed at Manchester for measuring recoiling metastable tar-
gets is shown schematically in Fig. 7 and is described in detail by Murray and
Hammond [47]. Essential for these experiments is a well-defined target momen-
tum, well-defined incident electron momentum, and a detector with narrow angular
resolution which can rotate accurately around the interaction region. Variations in
these parameters lead to a distribution in the final target momentum and hence a
decrease in experimental accuracy.
A supersonic expansion of the target gas is used to achieve a narrow velocity
spread. The target gas is emitted into a high vacuum from a high-pressure valve
operating with a pressure of 1–2 atm. The gas cools supersonically in the expansion,
and is skimmed by a small exponential skimmer [48] located 150 mm downstream
from the valve. The skimmer/nozzle combination produces an angular spread of 2◦
at the interaction region, constraining the target momentum vector to ±1◦ around
the x axis.
FIG. 7. Schematic of the experimental apparatus at Manchester. The supersonic gas jet is produced
in the lower source chamber and is skimmed before entering the interaction chamber. An electron gun
provides a well-collimated beam of electrons which interact with the target deflected and detected using
the metastable detector located on a vertical rotating table.
The velocity distribution of a supersonically expanding gas beam is related

to the Mach number M, which is a characteristic defined by the expansion. The
distribution is given by [49]
3
γ M 2 (v − vs )2

v
f (v) = c exp − , (5)
vs 2vs2
where

kT γ M 2
vs =
m{1 + [(γ − 1)/2] M 2 }
is the stream velocity, T is the temperature of the gas prior to expansion, γ is the
ratio of specific heat capacities (γ = 5/3 for monoatomic gases), c is a constant,
and k is Boltzmann’s constant. The velocity distribution is Gaussian, with an av-
erage velocity given by the stream velocity and a width inversely proportional
to the Mach number. For the helium experiments at Manchester, the Mach num-
ber is around 20, and so the average velocity is ∼1900 m/s with a distribution
of ±6%.
By contrast, the incident electron beam is produced in an electron gun which
has an energy resolution of ∼600 meV [50]. The electron gun is of a two-stage
electrostatic design, and produces a beam with a pencil angle of 2◦ and zero
beam angle at the interaction region. At 40-eV incident energy, the momentum
distribution of the incident electron beam is 3.4 × 10−24 kg m/s with a variation
of ±0.6%. This distribution can be neglected when compared to the much larger
variation in the target beam momentum.
The distribution of momenta in the target beam produces a corresponding un-
certainty in the metastable recoil angle which transforms into an uncertainty in the
corresponding electron-scattering angle. Using a Mach number of 20, the uncer-
tainty in the electron-scattering angle is found to be σ (θ e) ≈ ±3.5◦ for small recoil
angles, whereas this increases to σ (θe) = ±15◦ for a recoil angle α = 23◦ . The accu-
racy of differential cross sections measured using the target deflection technique is
therefore sensitive to the expansion, with higher Mach numbers yielding narrower
velocity profiles and hence greater accuracy.
The supersonic expansion occurs in the source chamber which is pumped by a
2500-liter/s diffusion pump. The pulsed Lasertechnics nozzle [51] opens a piezo-
electrically coupled Viton knife edge seal to release gas into the vacuum chamber
at a repetition rate of up to 150 Hz. The nozzle is driven by a 200- to 500-μs
pulse which can be varied from 10 to 100 V. It is necessary to adjust these driving
parameters to minimize extraneous bounce of the nozzle following initial opening
of the valve, as this decreases the load on the pumping system. By judicious
control of these parameters, the supersonic expansion can be adjusted for maximum
throughput while minimizing the velocity distribution of the beam.
The pulsed, supersonically cooled target beam is skimmed prior to entering

the interaction chamber using an exponential skimmer with a 1-mm-diameter
orifice [48]. The skimmer is constructed of rhodium-plated copper with a wall
thickness at the orifice of less than 10 μm. This very thin edge ensures that the
gas entering the interaction chamber suffers minimal deflection due to scattering
from the edge of the skimmer. Upon entering the interaction chamber, the gas
travels approximately 150 mm before interacting with a pulsed beam of electrons
emitted from the electron gun. The electron beam interacts orthogonally with the
gas beam, and a Faraday cup placed opposite to the gun monitors the yield of
emitted electrons.
Targets excited and deflected by the interaction are selected using a detector
located on a vertical rotating table which can be moved under computer control
with an accuracy of ± 0.01◦ . The angle of the detector is measured using an
opto-coupled array of LEDs and phototransistors, and the angle is calibrated using
apertured visible laser diodes located in the pulsed nozzle source, vertical table
rotation axis, and electron gun. The metastable target detector is constructed of
310-grade stainless steel, and has a conical input skimmer with a 1-mm entrance
orifice. A second 1-mm aperture located 70 mm behind the input aperture defines
the acceptance angle of the detector. Metastable targets which pass through both
apertures drift a further 100 mm before striking the front cone of a channel electron
multiplier. The resulting current pulse from the CEM is detected, amplified, and
fed to counting and noise-discriminating electronics external to the system.
The interaction chamber is constructed of nonmagnetic 310-grade stainless steel
and is internally lined with a double layer of μ-metal to reduce external magnetic
fields to less than 5 mG. All components inside the chamber are constructed of
nonmagnetic materials including 310-grade stainless steel, advance (constantan)
sheet, aluminum, copper, and PTFE. A full description of the apparatus can be
found in [47].
A time-of-flight measurement of the selected targets determines their momen-
tum following excitation. To facilitate this, the experiment is pulsed at a rate
between 100 and 150 Hz. Once the angle of the detector is adjusted, the experi-
mental cycle commences by opening the nozzle, releasing high-pressure gas into
the source chamber, where it cools supersonically. The edge of the nozzle driv-
ing pulse also starts a multichannel scaler. Helium takes around 300 μs after the
nozzle is opened for the maximum density of the target beam to pass through
the interaction region, and at this time the electron gun is turned on for 4 μs.
The electron beam interacts with the gas which is deflected. A small, angularly
resolved sample of deflected targets is selected by the analyzer, and these are
detected by the channel electron multiplier. The CEM pulse is amplified and dis-
criminated against noise before being fed to the multichannel scaler, where it is
recorded.
The experiment is repeated many times to accumulate a time-of-flight spectrum.
After a preset number of nozzle pulses, the resulting time-of-flight spectrum is
recorded, the analyzer moves to a new detection angle, and the sequence is repeated
until a complete set of results is obtained over all recoil angles. The time-of-flight
spectrum yields the speed of the atoms (and hence the magnitude of their momen-
tum), since the detector rotates at a constant radius around the interaction region.
C. DIFFERENTIAL CROSS-SECTION MEASUREMENTS

Figure 8 shows examples of time-of-flight spectra obtained from the manifold of
21,3S metastable states of helium. The incident electron energy was 40 eV, and the
signal accumulated for 250,000 nozzle pulses at each detection angle. The dual
nature of the detected signal can be seen. Two Gaussian peaks are fitted to the data,
allowing the signal under each peak to be determined. This relates directly to the
probability of excitation of the metastable atoms, allowing the differential cross
section to be established.
To estimate this cross section, time-of-flight spectra were accumulated at 1◦
intervals from α = 4◦ through to α = 28◦ . Figure 9 shows the result of these mea-
surements, where the peak amplitude, position, and width are plotted as functions
of detector angle. At lower and higher deflection angles the peaks are difficult to
distinguish, and so there is an increased uncertainty in the measurements at these
angles. The dual nature of the result is seen in Fig. 9b, and compares favourably
with the prediction of Fig. 5b.
The differential cross section (DCS) for electron excitation of a target can be
written as
dσ
DCS = (6)
d
where d is the solid angle at spherical angles (θe, φ e) with respect to the incident
beam direction. Since the experiments measure the excited target yield, it is neces-
sary to relate the target momentum to the correlated electron-scattering direction.
This is accomplished by rearranging Eq. (4) :

−1 Pai − Pa f sin θa
θe = tan (7)
Pei − Pa f cos θa
The value of Paf is obtained using the expression
m a xint
Pa f = (8)
tint
where ma is the mass of the target and xint is the distance from the interaction region
to the detector; tint is the time the atom takes to reach the detector from the interac-
tion region obtained from Fig. 9. The values of Pai and Pei were calculated by fitting
Eq. (4) to the data of Fig. 9b using a least-squares fit. The calculated incident energy
FIG. 8. Selected examples of the time-of-flight spectrum from helium with E inc ≈ 40 eV. For details, see text.
FIG. 9. Data obtained from fitting Gaussian peaks to the time-of-flight spectra obtained when
exciting helium at around 40 eV incident electron energy. Part (a) shows the fitted peak heights and
Part (c) the associated widths. Part (b) indicates the peak time of arrival from opening the gas nozzle.
The dual nature of the deflection process can be seen.
of the electron beam was 40.8 eV, and the momentum of the incident target beam
was found to be Pai = 1.3 × 10−23 kg m/s, corresponding to a speed of 1960 m/s.
The relative magnitude of the differential cross section is given by the area under the
Gaussian peaks fitted to the data, and this relates directly to the relative differential
cross section measured using conventional electron spectrometers.
FIG. 10. (a) DCS measurements exciting the 23,1 S and 23 P states of helium for an incident electron
energy of 40 eV using the magnetic angle changing technique of Ref. [46]. The results from θe = 0◦
to θe = 180◦ are normalized to the CCC calculations of Ref. [53]. (b). Estimated relative DCS for
excitation to the 23,1 S metastable states of helium for an incident electron energy of 40.8 eV obtained
from the measurements in Fig. 9. Also shown are the summed differential cross sections from excitation
of the 23,1 S metastable states and 23 P states of helium shown in Fig. 10a.
A comparison between relative cross sections measured using electron spec-

troscopy and the recoil atom technique is shown in Fig. 10. The results of Cubric
et al. [46] using the magnetic angle changing technique are shown in Fig. 10a, the
results being placed on an absolute scale by normalizing to the convergent close
coupled (CCC) calculation of Fursa and Bray [53]. The singlet cross section is
seen to be dominant for forward scattering, with a deep minimum at scattering
angles θ e ≈ 50◦ . The 23P cross section is reasonably uniform across all scatter-
ing angles, whereas the 23S cross section varies by a factor of ∼60 as the scattering
angle is varied.
Figure 10b shows the measured DCS obtained from the recoil measurements
of Fig. 9. The cross section is forward peaked, with a minimum around θ e = 90◦ .
The uncertainty in the deduced electron-scattering angle is obtained from the peak
width, which is directly proportional to the variation in velocity of the incident
target beam. The calculation does not consider effects due to noncoplanar contri-
butions (i.e., when φ = 0◦ ), but a more detailed study indicates that these effects
are small for the acceptance angle of the detector and at this energy. Noncoplanar
contributions must be considered when the incident energy is close to threshold for
excitation of the state, and when the detector solid angle is significantly widened
to increase yield.
Cascades from higher-lying states which feed the metastable states produce
additional contributions to the yield. These contributions depend on the differential
cross section for excitation of these higher states, together with the branching ratio
for decay back to the metastable state. It has been estimated [52,54] that cascade
contributions to the 21S0 metastable state play only a minor role, since most higher
lying singlet states decay preferentially to the 11S0 ground state. By contrast, all
upper triplet states decay to the 23S1 metastable state and so there is significant
contribution from these decay routes. Harries [52] used the data of Cubric et al.
[46] and Trajmar et al. [54] to estimate this contribution by assuming that the
excitation cross section for the n3L series is proportional to n ∗−3 , where n ∗ = n−δ l
and δ l is the associated quantum defect. This analysis (which did not include
angular contributions) indicates that up to 80% of the measured 23S1 yield is due
to cascade contributions at this energy.
To establish a comparison between the recoil cross-section results and those
using the magnetic angle changing technique, it is assumed that all singlet states
apart from the 21S0 state decay directly to the ground state. Further, it is assumed
that the only triplet states significantly excited at this energy are the 23S1 and 23P
manifold. Figure 10a shows the measured cross sections for excitation of each of
these states, and so it is possible to sum these results to predict the cross section
for the manifold of metastable targets.
Figure 10b also shows the result of summing these contributions and compares
this with the results from the recoil measurements. An arbitrary normalization
has been chosen to yield best overall agreement, since the recoil data are not
measured on an absolute scale. The recoil data are seen to compare well with
the summed signal from [46]. The major contribution to the cross section in the
forward direction arises from the 21S0 state, whereas at intermediate scattering
angles, from θ e = 30◦ to θ e = 100◦ , cascade contributions from the 23P states
dominate. At higher scattering angles the singlet state once more dominates, until
at θ e = 180◦ the 21S0 and 23S1 states have equal probability of excitation, with
cascade from the 23P states providing only a small contribution. These results
indicate that the approximations regarding the cascade contributions discussed
above are reasonable.
Although the recoil atom measurements agree well with the electron-scattering
measurements, the technique is restricted because the spectrometer can resolve
neither the cascade contribution nor the individual metastable state contribution
excited by electron impact. The recoil technique is therefore most applicable
when the states are widely spaced in energy, and this is discussed further in
Sect. III. For targets with closely lying states, the results are difficult to inter-
pret since many different states contribute. A different method is therefore re-
quired to study these interactions. Section IV describes such a technique, where
a high-resolution laser probes individual states within the metastable manifold.
These experiments produce results of far greater detail than is possible using con-
ventional electron-scattering experiments, and have the potential to provide new
information on electron-impact excitation of many different molecular and atomic
species.
III. Doubly Excited States Studied via the Fluorescence

Decay Product: Recoiling Excited Atoms
A. PRINCIPLES
In Sect. II the kinematics of interactions between electrons and targets which pro-
duce metastable excited targets was outlined. The yield of metastable targets as a
function of recoil angle and time-of-flight was explored in the laboratory frame,
and experimental results confirming the predicted distributions were shown. In
this section a method is described in which the angular and time-of-flight distribu-
tions are utilized to enable observation of targets excited to doubly excited states
which decay by fluorescence. The results demonstrate that doubly excited targets
can be observed without the continuum background signal which has plagued
observations for decades.
In previous studies, doubly excited targets formed in electron-impact experi-
ments have been difficult to study for three principle reasons: (1) excitation cross
sections are small in comparison to singly excited state cross sections; (2) signal
relating to doubly excited states is almost invariably embedded in the continuum
with which interference occurs; and (3) doubly excited states lie in an excitation
energy region which also contains triply excited negative ion states with which
they can be confused. The majority of previous work has involved detection of
either the scattered electron, the ejected electron, or the positive ion arising from
the excitation and autoionizing decay of the doubly excited states. The earliest
such electron-impact experiments [55–59] reported observations of the (2s2)1S,
(2s2p)3P, and (2p2)1D states in helium, which are optically forbidden in excitation
from the ground state but which can autoionize. These measurements comple-
mented photoabsorption measurements of Madden and Codling [60], who reported
the first observation of doubly excited states. All of these techniques are particularly
sensitive to the energetically broad doubly excited states that can autoionize.
For doubly excited states that cannot autoionize, experiments have to be directed
toward the scattered electron or to the decay products. Burrow [61] reported the
observation of the (2p2)3P state which is stable against autoionization, with a
total cross section of 4.4 × 10−21 cm2 at 0.11 eV above threshold. Fluorescence
measurements of the (2p2)3P state [62], made by observing the XUV radiation at
32 nm arising from the transition (2p2)3P to (1s2p)3P [63], enabled the observation
of the decay of highly excited negative-ion states via autodetachment into the
(2p2)3P state. A subsequent experiment [64] queried Burrow’s identification of the
(2p2)3P state.
Metastable singly excited atoms are also produced in the decay by fluorescence
of doubly excited atoms. Shpenik and co-workers [65] reported observations of
electron-impact-formed doubly excited states in helium using the metastable atom
recoil technique in which the yield of metastable atoms as a function of incident
electron energy and atom recoil were recorded. The doubly excited states were
observed as interference profiles with, and dominated by, the very large yield of
directly excited metastable atoms. In other experiments the decay of triply excited
negative-ion states has been observed by autodetachment to metastable states.
Defrance et al. [32] used the technique described in Sect. II, measuring metastable
atom recoil as a function of incident electron energy, atom scattering angle,
and metastable atom time of flight to perform a measurement of the (2s22p)2P
negative-ion state of helium. Trantham et al. [66] explored the structure of the
negative-ion states using both metastable atom yield and electron spectroscopy.
In other inert gas targets, the metastable atom yield has been used to study highly
excited states [25, 67].
In photoexcitation studies the observation of doubly excited states in helium
via photon and metastable atoms as fluorescence decay products has been pi-
oneered by Hammond and co-workers [68–70]. The use of the radiative decay
route was shown to provide excellent sensitivity to the energetically narrow dou-
bly excited states without the presence of a continuum background signal. Related
high-energy-resolution experiments have explored the branching ratio between the
autoionization and fluorescence decay routes [71–73]. These experiments provided
the motivation for extending electron-impact studies using metastable atom recoil
spectroscopy.
The primary goal of these new experiments is to distinguish between singly
excited metastable atoms formed by direct excitation and singly excited metastable
atoms formed by fluorescence from doubly excited states [74]. It is also desirable
to isolate the doubly excited state signal from that arising from triply excited
negative-ion states.
The principle underlying these experiments is as follows. The momentum trans-
ferred to a target excited by electron collision is represented by Fig. 4c. The mo-
menta of the metastable atoms, measurable with the apparatus described, fall on
the upper dashed circle shown in this figure. For excitation near threshold where
Pef is small, the circle is of small diameter. As the impact energy increases, Pef
increases and so the diameter of the circle also increases. When the incident elec-
tron energy is just sufficient to excite a doubly excited state, the end points of the
final momenta Paf of the doubly excited atoms are therefore represented as lying
on a small-diameter circle. However, at the same incident electron energy it is also
possible to form singly excited atoms with final momenta represented by a second,
larger-diameter circle which encloses that describing the doubly excited atoms. The
momentum change of the atom following emission of a photon in the fluorescent
decay process is negligible, so that the metastable atom decay product retains the
final momentum signature of the doubly excited atom from which it forms. Thus,
in general, at atom recoil angles that intersect both circles, four well-defined mo-
menta exist. The lowest and highest momenta are from directly excited metastable
atoms, and the two intermediate momenta are from the doubly-excited-state decay
product. It is therefore possible to separate the metastable atoms arising from the
two different formation processes.
B. EXPERIMENTAL TECHNIQUES
The apparatus described in Sect. II provides the means to study the metastable
product of the decay by fluorescence of doubly excited states. A metastable atom
detector with a wide acceptance angle replaces the detector described in Sect. II.
This new detector is placed closer to the interaction region and has a larger entrance
to provide a wider acceptance angle. Between the entrance aperture and the channel
electron multiplier a series of biased open aperture plates [69, 70] prevent scattered
electrons and positive ions from striking the CEM. The increase in detection solid
angle is necessary because the yield of metastable atoms from this decay is very
low. This detector also provides excellent sensitivity at threshold, since it can
detect a broader range of atoms with contributions also from events out of the
scattering plane.
C. OBSERVATIONS OF MOMENTUM-ANALYZED DOUBLY-EXCITED-STATE

FLUORESCENCE DECAY PRODUCTS
Demonstration [52, 74] of the separation of metastable atoms arising from
decay by fluorescence of doubly excited states from those directly excited is shown
in Fig. 11, where the yield of metastable atoms arising from electron impact at
63.3 eV is plotted as a function of atom scattering angle and time of flight. The
large-diameter oval structure which dominates the data represents the direct exci-
tation processes. The low-intensity structure lying within this oval at 21◦ and 58 μs
represent metastable atoms arising from the decay of doubly excited states. These
data compare well with the results shown in Fig. 5b, where the signal arising from
the decay of doubly excited states appears as a second, smaller oval lying within
that due to directly excited atoms.
FIG. 11. Time of flight of metastable helium atoms as a function of atom recoil angle at an incident
electron energy of 63.3 eV. The oval structure represents detection of directly excited metastable atoms.
The structure centered at 21◦ and 58 μs represents atoms arising from fluorescence decay of doubly
excited states. A logarithmic gray scale represents yield.
The yield of metastable helium atoms can also be measured at a given recoil
angle as a function of incident electron energy and time of flight. The data, shown
in Fig. 12, can be divided into two distinct regions. In the first region, metastable
atoms arising from directly excited states lie in two bands at approximately 45 and
72 μs, corresponding to electron scattering angles of 120◦ and 70◦, respectively. The
second region lies between these bands and represents metastable atoms arising
from fluorescence decay of doubly excited states. This region changes rapidly as
the incident energy increases from 60 to 65 eV. At 80 eV the structure becomes
two bands at approximately 52 and 67 μs, corresponding to electron-scattering
angles of 100◦ and 60◦ , respectively.
Doubly excited state excitation functions can be extracted from the two-
dimensional time-of-flight data in Fig. 12. This is achieved by summing the
yield lying between the dominant fast and slow metastable atom yields. The re-
sults of this summation are shown in Fig. 13. Signal arising from doubly excited
atom decay is clearly apparent in the region below the He+ (N = 2) ionization
threshold. No similar structure can be observed below higher ionization thresh-
olds, in marked contrast to the photoexcitation results of Sokell et al. [68]. The
structure lying below the He+ (N = 2) ionization threshold represents doubly ex-
cited states that decay by fluorescence. The states that can appear here are states
FIG. 12. Time of flight of metastable atoms as a function of incident electron energy at an atom
recoil angle of 20◦ . Metastable atoms arising from fluorescence decay of doubly excited states lie in
the 50- to 63-μs region of the spectrum. A logarithmic gray scale is used to represent yield.
FIG. 13. Doubly excited state excitation function extracted from the time-of-flight spectrum shown
in Fig. 12. The excitation energies of known doubly excited states are marked.
that are either parity favored or unfavored to autoionize. It is likely that states with
a lower probability to autoionize will be more likely to fluoresce. However, the
energy resolution used in these measurements is insufficient to allow individual
states to be identified. The residual yield of directly excited metastable atoms
results in an underlying background which decreases as the electron-impact energy
increases.
This new capability to resolve the metastable decay product of doubly excited
states allows these states to be studied free from a large continuum background and
without the complication due to interference. In addition, triply excited negative-
ion states that autodetach to produce singly excited neutral atoms [32, 66] can-
not be distinguished from directly excited atoms and so are separated from the
decay product of doubly excited atoms in momentum space. Only triply excited
negative-ion states that autodetach into doubly excited states and that subsequently
fluoresce can contribute to the yield detected here. At present the technique can
be used for light targets where the momentum definition of the incident beam
allows sufficient angular and temporal separation of the recoiling targets. In the
future, laser cooling which produces a well-defined momentum will enable de-
tailed measurements of doubly excited states to be conducted for a wide range of
targets.
IV. Stepwise Laser Probing of Deflected Metastable Targets
A. PRINCIPLES
Sections II and III show how differential cross sections can be extracted from
the deflected target yield following electron impact. Section II describes how
this is applied to direct excitation processes, whereas Sect. III indicates how
the technique can be applied to investigate doubly excited states. In both cases,
cascade contributions play a role. For direct excitation, these contributions reduce
the information which can be obtained, whereas for doubly excited states cascade
contributions are exploited to provide information about these states.
It is of interest to maximize the information obtained from an experiment, and
so it is beneficial to consider more sophisticated methods that allow extraction of
additional information. The technique adopted at Manchester uses stepwise laser
excitation to probe metastable targets following electron excitation. This allows
details to be obtained about closely lying states within the metastable manifold,
since the high spectral resolution of the laser is exploited. The combined electron-
impact and laser excitation techniques provide details about the reaction which
would be impossible to obtain using either technique separately.
The stepwise excitation scheme adopted is shown in Fig. 14. The incident
electron scatters from the reaction, deflecting and exciting the target. The target
FIG. 14. Stepwise laser excitation scheme following electron-impact excitation.
is further excited to a higher-lying state using laser radiation tuned to be resonant

between these states. Properties of the upper state are then measured, either by
observing the fluorescence from the laser excited state or by field ionizing the
target and detecting the resulting ions.
Since the laser radiation transfers information from the electron-excited state
to the upper state in a controlled way, this information can be extracted. A further
advantage is that the laser has very high spectral resolution (typically μeV to neV).
Hence, by varying the laser wavelength, different states within the metastable
manifold can be analyzed. This resolution is a factor of 103–108 better than is
possible with even the most sophisticated electron spectrometers.
The stepwise excitation scheme has been exploited in a number of different
ways, principle among these being the observation of fluorescence from the laser-
excited target. This has been used to evaluate partial differential cross sections
for both metastable and nonmetastable atoms excited by electron impact. This is
achieved by observing the polarization of the emitted fluorescence from the laser-
excited state [75–77]. Such experiments are often performed close to threshold so
that cascade contributions are eliminated.
An example of this technique was the first test of the Percival–Seaton hypothesis,
which predicts that the nuclear spin does not play a role when a target is excited by
electron impact [78, 79]. In these experiments a continuous-wave (CW), tunable,
single-mode ring dye laser resonantly excited the 61P state of mercury to the 61D
state from which fluorescence was monitored. The laser had sufficient resolution
to isolate individual isotopes, a technique impossible by other means. The partial
differential cross sections were measured as a function of incident energy for
isotopes with different nuclear spins. The Percival-Seaton hypothesis was found
to be valid at higher incident energies, but was found to break down as threshold
was approached.
A more sophisticated experiment has been performed by the Griffith group
[80–85], who measured the polarization of light emitted from the laser-excited state
in coincidence with the scattered electron. A well-defined beam of electrons excited
mercury to the 61P1 state and the scattered electron momentum was determined
using a cylindrical electron analyzer. The target was illuminated with radiation
from a stabilized CW ring dye laser operating around 579 nm, which excited
the target from the 61P1 state to the 61D2 state. The wavelength was chosen to
excite only the I = 0 isotope, thereby eliminating spectroscopic contributions
from other isotopes. The laser polarization was also varied, allowing a complete
set of scattering parameters (ρ 00, L ⊥ , Plin, and γ ) to be evaluated for the 61P1 state
without the need to measure the in-plane P4 Stokes parameter [86].
Stepwise techniques have also been used to study highly excited Rydberg atoms.
These atoms can be produced using one or more lasers tuned to excite the target to
high principal quantum numbers. Using high-resolution laser techniques, atoms
with principal quantum numbers greater than 300 have been produced [87]. The
highly excited electrons in these atoms behave as quasi-free particles when far
from the nucleus, and so experiments can be performed which use these electrons
for very-low-energy scattering experiments from other targets [88].
If an electric field is applied during the laser excitation process, it is possible to
Stark mix states of high n to produce atoms with high angular quantum numbers.
These “planetary” atoms have very long lifetimes, since the electrons spend most
of their time far from the nucleus. Planetary atoms are extremely sensitive to
small external perturbations which promote ionization. Since ions produced by
these perturbations can be detected with high efficiency, these targets have been
proposed as sensitive probes of low-intensity infrared radiation for astronomical
applications [89].
Whereas it is usual to observe the fluorescence in stepwise experiments, this can
have disadvantages since the fluorescent photon may not be emitted in the direction
of the detector. In addition, detectors may not be available that are sensitive to the
emitted radiation. This problem increases for coincidence studies, since both the
electron and the photon from a correlated scattering event must enter the detectors
simultaneously to register a true coincidence. The resulting coincidence yield is
therefore low, and long accumulation times are required.
An alternative detection method is to laser-excite the target to a high Rydberg
state which is then field ionized, the resulting ion being detected. The advantage of
this technique is that field ionization is efficient, and it is possible to collect all ions
that are produced. A disadvantage is that laser excitation to the Rydberg states is not
efficient, the probability of excitation being proportional to laser power density.
A tunable high-power pulsed laser is usually required, which has much lower
spectral resolution compared to the CW ring lasers described above. The repetition
rate of these lasers is also low, and so the experiment must be operated in a
pulsed mode with the target source, electron gun, laser, detector, and electronics
all being pulsed. The apparatus described in Sect. II is designed for these types of
experiments.
B. EXPERIMENTAL TECHNIQUES
To implement these experiments, the apparatus described in Fig. 7 is modified
to include a field ionization detector which encompasses the interaction region.
The upper rotating detector is also modified to include a field ionization detector,
allowing momentum selected targets to be probed by the laser while simultaneously
monitoring the metastable yield. A laser which produces a high-power laser beam
is needed, and this is provided by a YAG pumped doubled dye laser.
Figure 15 depicts the laser system used at Manchester. A YAG laser operating at
532 nm is Q-switched at 20 Hz using an external high-stability crystal-controlled
clock [89]. The YAG laser pumps a dye laser, the oscillator being pumped by 3% of
the input light and the amplifier by 97% of this radiation. A diffraction grating in the
oscillator cavity selects the laser wavelength. When radiation from the oscillator
passes through the amplifier cell, stimulated emission produces a bright pulse of
radiation with a peak power around 3 MW. The dye laser radiation is doubled to
produce a UV beam with a peak power around 100 kW and a temporal width
of ∼5 ns.
FIG. 15. The laser system that produces tunable pulsed radiation for the stepwise experiments.
The UV laser radiation is directed inside the vacuum system using optical fiber.
This has the advantage that the laser beam can be focused onto the interaction
region while the system is open. Upon closing the vacuum system, it is only
necessary to couple the laser radiation into the fiber for the beam to illuminate the
interaction region. The electron gun and Faraday cup are also shown in Fig. 15, and
it can be seen that the laser–target interaction region is displaced from the electron–
target interaction region. The field ionization plates for the lower detector which
encompass the interaction regions are also illustrated.
The laser–target and electron–target interaction regions are displaced, as this
allows highly excited or ionized atoms created by electron impact to be eliminated
from the field ionization signal. Measurements can then be conducted for electron-
impact energies above the ionization threshold, where the large yield of ions created
directly by electron impact would swamp the stepwise laser signal if the two
interaction regions were coincident.
The EHT electrode of the field ionization detector is pulsed up to a maximum of
3 kV to field ionize the laser-excited targets. Ions which are created accelerate onto
the entrance of an ETP electron multiplier [20], and the output pulse is amplified
and directed to external counting electronics. The ion signal is monitored as a
function of laser wavelength and electron-impact energy [50, 47].
The electronic hardware needed to control these interactions is more com-
plex than described in Sect. II, since the timing is critical in these experiments.
Overall timing is governed by a custom-built delay generator [89] which can be
programmed manually or via a PC. The repetition rate of the experiment is dic-
tated by the 20-Hz maximum rate at which the laser can be operated. As before,
the timing sequence is initiated by opening the target gas nozzle. For helium, the
laser flashlamps are fired 150 μs after the nozzle is opened, in preparation for
Q-switching the YAG laser. Around 300 μs after the nozzle is opened, the peak
of the helium beam arrives at the electron–target interaction region. The electron
gun is then turned on for 4 μs and metastable targets are excited. Since highly
excited Rydberg atoms and ions may also be created, the field ionization plates
are fired 40 ns after the electron gun is switched off using an EHT pulse switched
on for 1 μs. This removes ions and Rydberg atoms from the gas beam, allowing
only metastable and ground state targets to drift to the laser–target interaction
region.
Metastable helium atoms take around 10 μs to drift to the laser–target interaction
region following the electron gun pulse, and at this time the laser Pockel cell is
triggered, initiating a laser pulse. Laser radiation resonantly excites the metastable
targets to high Rydberg states. At 50 ns after the laser is initiated, the EHT pulse
is fired for 2 μs and the laser-excited targets are field ionized and accelerated to
the detector. The detector signal is gated before being passed to a constant fraction
discriminator which feeds a timer counter. The signal is then recorded by the
computer.
The metastable signal detected by the upper detector (Fig. 7) is monitored during
the interaction, as this allows the electron beam to be focused and steered onto
the gas target beam. Prior to carrying out these measurements the upper and lower
detectors are rotated to the desired target recoil angle using the stepper motor
electronics controlled by the computer.
Stepwise experiments using the lower field ionization detector do not produce
differential cross sections, since there is no angle discrimination of the recoiling
targets. This detector therefore produces an integrated cross section, with a corre-
sponding high yield. To measure differential cross sections for metastable targets
probed by the laser it is necessary to laser excite and field ionize targets selected
by the upper detector. Field ionization plates are therefore also located in this de-
tector, and the fiber-optic coupler is relocated to illuminate atoms selected by this
detector [47, 50]. The yield of laser-excited field-ionized targets is low in these
experiments, and so long accumulation times are required.
The timing sequence for stepwise experiments in the upper detector is identical
to that for the lower detector, except that the laser pulse and second field ionization
pulse are delayed until the targets drift into the laser–target interaction region inside
the upper detector. For helium, this occurs around 100 μs after the electron beam
pulse is initiated. Since the laser selects individual states within the metastable
manifold, differential cross-section measurements for excitation of these states
can be determined.
C. RESULTS FROM LASER PROBING OF METASTABLE TARGETS

Stepwise experiments can be carried out in the lower detector, where there is no
angular selectivity of the excited targets and the yield is high, or in the upper
detector, where the excited targets are angle selected prior to laser excitation, with
a corresponding lower yield. Results from both of these detectors are described in
this section.
The atomic target which is easiest to study is helium, since this is the lightest
atom which can be produced in a supersonic expansion. Rydberg states with prin-
cipal quantum numbers from n = 20 to n = 80 can be accessed from the 21S0
state using the dye laser described above. Resonant laser radiation is produced by
doubling the light from the dye laser operating between 626 and 630 nm.
Figure 16 shows a typical spectrum obtained using the lower detector. The laser
excites the electron-impact-excited metastable 21S0 atom to the n1P1 state. Ions
are detected by field ionization from the n = 21 through to the n = 60 states in
this example. The spectrum was obtained by firing the laser 100 times for each
recorded channel of the spectrum. The incident energy of the electrons was 22 eV,
and the average beam current was 740 pA. A 1000-V, 2-μs field ionization pulse
was fired 50 ns after the laser beam passed through the metastable beam, and the
average UV laser power density was 10 mW into a 1-mm2 beam. For n = 30, an
FIG. 16. High Rydberg spectrum obtained by resonantly exciting the 21 S0 state to the n 1 P1 state,
with n = 21 to n = 60. The laser wavelength is plotted on the x axis.
average of 3.5 ions were detected for each laser pulse, indicating the efficiency of
the detection process in these experiments.
The variation in the observed intensity shown in the figure is due to a number
of factors, including the laser excitation efficiency, laser intensity profile, field
ionization efficiency, and detector efficiency. Since the ions are all created with
almost identical kinetic energy following field ionization, the detector efficiency
remains constant as the spectrum accumulates. The ETP detector efficiency peaks
for ions accelerated to ∼10 keV, but in these experiments the anode was set to
2.5 kV and so the ion energy at the detector was around 3 keV. The ETP detec-
tor was therefore not operating at maximum efficiency when these results were
obtained.
The efficiency of laser excitation of the Rydberg state depends on a number
of factors, including the absorption and stimulated emission characteristics of
the laser–atom interaction together with the radial overlap integral between the
21S0 and n1P1 states. The laser–atom interaction can be modeled since the high
Rydberg states are hydrogenic in nature, with corresponding well-defined wave-
functions [90]. These models indicate that the laser excitation probability decreases
as n → ∞. By contrast, the field ionization efficiency increases as n increases. This
process therefore partly compensates for the decrease in efficiency due to the laser
interaction, allowing higher-n states to be detected. Correspondingly, the low yield
when n decreases toward n = 21 is due to the lower probability of field ionization
from these deeper-lying states.
One striking feature in this spectrum is the absence of ions created when the
laser is off-resonance. The signal-to-noise ratio is therefore very high. This is
important for stepwise experiments carried out in the upper detector, where the
density of metastable targets is low, since it is the high signal-to-noise ratio that
allows the field-ionized signal to be measured in this detector.
The excitation energy of the states shown in Fig. 16 follows the regular n−2
energy dependence which can be derived from the Rydberg formula. This is ex-
pected for helium, as LS coupling is strictly observed, and the states are well
separated in energy. Helium therefore provides a test bed to calibrate the appa-
ratus, and is used extensively throughout the experimental program to check the
operation of the instrument.
By contrast, Fig. 17 shows the spectrum obtained using argon as a target with
laser radiation around 295 nm. In this study the incident electron energy was
14 eV, the atom being excited from the ground state to metastable states with
excitation energies around 11.6 eV. Argon has two ionization limits, depending on
the coupling of the valence electron to the ionic core. The 2P3/2 core ionization
limit is at 15.73 eV, whereas the 2P1/2 core limit is at 15.91 eV.
Moore [91] indicates that jl coupling is required to describe the states
of argon, and this coupling scheme is adopted here. The electron-excited
metastable state accessed by the laser is the (nominal) 3P2 state at 11.52 eV,
which is designated as 4s[3/2]0J =2 in jl coupling. The laser radiation resonantly
couples this state to higher-lying states with selection rules being restricted
to J, K = 0, ±1. Rydberg states that can be resonantly excited are there-
fore the np[3/2] J =2,1, np[5/2] J =3,2, np[1/2] J =1, np ′ [3/2] J =2,1, np ′ [1/2] J =1,
n f [3/2] J =2,1, n f [5/2] J =3,2 , and n f ′ [5/2] J =3,2 states, where the prime indicates
the 2P3/2 ionic core has changed to the 2P1/2 core during the transition.
Figure 17 shows the Rydberg series observed in argon, again using the lower
field ionization detector. The assignments are due to Pellarin et al. [92] who used
a multichannel quantum defect theory to determine the energies of the Rydberg
states in this region. Dominant in the spectrum are transitions to the 10p′ [3/2] J =2
and 10p′ [1/2] J =1 core changing states. These are the most energetic 2P1/2 core
states which lie below the first ionization limit, and so these can be field ionized
and detected. The strength of these transitions is due to the large overlap between
the metastable state wavefunction and the n = 10 state wavefunctions compared to
the corresponding overlap with states of n > 29.
The argon spectrum illustrates the difficulty of resolving individual high
Rydberg states using the pulsed dye laser. Pellarin et al. [92] measured their spec-
trum using a CW dye laser, easily resolving individual states up to n = 70. This was
possible because the resolution of the CW laser was around 100 neV, in contrast to
the resolution of the pulsed dye laser, which is around 20 μeV. The pulsed system
has the advantage that the signal is more intense than that obtained using a CW
source, but the resolution of the pulsed laser restricts the experiment to probe states
whose energy difference is larger that this. These limitations are further discussed
in Sect. V.
Restrictions due to energy resolution are compounded when studying molecular
targets, since for each electronic transition in a molecule there is an associated
FIG. 17. High Rydberg spectrum obtained by resonantly exciting from the 33 P2 state of argon to high Rydberg states. The laser
wavelength is plotted on the x axis. Transition assignments are due to Pellarin et al. (1988) [Ref. 92].
FIG. 18. H2 Rydberg state spectrum excited from the c3

*
u metastable state manifold by resonant
laser radiation between 350 and 365 nm.
rotational and vibrational series due to the motion of the atoms around their center
of mass. Direct electron-impact excitation studies of molecules are restricted by
the resolution of the spectrometer, which is at best around 20 meV. It has been
impossible to resolve individual rotational states for any molecular species other
than H2, and only a limited number of targets have vibrational energy spacings
large enough to be resolved directly using electron impact [93].
These difficulties can be resolved using stepwise laser techniques, since it is
possible to access individual vibrational and rotational states created by electron
impact. Experiments conducted so far with the techniques described here have
been from the c3u metastable states of hydrogen, although a preliminary study
of nitrogen has also been carried out.
One complication with these studies is the diversity of high Rydberg states which
can be accessed. Fig. 18 is an example of a small portion of the field ionization
spectrum obtained from H2 with the laser tuned between 350 and 365 nm, showing
the complexity of the spectrum to be analyzed. Harries [52] has calculated this
spectrum by assuming that the Rydberg states of H2 obey Hunds case (d). In this
case the valence electron is treated as being in the Coulomb field of the ionic core
and the analysis follows that for an atomic target, with appropriate factors being
added to account for rovibrational motion of the core [94].
Figure 19 presents a more detailed plot of the spectrum between 361 and
362 nm, together with the prediction of Harries [52]. The correspondence between
FIG. 19. The H2 Rydberg state spectrum between 361 and 362 nm, showing corresponding as-
signments from Ref. [52]. Transitions are to the (n = 8, v = 1) Rydberg states, where v = 0.J +
indicates the nuclear orbital angular momentum of the ionic core and N = L + J + is the total angular
momentum excluding spin. L is the total electronic orbital angular momentum.
experiment and theory is quite good in this region, since the states are well sepa-
rated and perturbations are small. Transitions with N = J +−1 appear to be missing,
most noticeable being the absence of the (1, 0) transition. The relative strengths of
the transitions are predicted reasonably well, and the wavelengths also fit the data.
By assigning transitions in the high Rydberg spectrum, the electron-excited
rovibrational state in the c3u metastable manifold can be studied for the first time.
As an example, preliminary results for the electron-impact excitation functions of
the (v, N ) = (1, 1), (2, 1), and (3, 1) ro-vibrational metastable H2 states are shown
in Fig. 20. These measurements are presently being refined, and these results will
be published in the near future [95].
To complete this section, results from stepwise excitation of metastable targets
selected by the upper detector are described. Since this detector selects a narrow
angular range of deflected targets, the density of targets which enter the detector
is low. Stepwise excitation and subsequent field ionization detection inside this
analyzer therefore require long accumulation times.
The motivation for these experiments is to allow differential cross sections to
be measured as described in Sect. II, with the added benefit that individual states
within the metastable manifold can be studied using the stepwise laser process. Fur-
ther, since the scattering angle of electrons which deflect the selected targets can be
FIG. 20. Preliminary studies of excitation functions of individual (v, N ) states in the c3 u
*
metastable manifold of states excited by electron impact. The incident electron energy is plotted on the
x axis and the normalized field ionization signal is plotted on the y axis. The excitation functions are
obtained from stepwise laser excitation of these states to the n = 8 Rydberg states.
ascertained from Eq. (4), an energy-selected electron analyzer placed at this angle
will detect the corresponding electrons. By using coincidence techniques between
the energy-selected scattered electrons and the field-ionized laser-excited targets,
cascade contributions can be eliminated without appreciable loss of efficiency.
Experiments using the coincidence technique have yet to be performed, but
the principle has been demonstrated as shown in Fig. 21. In this case, helium
was chosen as the target, since the laser wavelengths for excitation from the 21S0
metastable state to the n1P1 states are well known (Fig. 16). The detection angle
was 6◦ , corresponding to an electron scattering angle θ e = 24◦ for the fast peak
and θ e = 17◦ for the slow peak, with an uncertainty around ±5◦ in each angle. The
301P1 Rydberg state was excited by the laser, since this produces maximum yield in
the field ionization signal. Signal from the manifold of metastable states is shown
for comparison with the field ionization signal where only the 21S0 metastable
state is selected. The time difference between these signals is due to the field
ionization detector being positioned upstream from the metastable detector. The
field ionization signal was measured as a function of time by delaying the laser
pulse to select different portions of the metastable targets passing adjacent to the
detector. Gaussian peaks are fitted to the signal as shown. The field ionization signal
was collected at each data point for 2 × 104 laser pulses, and the maximum number
of ions detected for each laser pulse was around 1 count for 1000 laser pulses. This
compares with the signal in the lower detector, where over 3500 ions would be
detected for an equivalent number of laser pulses. The signal in the upper detector
is significantly reduced due to angle selection of the deflected metastable atoms.
FIG. 21. Field ionization signal and metastable signal in the upper detector. For details, see text.
Signal from the upper detector can be compared to the results of Cubric et al. [46]
for the 21S0 metastable state, assuming that cascades do not play a significant role.
Cubric et al. found the ratio of differential cross sections for θ e = 17◦ compared
to θ e = 24◦ to be 2.94:1. From the results presented in Fig. 21, the stepwise recoil
experiments produce a ratio of 3.11:1, with an uncertainty of ±0.74. This is in
favorable agreement with the more accurate electron-scattering data.
The agreement between these different techniques shows that the stepwise recoil
experiments are a valid method for obtaining differential cross sections. The recoil
results depend on the velocity variation of the target atoms, and are not as accurate
as those obtained using electron scattering. The main advantage of this technique
is that targets can be studied with states too closely spaced to be analyzed using
conventional methods. The resolution of the laser allows cross sections to be
obtained from individual states as in H2, providing new information which cannot
be obtained by other means.
V. Conclusions and Future Experiments
The experiments described here indicate that the recoil technique pioneered by
Bederson and co-workers [8] continues to provide new information about electron-
impact excitation of a wide variety of targets. The technique has been demonstrated
for evaluation of differential cross sections for atoms excited to metastable states,
and has been shown to provide information about doubly excited states which
decay by fluorescence to metastable states. These latter experiments are currently
being refined, and it is expected that they will provide new information about these
highly excited states.
The principal difference between the experiments described here and the earlier
recoil experiments is the use of laser probing of the excited state. These techniques
allow molecular rotational levels excited by electron impact to be studied for the
first time, and preliminary studies in this area are promising. Electron excitation
functions for these states have been measured for hydrogen, and the first mea-
surements of laser probing of excited recoil targets has been presented. These
results compare well with measurements performed using conventional electron
spectroscopy, and indicate that such measurements are feasible for more complex
targets.
Limitations of the recoil technique have also been shown, and future experi-
ments are being planned to reduce uncertainty due to these restrictions. One of the
main limitations is due to the velocity distribution of the target beam, which pro-
duces a corresponding uncertainty in the calculated electron-scattering direction.
New methods are being developed to reduce this distribution by applying laser
cooling methods. The average velocity and associated distribution can be reduced
using these techniques, with a corresponding increase in sensitivity.
The uncertainty in the electron-scattering direction can also be reduced by
adopting coincidence techniques between the recoil target and the corresponding
electron, since the correlated electron direction is well known from conservation of
momentum and energy. Selection of the momentum of these electrons also removes
cascade contributions, with only a small reduction expected in the counting rate.
The advantages of stepwise laser probing can still be applied in these experiments.
A further limitation of these experiments is due to the resolution of the pulsed
laser, which is severely limited when compared to excitation using a CW laser
source (Fig. 17). Results from laser excitation of H2 (Fig. 18) also show these
limitations, since accurate identification of the transitions to the Rydberg states of
the molecule must be made. Harries [52] has identified some transitions in H2, but
for these experiments to be more generally applicable it is necessary to reconsider
the experimental technique.
These difficulties can be eliminated by adopting a two-step laser process as
shown in Fig. 22. In this case a tunable CW laser excites the metastable target
to an intermediate state, and the pulsed laser further excites the target to a high
Rydberg state from which it is field ionized. This two-step process has a number
of advantages. First, the efficiency of the CW excitation process is high, with up
to 50% of the metastable state population being transferable to the intermediate
state [96]. Second, a stabilized CW ring laser has a linewidth variation of less than
one part in 109. This radiation can therefore not only select individual rotational
and vibrational states in the excited molecular manifold, but can also select targets
FIG. 22. Two-step laser excitation from a metastable target excited by electron impact.
with different isotopic masses [78]. The intermediate states of many molecules are
well known from spectroscopic studies [97], unlike the Rydberg states, which have
not been studied in detail. The results of these spectroscopic studies can therefore
be exploited when determining the electron-excited states under study.
Following CW laser excitation in this scheme, the pulsed laser transfers infor-
mation about the metastable target to the upper Rydberg state, from where it is
field ionized. It is therefore no longer necessary to accurately identify the Ryd-
berg states, since the metastable target state has already been selected. The pulsed
laser and field ionization techniques are exploited to provide an efficient detection
channel with high signal-to-noise ratio as described above.
The disadvantage of this technique is that two lasers are required, with a cor-
responding increase in initial setup and running costs. The experiments are also
more complex, but it has been demonstrated that these difficult experiments are
now possible using current technology. The advantages of using resonant two-
photon excitation from the electron-excited state are expected to be significant,
and it is envisaged that such experiments will be carried out in the near future.
VI. Acknowledgments
We would like to thank the Engineering and Physical Sciences Research Council
in the United Kingdom for funding this work, and would also like to thank our
colleagues at Manchester for providing ideas, advice, and support. In particular,
thanks go to John Reardon, James Harries, and Rob Chandler, who worked as
Ph.D students during the construction and implementation of the apparatus. We
would also like to thank the mechanical and electronic technicians in the Schuster
Laboratory for providing continuing expertise and support.
VII. References
1. Schulz, G. J. (1973). Rev. Mod. Phys. 45, 378.
2. Brunt, J. N. H., Read, F. H., and King, G. C. (1977). J. Phys. E 10, 134.
3. Joyez, G., Comer, J., and Read, F. H. (1973). J. Phys. B 6, 2427.
4. Brunt, J. N. H., King, G. C., and Read, F. H. (1977). J. Phys. B 10, 1289.
5. Pichou, F., Huetz, A., Joyez, G., Landau, M., and Mazeau, J. (1976). J. Phys. B 9, 933.
6. Cvejanovic, S., and Read, F. H. (1974). J. Phys. B 7, 1180.
7. Hammond, P., Read, F. H., Cvejanovic, S., and King, G. C. (1985). J. Phys. B 18, L141.
8. Rubin, K., Pezel, J., and Bederson, B. (1960). Phys. Rev. 117, 151.
9. Ying, C. H., Perales, F., Vuskovic, L., and Bederson, B. (1993). Phys. Rev. A 48, 1189.
10. Jaduszliwer, B., Shen, G. F., Cai, J. L., and Bederson, B. (1985). Phys. Rev. A 31, 1157.
11. Jiang, T. Y., Shi, Z., Ying, C. H., Vuskovic, L., and Bederson, B. (1995). Phys. Rev. A 51, 3773.
12. Vuskovic, L., Zuo, M., Shen, G. F., Stumpf, B., and Bederson, B. (1989). Phys. Rev. A 40, 133.
13. Dorner, R., Mergel, V., Zhaoyuan, L., Ullrich, J., Spielberger, L., Olson, R. E., and Schmidt-
Böcking, H. (1995). J. Phys. B 28, 435.
14. Dörner, R., Bräuning, H., Jagutzki, O., Mergel, V., Achler, M., Moshammer, R., Feagin, J. M.,
Osipov, T., Bräuning-Demian, A., Spielberger, L., McGuire, J. H., Prior, M. H., Berrah, N., Bozek,
J. D., Cocke, C. L., and Schmidt-Böcking, H. (1998). Phys. Rev. Lett. 81, 5776.
15. Mergel, V., Achler, M., Dörner, R., Khayyat, K., Kambara, T., Awaya, Y., Zoran, V., Nyström,
B., Spielberger, L., McGuire, J. H., Feagin, J., Berakdar, J., Azuma, Y., and Schmidt-Böcking, H.
(1996). Phys. Rev. Lett. 80, 5301.
16. Ullrich, J., Moshammer, R., Dörner, R., Jagutzki, O., Mergel, V., Schmidt-Böcking, H., and Spiel-
berger, L. (1997). J. Phys. B 30, 2917.
17. Mergel, V., Dörner, R., Achler, M., Khayyat, K., Lencinas, S., Euler, J., Jagutzki, O., Nüttgens, S.,
Unverzagt, M., Spielberger, L., Wu, W., Ali, R., Ullrich, J., Cederquist, H., Salin, A., Wood, C. J.,
Olson, R. E., Belkic, D., Cocke, C. L., and Schmidt-Böcking, H. (1997). Phys. Rev. Lett. 79, 387.
18. Photonis, Inc., Avenue Roger Roncier, Z.I. Beauregard, B.P. 520, 19106 Brive Cedex, France.
19. Galileo Corporation, P.O. Box 550,Sturbridge, MA 01566, USA.
20. ETP Electron Multipliers Pty Ltd, Ermington, Sydney, NSW, Australia.
22. Newman, D. S., Zubek, M., and King, G. C. (1985). J. Phys. B 18, 985.
24. Buckman, S. J., Hammond, P., Read, F. H., and King, G. C. (1983). J. Phys. B 16, 4039.
25. Fabrikant, I. I., Shpenik, O. B., Snegursky, A. V., and Zavilopulo, A. N. (1988). Phys. Rep. 159, 1.
26. Buckman, S. J., and Clark, C. W. (1994). Rev. Mod. Phys. 66, 2.
27. Shpenik, O. B., Zavilopulo, A. N., Snegursky, A. V., and Fabrikant, I. I. (1984). J. Phys. B 17, 887.
28. Bhadra, R., Callaway, F., and Henry, R. J. M. (1979). Phys. Rev. A 19, 1841.
29. Fon, W. C., Berrington, K. A., Burke, P. G., and Kingston, A. E. J. Phys. B 12, 1861.
30. Fon, W. C., Berrington, K. A., and Kingston, A. E. (1980). J. Phys. B 13, 2309.
31. Pearl, J. C., Donnelly, D. P., and Zorn, J. C. (1969). Phys. Lett. 30A, 145.
32. Defrance, A., Hagene, M., and Pasquerault, D. (1984). J. Physique Lett. 45, L427.
33. Oshima, S., Kondow, T., Fukuyama, T., and Kuchitsu, K. (1990). Chem. Phys. Lett. 169, 331.
34. Metcalf, H. J., and van der Straten, P. (1999). “Laser Cooling and Trapping.” Springer-Verlag, New
York.
35. Hoogerland, M. D., Driessen, J. P. J., Vredenbregt, E. J. D., Megens, H. J. L., Schuwer, M. P.,
Beijerinck, H. C. W., and van Leeuwen, K. A. H. (1996). Appl. Phys. B 62, 323.
36. Hoogerland, M. D., Milic, D., Lu, W., Bachor, H., Baldwin, K. G. H., and Buckman, S. J. (1996).
Austral. J. Phys. 49, 567.
37. Lu, Z. T., Corwin, K. L., Renn, M. J., Anderson, M. H., Cornell, E. A., and Weiman, C. E. (1996).
Phys. Rev. Lett. 77, 3331.
38. Bjorkholm, J., Freeman, R., Ashkin, A., and Pearson, D. (1976). Phys. Rev. Lett. 41, 1361.
39. Baudon, J., Mathevet, R., and Robert, J. (1999). J. Phys. B 32, R173.
40. Raab, E., Prentiss, M., Cable, A., Chu, S., and Pritchard, D. (1987). Phys. Rev. Lett. 59, 2631.
41. Ketterle, W., and van Druten, N. J. (1996). Adv. Atomic Mol. Opt. Phys. 37, 181.
42. Davis, K., Mewes, M.-O., Andrews, M., van Druten, M., Durfee, D., Kurn, D., and Ketterle, W.
(1995). Phys. Rev. Lett. 75, 3969.
43. Read, F. H., and Channing, J. M. (1996). Rev. Sci. Instum. 67, 2372.
44. Zubek, M., Gulley, N., King, G. C., and Read, F. H. (1996). J. Phys. B 29, L239.
45. Allen, M. (2000). J. Phys. B 33, L215.
46. Cubric, D., Mercer, D. J. L., Channing, J. M., King, G. C., and Read, F. H. (1999). J. Phys. B 32,
L45.
47. Murray, A. J., and Hammond, P. (1999). Rev. Sci. Instum. 70, 1939.
48. Beam Dynamics, Inc, 250 Prairie Center Drive, Suite 212, Eden Prairie, MN 55344, USA.
49. Anderson, J. B., and Fenn, J. B. (1965). Phys. Fluids 8, 780.
50. Reardon, J. P. (1995). Ph.D. thesis University of Manchester, UK.
51. Lasertechnics, Inc, Albuquerque, NM 87113, USA.
52. Harries, J. R. (1999). Ph.D. thesis University of Manchester, UK.
53. Fursa, D. V., and Bray, I. (1997). J. Phys. B 30, 757.
54. Trajmar, S., Register, D., Cartwright, D., and Csanak, G. (1992). J. Phys. B 25, 4889.
55. Simpson, J. A., Mielczarek, S. R., and Cooper, J. (1964). J. Opt. Soc. Am. 54, 269.
56. Simpson, J. A., Chamberlain, G. E., and Mielczarek, S. R. (1965). Phys. Rev. 139, A1039.
57. Rudd, M. E. (1964). Phys. Rev. Lett. 13, 503.
58. Rudd, M. E. (1965). Phys. Rev. Lett. 15, 580.
59. Oda, N., Nishimura, F., and Tahira, S. (1970). Phys. Rev. Lett. 24, 42.
60. Madden, R. P., and Codling, K. (1963). Phys. Rev. Lett. 10, 516.
61. Burrow, P. D. (1970). Phys. Rev. A 2, 1774.
62. Westerveld, W. B., Kets, F. B., Heideman, H. G. M., and van Eck, J. (1979). J. Phys. B 12, 2575.
63. Tech, J. L., and Ward, J. F. (1971). Phys. Rev. Lett. 27, 367.
64. van Linden, H. B., van den Heuvell, W., van de Water, H., Heideman, G. M., van Eck, J., and
Moorman, L. (1980). J. Phys. B 13, 2475.
65. Snegursky, A. V., Yu Remeta, E., Zavilopulo, A. N., and Shpenik, O. B. (1994). J. Phys. B 27, 1589.
66. Trantham, K. W., Jacka, M., Rau, A. R. P., and Buckman, S. J. (1999). J. Phys. B 32, 815.
67. Dassen, H. W., Gomez, R., King, G. C., and McConkey, J. W. (1983). J. Phys. B. 16, 1481.
68. Sokell, E., Wills, A. A., Hammond, P., MacDonald, M. A., and Odling-Smee, M. K. (1996). J. Phys.
B 29, L863.
69. Odling-Smee, M. K., Sokell, E., Hammond, P., and MacDonald, M. A. (1998). Proceedings of
VUV-12, Program and abstracts book
70. Odling-Smee, M. K., Sokell, E., Hammond, P., and MacDonald, M. A. (2000). Phys. Rev. Lett. 84,
2598.
71. Rubensson, J. E., Sathe, C., Cramm, S., Kessler, B., Stranges, S., Richter, R., Alagia, M., and
Coreno, M. (1999). Phys. Rev. Lett. 83, 947.
72. Lablanquie, P., Penent, F., Hall, R. I., Eland, J. H. D., Bolognesi, P., Cooper, D., King, G. C., Avaldi,
L., Camilloni, R., Stranges, S., Coreno, M., Prince, K. C., Muehleisen, A., and Zitnik, M. (2000).
73. Gorczyca, T. W., Rubensson, J. E., Sathe, C., Strom, M., Agaker, M., Ding, D., Stranges, S., Richter,
R., and Alagia, M. (2000). Phys. Rev. Lett. 85, 1202.
74. Harries, J. R., Odling-Smee, M. K., Murray, A. J., and Hammond, P. Submitted for publication.
75. Zetner, P. W., Westerveld, W. B., King, G. C., and McConkey, J. W. (1986). J. Phys. B 19, 4205.
76. Hanne, G. F., Nickich, V., and Sohn, M. (1985). J. Phys. B 18, 2037.
77. Webb, C. J., MacGillivray, W. R., and Standage, M. C. (1985). J. Phys. B 18, 1701.
78. McLucas, C. W., MacGillivray, W. R., and Standage, M. C. (1982). Phys. Rev. Lett. 48, 88.
79. MacGillivray, W. R., and Standage, M. C. (1988). Phys. Rep. 168, 1.
80. Murray, A. J., Webb, C. J., MacGillivray, W. R., and Standage, M. C. (1989). Phys. Rev. Lett. 62,
411.
81. Murray, A. J., MacGillivray, W. R., and Standage, M. C. (1990). J. Phys. B 23, 3373.
82. Murray, A. J., MacGillivray, W. R., and Standage, M. C. (1991). J. Mod. Opt. 38, 961.
83. Murray, A. J., MacGillivray, W. R., and Standage, M. C. (1991). Phys. Rev. A 44, 3162.
84. Murray, A. J., Pascual, R., MacGillivray, W. R., and Standage, M. C. (1992). J. Phys. B 25, 1915.
85. Masters, A. T., Murray, A. J., Pascual, R., and Standage, M. C. (1996). Phys. Rev. A 53, 3884.
86. Andersen, N., Gallagher, G. W., and Hertel, I. V. (1988). Phys. Rep. 65, 1.
87. Frey, M. T., Dunning, F. B., Reinhold, C. O., and Burgdörfer, J. (1997). Phys. Rev. A 55, R865.
88. Gallagher, T. F. (1994). ). Rydberg Atoms. Cambridge University Press (Cambridge, UK).
89. Murray, A. J., and Hammond, P. (1999). Meas. Sci. Technol. 10, 225.
90. van Regemorter, H., Dy, H. B., and Prud’homme, M. (1979). J. Phys. B 12, 1053.
91. Moore, C. E. (1971). Atomic energy levels. National Standards Reference Data Series 35-V1.
92. Pellarin, M., Vialle, J.-L., Carré, M., Lermé, J., and Aymar, M. (1988). J. Phys. B 21, 3833.
93. Daviel, S., Wallbank, B., Comer, J., and Hicks, P. J. (1982). J. Phys. B 15, 1929.
94. Herzberg, G. (1950). “Spectra of Diatomic Molecules.” Van Nostrand, London.
95. Harries, J., Chandler, R., Murray, A. J., and Hammond, P. J. Phys. B, in press.
96. Allen, L., and Eberly, J. H. (1987). “Optical Resonance and Two level Atoms” Dover, New York.
97. Dieke, G. H. (1958). See, for example, J. Mol. Spectrosc. 2, 494.
QUANTUM NOISE OF
SMALL LASERS
J. P. WOERDMAN and M. P. VAN EXTER
Huygens Laboratory, Leiden University, 2300 RA Leiden, The Netherlands
N. J. VAN DRUTEN
Department of Applied Physics, Delft University of Technology, Lorentzweg 1
2628 CJ Delft, The Netherlands
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
II. Overview of Threshold Behavior. . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
III. Good-Cavity versus Bad-Cavity Regime . . . . . . . . . . . . . . . . . . . . . . . 211
IV. Spontaneous Emission Factor β . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
V. Cavity QED and β . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
VI. Magnitude of β . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
VII. Relaxation Oscillation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
VIII. Disappearance of Fluctuation Threshold . . . . . . . . . . . . . . . . . . . . . . . 227
IX. Petermann Excess Quantum Noise. . . . . . . . . . . . . . . . . . . . . . . . . . . 236
X. Finite Number of Atoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
XI. Concluding Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
XII. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
XIII. References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
I. Introduction
The theme of this chapter is relevant in view of the present drive toward smaller
lasers. Here one may think of vertical-cavity semiconductor lasers (Chang-
Hasnain, 1998), microdisk and microring semiconductor lasers (Mohideen et al.,
1994; Zhang et al., 1996) photonic bandgap lasers (Lee et al., 1999), rare-
earth microchip lasers (Becher and Boller, 1999), rare-earth microsphere lasers
(Sandoghdar et al., 1996), dye microsphere lasers (Tzeng et al., 1984), micropla-
nar dye lasers (De Martini et al., 1992), one-atom lasers, and one-photon lasers
(An et al., 1994; Löffler et al., 1997; Meyer et al., 1997). Conversely, one may ask
the question how macroscopic laser action emerges out of a one-atom, one-photon
device when the atom and photon numbers are increased.
In this chapter we focus on the question of how the concept “small” should be
quantified; this allows us to compare various lasers in a unified framework. This is
in fact the realm of mesoscopic lasers, i.e., lasers which are between macroscopic
and microscopic. In macroscopic lasers the degrees of freedom of electromagnetic

ISBN 0-12-003847-1/ISSN 1049-250X/01 $35.00
206 J. P. Woerdman et al.
field and gain medium are effectively continuous variables; in mesoscopic lasers
they are discrete variables, with the one-photon, one-atom laser as ultimate limit.
The discreteness introduces quantum noise, the more so the smaller the number
of degrees of freedom. Since the quantum noise depends on details of the system,
the universality of the macroscopic laser gets more and more lost the deeper one
penetrates into the mesoscopic domain (Rice and Carmichael, 1994). Starting
from an ideal, amplitude-stabilized coherent oscillator, the route branches into a
manifold of parallel paths to a variety of nonlinear amplifiers of quantum noise.
These generally still allow a semiclassical description. Much further along the way,
a true quantum description becomes gradually essential (Brecha et al., 1995), in
particular at the end of the route where we reach the cavity quantum electrodynamic
(QED) limit in the form of a one-atom, one-photon device. The latter may come in
many varieties (Löffler et al., 1997; An et al., 1994; Kimble, 1994); a spectacular
example is a deterministic device that emits a single photon upon command (Kim
et al., 1999). The transition between the semiclassical limit and the quantum limit
is of course very vague; this is the realm of mesoscopic physics.
There is a close analogy with the mesoscopic aspects of a material phase tran-
sition. The laser threshold transition of a macroscopic laser can be seen as a
thermodynamic phase transition in a classical macroscopic system (Sargent et al.,
1974). When the system involved is made smaller, the universal nature of any
phase transition is affected by finite-size effects: e.g., the melting temperature of
a cluster of, say, 60 H2O molecules differs from the liquid-phase melting temper-
ature (Berry, 1994). This difference is nonuniversal: when considering molecules
other than H2O, the percentage “correction” of the melting temperature will de-
pend on the specific molecule involved. For sufficiently small clusters, a quantum
description becomes essential, certainly so at the end of the route, where we deal
with a single molecule. Note, however, that a cluster such as C60 is still an almost
classical body because of its many internal degrees of freedom and their possible
couplings to the environment (Arndt et al., 1999).
Some popular notions that hold true for a macroscopic laser are as follows
(Wiseman, 1997):
1. The threshold condition occurs when the round-trip gain equals the round-
trip loss.
2. The laser linewidth does depend on the cavity linewidth but not on the gain
linewidth.
3. The photon statistics of a laser change when the laser goes through threshold.
In fact, as we will see, these statements generally break down for small lasers, long
before we hit the one-photon, one-atom limit.
A lot of what we have said above is not really new, in the sense that much of
it can be found in the semiclassical laser theories developed around 1970 (Haken,
1966; Lax, 1966; McCumber, 1966; Risken, 1970; Seybold and Risken, 1973;
Sargent et al., 1974). In those days, however, theory was far ahead of experiment,
QUANTUM NOISE OF SMALL LASERS 207
so it was mainly the generic theoretical results that made a lasting impression;
the “special” cases that, in our present language, correspond to routes into the
mesoscopic domain were often forgotten (in fact, some of the theory papers that
stimulated us were hardly ever cited). It did not help either that sometimes such a
“special” case was labeled as experimentally unrealistic (that is, around 1970).
A direct comparison of laser theories (old or new) is impossible once we are in
the mesoscopic domain, since the theories start from different models and thus lead
in principle to different results; generic results are found only in the macroscopic
limit. Since these models are approximations of physical reality regarding level
scheme, pumping mechanism, hierarchical ordering of relaxation rates, etc., it is
only by comparison with experiments that we can decide whether a certain model
description is realistic for a certain laser. Moreover, for a given laser model, theories
may be developed with various levels of sophistication. All this explains why there
are so many laser theories.
It is not our ambition to give a systematic overview of this confusing situation.
Instead, we will ad hoc draw on existing theories (including our own), compare
them with experiments, and point to cases where understanding is lacking. We will
emphasize throughout a mesoscopic perspective.
II. Overview of Threshold Behavior
We start with a very simple laser model, namely, a single-mode laser based on a
homogeneously broadened gain transition. We assume an ideal four-level scheme,
i.e., negligible population in the lower laser level (see Fig. 1). For such a laser the
rate equations for the number of intracavity photons s and the number of inverted
atoms N are
ṡ = βγ N (s + 1) − 2κs, (1)

Ṅ = S − βγ s N − γ N , (2)
FIG. 1. Level scheme of a laser. The upper level is pumped at a rate S, the upper level decays at a
rate γa , and the lower level decays at a rate γb . We assume γb ≫ γa so that the inversion decay rate
γ equals γa . A fraction β of the spontaneous emission is in the laser mode which is populated by
s photons. The laser cavity has an intensity outcoupling rate Ŵc and a field outcoupling rate κ, with
Ŵc = 2κ. The dipole dephasing rate of the lasing transition a → b is denoted as γ⊥ .
where S is the pump rate, κ is the cold-cavity damping rate of the field, and γ is
the inversion damping, which equals, for an ideal four-level laser, the upper-level
decay rate. Note that 2κ is the damping rate of the intracavity intensity or photon
number; later in the chapter we will often refer to 2κ as Ŵc , as in van Exter et al.
(1995). Note also that in writing Eqs. (1) and (2) we have assumed an (effectively)
frequency-independent gain, i.e., the polarization of the gain medium (which has
a damping rate γ⊥ ) has been adiabatically eliminated. This is allowed as long
as γ⊥ ≫ κ. The cross-terms Ns and sN in Eqs. (1) and (2) account for the gain
saturation; this nonlinearity is essential for the laser dynamics. The parameter β in
Eqs. (1) and (2) is of a phenomenological nature: it is the fraction of spontaneous
emission of an inverted atom that is emitted in the laser mode (a fraction 1 − β
“escapes”). In writing Eqs. (1) and (2) it has been assumed that β ≪ 1; i.e., we stay
within a semiclassical description where spontaneous emission noise is accounted
for in a perturbative sense. Setting the time derivatives in Eqs. (1) and (2) equal to
zero, we find for the steady-state photon number s0 = s (see also Siegman, 1986)
1
s0 = [(M − 1) + (M − 1)2 + 4M β], (3)
2β
where M ≡ S/Sthr is the dimensionless pump parameter and Sthr ≡ 2κ/β is the
threshold pump rate.
We have plotted Eq. (3) in Fig. 2 for the case β = 10−8 . Far above threshold
(M ≫ 1), we find
s0 = (M − 1)β −1 , (4)
FIG. 2. Steady-state photon number, s0 , as a function of the pumping parameter M ≡ S/Sthr , for a
laser with β = 10−8 . Regime III corresponds to linearized semiclassical theory, regime II to nonlin-
earized semiclassical theory, and regime I to quantum theory.
FIG. 3. Steady-state photon number s0 versus S/ Ŵc , where S is the pumping rate and Ŵc the cavity
intensity decay rate for a laser with β = 10−8 , 10−6 , 10−4 , 10−2 , and 1.
precisely at threshold (M = 1)
s0 = β −1/2 , (5)
and somewhat below threshold (M < 1),

1
s0 = . (6)
1−M
The “sharpness” of the threshold transition, defined as the width of d ln(s)/d lnM,
is ∼β −1/2 , when expressed in the dimensionless parameter M. For β → 1 the
jump is more and more smoothened due to spontaneous emission noise; for β = 1
it disappears completely, leading to a thresholdless laser (Fig. 3). Thus β is a
parameter that characterizes the “system size”; the concept of the laser threshold is
well defined only in the “thermodynamic” limit β −1 → ∞ (Rice and Carmichael,
1994).
For the steady-state inversion number N0 we find
N0 M s0
= = , (7)
Nthr 1 + βs0 s0 + 1
where Nthr = 2κ/βγ is the inversion number at threshold. This implies that the
equilibrium round-trip gain is smaller than the round-trip loss, the more so the
smaller the photon number s0 is, i.e., generally the more so the smaller the laser
is. The gain deficit is made up for by the spontaneous emission input in the mode.
In the limit β → 1 the concept of round-trip gain has lost its significance, since
stimulated and spontaneous emission are no longer meaningful concepts when the
active medium can interact with only one mode of the electromagnetic field.
The merit of introducing the phenomenological parameter β is that it leads to
an expression for s0 as simple as Eq. (3). This expression has generic validity,
i.e., it survives in other models (An and Feld, 1997; Sommers, 1982; Koganov
and Shuker, 1998; Protsenko et al., 1999). If the variables of the gain medium can
be adiabatically eliminated (i.e., κ ≪ γ , γ⊥ ), we are left with the laser field as
the only independent variable which “slaves” the gain variables. In this case of a
so-called class-A laser, β is the only threshold parameter. Equation (3) represents
then a universal route through the mesoscopic domain, from the macroscopic to
the microscopic limit, as far as the steady-state photon number is concerned.
After the above discussion of the steady-state behavior we give now an impres-
sion of the multitude of theories of laser fluctuations. For this case we add noise
sources to Eqs. (1) and (2) and allow ṡ = 0 and Ṅ = 0. Semiclassical laser theories
assume s0 ≫ 1. Quantum corrections thereupon (due to the operator nature of the
field) are of the order of the inversion photon number at threshold, sthr ∼ β −1/2
(Risken, 1970; Seybold and Risken, 1974). As long as sthr ≥ 10 (i.e., β ≤ 10−2 ),
semiclassical theory is basically sufficient. Even for β as large as 0.1, the quan-
tum corrections, despite being large, are not yet dominant. (Coincidentally, these
numbers are not so different from mesoscopic cluster physics, as discussed in
Sect. I).
The labels I, II, and III in Fig. 2 are useful to mark the validity of various class-A
laser theories (Fig. 2 does this for β = 10−8 ). Restricting ourselves first to the
semiclassical approximation, we find that far above threshold (M − 1 ≫ β −1/2 )
the inversion is pinned, leading to a strong reduction of the amplitude fluctuations.
The fluctuations can then be linearized around the steady state, which greatly
simplifies the situation. Such theories (Sargent et al., 1974; Haken, 1966; Kolobov
et al., 1993) are valid in regime III. Close to threshold (regime II) the linearization
breaks down and the full semiclassical equations have to be used (Risken, 1970;
Seybold and Risken, 1974; Gnutzmann, 1998).
Far below threshold the photon number becomes sufficiently small that a semi-
classical description is inadequate. This applies to regime I; a full quantum de-
scription is then required (Löffler et al., 1997). When we increase the value of β,
as in Fig. 3, the (dashed) border lines between the regimes I, II, and III in Fig. 2
shift upward until for β = 1 a quantum theory is required over the full pumping
range; the essentially classical concepts of threshold, amplitude stabilization, and
inversion pinning have then disappeared (Rice and Carmichael, 1994).
If we do not deal with a class-A laser, i.e., if the inversion and /or polarization
of the gain medium cannot be adiabatically eliminated, Eq. (3) remains valid but
the fluctuation properties near threshold will, at least in principle, also depend
on other parameters. A number of examples of this will appear in the rest of the
chapter.
III. Good-Cavity versus Bad-Cavity Regime

As we saw in Sect. II, in a conventional laser the optical field is a slow variable that
slaves the inversion and the polarization of the gain medium, κ ≪ γ , γ⊥ (class-A
laser). In practice, microlasers rarely belong to class A, and theories that make
the class-A assumption therefore have limited usefulness (Pedrotti et al., 1999;
Protsenko et al., 1999). The reason is that, when moving to microlasers, the decay
rate of the intracavity intensity,

c
Ŵc = 2κ = − ln(R1 R2 ), (8)
2L
generally increases, since in practical cases the decrease in cavity length L is

usually not compensated by an increase of the mirror reflectivity product R1 R2 .
The violation of the class-A condition can take several forms. One example is the
class-B laser, κ ≫ γ , which is discussed in detail in Sects. VII and VIII. Another
example is the so-called bad-cavity laser, κ > γ⊥ , as opposed to the good-cavity
laser (κ < γ⊥ ) which is implied by class A. Note that an ideal four-level laser
scheme implies γb ≫ κ (see Fig. 1) and thus leads automatically to the good-
cavity condition, since γ⊥ = γ + γb + 2γcoll ≫ κ. In this section we compare the
good-cavity and the bad-cavity regimes.
In the bad-cavity regime the gain linewidth is narrower than the cavity linewidth;
this has dispersive consequences which can be expressed via the group refractive
index n gr . Using the Kramers–Kronig relations, we find (van Exter et al., 1995)
dn κ
n gr ≡ n + ω = n 0gr + , (9)
dω γ⊥
where n 0gr is the group refractive index of the unpumped gain medium (n 0gr = 1
for a gas laser) and where the term κ/γ⊥ represents the anomalous dispersion
associated with the pumped, resonant gain medium. The effective optical length
n gr L of a bad-cavity laser can thus be much larger than the optical length of the
unpumped cavity (n 0gr L) if n gr ≫ n 0gr ; this shows up in the phenomenon of mode
pulling (Casperson and Yariv, 1973).
Operation in the bad-cavity instead of the good-cavity regime has a dramatic
effect on the linewidth of the laser. The quantum-limited linewidth due to diffusion
of the phase of the laser light is given by the generalized Schawlow–Townes
expression (Haken, 1966; Lax, 1966; Kolobov et al., 1993; Kuppens et al., 1994;
van Exter et al., 1995; Wiseman, 1999),
1 κγ⊥ β κγ⊥
ωℓ = = , (10)
s0 γ ⊥ + κ (M − 1) γ⊥ + κ
where ωℓ is the full linewidth at half-maximum, expressed in angular frequency.

In the good-cavity limit κ ≪ γ⊥ , which is almost always (implicitly) assumed, this
reduces to the “standard” Schawlow–Townes expression found in many textbooks
(Siegman, 1986; Sargent et al., 1974),
κ
ωℓ = . (11)
s0
When expressed in the intensity decay rate Ŵc instead of the field decay rate κ,
Eq. (11) reads ωℓ = Ŵc /2s0 . In the bad-cavity limit the result is
γ⊥
ωℓ = . (12)
s0
Equation (12) is in fact the famous formula in the form originally given by
Schawlow and Townes (1958). Equations (11) and (12) are also valid below thresh-
old if multiplied at the right-hand side by a factor of 2, which accounts for the
amplitude fluctuations which are not suppressed since below threshold there is no
gain saturation that can stabilize the intensity. When proceeding through threshold
from below, the linewidth decreases abruptly due to the sudden increase of s0 by
a factor of order β −1 . For a “β = 1” laser the threshold has disappeared and the
linewidth decreases continuously as a function of the pump strength (Yamamoto
and Imamoglu, 2000).
Figure 4 shows an experimental example of the good-cavity limit of the
Schawlow–Townes linewidth for a semiconductor laser. Such a laser has typically
FIG. 4. Inverse linewidth of a Hitachi HLP1400 AlGaAs semiconductor laser plotted as a function
of the total output power from both facets. (From van Exter et al., 1992, copyright 1992 IEEE.)
γ⊥ ≈ 1013 s −1 and κ ≈ 1011 s −1 (van Exter et al., 1992). The measured linewidth
is inversely proportional to the output power Pout = 2κhνs0 . For completeness we
note that the fit in Fig. 4 was not based on Eq. (11) but on
κ
ωℓ = n sp (1 + α 2 ). (13)
s0
Here n sp ≡ Na /(Na − Nb ) is the incomplete inversion factor that accounts for the
fact that we do not deal with an ideal four-level laser (see Fig. 1) and α is the
linewidth enhancement parameter that accounts for detuning of the cavity mode
and the gain profile. (For simplicity we neglect these complications in the rest of
the chapter.)
Figure 5 shows an experimental example of the bad-cavity limit of the
Schawlow–Townes linewidth for a short HeNe 3.39-μm laser (Kuppens et al.,
1994); the plot shows data for the product νℓ Pout versus Ŵc2 and the solid and
dash-dotted fit curves represent Eq. (10), with and without including the n sp factor
as in Eq. (13), using s0 = Pout / Ŵc hν, Ŵc = 2κ, and n sp = 1.3. Using the standard
Schawlow–Townes formula from the textbooks (i.e., implicitly assuming good-
cavity conditions) leads to the dashed line in Fig. 5 as a fit curve, i.e., a completely
meaningless fit.
FIG. 5. Linewidth–power product of a HeNe 3.39-μm laser at zero detuning as a function of Ŵc2 ,
where Ŵc is the cavity decay rate. The points are measured data, corresponding with different combi-
nations of mirrors with reflectivities R = 8%, 30%, 70%, 90%, and 98%. The dashed line is based
on Eq. (11), whereas the solid and dash-dotted curves are based on Eq. (10) multiplied by n sp , with
n sp ≡ 1 and n sp = 1.3, respectively. (From Kuppens et al., 1994, copyright 1994 APS.)
Interestingly, one can derive the general result Eq. (10) from the good-cavity
result Eq. (11), namely, by appreciating that the optical length of the cavity, and
thus the photon cavity lifetime, are proportional to n gr as given by Eq. (9). We
should therefore introduce a “dressed” cavity loss rate κ/n gr in Eq. (11) instead of
the “bare” cavity loss rate κ. This immediately leads to Eq. (10).
A side issue when pursuing the bad-cavity limit is how far R1 R2 in Eq. (8) can
be reduced before the laser stops being a laser and turns into a source of amplified
spontaneous emission only. By requiring that the reflective feedback fraction of
the intracavity intensity by the mirrors be larger than the fraction of spontaneous
emission into the laser mode and assuming saturation of the inversion, we find as
condition
R1 R2 > β 2 . (14)
Thus β appears again in this context. Equation (14) should not be confused with
the threshold condition; in fact, Eq. (14) distinguishes between feedbackless
amplified spontaneous emission and a subthreshold laser.
IV. Spontaneous Emission Factor β

As we have seen, the threshold behavior of a class-A laser, in which the gain
medium is slaved to the field, is governed by the spontaneous emission factor β. A
calculation of β deals basically with mode counting, i.e., with the cavity geometry.
An accurate calculation is often lengthy and tedious. It involves a summation or in-
tegration of the atom–field coupling over all optical modes of a three-dimensional
cavity and thereby touches the heart of quantum electrodynamics. However, as out-
lined by van Exter et al. (1996), the theory can be greatly simplified by introducing
a few assumptions.
Figure 6 sketches a typical system for which we will calculate β. The emitters
in the active medium are for convenience denoted by “atoms” with transition
dipoles μ12 , although they might equally well be molecules, electrons, excitons,
or any other entities interacting with light. The atoms radiate spontaneously in a
multitude of directions and frequencies. It is now our task to determine the fraction
β of photons emitted into a single discrete mode, which occupies inside the cavity
an optical volume Vcav and subtends a space angle cav . This mode has a finite
spectral width due to outcoupling at the mirrors. The remaining optical modes are
treated as a continuum.
The calculation simplifies considerably when a quasi-one-dimensional
approach is taken, in which the optical continuum is subdivided into modes that
interact with the discrete cavity mode, through the mirrors, and modes that do not.
The former part of the continuum is associated with the “longitudinal” modes, as
they have their wave vectors oriented more or less along the cavity axis, whereas
FIG. 6. Typical geometry for which β is calculated. (From van Exter et al., 1996, copyright 1996
APS.)
the latter part may then be associated with the “transverse” modes. This separa-
tion is depicted in Fig. 7. The various arrows show the pump as energy supplier,
additional loss of atomic coherence due to collisional dephasing (rate γcoll ), spon-
taneous emission to the transverse-optical continuum, coherent coupling to the
discrete cavity mode with a coupling constant g ≡ μ12 E 0 /h, where E 0 is the vac-
uum field, and optical loss out of this mode (loss rate κ). We restrict our attention
to the weak-coupling regime, i.e., g ≪ κ (see also Sect. V). In that case the co-
herent coupling and cavity loss combine to an effective exponential decay from
the excited atomic state to the longitudinal-optical continuum outside the cavity
(Jin et al., 1994; Sachdev, 1984). The spontaneous emission factor β is then
equal to the relative contribution of the latter decay to the total spontaneous emis-
sion rate γ , which also includes (and for β ≪ 1 is dominated by) decay to the
FIG. 7. Schematic diagram of the couplings between the atomic excitation, the discrete cavity mode,
and the optical continuum. (Adapted from van Exter et al., 1996, copyright 1996 APS.)
FIG. 8. Spectra of the spontaneous emission and the cavity resonance. (Adapted from van Exter
et al., 1996, copyright 1996 APS.)
transverse-optical continuum. For the atomic system the radiative decay (as T1 pro-
cess) and collision dephasing (as pure T2 process) combine to γ⊥ = γ + 2γcoll .
In the literature one finds basically two different approaches to calculate β.
Although this is often not clearly stated, these approaches are valid in two mu-
tually excluding regimes, namely, the good-cavity (κ ≪ γ⊥ ) and the bad-cavity
(κ ≫ γ⊥ ) regime. The cavity linewidth κ and the atomic linewidth γ⊥ are indi-
cated in Fig. 8.
In the good-cavity regime the cavity mode has a relatively well-defined fre-
quency. The standard approach is then to assume lateral confinement, by setting
the optical quantization volume Vquant equal to the volume Vcav taken up by the
cavity mode, and simply counting the number of modes p with eigenfrequencies
falling underneath the atomic spectral profile. At resonance this “mode count”
gives (Yamamoto et al., 1991, Baba et al., 1991)
1 1 λ3 ω
β= = 2
× × , (15)
p 4π Vcav γ⊥
where λ and ω are the optical wavelength and optical frequency. Equation (15)
shows that β is basically the inverse of the modal volume, in units of λ3 , multiplied
by the Q factor of the atomic transition. Therefore, it is not only the cavity volume
Vcav that determines how “mesoscopic” a laser is; the atomic linewidth γ⊥ is as
important. The product of Vcav and γ⊥ determines how many free-space modes are
on “speaking terms” with the active medium.
As an aside, there is an interesting historical anecdote connected with Eq. (15)

(see Gutiérrez and Yáñez, 1997). The problem of counting normal modes in a
cavity was apparently introduced by Lorentz during a guest lecture in Göttingen,
addressing black-body radiation. He posed to the mathematicians in the audience
the problem of proving that the mode count is independent of the shape of the
cavity and thus depends only on the volume (if the latter is large enough). Hilbert,
who was in the audience, predicted that this conjecture would not be solved during
his lifetime. Surprisingly, Weyl, a former student of Hilbert, only 2 years later was
able to prove the conjecture.
In the bad-cavity regime (κ ≫ γ⊥ ) the cavity modes are spectrally broad and
it is common to consider the problem from the atomic point of view. When the
emitted light is followed on its path between the mirrors, interference is found to
lead to radiation enhancement for some emission angles and suppression for others
(Yamamoto et al., 1992; Björk et al., 1991, 1993). As κ ≫ γ⊥ this modification
will be more or less constant over the full spontaneous emission spectrum, and for
a very thin (≪ λ) active medium positioned in an antinode of the optical field one
finds at resonance (Matinaga et al., 1993; Yamamoto et al., 1992)

cav 4
β= , (16)
8π 1− R
where cav is the solid angle subtended by the cavity mode and R is the intensity
reflectivity of the cavity mirrors. For a much thicker active medium, which extends
over many nodes and antinodes, the “position-averaged” β is a factor of 2 smaller
than the value given by Eq. (16).
Although this is not at all obvious, Eqs. (15) and (16) are limiting cases of
a single, more general form. This form can be derived by making the following
assumptions:
1. The radiating dipoles μ are randomly oriented in space.
2. The active medium is spread out over many nodes and antinodes in a volume
not exceeding the optical mode volume Vcav .
3. The total spontaneous emission rate is equal to its value in free space.
We first discuss these assumptions in some detail, before giving the resulting
expression for β. Assumption 1 implies that the vector nature drops out of the
problem, and assumption 2 implies that the active dipoles interact on average with
the same electric-field strength. Together they allow us to eliminate the atomic
transition dipole, μ (see Fig. 6), thus clearly showing the geometric nature of the
problem. If desired, some of the restrictions can be easily removed. For instance,
β can be increased by preferential alignment of the dipoles μ along the field
polarization; this increase is at most a factor 3. Preferential positioning of the
dipoles in the antinodes of the field yields another factor 2.
Assumption 3, which takes the total spontaneous emission rate γ to be more or

less equal to its free space value, is a crude assumption, as we know that the cavity
leads to enhanced spontaneous emission in some modes and we are in fact trying
to calculate this enhancement and its effect on β. However, the spontaneous emis-
sion will generally be enhanced only for relatively few modes, whereas it remains
unaffected or will even be suppressed for other modes. Therefore, the cavity leads
mainly to a directionality, i.e., a redistribution over phase space of the emitted pho-
tons, but its effect on the total emission rate γ is generally small and assumption 3
is approximately valid for cavities with relatively small β values (e.g., β < 0.1).
Experimentally, it has been found very difficult to realize β > 0.1; see Sect. VI.
These assumptions allow a calculation of β based on a dipolar atom–field
interaction Hamiltonian combined with Fermi’s “golden rule” for decay into a
continuum of final states. We introduce β(ω) for monochromatic emitters,
1 λ3 + ω , (κ/2)2
β(ω) = (17)
4π 2 Vcav κ (ω − ωcav )2 + (κ/2)2
and integrate β(ω) over the free-space emission spectrum L(ω),

β = dω β (ω)L(ω), (18)
with
γ⊥ /(2π )
L(ω) = . (19)
(ω − ωatom )2 + (γ⊥ /2)2
At resonance (ωcav = ωatom ), the final result becomes
1 λ3 ω
β= , (20)
4π 2 Vcav κ + γ⊥
which generalizes Eq. (15) and reduces to it in the good-cavity regime, κ ≪ γ⊥ .

Substituting now Vcav = AL, in Eq. (20), where A is the mode cross-sectional area
and L is the cavity length, and introducing the diffraction result for the solid angle
of the cavity modes,
2λ2
cav ≈ , (21)
A
we find for the resonant case ωcav = ωatom ,

cav 2ω f sr /π cav 2 κ
β= = . (22)
8π κ + γ⊥ 8π 1− R κ + γ⊥
Here ωf sr is the free spectral range of the longitudinal cavity modes; note that
1− R ωf sr
κ= ωf sr = , (23)
π F
where F is the cavity finesse. Equation (22) generalizes Eq. (16), and reduces to
it in the bad-cavity regime, κ ≫ γ⊥ , apart from a factor of 2. This factor arises
since we have assumed the atoms to be evenly distributed over the cavity volume,
i.e., over nodes and antinodes, instead of being concentrated in the antinodes as
assumed for Eq. (16).
Whether Eq. (20) or Eq. (22) is more convenient depends in practice on the
geometry of the cavity. Anyway, it is now obvious that to obtain a large β, both the
cavity decay rate κ and the atomic decay rate γ⊥ have to be small.
The generalization of Eq. (15) into Eq. (20) can also be seen differently, namely,
by introducing the group refractive index n gr in the mode-counting recipe leading to
Eq. (15). The density of modes in a space filled with the gain medium is (Verdeyen,
1989)
ω2
ρ(ω) = n 2 n gr . (24)
π 2 c3
This leads to an extra factor n gr in Eq. (15); when using Eq. (9) we obtain again
Eq. (20) (Kuppens et al., 1994; van Exter et al., 1995). Note that off-resonance
tuning, i.e., ωcav = ωatom , will lead to different results.
Instead of “dumping” the bad-cavity aspects into an effective β, as was done
in Eqs. (20) and (22), one may also stick with the geometrically defined β as in
Eq. (15). In that case the threshold behavior of the laser can be said to depend
on two parameters, namely, the geometric β and the ratio γ⊥ /κ, instead of one
parameter, i.e., the effective β.
V. Cavity QED and β

As mentioned in Sect. II, rate equation theories of the laser have been developed
on several levels of sophistication. The (phenomenological) spontaneous emission
factor β can be recognized in such theories as a relatively simple algebraic function
of the evolution rates of field and gain medium. The rates that may potentially ap-
pear in β are κ, γ , γ⊥ , and g (Kimble, 1994). The various mesoscopic routes from
a macroscopic laser to a one-photon, one-atom microscopic device correspond in
cavity QED parlance to different hierarchical orderings of these rates.
In the simplest possible macroscopic case, namely, a single-mode, homoge-
neously broadened, four-level class-A laser as considered in Sect. IV, one finds
(Rice and Carmichael, 1994; Jin et al., 1994)
2g 2
β= (25)
γ (γ⊥ + κ)
which is valid as long as β ≪ 1. Although Eq. (25) looks very different from
Eq. (20) or Eq. (22), it is in fact identical; this follows by substitution of g =
μ12 E 0 /h in Eq. (25), using E 0 = (hω/ǫ0 Vcav )1/2 for the vacuum field and γ =
ω3 μ212 /π ǫ0 hc3 for the spontaneous emission rate into free space. In cavity QED the
parameter (β)−1 is known as the “critical” photon number scrit (Kimble, 1994); in
that context the phenomenological interpretation of β as the fraction of spontaneous
emission in the laser mode is irrelevant.
Equations (20), (22), and (25) remain valid as long as they predict values β ≪ 1.
However, they lose their validity, at least if we want to interpret β as the fraction
of spontaneous emission in the laser mode, when the radiative properties of the
active atoms are seriously affected by the electromagnetic boundary conditions
represented by the cavity. Incorporation of this aspect implies tedious calculations
for each specific cavity configuration. Instead, we will keep using the free-space ex-
pression γ as the spontaneous emission rate of an atom in the cavity, but normalize
the resulting value of β “by hand,” by introducing
β
β∗ = (26)
1+β
as the spontaneous emission factor of a cavity QED laser (van Exter et al., 1996; Jin
et al., 1994; An and Feld, 1997). Henceforth we will drop the asterisk as superscript
and refer to a “β = 1” laser. The ad hoc procedure contained in Eq. (26) does of
course no justice to the full richness of a cavity QED laser, but it catches the
semiclassical aspects of such a laser. It is, for instance, a valid description of a
one-atom laser with random atomic injection and small Rabi rotation; a one-atom
laser shows nonclassical photon statistics only if the atoms are injected regularly
or if the Rabi rotation during an atomic transit is at least of the order of 1 rad
(An and Feld, 1997).
It is interesting to see how the magnitude of β compares with the usual cate-
gorization of the weak- and strong-coupling regimes of cavity QED. In the weak-
coupling limit,
g ≪ κ, γ , γ⊥ , (27)
cavity QED effects are completely absent and a semiclassical description is

sufficient. The strong-coupling limit is defined by
g ≫ κ, γ , γ⊥ (28)
and leads to oscillatory Rabi exchange of excitation between field and gain medium;
the field acquires then a strongly nonclassical signature (Löffler et al., 1997; Meyer
et al., 1997; Kimble, 1994). Using Eqs. (25) and (26) and imposing as condition
β ∗ = 1, we find that a “β = 1” laser requires
g 2 ≫ γ (γ⊥ + κ). (29)
This may be called the strong-field limit (Brecha et al., 1995); it says that a sin-
gle photon is able to saturate the inversion. This must not be confused with the
strong-coupling limit Eq. (28); the latter is not necessary to realize a “β = 1” laser
(Yokoyama and Brorson, 1989). On the other hand, we cannot afford to stay in the
weak-coupling limit Eq. (27) if we want to realize a “β = 1” laser; g must be larger
than at least one of the trio γ , γ⊥ , κ. A theoretical study of the case g ≥ κ has
been reported, under the heading “mesomaser” (Balconi et al., 1995); in Sect. VIII
we will encounter the case g ≥ γ . Other connections between cavity QED theory
and mesoscopic lasers will be pointed out in Sect. X.
VI. Magnitude of β
Which β values can be realized in practice? We start with some conventional lasers.
For a typical gas laser such as the HeNe 633-nm laser, one has β ≈ 10−8 (Siegman,
1986). A small, high-gain HeXe gas laser at 3.51 μm can have β = 10−5 –10−6
(van Eijkelenborg et al., 1998). Index-guided edge-emitting semiconductor lasers
have β ≈ 10−5 ; in case of gain guiding one has β ≈ 10−4 (Petermann, 1979;
Arnold et al., 1983; Streifer et al., 1982). Semiconductor vertical-cavity surface
emitting lasers have β = 10−3 –10−4 (Shin et al., 1997). All these β values have
been deduced from the sharpness of the threshold transition and have been found
to be consistent with Eqs. (20) and (22).
Special microring and microdisk semiconductor microlasers have been reported
to have β values in excess of 0.1 (Mohideen et al., 1994; Zhang et al., 1996). How-
ever, these values have not been corroborated by study of the fluctuation prop-
erties but are based on steady-state output versus pump characteristics only; this
method becomes rather unreliable when β approaches unity. In an atomic system a
“β ≈ 1” laser has already been realized, be it effectively in the semiclassical limit
(An et al., 1994; An and Feld, 1997). In a nonlaser context β-values very close to
unity have been realized at microwave frequencies in cavity QED experiments in
the strong-coupling regime, i.e., corresponding to strongly nonclassical features
(Rempe et al., 1990). More recently, this work has been extended into the optical
domain (Hood et al., 1998; Münstermann et al., 1999).
Semiconductor lasers are very many orders of magnitude smaller than gas lasers
and yet their β values are only a few orders of magnitude larger than for gas lasers.
This is of course a consequence of the fact that not only the cavity volume but
also the spectral width of the gain transition enters into Eq. (20); the ratio of these
spectral widths is typically 2 × 1013 Hz/2 × 109 Hz ≈ 104.
Continuing this line of thought, we have tried to obtain higher β values by
using a solid-state “atomic” gain medium, such as Nd3+ :YAG or Nd3+ YVO4.
These systems offer a much larger gain coefficient than a gas laser, i.e., they
can be made much smaller; furthermore, their (homogeneous) linewidth is still
modest (γ⊥ /π ≈ 220 GHz at 300 K). When substituting material parameters as
given in the Casix crystal guide (1997/1998) into Eq. (20) for these systems,
one finds that β ≈ 0.1 requires Vcav ≈ 100 μm3. It seems feasible to realize
such a laser since the gain coefficient of fully inverted Nd3+ :YVO4 is ∼2%/μm
(at 300 K), so a cavity length of 10 μm and a mirror reflectivity product
R1 R2 ≥ 80% would be sufficient. This implies a mode diameter of ∼3 μm. In
this numerical example we have κ ≈ 250 × 109 s−1 and γ⊥ ≈ 780 × 109 s−1 in
Eq. (20), i.e., we are between the good-cavity and bad-cavity regimes. Further
decreasing γ⊥ (e.g., by cooling) does not substantially increase β, unless we si-
multaneously decrease κ. The only published work along these lines (Sandoghdar
et al., 1996) used a Nd3+ -doped-glass microsphere; a value of β ≈ 4 × 10−2 was
deduced by using a mode-counting formula like Eq. (15).
VII. Relaxation Oscillation

The relaxation oscillation of a laser is an oscillatory exchange of excitation between
the optical intensity and the inversion. Relaxation oscillations are well developed
when the inversion damping is much smaller than the cavity photon damping, i.e.,
γ ≪ κ, also called a class-B laser. Practical microlasers tend to be in this class,
since their small length leads to a large value of κ (in practical microlasers the
small length is at most partly compensated by an increase in mirror reflectivities).
Furthermore, the spontaneous emission noise source that excites the relaxation
oscillation becomes increasingly stronger when the laser becomes smaller (more
precisely, when β increases). Therefore, microlaser theories that neglect the relax-
ation oscillation (Jin et al., 1994; Protsenko et al., 1999; Yamamoto and Imamoglu,
1999) are of limited use.
Naively, one might think that for the case γ ≪ κ the field can be adiabatically
eliminated, leaving only the inversion as independent degree of freedom. This
would then be in the same spirit as the reduction to the field as the only independent
degree of freedom for the case γ ≫ κ. However, this is not correct, as can already
be guessed from the fact that the evolution equations for photon number and
inversion number are not symmetric [cf. Eq. (1) and Eq. (2)]. The key point is
that the adiabatic elimination of the field in a class-B laser (γ ≪ κ) leads not to
a stable solution but to a limit cycle solution, i.e., the relaxation oscillation.
The relaxation oscillation is a semiclassical concept, i.e., it refers to a laser
with many photons and many atoms. The theoretical description starts from the
Maxwell–Bloch equations; we adiabatically eliminate the polarization, i.e., we
assume the good-cavity regime, and rewrite the equations in terms of the optical
intensity (instead of the optical field). This leads to the set of rate equations already
given as Eqs. (1)–(2). By linearizing around the steady state one finds (McCumber,
1966; Lax, 1966)

d s −Ŵs γ βs0 s fs
= + , (30)
dt N −Ŵc −ŴN N fN
where s ≡ (s − s0 ) ≪ s0 , N ≡ (N − N0 ) ≪ N0 , and where we have added

Langevin noise sources f s and f N . The assumption N ≪ N0 is an excellent
approximation since for a class-B laser the damping time of the inversion is much
larger than that of the cold cavity, (Mγ )−1 ≫ κ −1 , so that the effect of the (poten-
tially wild) intensity fluctuations on the inversion is almost completely averaged
out. Note, however, that it is incorrect to set N ≡ 0, since this will “kill” the
relaxation oscillation (Protsenko et al., 1999). The assumption s ≪ s0 will be
relaxed in Sect. VIII. Following Lax (1966), we have introduced in Eq. (30) damp-
ing rates of the fluctuations, namely, a “photonic” damping Ŵs and an “atomic”
damping ŴN , defined as
Ŵc βŴc
Ŵs = ≈ , (31)
s0 + 1 M −1
ŴN = γ {1 + β(s0 + 1)} ≈ Mγ , (32)
where we assume s0 ≫ 1. Note that Ŵs is the net damping rate of the intensity
fluctuations in the “hot” cavity, i.e., including the gain medium. We emphasize
that the fluctuations in the photon number and the inversion are both intrinsically
damped, so that the steady state of a class-B laser is indeed a stable solution. Note
also that photonic damping plays no role for a class-A laser, since the condition
γ > κ = 21 Ŵc implies in almost any practical situation that Ŵs ≪ ŴN [see Eqs. (31)
and (32)]. Very close to threshold, however, photonic damping does play a role in
a class-A laser (see Oraevsky, 1990).
The linearized matrix equation (30) can be converted into a second-order dif-
ferential equation with a deterministic part,
d 2 s ds 2
2
+ (ŴN + Ŵs ) + ωr o + ŴN Ŵs s = 0, (33)
dt dt
where
ωr2o = Ŵc γ βs0 ≈ Ŵc γ (M − 1). (34)
This is the equation of a damped harmonic oscillator with a damping rate γr o ,
γr o = 21 (ŴN + Ŵs ), (35)
and a resonance frequency ωres ,

ωres = ωr2o − 14 (ŴN − Ŵs )2 . (36)
Thus, the actual oscillation frequency is given not by Eq. (34) but by Eq. (36). Often
(but not always!) we deal with γr o ≪ ωr o , so that ωres ≈ ωr o . Note that close to
threshold the oscillation can become overdamped when γr o > ωr o .
In conventional lasers the contribution of the photonic damping Ŵs to the relax-
ation oscillation damping γr o is usually very small, i.e., Ŵs ≪ ŴN . When a laser is
made very small, the photonic damping may become important (see Sect. VIII).
Thus “small” lasers are not only distinct from “large” lasers by a greater promi-
nence of the relaxation oscillation but also by a different nature of the damping of
the relaxation oscillation. In the limit of a thresholdless laser (β = 1) the relaxation
oscillation must disappear since, in the absence of a threshold, pump energy is con-
verted directly to laser output, without intermediate storage in the gain medium
(Yokoyama and Brorson, 1989). Another way to appreciate this is that the relax-
ation oscillation has a 90◦ phase shift between photons and atoms, i.e., total energy
is not conserved, whereas β = 1 implies energy conservation with respect to the
photon/atom balance (180◦ phase shift).
The amplitude spectrum s(ω) of the intensity noise can be obtained by Fourier
transformation of Eq. (30), leading to
[γ (1 + βs0 ) − iω] f s (ω) + γ βs0 f N (ω)
s(ω) = . (37)
ωr2o + ŴN Ŵs − ω2 − 2iωγr o
Note that for ω ≈ 0 (specifically, ω ≪ ωr o ) the class-B description of Eq. (36)

should reduce to class A; using d/dt = −iω ≈ 0 and Eq. (36), we find
γ (1 + βs0 ) f s (ω) + γ βs0 f N (ω)
s(ω ≈ 0) = . (38)
2κγ βs0 + ŴN Ŵs
When the relaxation oscillations are well resolved, i.e., ωr o ≫ γr o , we can

expand Eq. (37) around ω ≈ ωr o , where the denominator has a minimum. For these
high frequencies the prefactor in front of f s (ω) by far outweighs the prefactor in
front of f N (ω). In that case we can neglect the inversion noise, and integrate only
the −iω f s part of the noise power over frequency to find
+∞
1

dω Ŵs
|s(ω)|2 ≈ | f s (ω)|2 = s02 , (39)
−∞ 2π 4γr o Ŵs + ŴN
where we have used | f s (ω)|2 = 2κs0 .

Also for low frequencies (ω ≪ ωr o ), it is allowed to neglect the inversion noise
f N (t) if we remain relatively close to threshold so that spontaneous-emission
noise dominates, i.e., βs0 ≪ 1. An experimental example of this latter case is
shown by the solid circles in Fig. 9, for a small (L ≈ 10 cm) HeXe 3.51-μm
laser oscillating in a single mode. We have plotted the power spectrum of the low-
frequency intensity noise versus the steady-state photon number (s0 ) in the cavity.
The fitted curve represents Eq. (38) with f N ≡ 0, taking into account corrections
for the bad-cavity effect (see Sect. III) and the nonideal nature of the Xe 3.51-μm
four-level scheme. This leads to the expression (van Eijkelenborg, 1998)
s0 2

|s(ω ≈ 0)| 2
≈
2
s0
sthr
+ , (40)
κβ s0 sthr
FIG. 9. Intensity-noise strength at low frequency versus the number of photons s0 in the laser cavity;
σ ≡ s − s0 stands for a small excursion of the photon number. We show results for both the stable
cavity (filled circles) and unstable cavity (open circles). The dashed curves are fits to Eq. (40), which
yields the threshold photon number sth for each case. We find sth = 858 ± 60 for the stable-cavity
laser (filled circles) and sth = (15.9 ± 1.5) × 103 for the unstable-cavity laser (open circles). (From
van Eijkelenborg et al., 1998, copyright 1998 APS.)
FIG. 10. Spectrum of a Hitachi HLP1400 AlGaAs semiconductor laser operating at a total output
power of 3.2 mW. The spectrum was obtained by heterodyning the laser output with a tunable local
oscillator. Plotted is the spectral intensity versus the frequency of the local oscillator. In (a) the central
line is fitted to a Lorentzian curve with a HWHM of 34 MHz; (b) gives a different representation of the
same data showing the side bands resulting from relaxation oscillations. (From van Exter et al., 1992,
copyright 1992 IEEE.)
with sthr as the photon number precisely at threshold (M = 1), sthr ∝ β −1/2 with a
prefactor of order unity.
The inversion noise f N plays an important role only high above threshold, when
it dominates the low-frequency intensity fluctuations. When the inversion noise is
suppressed, by using a quiet pump source, the low-frequency intensity fluctuations
disappear completely (“intensity noise squeezing”; see Lathi and Yamamoto, 1999,
and references therein). We will come back to this issue in Sect. IX.
Relaxation oscillations appear not only in the intensity–noise spectrum but also
in the field–noise spectrum, i.e., the optical spectrum of the laser. In this case
a theoretical description starts again from the Maxwell–Bloch equations, which
are now reduced to rate equations for the field E and the inversion N (instead of
s and N ). Amplitude modulation of the field E by the relaxation oscillation then
leads to sidebands spaced by ωr o from the central laser frequency; in the linearized
description there are two (weak) sidebands, on either side of the laser frequency.
An experimental example is shown in Fig. 10, for a semiconductor diode laser;
in such a laser the relaxation-oscillation sidebands have typically a strength of
∼1% of the central carrier; see Fig. 10b. Interestingly, the two sidebands have
different strengths. This is due to the fact that the oscillating inversion leads not
only to amplitude modulation (AM) but also to phase or frequency modulation
(FM) of the field. In fact, FM modulation occurs when there is a detuning of the
cavity mode and the gain profile [α = 0 in Eq. (13)]. The AM and FM modulation
interfere constructively in the red sideband and destructively in the blue side-
band. A semiconductor laser is effectively detuned, since its spectral gain curve is
asymmetric.
Semiclassical relaxation oscillations must somehow transform into one-photon,
one-atom Rabi oscillations when the laser device is reduced to a single atom
interacting with a single field mode. It is interesting to speculate on the na-
ture of this route. It seems likely that the concept of collective Rabi oscillation
(see Zhu et al., 1990; Brecha et al., 1995) is the link between the √ two limiting
cases. In this intermediate regime the dipolar
√ coupling is given by g N , leading
to a well-resolved Rabi oscillation if g N > κ, γ⊥ . In a linearized description
this is expected to correspond to the semiclassical relaxation oscillation; we are
presently studying the nature of the transition.
VIII. Disappearance of Fluctuation Threshold
As argued in Sect. VII, the relaxation oscillation becomes stronger when a laser is
miniaturized; therefore, at some point, we will violate the linearization condition
s ≪ s0 . The onset of this breakdown can be predicted with the linearized theory
as presented in Sect. VII; we follow here the treatment by van Druten et al. (2000).
Frequency integration of Eq. (37) leads to (Lax, 1966)
s 2 ωr2o

ŴN
=1− ≡ 1 − (B)(A), (41)
s02 ŴN + Ŵs ωr2o + ŴN Ŵs
where we have introduced the factors A and B for easy reference. Equation (41) is
related to the second-order reduced factorial moment Q 2 and the Fano factor F of
the photon distribution,
F −1 s(s − 1) − s02 s 2

= Q2 = 2
≈ , (42)
s0 s0 s02
where we assumed s0 ≫ 1 in order to neglect the subtleties associated with the

difference between s0 and s0 + 1. The Fano factor normalizes the intensity fluc-
tuations to the shot-noise level, F = 1 for a Poissonian distribution. The reduced
factorial moment Q 2 normalizes the intensity fluctuations to those in a single
mode of a thermal field. For example, Q 2 = 1 for the Planck distribution in a
single optical mode, while Q 2 = 0 for a Poissonian distribution.
Although Eq. (41) has been derived for a class-B laser, it can be reduced to
recover the well-known result for a class-A laser by using Ŵs ≪ ŴN , i.e., B ≈ 1 in
Eq. (41); the behavior of Q 2 is then determined by the factor A ≤ 1 in Eq. (41),
M 1
Q2 ≈ , (43)
M − 1 s0
where we assumed that s0 ≫ β −1/2 . Thus, for a class-A laser, Q 2 ≈ 0 above thresh-
old, approaching the shot-noise limit (1/s0 ) for M → ∞. In fact, if we pass through
the threshold of a class-A laser from below, Q 2 jumps from ∼1 to ∼0 in an M in-
terval around M = 1 of order β 1/2 ; equivalently, the Fano factor F peaks at M = 1
at a value Fmax ≈ 21 β −1/2 , the width of the peak being of order β 1/2 (Rice and
Carmichael, 1994). Figure 11 illustrates this behavior of Q 2 and F as a function
of the pumping parameter M in a conventional (i.e., class-A) laser.
The appeal of Eq. (41) is that it shows directly that the noise level only drops
appreciably below the thermal level Q 2 = 1 when two conditions are met, namely,
A → 1 and B → 1. For a class-B laser we still have A ≤ 1 in Eq. (41), but now
also the possibility B < 1 arises, namely, when the photonic damping exceeds the
inversion damping, Ŵs > ŴN . Using Eqs. (31) and (32), this condition translates
into
β > M(M − 1), (44)

FIG. 11. Steady-state photon number s0 , second-order moment Q 2 , and Fano factor F of a conven-
tional laser (i.e., class A) as a function of the pumping parameter M.
where we have defined the dimensionless ratio
Ŵc
= , (45)
γ
and assume, as before, the good-cavity regime. Thus, for a class-B laser (i.e.,
≫ 1), the possibility arises that the photon statistics remains thermal (Q 2 ≈ 1)
until far above threshold, as long as Eq. (44) is satisfied [here we assume that
the linearized Eq. (30), which is the basis of our analysis, remains approximately
valid]. Correspondingly, the peak in the Fano factor may shift far beyond M = 1.
We can now associate the threshold with either the kink in the steady-state output
[see Eq. (3)] or with the peak in the Fano factor. For a class-A laser the two
definitions coincide and are both governed by β. For a class-B laser, however, the
steady-state threshold is governed by β and the fluctuation threshold by β and ;
FIG. 12. The solid curve shows the Fano factor F versus pump parameter M for a class-B laser with
β = 10−5 and = 102 ; see Eq. (45) for the definition of . For comparison, the dashed curve shows
the Fano factor of a class-A laser with the same β.
the spontaneous emission events in the lasing mode (at a rate κ) now build up
during the damping time of the relaxation oscillation (∼γ−1 ), yielding a much
more pronounced effect on the intracavity intensity. This integration effectively
increases the fraction of spontaneous emission in the laser mode from β into β,
as far as the fluctuations of the intracavity intensity are concerned.
The behavior of the Fano factor of a class-B laser is easily derived from Eqs. (41)
and (42). The Fano peak acquires a width of order (β)1/2 and occurs at M ≈ 1 +
(β)1/2 . This behavior of the Fano factor is illustrated in Fig. 12 for the case β =
10−5 and = 102 , i.e., β = 10−3 . For β ≫ 1 the laser becomes effectively
thresholdless as far as the intensity fluctuations are concerned; i.e., the inversion
decay into the nonlasing modes has negligible effect upon the fluctuation dynamics.
Note that the fluctuation threshold as discussed in this section refers to the
frequency-integrated noise [cf. Eq. (42)]. This is to be contrasted with the low-
frequency noise as given by Eq. (40) (see also Fig. 9). In the low-frequency limit
there is no distinction between a class-A and a class-B laser and the only thresh-
old parameter is β. As a consequence, when studying the two thresholds of a
class-B laser we may either compare the steady-state threshold with the frequency-
integrated fluctuation threshold, or we may compare the low-frequency fluctuation
threshold with the frequency-integrated fluctuation threshold.
The nature of the fluctuation threshold depends thus not only on the class-A
versus class-B nature (i.e., the value of ), but also on the value of β. This has
been illustrated in the phase diagram of Fig. 13. The macroscopic regime cor-
responds to class-A threshold behavior, characterized by a width β 1/2 . In this
case the atomic damping dominates over the photonic damping, even around the
steady-state threshold. For β > −2 we enter into a region where the photonic
damping exceeds the atomic damping, but in a narrow range above the steady-state
threshold only. The noise threshold is then broadened (but still relatively sharp).
FIG. 13. Schematic display of the various regimes of the fluctuation threshold of a laser. The
parameter is defined by Eq. (45). The solid lines correspond to β = −2 and β = −1 . In
the macroscopic regime the spontaneous emission-driven relaxation oscillation is strongly damped,
in the mesoscopic regime the relaxation oscillation is present but is relatively weak, and in the mi-
croscopic regime it is completely dominating (100% modulation depth). The arrows indicate possible
routes toward a β = 1 laser. (Adapted from van Druten et al., 2000, copyright 2000 APS.)
Following Hofmann and Hess (2000), and in the spirit of the theme of this chapter,
we call this the mesoscopic regime. Finally, for β > −1 the photonic damp-
ing exceeds the atomic damping until far above threshold, so that the intensity
fluctuations have effectively become thresholdless. This may be called the micro-
scopic regime. In this regime the Fano peak has a width of order unity and a peak
value Fmax ≈ (/4β)1/2 . The arrows in Fig. 13 illustrate that one may reach the
fully thresholdless regime β = 1 either directly, always staying in the macroscopic
regime, or via the mesoscopic and microscopic regimes, where the fluctuation
threshold disappears first. The latter route seems more relevant, since small lasers
tend to be class B in the first place.
So far we have concentrated on the regime β ≪ 1, where a semiclassical treat-
ment is sufficient. This means that the extension toward β = 1 of the different
regimes in Fig. 3 is an extrapolation. The validity of this extrapolation will have
to be verified by a more rigorous (quantum) theory. In this regard it is interesting
that from a quantum birth–death model for a β = 1 laser with > 1, Rice and
Carmichael (1994) numerically find scaling laws that closely resemble our analyt-
ical results, strongly suggesting that the extrapolation of our results toward β = 1
is meaningful. For instance, for β = 1 they find Fmax ≈ 0.61/2 at s0 ≈ 1.61/2
(see the discussion of Fig. 8b in their paper, where λ ≡ ), directly analogous to
our results in the microscopic regime Fmax = 21 1/2 at s0 = 1/2 for β = 1).
Surprisingly, the condition β > 1 for thresholdless intensity fluctuations can
be simply rewritten in cavity QED terminology. By using Eq. (25) we can rewrite
FIG. 14. Nd3+ microlaser (not to scale). The laser cavity is formed by the concave surface of the
output coupler and the pump-side surface of the Nd:YVO4 laser crystal. The (optical) cavity length is
∼0.5 mm. The dielectric coatings are indicated, HR for highly reflective, AR for antireflection coating.
(From van Druten et al., 2000, copyright 2000 APS.)
Eq. (44) as a condition on the atom-field coupling g, i.e., g > γ . Thus, the thresh-
oldless nature of the intensity fluctuations can be seen as some kind of remnant of
the vacuum Rabi oscillations of cavity QED; purely weak coupling would imply
that g is smaller than all other relaxation rates, i.e., also g < γ . Presently we are
studying the nature of this connection.
We have performed an experiment on a Nd3+ :YVO4 microlaser to investigate
these issues (van Druten et al., 2000). The configuration of the microlaser cavity is
shown in Fig. 14 and the essence of its level scheme has already been displayed in
Fig. 1. As a complication, the Nd3+ ion does not have an ideal four-level structure,
i.e., the decay rate of the lower laser level, γb , is not larger than the cavity decay
rate Ŵc , although it is very much larger than the decay rate of the upper laser
level (in fact, γb = 1.6 × 109 s−1 , γ = 1.3 × 104 s−1 , and Ŵc = 7 × 1010 s−1 ).
As a consequence, the inversion damping is not given by Eq. (32), but by (see van
Druten et al., 2000; Becher and Boller, 1999; Böhm et al., 1999)

Ŵc
ŴN = γ M + γ (M − 1) , (46)
γb
where the second term generally dominates since γb ≪ Ŵc . Physically speaking,

the response rate of the inversion becomes faster than γ M even if the lower laser
level is hardly populated, since the response of the population of the lower level
is very much faster than that of the upper level, γb ≫ γ . Equivalently, the second
term in Eq. (46) can be seen as a form of “nonlinear gain” or “gain compression,”
a well-known concept in the context of semiconductor laser physics (Becher and
Boller, 1999).
As shown in Fig. 15a, our Nd3+ :YVO4 laser has a sharp steady-state thresh-
old from which we deduce its β value, β = 7 × 10−6 . From the measured
FIG. 15. Diagnostics of Nd3+ microchip laser. (a) Intracavity photon number s0 versus pumping
parameter M; the inset gives a blow-up of the threshold region. (b) Relaxation oscillating frequency
ωr o versus pumping parameter M. (Adapted from van Druten et al., 2000).
relaxation-oscillation frequency versus pump parameter (see Fig. 15b), we deduce

Ŵc = 7 × 1010 s−1 . Using Eqs. (31) and (46), we predict that the point ŴN = Ŵs is
reached around M ≈ 2, so the intensity fluctuations should be thermal for M ≤ 2.
This is illustrated in Fig. 16, which shows experimental time traces of the out-
put intensity for M = 6.6, M = 1.9, and M = 1.03. A reasonably stable output
with a modest relaxation oscillation (see Fig. 16a) changes into a strong, anhar-
monic relaxation oscillation (see Fig. 16c) and then into a highly irregular output
(see Fig. 16e). Note that M = 1.03 is still clearly above the steady-state threshold,
since β = 7 × 10−6 . We emphasize that these time traces are not chaotic; they cor-
respond to damped relaxation oscillations, continuously excited by spontaneous
emission (technical noise is negligible at these time scales). Chaotic time depen-
dence requires three independent variables instead of the two that we have here
(intensity and inversion); see Arecchi (1987). Figures 16b, 16d, and 16f show the
corresponding frequency spectra of the output intensity; higher harmonics of the
relaxation-oscillation appear and blend into a structureless continuum.
The experimental results for Q 2 versus M are illustrated in Fig. 17; instead
of a step function at M = 1, we observe the expected gradual decrease. In this
regime the inversion dynamics are incapable of stabilizing the laser output, which
is then very similar to that of a subthreshold oscillator. Surprisingly, the linearized
Eq. (41), plotted as the solid curve, gives a quite good representation of the data,
hardly worse than the numerical solution of the nonlinearized theory (dashed
curve).
FIG. 16. Typical experimental time traces (left) and RF spectra (right, solid lines) of the output of
the Nd3+ microlaser, for different pump parameters M (a), (b): M = 6.6; (c), (d): M = 1.9; (e), (f):
M = 1.03. The RF spectra are normalized to the average output power, to yield the relative intensity
noise (RIN). The dashed lines in the spectra are the result of numerical integration of the laser Maxwell–
Bloch equations. (From van Druten et al., 2000, copyright 2000 APS.)
FIG. 17. Second-order moment Q 2 of intensity fluctuations [see Eq. (42)] of Nd3+ microlaser versus
pump parameter M. The dots are experimental data, the full curve represents the linearized analytical
theory [Eq. (41)], and the dashed curve results from numerical integration of the laser Maxwell–Bloch
equations. The inset gives the same data but now expressed in the Fano factor F [see Eq. (42)]. (Adapted
from van Druten et al., 2000, copyright 2000 APS.)
Also very interesting is the shape of the photon distribution function; Fig. 18
shows an example for M = 7.28. Although the average photon number in this
experiment is as large as s0 = 4.1 × 105 , the distribution function has the shape of
a Poisson distribution with s0 ≈ 27. This corresponds again to the “fluctuation β”
of this laser being much larger than the conventional β, in this case by a factor
of 104 . For smaller M values the Poisson distribution in Fig. 18 develops in an
approximately exponential (thermal) distribution; see Fig. 19.
The good agreement between our experimental, theoretical, and numerical re-
sults shows that the linearized semiclassical treatment is surprisingly robust, even
when the fluctuations in intensity become comparable to the average. The latter is
by definition the case when Q 2 ≈ 1, i.e., precisely in the regime of interest. Never-
theless, the linearized theory must break down somewhere along the route to larger
values of β; at present β ≈ 1 has been achieved experimentally (van Druten
FIG. 18. Photon-number distribution of Nd3+ microlaser for a pump parameter M = 7.28. The
average photon number is s0 ≈ 4.1 × 105 . However, the shape of the photon distribution is that of a
Poissonian with on average only ∼27 photons in the mode. (Adapted from Lien et al., 2000, copyright
2000 APS.)
FIG. 19. Photon-number distribution of Nd3+ microlaser for a pump parameter M = 1.04. The
fit curve is close to an exponential, showing the almost thermal nature of the laser emission above
threshold. (Adapted from Lien et al., 2000, copyright 2000 APS.)
et al., 2000), but β ≈ 10 seems feasible. One may still profit from linearization
of inversion excursions, N ≪ N0 , even when linearization of the photon num-
ber excursions (s ≪ s0 ) is no longer allowed. By separation of the time scale of
the relaxation oscillation itself and that of its slowly varying amplitude, one can
pursue analytical solutions farther into the nonlinear regime; this corresponds to
the concept of “quasi-energy” (Paoli et al., 1988).
Are there also other lasers besides Nd3+ microlasers where these phenomena
occur? It is interesting to have a look at semiconductor lasers; see also Hofmann and
Hess (2000), where the same issue is addressed. For an edge-emitting semiconduc-
tor laser one has typically β ≈ 10−5 and = Ŵc /γ = 1011 s−1 /109 s−1 ≈ 102 ;
thus β ≈ 10−3 , so the Fano peak should become asymmetric and wider by a factor
1/2 = 10 times wider than the steady-state threshold region (cf. Fig. 12). For a rea-
sonably small oxide-confined VCSEL one has a larger β, namely, β ≈ 10−3 (Shin
et al., 1997); however, one has the same value of = Ŵc /γ , namely, ∼102 , so the
relative increase in the Fano peak width is again a factor of 10. Verification of this
effect requires careful experimentation and in particular the exclusion of spurious
side effects; to our knowledge such an experiment has not yet been reported.
IX. Petermann Excess Quantum Noise

The quantum noise of a laser depends on (among other things) the cold cavity
loss rate Ŵc , either directly [cf. Eq. (11)] or indirectly [cf. Eq. (41)], with expres-
sions for ŴN , Ŵs , and ωr o substituted in these equations. It has been pointed out by
Petermann (1979) that there is a catch to this: different cavity configurations with
the same loss rate κ can lead to very different quantum noise. Siegman (1989)
has shown that the most universal explanation for this is that the eigenmodes of
the cavity may become nonorthogonal due to loss; the eigenmode with the lowest
loss is the “lasing” eigenmode. The nonorthogonality of the cavity eigenmodes
causes noise in the other modes to project into the lasing mode. Petermann excess
noise is thus intrinsically a multimode phenomenon; in practice, however, it oc-
curs in cases that are usually labeled as “single-mode,” since the intensity in the
lasing mode greatly exceeds that in the other modes. Although loss is essential,
the degree of nonorthogonality is not at all uniquely determined by the value of
the round-trip loss (=round-trip gain), but instead by spatial inhomogeneity of
loss or gain in the cavity, and the same therefore applies to the quantum noise
of the laser output. Nonorthogonality due to spatial inhomogeneity may refer to
transverse eigenmodes (Petermann, 1979; Siegman, 1989) or longitudinal eigen-
modes (Hamel and Woerdman, 1989, 1990). Nonorthogonality may also occur
in polarization eigenmodes (van der Lee et al., 1997); in that case it is due to
polarization inhomogeneity (i.e., polarization anisotropy) of the laser cavity. The
equations used so far in this chapter assumed implicitly eigenmode orthogonality;
in the present section we will develop some feeling when and how this must be
generalized. For completeness we note that the K factor is an important concept
for any open quantum system, not just a laser. It arises in the general theory of
scattering resonances, where it enhances the scattering strength of a quasi-bound
state embedded in a continuum (see Frahm et al., 2000).
A first example is a laser based on an unstable cavity. In spite of its name, an
unstable-cavity laser is a perfectly bona-fide laser that can be easily made to emit at a
single frequency in a single spatial mode. In fact, the modal discrimination is much
better than in stable-cavity lasers. A typical example of an unstable-cavity laser is a
Fabry–Perot laser consisting of two convex mirrors (Siegman, 1986); a distributed
variety thereof has two plane mirrors with an index-antiguiding waveguide
between (Petermann, 1979).
The transverse eigenmodes of an unstable resonator are nonorthogonal. This
increases the spontaneous emission by the Petermann excess noise factor K; i.e.,
in a semiclassical treatment β is replaced by Kβ, with K ≥ 1, each eigenmode
having its own K factor (Siegman, 1989). Due to the mode nonorthogonality, the
(enhanced) spontaneous emission in the various eigenmodes is correlated in such
a way that the total spontaneous output of the device is not affected (Haus and
Kawakami, 1985); this is to be expected in view of the fluctuation-dissipation
theorem. Above threshold, however, the laser mode is much stronger than all other
modes, so that its K-factor survives. The value of this K factor depends on the
cavity geometry; for a mirror-based unstable cavity it depends on the transverse
magnification and the Fresnel number. Usually, K has to be calculated numerically;
analytical solutions are available when the mirrors have a Gaussian reflectivity
profile (Doumont et al., 1985).
K values as high as 500 have been measured, and K ≈ 104 looks feasible in
a practical cavity (Cheng et al., 1996; van Eijkelenborg et al., 1996; van der Lee
et al., 1997, 2000a; Karman et al., 1998). Since Kβ > 1 seems within reach in a
reasonably small laser (β > 10−4 ), the question arises whether this implies some
form of thresholdless lasing (see below).
Petermann excess noise can be included in the laser rate equations, e.g., Eqs. (1)
and (2), in a heuristic way. We assume that the atoms in the gain medium have
p = β −1 photon-emission channels available for spontaneous emission. We simply
account for possible mode nonorthogonality by giving one of the photon emission
channels, i.e., the lasing mode, a K-times higher weight than the others. Note that
this weight factor applies only to the spontaneous emission rate into the lasing
mode, not to the stimulated emission rate. This recipe leaves Eqs. (1) and (2)
unchanged apart from replacing (s + 1) in Eq. (1) by (s + K ), i.e., it is as if there
are K noise photons in the laser mode instead of the usual 1. This result can also be
derived more rigorously from the Maxwell–Bloch equations (Dutra et al., 1999).
Since in our model Kβ takes the place of β, we have to assume Kβ ≪ 1, similar
to the assumption β ≪ 1 required for the validity of Eqs. (1) and (2).
Almost all work so far on the K factor has dealt with its effect on the phase
noise. The optical laser linewidth, as given by Eq. (11), is simply multiplied by K,
κ Ŵc
ωℓ = K =K , (47)
s0 2s0
where we assume the good-cavity limit. Figure 20 shows an experimental result for
a small unstable HeXe 3.51-μm laser where K ≈ 200 was deduced from a mea-
surement of the optical linewidth (van Eijkelenborg et al., 1996). Figure 21 shows
an experimental result for the linewidth of a stable HeXe 3.51-μm laser where the
polarization modes have been made nonorthogonal by using intracavity polariza-
tion optics; the linewidth (i.e., K ) is seen to diverge when the polarization modes
become parallel (van der Lee et al., 2000a). This polarization variety of Petermann
excess noise is not fundamentally different from the transverse eigenmode variety;
it has the experimental advantage, however, that the degree of nonorthogonality
can be easily adjusted. A theoretical advantage is that it allows analytical theory,
since only two modes are involved.
Equation (47) has been derived and experimentally verified for the usual case
that ωℓ ≪ Ŵc , i.e., s0 ≫ K . For s0 ≤ K one expects that the cavity acts as a
linear spectral filter so that ωℓ ∼ Ŵc . Thus, Schawlow–Townes narrowing should
set in for s0 > K instead of s0 > 1. The same statements apply to the bad-cavity
limit when K is replaced by γ⊥ . These predictions have not yet been verified
experimentally.
The effect of the K factor on the laser intensity noise has been studied recently
(van Eijkelenborg et al., 1998). Theoretically, one finds from Eqs. (1) and (2), after
FIG. 20. Resonance behavior of the Petermann excess noise factor of a HeXe 3.51-μm unstable-
cavity laser as a function of the equivalent Fresnel number Neq : (a) theoretical; (b) experimental. The
curves in (a) and (b) have been added to guide the eye. (From van Eijkelenborg et al., 1996, copyright
1996 APS.)
replacing (s + 1) by (s + K ), for the steady-state photon number
1
s0 = [(M − 1) + (M − 1)2 + 4M Kβ]. (48)
2β
Comparison with Eq. (3) shows that the abruptness of the threshold transition is
now governed by Kβ instead of β. This is illustrated in Fig. 22, which shows the
cases Kβ = 10−6 , 10−4 , and 10−2 . Theory for Kβ ≥ 1 is not available, but we
anticipate that the threshold in Fig. 22 will simply disappear, as it does in Fig. 3
for β → 1. This implies thresholdless lasing, but of a different nature than in
Fig. 3 since the threshold, when fading away in Fig. 22, does not move toward a
vanishing pump rate but stays at a fixed pump rate. In other words, thresholdless
lasing corresponding to Kβ > 1 is naturally inefficient, contrary to the β = 1
thresholdless laser, which has 100% efficiency.
FIG. 21. Petermann excess noise factor in the optical linewidth of a HeXe 3.51-μm laser with
nonorthogonal polarization modes. The angle between the eigenpolarizations of the cavity is α. The
squares are experimental points at an output power of 9 μW. The dashed theoretical curve represents the
prediction of mode-nonorthogonality theory and the solid theoretical curve incorporates, apart from
the (linear) mode-nonorthogonality theory, the effect of saturation. (Adapted from van der Lee et al.,
2000a, copyright 2000 APS.)
Experimentally, the effect of the K factor on the intensity noise spectrum is

shown in Fig. 9, in which we compare the low-frequency intensity noise of a
stable- and unstable-cavity HeXe 3.51-μm lasers. The peaks in Fig. 9 occur at the
respective thresholds; the peak positions are consistent with the values of Kβ for
the two lasers (van Eijkelenborg et al., 1998). Figure 23 shows the effect of the
K factor on the intensity noise of a HeXe 3.51-μm laser for the polarization variety
FIG. 22. Laser threshold characteristics in the presence of Petermann excess noise. The intracavity
photon number s0 is plotted versus the dimensionless pump parameter M. The drawn curves are
calculated from Eq. (48) using β = 10−6 and K = (a) 1, (b) 102 , and (c) 104 . The presence of excess
noise smoothens the input–output curve. (From van Eijkelenborg et al., 1998, copyright 1998 APS.)
FIG. 23. Petermann excess noise factor in the low-frequency (0.5-MHz) intensity noise of a HeXe
3.51-μm laser with nonorthogonal polarization modes. The angle between the eigenpolarizations of
the cavity is α. The fit curve corresponds to mode-nonorthogonality theory. (Adapted from van der Lee
et al., 2000a, copyright 2000 APS.)
of Petermann excess noise; the noise strength diverges when the polarization modes
become parallel (van der Lee et al., 2000a).
The K factor has been shown to be frequency dependent due to the dynamics
of the gain medium. In other words, the excess quantum noise appearing in the
intensity noise spectrum is “colored” as opposed to true quantum noise, which
is “white.” This is illustrated in Fig. 24 (van der Lee et al., 1998). This spectral
coloring is often Overlooked, since one generally assumes one mode to dominate
FIG. 24. Coloring of the Petermann excess quantum noise, measured in the enhancement of the
intensity noise of a stable-cavity HeXe gas laser, for two different values of the nonorthogonality
of the polarization eigenmodes, i.e., two different values of cos α. The dashed curves are fits to the
experimental data. Note that the bandwidth of the coloring is smallest for the case of largest excess
noise factor. (Adapted from van der Lee et al., 1998, copyright 1998 APS.)
over all others and thus neglects the dynamics of weak nonorthogonal side modes.
It is the correlated dynamics in these side modes that project into the measurement
direction and can thereby partially cancel the excess noise. These projections can
be relatively strong, as they correspond to a type of heterodyning and are therefore
first-order in the side-mode amplitude. Spectral coloring shows if one explicitly
takes into account the time dependence of all side modes and thereby proceeds
beyond the standard geometric picture of excess noise (van Exter et al., 2000).
The “nonorthogonality theory” of Petermann excess noise is a semiclassical
theory; it has the advantage of giving a very simple “geometric” description. It has
the disadvantage that it cannot be straightforwardly quantized, since the complex
amplitudes of a set of classical nonorthogonal modes cannot be turned into a set
of commuting operators. Grangier and Poizat (1998, 1999) have shown how this
problem can be circumvented by introducing appropriate “vacuum modes” that
allow one to recover the unitarity of the input–output scattering matrix.
Very recently, attention has been drawn to the fact that Petermann excess noise
may have disastrous effects on the intensity noise squeezing in a quietly pumped
semiconductor laser as discussed in Sect. VII (van der Lee et al., 2000b). The cru-
cial point is that nonorthogonality leads to effects in first order of the nonlasing-
mode amplitudes, in contrast to many multimode effects considered previously that
are second-order effects (Lathi and Yamamoto, 1999). Thus, mode nonorthogonal-
ity leads to homodyning of the noise of side modes into the lasing mode; this can
destroy squeezing even when the side modes are suppressed, intensity-wise, far
below the shot noise of the lasing mode. It can be shown that above a critical value
of the K factor (K crit = 32 ) the laser intensity noise cannot be brought below shot
noise even for perfectly quiet pumping. In real semiconductor lasers the existence
of mixed gain-index guiding leads to nonorthogonality of the transverse modes
and implies K > 1; the critical value of 23 may be easily exceeded, depending on
the details of the device architecture.
The value K crit = 32 may seem strange, since rarely in physics does a factor
3
2
appear. One must realize, however, that the number of “extra” noise photons
in the laser mode is K − 1. When this number is of the order of the shot noise,
i.e., K − 1 = O(1), one must expect that squeezing due to quiet pumping becomes
impossible, since quiet pumping affects only the “original” noise photon in the laser
mode and not the projected noise from the other modes. The actual calculation
gives K crit − 1 = 21 , where the “ 12 ” is due to the 3 dB extra noise above shot noise
associated with the (linear) phase-insensitive amplification of the nonlasing mode
(Haus and Mullen, 1962).
X. Finite Number of Atoms

We now address the role of the number of atoms in the mode volume of a laser,
Nat , instead of the number of photons, s. The latter was continuously emphasized
in the previous sections; as we have seen, all kind of deviations from the behavior
of a conventional laser may occur even if the photon number remains much larger
than unity.
Does a similar statement also hold for the atom number Nat ? In fact, it does,
namely, when Nat becomes of order Nthr , where Nthr is the threshold inversion
required for lasing, Nthr = Ŵc /βγ = β −1 in the good-cavity case. In the usual
laser theories it is assumed that Nat ≫ Nthr ; in that case the pumping process can
be considered as Poissonian and can be simply described by a pumping rate R
in the rate equation for the inversion [cf. Eq. (2)] (Protsenko et al., 1999). This
assumption is so common that it is hardly ever made explicit. Nevertheless, when
making a laser smaller and smaller, the condition Nat ≫ Nthr may break down.
This has been discussed (see Enomoto et al., 1996) in work on pulsed microcavity
dye lasers; it was pointed out that the S-shaped output–input curve (as shown in
Fig. 2) is greatly “softened” when the number of dye molecules in the mode volume
is only marginally larger than the required threshold inversion number. In other
words, a naively deduced β value based on this S curve would be far too large. This
issue has been addressed theoretically, studying the transition from Nat ≫ Nthr to
Nat ≪ Nthr (Koganov and Shuker, 1998). This analysis confirms that the threshold
indeed disappears for Nat = Nthr . Thus, we deal here with yet another variety of
the thresholdless laser.
In a cavity QED context (not dealing with lasers), the importance of the atom
number Nat and the photon number s is well known (Kimble, 1994). In that con-
text the so-called critical (or saturation) value of Nat is β −1 and the critical
(or saturation) value of s is given by scrit = β −1 . The emphasis is then on cases
where Nat ≈ β −1 ≈ 1 and s ≈ β −1 ≈ 1.
Experimentally, the situation that Nat ∼ Nthr can be realized with a
continuous-wave (CW) pumped Nd3+ :YVO4 microlaser, which has [Nd3+ ] =
Nat ≈ 1 × 1020 cm−3 . Can all Nd3+ ions (in the mode volume) be indeed inverted,
using a practical pump, in this case a Ti:Sa laser at 809 nm? This requires that
the pump intensity is much larger than the saturation intensity of the pump transi-
tion. Based on the absorption coefficient of Nd3+ :YVO4 of 31 cm−1 at the pump
laser wavelength and an upper-level lifetime of 90 μs (data taken from the Casix
Crystal Guide 1997/1998), we estimate the saturation intensity as 104 W/cm2 .
Full inversion requires then, say, 105 W/cm2 as pump intensity, e.g., 100 mW in a
10-μm-diameter pump spot, which is feasible. Assuming full inversion and using
a stimulated-emission cross section of 2.5 × 10−18 cm2 , we calculate a gain co-
efficient of 2.5%/μm. Therefore a fully inverted Nd3+ :YVO4 microlaser with a
chip thickness of 40 μm and a mirror-reflectivity product of R1 R2 = e−2 satisfies
the criterion Nat = Nthr .
It is straightforward to estimate Nthr for the general case (i.e., beyond the nu-
merical example given above). We have Nthr = Ŵc /βγ = β −1 in the good-cavity
case. We conclude that finite-atom-number effects will occur when the atom num-
ber is reduced to roughly β −1 . It would be very interesting to study the fluctuation
FIG. 25. Level scheme of a laser with emphasis on the recycling of the active atoms (cf. Fig. 1).
properties of such a device, since its pump process is highly non-Poissonian. A

theory for these nonclassical fluctuation aspects does not seem to be available.
The finite-atom-number effect discussed here must not be confused with another
effect, namely, intensity noise squeezing due to depletion of the atomic ground
state by the pump (Hart and Kennedy, 1991; Ritsch and Zoller, 1992; Koganov
and Shuker, 1999). This effect is a consequence of the conservation law of level
populations in a single atom, i Ni = 1 (in contrast, the finite-atom-number effect
is based on the conservation law for the total number of atoms). The dynamical
recycling of the active laser atoms to the upper laser level after having emitted
a laser photon (see Fig. 25) leads to intensity noise squeezing. This condition
requires in practice impossibly high pump parameters, e.g., M > 105 in case of
a Nd3+ :YVO4 laser. Kolobov et al. (1993) also discovered this noise suppression
due to dynamical recycling.
XI. Concluding Summary

We have discussed the various aspects that appear when a laser is made “small.”
What counts is not the physical size of the device, but the number ( p) of degrees
of freedom of the optical field. This leads to the spontaneous emission factor
β (= p −1 ); this is the key geometric parameter that characterizes the threshold
behavior of the laser. For a conventional (i.e., “class-A”) laser, the variables of the
gain medium are slaved to the cavity field and β is easily calculated. When the
size of a laser is reduced, the variables of the gain medium become increasingly
independent. Particularly relevant then is the case that the inversion decay rate
is much smaller than the cavity decay rate (“class-B” laser) and/or the case that
the cavity decay rate is much larger than the gain linewidth (“bad-cavity” laser).
For these cases, apart from β, also another dimensionless parameter, this time
associated with the gain medium, determines the threshold behavior. This is γ⊥ /κ
for a bad-cavity laser [cf. Eq. (25)], or = Ŵc /γ = 2κ/γ for a class-B laser
[cf. Eq. (45)], or Ŵc /γb for a nonideal four-level laser [cf. Eq. (46)].
Small lasers, even when emitting in a single spatial mode, are much less coherent
than large lasers. For the case of a class-B laser this is associated with the fact that
the spontaneous emission-driven relaxation oscillation becomes very strong. As
a function of pump rate, the average output intensity shows a sharp threshold,
but the intensity fluctuations do not. Under quite generic conditions the intensity
fluctuations of a microlaser may show highly super-Poissonian statistics, due to
the very weak damping of the spontaneous emission-driven relaxation oscillation.
This classification corresponds to the mesoscopic and microscopic regimes
in the phase diagram of Fig. 13. It should be pointed out that the distinction
between these regimes is not identical to the distinction between class-A and class-
B lasers. The latter distinction is based purely on (=Ŵc /γ ); < 1 for class-A
lasers and > 1 for class-B lasers. Whereas class-A lasers always operate in
the conventional, “macroscopic” regime, class-B lasers may operate in any of the
three regimes, depending on the value of β. The current trend toward smaller laser
devices leads to lasers with increased β and ( increases because typically γ is
a material property of the gain medium, while κ increases with decreasing cavity
length). Thus, this trend will naturally lead to lasers that have “mesoscopic” or
even “microscopic” intensity fluctuations.
From the point of view of applications, it is not necessarily a disadvantage that a
small laser is naturally noisy/incoherent. On the contrary, for optical recording and
reading, an incoherent light source is preferred, since it is much less susceptible to
noise due to (unavoidable) optical feedback. The light source has to emit, however,
in a single spatial mode, as is the case for the microlasers discussed in this chapter.
We have elucidated the role of Petermann excess noise due to mode non-
orthogonality; this may make a laser much noisier than would naively correspond
to its cavity loss rate. Particularly relevant is that this may impede intensity noise
squeezing in quietly pumped semiconductor lasers. Fully or partly gain-guided
laser devices are very unfavorable in this respect; this is unfortunate, since a cer-
tain amount of gain guiding is unavoidable in any efficient device since the gain
must be localized, i.e., must not extend beyond the volume of the oscillating mode.
As a unifying remark, Petermann excess noise can be seen as a consequence
of inhomogeneity of the laser device. Transverse spatial inhomogeneity results
from having a cavity with convex mirrors, or from waveguiding by an index or
gain profile; longitudinal spatial inhomogeneity results from strong outcoupling.
Both varieties may lead to excess noise (K > 1). Polarization inhomogeneity
results from polarization-dependent elements inside the cavity; this may also
lead to excess noise (K > 1). Finally, spectral inhomogeneity, in the form of a
detuning of the cavity mode and the gain peak, also leads to excess noise via the α
factor [cf. Eq. (13)]. Formally, the α factor can be seen as a variety of the K factor
(van Exter et al., 2000).
In the laser literature one finds much emphasis on the value of the photon num-
ber s in its route s = ∞ → s = 1. The importance of the atom number Nat in the
similar route Nat = ∞ → Nat = 1 has only recently been appreciated in a laser
context, and the nonclassical consequences of the pump correlations imposed
by Nat ≈ Nthr ≫ 1, where Nthr is the threshold inversion, await experimental
verification.
We have encountered a variety of thresholdless lasers. Conditions for thresh-
oldless lasing are β = 1 (Sect. II), β ≫ 1 (Sect. VIII), Kβ ≫ 1 (Sect. IX) or
Nat = Nthr (Sect. X). It will be very interesting to study the fluctuation behavior
for these various cases, both theoretically and experimentally.
Finally, we come back to the common wisdom as expressed in the statements
1–3 in Sect. I. For small lasers these statements should be amended as follows:
1. For a small laser the threshold condition is gain < loss; the difference is
made up by spontaneous emission, which becomes increasingly important
for smaller lasers.
2. The laser linewidth depends also on the gain linewidth when the cavity
linewidth becomes larger than the gain linewidth (bad-cavity laser); this is a
natural situation for a sufficiently small laser.
3. The photon statistics of a sufficiently small class-B laser do not change when
the pump exceeds threshold due to strong relaxation oscillations, sufficiently
small being quantified by Eq. (44).
XII. Acknowledgments
We acknowledge the Foundation for Fundamental Research of Matter (FOM) and
the European Community (TMR Network Microlasers and Cavity QED) for fi-
nancial support. The research of NJvD was made possible by the Koninklijke
Nederlandse Akademie van Wetenschappen.
XIII. References
An, K., and Feld, M. S. (1997). Phys. Rev. A 56, 1662.

An, K., Childs, J. J., Dasari, R. R., and Feld, M. S. (1994). Phys. Rev. Lett. 73, 3375.
Arecchi, F. T. (1987). In “Instabilities and Chaos in Quantum Optics” (F. T. Arecchi and R. G. Harrison,
Eds.), p. 9. Springer, Berlin.
Arndt, M., Nairz, O., Vos-Andreae, J., Keller, C., van der Zouw, G., and Zeilinger, A. (1999). Nature
401, 680.
Arnold, G., Petermann, K., and Schlosser, E. (1983). IEEE J. Quant. Electron. 19, 974.
Baba, T., Hamano, T., Koyama, F., and Iga, K. (1991). IEEE J. Quant. Electron. 27, 1347.
Balconi, C., Casagrande, F., Lugiato, L. A., Lange, W., and Walther, H. (1995). Opt. Commun. 114,
425.
Becher, C., and Boller, K. J. (1999). J. Opt. Soc. Am. B 16, 286.
Berry, R. S. (1994). In “Clusters of Atoms and Molecules” (H. Haberland, Ed.), p. 187. Springer,
Berlin.
Björk, G., Machida, S., Yamamoto, Y., and Igeta, K. (1991). Phys. Rev. A 44, 669.
Björk, G., Heitmann, H., and Yamamoto, Y. (1993). Phys. Rev. A 47, 4451.
Böhm, R., Baev, V. M., and Toschek, P. E. (1997). Opt. Commun. 134, 537.
Brecha, R. J., Orozco, L. A., Raizen, M. G., Xiao, M., and Kimble, H. J. (1995). J. Opt. Soc. Am. B 12,
2329.
Casperson, L., and Yariv, A. (1970). Appl. Phys. Lett. 17, 259.
Chang-Hasnain, C. J. (1998). Optics & Photonics News, May, 35.
Cheng, Y. J., Fanning, C. G., and Siegman, A. E. (1996). Phys. Rev. Lett. 77, 627.
De Martini, F., Cairo, F., Mataloni, P., and Verzegnassi, F. (1992). Phys. Rev. A 46, 4220.
Doumont, J. L., Mussche, P. L., and Siegman, A. E. (1989). IEEE J. Quant. Electron. 25, 1960.
Dutra, S. M., Joosten, K., Nienhuis, G., van Druten, N. J., van der Lee, A. M., van Exter, M. P., and
Woerdman, J. P. (1999). Phys. Rev. A 59, 4699.
Enomoto, T., Sasaki, T., Sekiguchi, K., Okada, Y., and Ujihara, K. (1996). J. Appl. Phys. 80, 6595.
Frahm, K. M., Schomerus, H., Patra, M., and Beenakker, C. W. J. (2000). Europhys. Lett. 49, 48.
Gnutzmann, S. (1998). Eur. Phys. J. D 4, 109.
Grangier, P., and Poizat, J.-P. (1998). Eur. Phys. J. D. 1, 97.
Grangier, P., and Poizat, J.-P. (1999). Eur. Phys. J. D. 7, 99.
Gutiérrez, G., and Yáñez, J. M. (1997). Am. J. Phys. 65, 739.
Haken, H. (1966). Z. Physik 190, 327.
Hamel, W. A., and Woerdman, J. P. (1989). Phys. Rev. A 40, 2785.
Hamel, W. A., and Woerdman, J. P. (1990). Phys. Rev. Lett. 64, 1506.
Hart, D. L., and Kennedy, T. A. B. (1991). Phys. Rev. A 44, 4572.
Haus, H. A., and Kawakami, S. (1985). IEEE J. Quant. Electron. 21, 63.
Haus, H. A., and Mullen, J. A. (1962). Phys. Rev. 128, 2407.
Hood, C. J., Chapman, M. S., Lynn, T. W., and Kimble, H. J. (1998). Phys. Rev. Lett. 80, 4157.
Hofmann, H. F., and Hess, O. (2000). Phys. Rev. A, 62, 063807.
Jin, R., Boggavarapu, D., Sargent III, M., Meystre, P., Gibbs, H. M., and Khitrova, G. (1994). Phys.
Rev. A 49, 4038.
Karman, G. P., Lindberg, Å. M., and Woerdman, J. P. (1998). Opt. Lett. 23, 1698.
Kim, J., Benson, O., Kan, H., and Yamamoto, Y. (1999). Nature 397, 500.
Kimble, H. J. (1994). In “Cavity Quantum Electrodynamics” (P. R. Berman, Ed.), p. 203. Academic,
New York.
Koganov, G. A., and Shuker, R. (1998). Phys. Rev. A 58, 1559.
Kolobov, M. I., Davidovich, L., Giacobino, E., and Fabre, C. (1993). Phys. Rev. A 47, 1431.
Kuppens, S. J. M., van Exter, M. P., and Woerdman, J. P. (1994). Phys. Rev. Lett. 72, 3815.
Lathi, S., and Yamamoto, Y. (1999). Phys. Rev. A 59, 819.
Lax, M. (1966). In “Brandeis University Summer Institute in Theoretical Physics,” (M., Chretien, E. P.,
Gross, and S. Deser, Eds.), Vol. 2, p. 271. Gordon and Breach, New York.
Lee, R. K., Painter, O. J., Kitzke, B., Scherer, A., and Yariv, A. (1999). Electron. Lett. 35, 569.
Lien, Y., de Vries, M. S., van Druten, N. J., van Exter, M. P., and Woerdman, J. P. (2001). Phys. Rev.
Lett. 86, 2786.
Löffler, M., Meyer, G. M., and Walther, H. (1997). Phys. Rev. A 55, 3923.
Matinaga, F. M., Karlsson, A., Machida, S., Yamamoto, Y., Suzuki, T., Kadota, Y., and Ikeda, M. (1993).
Appl. Phys. Lett. 62, 443.
McCumber, D. E. (1966). Phys. Rev. A 141, 306.
Meyer, G. M., Briegel, H.-J., and Walther, H. (1997). Europhys. Lett. 37, 317.
Mohideen, U., Hobson, W. S., Pearton, S. J., Ren, F., and Slusher, R. E. (1994). Appl. Phys. Lett. 64,
1911.
Münstermann, P., Fischer, T., Maunz, P., Pinkse, P. W. H., and Rempe, G. (1999). Phys. Rev. Lett. 82,
3791.
Oraevskii, A. N. (1990). JETP Lett. 51, 296.
Paoli, P., Politi, A., and Arecchi, F. T. (1988). Z. Physik B 71, 403.
Pedrotti, L. M., Sokol, M., and Rice, P. R. (1999). Phys. Rev. A 59, 2295.
Petermann, K. (1979). IEEE J. Quant. Electron. 15, 566.
Protsenko, L., Domokos, P., Lefévre-Seguin, V., Hare, J., Raimond, J. M., and Davidovich, L. (1999).
Phys. Rev. A 59, 1667.
Rempe, G., Schmidt-Kaler, F., and Walther, H. (1990). Phys. Rev. Lett. 64, 2783.
Rice, P. R., and Carmichael, H. J. (1994). Phys. Rev. A 50, 4318.
Risken, H. (1970). In “Progress in optics,” (E. Wolf, Ed.), Vol. VIII, p. 239. Amsterdam, North-Holland.
Ritsch, H., and Zoller, P. (1992). Phys. Rev. A 45, 1881.
Sachdev, S. (1984). Phys. Rev. A 29, 2627.
Sandoghdar, V., Treussart, F., Hare, J., Lefévre-Seguin, V., Raimond, J. M., and Haroche, S. (1996).
Phys. Rev. A 54, R1777.
Sargent, M., Scully, M. O., and Lamb, Jr., W. E. (1974). “Laser Physics.” Addison-Wesley, London.
Schawlow, A. L., and Townes, C. H. (1958). Phys. Rev. 112, 1940.
Seybold, K., and Risken, H. (1974). Z. Physik 267, 323.
Shin, J. H., Ju, Y. G., Shin, H. E., and Lee, Y. H. (1997). Appl. Phys. Lett. 70, 2344.
Siegman, A. E. (1986). “Lasers.” University Science Books, Mill Valley, CA.
Siegman, A. E. (1989). Phys. Rev. A 39, 1253 and 1264.
Sommers, Jr., H. S. (1982). J. Appl. Phys. 53, 156.
Streifer, W., Scrifes, D., and Burnham, R. (1982). Appl. Phys. Lett. 40, 305.
Tzeng, H.-M., Wall, K. F., Long, M. B., and Chang, R. K. (1984). Opt. Lett. 9, 499.
van der Lee, A. M., van Druten, N. J., Mieremet, A. L., van Eijkelenborg, M. A., Lindberg, Å. M., van
Exter, M. P., and Woerdman, J. P. (1997). Phys. Rev. Lett. 79, 4357.
van der Lee, A. M., van Exter, M. P., Mieremet, A. L., van Druten, N. J., and Woerdman, J. P. (1998).
van der Lee, A. M., Mieremet, A. L., van Exter, M. P., van Druten, N. J., and Woerdman, J. P. (2000a).
Phys. Rev. A 61, 033812.
van der Lee, A. M., van Druten, N. J., van Exter, M. P., Woerdman, J. P., Poizat, J.-Ph., and Grangier,
Ph. (2000b). Phys. Rev. Lett. 85, 4711.
van Druten, N. J., Lien, Y., Serrat, C., Oemrawsingh, S. S. R., van Exter, M. P., and Woerdman, J. P.
(2000). Phys. Rev. A 62, 053808.
van Eijkelenborg, M. A., Lindberg, Å. M., Thijssen, M. S., and Woerdman, J. P. (1996). Phys. Rev. Lett.
77, 4314.
van Eijkelenborg, M. A., van Exter, M. P., and Woerdman, J. P. (1998). Phys. Rev. A 57, 571.
van Exter, M. P., Hamel, W. A., Woerdman, J. P., and Zeijlmans, B. R. P. (1992). IEEE J. Quant.
Electron. 28, 1470.
van Exter, M. P., Kuppens, S. J. M., and Woerdman, J. P. (1995). Phys. Rev. A 51, 809.
van Exter, M. P., Nienhuis, G., and Woerdman, J. P. (1996). Phys. Rev. A 54, 3553.
van Exter, M. P., van Druten, N. J., van der Lee, A. M., Dutra, S. M., Nienhuis, G., and Woerdman, J. P.
(2001). Phys. Rev. A 63, 043801.
Verdeyen, J. T. (1989). “Laser Electronics.” Prentice-Hall, Englewood Cliffs, NJ.
Wiseman, H. M. (1997). Phys. Rev. A 56, 2068.
Wiseman, H. M. (1999). Phys. Rev. A 60, 4083.
Yamamoto, Y., Machida, S., and Björk, G. (1991). Phys. Rev. A 44, 657.
Yamamoto, Y., Machida, S., and Björk, G. (1992). Opt. Quantum Electron. 24, S215.
Yamamoto, Y., and Imamoglu, A. (1999). “Mesoscopic Quantum Optics.” Wiley, New York.
Yokoyama, H., and Brorson, S. D. (1989). J. Appl. Phys. 66, 4801.
Zhang, J. P., Chu, D. Y., Wu, S. L., Bi, W. G., Tiberio, R. C., Tu, C. W., and Ho, S. T. (1996). IEEE
Photonics Technology Lett. 8, 968.
Zhu, Y., Gauthier, D. J., Morin, S. E., Wu, Q., Carmichael, H. J., and Mosberg, T. W. (1990). Phys. Rev.
Lett. 64, 2499.
Index
A forward, backward, and general

electron-scattering conditions in
Absorption rate coefficient, definition, 71 planar geometry, 170
Air Force Geophysical Laboratory general coordinate system in laboratory
(AFGL), spectral database, 156 frame of reference, 168
AlGaAs semiconductor laser magnetic angle changing technique, 173
inverse linewidth as function of total principles, 167–173
output power, 212 pulsed supersonically cooled target
spectrum, 226 beam, 176
Alkali dimers, long-range wells, 59 relationship between electron-scattering
Amplification, parametric, of atomic and angle and target recoil angle, 170–172
optical fields, 38–43 schematic of experimental apparatus at
Angular momentum operators, four-wave Manchester, 174
mixing, 31–32 supersonic expansion, 175
Anomalous density, finite temperatures, time-of-flight measurement, 176–177
21–22 time-of-flight spectrum from
Argon, high Rydberg spectrum, helium, 178
194, 195 velocity distribution of supersonically
Atomic deflection expanding gas beam, 175
advantage of measuring deflected target Atom optics
yield, 172–173 analogy with nonlinear optics, 2
comparing recoil cross-section and developments, 1–2
magnetic angle changing technique early experiments, 2
results, 181–182 Atoms, finite number, small laser, 242–244
conservation of energy and momentum, Aurorae
168–169 excited atmospheric species losing
DCS measurements exciting states of energy, 142–143
helium, 180 Galileo orbiter revolving around Jupiter
deflection angle as function of incident and satellites, 141
electron energy for helium, 172 giant planets, 139–143
deflection of helium as function of H+3 and H2 emission lines on Jupiter, 140
electron scattering angle, 171 pure rotational excitations of molecules
detector for exciting and deflecting by electrons, 142
targets, 176
differential cross section (DSC), 177
differential cross-section measurements, B
177–182
dual nature of reaction, 170, 171 Bad-cavity regime
experimental recoil techniques, laser, 211–214
173–177 spontaneous emission factor, 216–217
fitting Gaussian peaks to time-of-flight Ballistic expansion, temperature
spectra, 179 measurement, 95
249
250 INDEX
Bogoliubov quasi-particle approach, vibrational wavefunctions in 1g

perturbations from static, 19–20 potential, 65
Bohr–Sommerfeld, quantization Chebyshev expansions, vibrational
condition, 66 wavefunctions, 107
Bose–Einstein condensates Cold atomic source, Orsay experiment,
contour and three-dimensional 80–83
gray-scale renditions, 18 Cold molecules
four-wave mixing, 28–29 applications, 54
See also Mean-field theory detection, 84, 85
efficient ways, 56
formation rate, 111–114
C helium buffer gas cooling technique, 56
laser cooling techniques, 55–56
Canada–France–Hawaii Telescope rapid evolution of field, 56
(CFHT), auroral images of Jupiter, 133 Stark decelerator, 36, 56
Cavity QED, spontaneous emission factor, See also Photoassociation; Ultracold
219–221 molecules
Cavity resonance, spectra, 216 Cold target recoil ion momentum
Centre National de la Recherche spectroscopy (COLTRIMS)
Scientifique (CNRS), spectral ionization experiment, 165
database, 156 See also Recoil techniques
Cesium Collective atomic recoil laser (CARL),
Cs+2 ion signal, 86 ultracold regime, 38–39
Cs+2 ion signal vs. detuning of Collisional processes, cometary spectra,
photoassociation laser, 90 154–155
demonstrating use of photoassociation, Comets
99–100 collisional processes for cometary
detection, 84, 85 spectra, 154–155
energy-normalized wavefunction for comparing best-fit model and tail
scattering of two ground-state atoms, spectrum, 148
75, 76 CSHELL/IRTF spectra of comet
experimental and theoretical Hale–Bopp, 150
photoassociation rates, 110 infrared spectra, 149–151
ion spectrum, 88–91 isotopic abundances, 153–154
level spacing, 68–69 producing atoms, molecules, radicals
optical pumping scheme, 55 and ions, 145
photoassociation spectra, 84–91 radio spectra, 151–155
potential curves for Hund’s case, 63 spectroscopy, 145–155
potential curves for interaction between ultraviolet and visible spectra, 146–149
two ground-state atoms, 59–60 visible range, 146, 147
short-range behavior of wavefunctions See also Giant planets and comets
for two levels, 65 Condensate density, finite temperatures, 21
time evolution of Cs+ 2 ion signal, 92 Condensate modes, quantifying quantum
translational temperature, 93–94 entanglement, 35–37
trap-loss spectrum, 85, 87–88
tunneling effect, 114–115, 116
ultracold molecules by photoassociation, D
79–80
vapor-loaded MOT, 81–82 Databases, spectral, and needed
vibrational wavefunction, 93 improvements, 155–156
INDEX 251
de Broglie waves metastable atom recoil angular

four-wave mixing, 23–37 measurements, 166
See also Nonlinear optics observations of momentum-analyzed
Differential cross section (DCS) doubly excited state fluorescence
electron excitation of target, 177 decay, 184–187
fitting Gaussian peaks to time-of-flight principles of deflection technique,
spectra from helium, 179 167–173
measurements for helium, 180 principles of doubly excited targets,
time-of-flight spectra for helium, 178 182–184
Doubly excited states principles of stepwise laser probing,
experimental techniques, 184 187–190
function extracted from time-of-flight recoil in crossed-beam interaction, 164
spectrum, 186 recoiling excited atoms, 182–187
observations of momentum-analyzed, results from laser probing of metastable
fluorescence decay products, 184–187 targets, 192–199
principles, 182–184 stepwise laser probing of deflected
time-of-flight of metastable atoms as metastable targets, 187–199
function of incident electron techniques producing well-defined target
energy, 186 momentum, 166–167
time-of-flight of metastable helium as threshold metastable excitation of
function of atom recoil angle, 185 helium, 166
time-of-flight of metastable atoms as
E function of incident electron
energy, 186
Electric dipole interaction, matter-light time of flight of metastable helium as
coupling, 13 function of atom recoil angle, 185
Electric dipole transition moments, need two-step laser process, 200–201
for accurate, 108–109 Electron impact. See Atomic deflection
Electron-excited targets Equation of motion, finite temperatures,
atomic deflection using electron impact, 21–22
167–182
cold target recoil ion momentum F
spectroscopy (COLTRIMS), 165
differential cross-section measurements, Finite-atom-number, small laser, 242–244
177–182 Finite temperatures, mean-field theory,
doubly excited state excitation function 20–23
from time-of-flight spectrum, 186 Fluctuation threshold
doubly excited states via fluorescence diagnostics of Nd3+ microchip
decay, 182–187 laser, 233
experimental recoil techniques, 173–177 disappearance for small lasers, 227–236
experimental techniques for doubly Fano factor versus pump parameter for
excited targets, 184 class-B laser, 230
experimental techniques of stepwise frequency-integrated noise, 230
laser probing, 190–192 nature depending on class-A vs. class-B
experiments exploiting internal energy nature and β value, 230–232
of metastable targets, 165–166 Nd3+ microlaser, 232
experiments studying, 163–164 photon-number distribution of Nd3+
future experiments, 199–221 microlaser, 235, 236
inelastic electron-scattering schematic of various regimes of,
process, 165 of laser, 231
252 INDEX
Fluctuation threshold (contd.) reflection principle, 73–75

second-order moment of intensity stationary-phase method, 74
fluctuations of Nd3+ microlaser, 235
second-order reduced factorial moment G
and Fano factor of photon
distribution, 228 Galileo
steady-state photon number, carbon abundances in Jupiter, 136
second-order moment, and Fano isotope ratios by mass spectrometer
factor of conventional laser, 229 probe, 137
typical experimental time traces and RF revolving around Jupiter and its
spectra of Nd3+ microlaser, 234 satellites, 141
Fluorescence decay. See Doubly excited studying giant planets, 132
states Giant planets
Formation of molecules. See aurorae, 139–143
Photoassociation infrared spectra, 132–135
Four-wave mixing Jupiter, Saturn, Uranus, and
angular momentum operators, 31–32 Neptune, 131
Casimir operators, 32 molecular abundances from
Central-mode-side-mode correlation, 36 spectroscopic observations, 135–138
condensate wave function, 25 Titan, 143, 144
conservation of particle number, 30–31 ultraviolet and visible spectra, 132
dynamics of population exchange Giant planets and comets
between modes, 33–34 atmospheres, 130
evolution of population, 34 laboratory spectroscopic studies of
generation of fourth matter wave from visible region, 130–131
three initial waves, 28, 29 molecular emissions, 131
geometry, 30 spectral databases and improvements
Gross–Pitaevskii equation, 25 needed, 155–156
Hamiltonian, 30 spectroscopic observations, 130
illustration, 32–33 Goddard Space Flight Center (GSFC),
mean field analysis, 24–27 spectral database, 156
momentum and energy conservation Good-cavity regime
conditions, 26–27 laser, 211–214
original theory, 23–24 spontaneous emission factor, 216
population differences, 31 Gordon method, vibrational
possibility of obtaining quantum wavefunctions, 107
correlations between side modes, 35 Gross–Pitaevskii
quantifying quantum entanglement derivation, 15–19
between condensate modes, 35–36 energy functional, 17
quantum theory of atomic, 27–37 four-wave mixing, 25–26
scattering process, 27 nonlinear Schrödinger equation, 16
second-quantized Hamiltonian, 29 Ground state
side-mode-side-mode correlation, 37 formation of, molecules as short-range
spinor operators, 31 process, 69
time evolution of central mode-side making molecule, 78–80
mode correlation function, 36
two-mode correlation functions, 37 H
Franck–Condon factor
estimation of photoassociation Hale–Bopp
probability, 73 CSHELL/IRTF spectra, 150
INDEX 253
millimeter spectra containing emission Petermann excess noise factor in

lines, 152 low-frequency, 241
radio spectra, 151–153 resonance behavior of Petermann excess
Hamiltonian noise factor, 239
four-wave mixing, 30 Hubble Space Telescope (HST), studying
Hartree–Fock–Bogoliubov, 22–23 giant planets, 132
manybody theory, 8–9 Hyakutake
manybody to N-particle state, 11 measured and modeled A–X (0–0)
parametric amplification of optical and bands of CH and CD, 153
matter waves, 39 radio spectra, 151–153
second-quantized for four-wave Hydrogen (H2), Rydberg state spectrum,
mixing, 29 196, 197
second-quantized with bosonic
commutation relations, 13 I
two-body collisions, 14, 15
Hartree–Fock–Bogoliubov Hamiltonian, Inelastic electron-scattering process,
finite temperatures, 22–23 exploiting initial target momentum, 165
Hartree wave functions, mean-field Infrared Interferometer Spectrometer and
theory, 15 Radiometer (IRIS)
Heisenberg equation of motion, manybody far-infrared spectra of Jupiter and
theory, 9–10 Saturn, 137–138
Helium Jupiter in 1979, 132
atomic deflection angle as function of molecular abundances from
incident electron energy, 172 spectroscopic observations,
deflection of metastable state, 171 135–138
differential cross-section Voyager IRIS spectrum from equatorial
measurements, 180 region of Jupiter, 135
doubly excited state function from Infrared Space Observatory (ISO)
time-of-flight spectrum, 186 Saturn and Neptune spectra, 132–133
field ionization and metastable signals in short-wavelength spectrometer, 133–134
upper detector, 198, 199 Infrared spectra
fitting Gaussian peaks to time-of-flight Canada–France–Hawaii Telescope
spectra, 179 (CFHT), 133
laser probing results, 192–193 comets, 149–151
momentum of metastable state, 171 far-infrared spectra of Jupiter and
threshold metastable excitation, 166 Saturn, 137–138
time-of-flight of metastable, as function giant planets, 132–135
of atom recoil angle, 185 Infrared Interferometer Spectrometer
time-of-flight of metastable atoms as and Radiometer (IRIS), 132–133
function of incident electron Infrared Space Observatory (ISO),
energy, 186 132–133
time-of-flight spectra, 178 Infrared Telescope Facility (IRTF), 133
Helium buffer gas, cooling local thermodynamic equilibrium (LTE)
technique, 56 approximation, 134–135
HeNe laser, linewidth-power product as near-infrared spectra of Neptune and
function of cavity decay rate, 213 Saturn, 139
HeXe laser strong emission of V3P branch of CH4
coloring of Petermann excess quantum on Jupiter, 134
noise, 241 Infrared Telescope Facility (IRTF), auroral
Petermann excess noise factor, 240 images of Jupiter, 133
254 INDEX
International Ultraviolet Explorer (IUE), ultracold molecules via

studying giant planets, 132 photoassociation, 80
Ion spectrum, photoassociation, 88–91 vapor-loaded, 81–82
Isotopic abundances, comets, 153–154 Magnitude, spontaneous emission factor,
221–222
Manybody theory
J applying manybody Hamiltonian to
Jet Propulsion Laboratory (JPL), spectral N-particle state, 11
database, 156 continuity equation for quantum field
Jupiter operators, 11–12
carbon abundances, 136 Hamiltonian, 8–9
far-infrared spectra, 137–138 Heisenberg equation of motion, 9–10
H+ matter-light coupling via electric dipole
3 and H2 emission lines over polar
haze of auroral region, 140 interaction, 13
Infrared Interferometer Spectrometer mode expansion of Schrödinger field
and Radiometer (IRIS), 132 operator, 12–13
strong emission of V3 P branch of multiparticle evolution, 11
CH4, 134 N-particle wave function in terms of
Voyager IRIS spectrum from equatorial Schrödinger field creation operator, 10
region, 135 nonlinear optics, 8–14
See also Giant planets Schrödinger wave function, 9
simple harmonic oscillator, 13–14
two-body collisions, 14
L Mapped Fourier method
calculation of kinetic energy
Laser cooling operator, 104
applications, 54 Fourier expansion, 105
photoassociation, 55–56 interpolated wavefunction in x
Laser probing. See Stepwise laser probing variable, 106
Lasers. See Quantum noise of small lasers maximum momentum, 102
LeRoy–Bernstein law radial Schrödinger equation,
bound levels, 64–70 103, 104
fit for trap-loss spectrum, 87 vibrational wavefunction for Cs2, 106
interpretation and fitting vibrational wavefunctions, 101–107
experiments, 70 Matter-light coupling, electric dipole
Local thermodynamic equilibrium (LTE), interaction, 13
radiative transfer calculations, 134–135 Matter-wave amplifier
Long-range external wells characteristic input-output, 50
alkali dimers, 59 phase-coherent, 49–50
dipole-dipole interaction, 63–64 Matter-wave superradiance
explanation for existence, 61 antinormally ordered characteristic
Luminorefrigeration, laser cooling, 54 function, 47
atomic field operator, 44
M closed atomic system, 46
coupling coefficients of Hamiltonian, 43
Magnetic angle changing technique deriving equation of motion, 45–46
atomic deflection, 173 generating families of higher-order side
comparison to recoil cross-section modes, 45
results, 181–182 geometric dependence of single-atom
Magnetooptical trap device (MOT) gain, 47–48
INDEX 255
geometry of condensate, 43–44 Molecular potential curves, need for

number distribution for chaotic field, 47 accurate, 108–109
quantization of q values, 44–45 Molecules, formation. See
quasi-mode populations, 45 Photoassociation
single-atom gain, 48
superradiant matter-wave scattering, 44
unfair competition, 48–49 N
Mean-field theory 3+
Nd laser
anomalous Bose correlations, 20
diagnostics, 233
anomalous density, 21
photon-number distribution, 235, 236
Bose commutation relation, 16
schematic, 232
condensate density, 21
second-order moment of intensity
condensate wave function, 16
fluctuations, 235
deriving Gross–Pitaevskii equation,
typical experimental time traces and RF
15–19
spectra, 234
effective single-particle states, 15
See also Fluctuation threshold; Quantum
equation of motion for condensate
noise of small lasers
excitations, 21–22
Neptune
equation of motion for condensate wave
H2–H2 dimer features in near-infrared
function, 21
spectra, 139
finite temperatures, 20–23
Infrared Space Observatory (ISO),
four-wave mixing, 24–27
132–133
Gross–Pitaevskii energy functional, 17
See also Giant planets
Gross–Pitaevskii nonlinear Schrödinger
Nodal structure, zero-energy scattering
equation, 16
wavefunction, 75–77
Hartree–Fock–Bogoliubov Hamiltonian,
Noncondensate density, finite
22–23
temperatures, 21
Hartree wave functions, 15
Nonlinear optics
matter-wave four-wave mixing in
atom optics, 1–2
sodium Bose–Einstein condensate, 18
four-wave mixing of de Broglie waves,
noncondensate density, 21
23–37
order parameter, 16
manybody theory, 8–14
perturbations from static using
mean-field theory of Bose–Einstein
Bogoliubov quasi-particle approach,
condensates, 15–23
19–20
mixing of optical and matter waves,
self-consistent mean-field
37–50
approximation, 21
s-wave scattering, 3–8
Thomas–Fermi approximation, 19
Nonorthogonality theory, Petermann
total anomalous density, 22
excess noise, 242
total density, 22
Numerov approach, vibrational
zero temperature, 15–20
wavefunctions, 107
Mechanism, cold molecule formation,
91–92
Metastable atom recoil angular O
measurements, noble gases, 166
Metastable targets, experiments exploiting Optical and matter waves
internal energy, 165–166 atomic field initially consisting of
Molecular abundances, spectroscopic condensate below critical
observations, 135–138 temperature, 40–41
Molecular laser, proposal, 54 cross-correlation functions, 42–43
256 INDEX
Optical and matter waves (contd.) absorption rate coefficient, 71

effective atom-probe coupling analogy with Rydberg law, scaling law,
constant, 40 64–70
effective Hamiltonian, 39 approach using atomic physics
input-output characteristic of viewpoint, 72–73
matter-wave amplifier, 50 approximating potential matrix, 62–63
matter-wave superradiance, 43–49 approximations to simplify calculations,
mixing, 37–50 71–72
parametric amplification of atomic and Bohr–Sommerfeld quantization
optical fields, 38–43 condition, 66
phase-coherent matter-wave bottleneck in probability, 77
amplification, 49–50 C3 coefficient, 61
phase shift due to cross-phase calculations of overlap integrals, 77
modulation, 40 case of, into excited curve with
physics underlying parametric long-range behavior, 68
amplification, 38–39 choosing scattering lengths, 76–77
second-order equal-time intensity comparing experiment and theory for
correlation function, 42 rates, 109–111
two-mode correlations, 42 conservation of energy and angular
Optical pumping cycles momentum, 57
atomic cesium sample, 55 determining number of molecules in
cooling atoms, 54–55 experiment, 58–59
Orsay experiment dipole-dipole interaction dominating
cold atomic source, 80–83 long-range wells, 63–64
detection scheme, 85 effective quantum number, 67
experimental setup, 80–84 efficiency of reaction, 74–75
photoassociation laser, 83–84 energy-normalized wavefunction for
scattering of two ground-state Cs
P atoms, 75, 76
experiments, 57
Petermann excess quantum noise fine-structure effective operator, 61–62
coloring of, 241 formation of ground-state molecules in
dependence on cold cavity loss rate, short-range process, 69
236–237 formation of ultracold molecules, 56–57
effect of K factor on laser intensity Franck–Condon factor, 72
noise, 238–241 highly excited vibrational levels, 67–68
frequency dependence of K factor, ion spectrum, 88–91
241–242 laser cooling, 55–56
K factor, 237–238 LeRoy–Bernstein formula, 70
laser threshold characteristics in LeRoy–Bernstein law for bound levels,
presence of, 240 64–70
noise factor in low-frequency intensity level spacing, 74
noise of HeXe laser, 241 level spacing of Cs2, 68–69
nonorthogonality theory, 242 linking normalization factor to classical
Petermann excess noise factor in optical period of motion, 66–67
linewidth of HeXe laser, 240 local wavenumber and local de Broglie
resonance behavior of, of HeXe unstable wavelength, 66
cavity laser, 239 long range, 76–77
small lasers, 236–242 long-range alkali dimer molecules: pair
Photoassociation of atoms, 59–64
INDEX 257
long-range wells, existence of, 61 disappearance of fluctuation threshold,

magnetooptical trap device (MOT), 80 227–236
making molecule in ground state or finite number of atoms, 242–244
lower triplet state, 78–80 good-cavity vs. bad-cavity regime,
motion in potential with power-law 211–214
asymptotic behavior, 66 inverse linewidth of AlGaAs
new schemes for ultracold molecules, semiconductor laser as function of
116–117 total output power, 212
nodal structure of zero-energy scattering level scheme of laser, 207
wavefunction, 75–77 linewidth-power product of HeNe laser
orientation of dipoles, 61 as function of cavity decay rate, 213
overlap integral in semiclassical magnitude of β, 221–222
approach, 73–74 overview of threshold behavior, 207–210
potential curves for Hund’s case, 63 Petermann excess quantum noise,
potential curves for interaction between 236–242
two ground-state Cs atoms, 59–60 relaxation oscillation, 222–227
quantum formulation of transition spontaneous emission factor β, 214–219
between bound level and continuum, steady-state inversion number, 209
70–71 steady-state photon number as function
rate, 94–96 of pumping parameter for laser, 208
reflection principle for calculating steady-state photon number vs. pumping
Franck–Condon factors, 73–75 rate and cavity intensity decay
resonance condition, 58 rate, 209
short range, 77 Quantum theory, atomic four-wave mixing,
simple theory, 77–78 27–37
stationary-phase method for
Franck–Condon factor, 74 R
theoretical predictions for
photoassociation rates, 70–73 Radio spectra, comets, 151–153
trap-loss spectrum, 85, 87–88 Recoil techniques
two states by diagonalization, 62 atomic deflection, 173–177
ultracold Cs experiment, 79–80 cold target recoil ion momentum
uniform semiclassical (USC) spectroscopy (COLTRIMS), 165
wavefunction, 69–70 comparison to magnetic angle changing
vibrational motion in long- and results, 181–182
short-range regions, 64–66 limitations, 200
vibrational spacing as function of providing new information about
detuning, 68 electron-impact excitation, 199–200
Photoassociation laser, Orsay experiment, schematic of apparatus at
83–84 Manchester, 174
Planets. See Giant planets and comets two-step laser process eliminating
Population differences, four-wave difficulties, 200–201
mixing, 31 Recycling active atoms, level scheme of
Population exchange, studying dynamics, laser, 244
33–34 Reflection principle, calculating
Franck–Condon factors, 73–75
Q Relaxation oscillation
amplitude spectrum of intensity
Quantum noise of small lasers noise, 224
cavity QED and β, 219–221 damped harmonic oscillator, 223
258 INDEX
Relaxation oscillation (cont’d.) giant planets, 131–143

intensity-noise strength at low frequency heavenly bodies, 130
versus number of photons in laser laboratory studies in visible region,
cavity, 225 130–131
laser, 222–227 molecular abundances from
neglecting inversion noise for low observations, 135–138
frequencies, 225 Spinor operators, four-wave mixing, 31
photonic damping and atomic Spontaneous emission factor β
damping, 223–224 assumptions of general form, 217–219
spectra of spontaneous emission and calculation, 214–216
cavity resonance, 216 cavity mode in bad-cavity regime,
spectrum of AlGaAs semiconductor 216–217
laser, 226 cavity mode in good-cavity regime, 216
Resonance condition, photoassociation, 58 cavity QED and β, 219–221
R-matrix method, vibrational magnitude, 221–222
wavefunctions, 107 nature of fluctuation threshold
Rotational temperatures, cold molecules, depending on, 230–232
96–99 schematic of couplings between atomic
Rydberg atoms excitation, discrete cavity mode, and
alternative detection method, 189–190 optical continuum, 215
stepwise techniques to study highly spectra, 216
excited, 189 typical geometry for calculating, 215
See also Stepwise laser probing Stark decelerator, cooling technique, 56
Stepwise laser probing
absence of ions when laser is
S off-resonance, 193–194
advantage of UV laser radiation, 191
Saturn allowing molecular rotational levels, 200
far-infrared spectra, 137–138 alternative detection method, 189–190
H2–H2 dimer features in near-infrared deflected metastable targets, 187–199
spectra, 139 efficiency of laser excitation of Rydberg
Infrared Space Observatory (ISO), state, 193
132–133 electronic hardware to control
See also Giant planets interactions, 191
Scaling laws, interpretation and fitting experimental techniques, 190–192
experiments, 70 experiments using coincidence
Scattering length, concept, 3 technique, 198
Schrödinger field operator field ionization and metastable signals in
mode expansion, 12 upper detector, 199
N-particle wave function, 10 field ionization spectrum from H2,
Second-order equal-time intensity 196, 197
correlation function, definition, 42 high Rydberg spectrum, typical using
Self-consistent mean-field approximation, lower detector, 192–193
finite temperatures, 20–23 laser system producing tunable pulsed
Small lasers. See Quantum noise radiation, 190
of small lasers principles, 187–190
Spectroscopy results, 192–199
comets, 145–155 results from upper detector, 197–199
databases and improvements needed, Rydberg series in argon using lower
155–156 detector, 194, 195
INDEX 259
scheme following electron-impact U

excitation, 188 Ultracold molecules
studying highly excited Rydberg accurate electronic dipole transition
atoms, 189 moments, 108–109
test of Percival–Seaton hypothesis, accurate molecular potential curves,
188–189 108–109
Stimulated Raman photoassociation analytical treatment, 108
relevant energy levels, 97 applications, 54
resonance, 98 atomic lifetime, 87–88
ultracold molecules, 97–98 cesium photoassociation experiment,
Supersonic expansion 79–80
gas, 175 cold atomic source, 80–83
velocity distribution of gas, 175 comparing experiment and theory,
s-wave scattering 109–115
approximation, 7 Cs+2 ion signal, 86
elastic, 3–5 Cs+2 ion signal vs. detuning
expanding wave function in terms of photoassociation laser, 90
partial waves, 5–6 Cs2 1u pure long-range state, 89, 91
expansion in terms of spherical Cs2 0−g pure long-range state, 88–89
harmonics, 6–7 demonstrating use of photoassociation,
ultracold particles, 6 99–100
detection, 84, 85
detection scheme in Orsay
T experiment, 85
determining vibrational wavefunctions,
Temperatures 101–108
rotational and vibrational, 96–99 experimental setup in Orsay experiment,
translational, 93–94 80–84
Thomas–Fermi approximation, formation by photoassociation,
large-condensates, 19 56–57, 100
Time-of-flight formation rate, 111–114
measurements of atomic deflection, Gordon method, 107
176–177 ion spectrum, 88–91
spectra from helium, 178 LeRoy–Bernstein law, 87
temperature measurement, 94 magnetooptical trap device (MOT), 80
Titan, spectroscopy, 143, 144 mapped Fourier method, 101–107
Translational temperature maximum of detected ions, 98
cesium, 93–94 mechanism for formation, 93
temperature measurement through mechanism of formation, 91–92
ballistic expansion, 95 new schemes for making, 116–117
temperature measurement through time Numerov approach, 107
of flight, 94 photoassociation laser, 83–84
Trap-loss spectrum, photoassociation of photoassociation rate, 94–96
cesium, 85, 87–88 photoassociation rates, 109–111
Trapping atoms, applications, 54 photoassociation spectra, 84–91
Triplet state, making molecule, relevant energy levels of stimulated
78–80 Raman photoassociation transition, 97
Tunneling effect, ultracold molecules, R-matrix method, 107
114–115, 116 rotational and vibrational temperatures,
Two-body collisions, Hamiltonian, 14 96–99
260 INDEX
Ultracold molecules (contd.) V

scheme of experimental setup, 81
stimulated Raman photoassociation, Vibrational temperatures, cold molecules,
97, 98 96–99
temperature measurement through Vibrational wavefunctions
ballistic expansion, 95 analytical treatment, 108
temperature measurement through time determination for photoassociated
of flight, 94 molecules, 101–108
theoretical methods, 100–109 Gordon method, 107
time evolution of Cs+ mapped Fourier method, 101–107
2 ion signal, 92
total number of cold atoms in trap, 82 Numerov approach, 107
translational temperature, 93–94 R-matrix method, 107
trap-loss spectrum, 85, 87–88 Visible spectra
tunneling effect, 114–115, 116 comets, 146–149
vapor-loaded MOT, 81–82 giant planets, 132
vibrational wavefunction in outer well Voyager. See Infrared Interferometer
of 0− Spectrometer and Radiometer (IRIS)
g potential for Cs2, 93
zoom of Cs+ 2 ion signal vs. detuning
photoassociation laser, 90 Z
See also Cold molecules;
Photoassociation Zero-energy scattering wavefunction,
Ultraviolet spectra nodal structure, 75–77
comets, 146–149 Zero temperature
giant planets, 132 reflection principle, 73–75
Uranus. See Giant planets mean-field theory, 15–20
Contents of Volumes in This Serial
Volume 1 Volume 3
Molecular Orbital Theory of the Spin The Quantal Calculation of Photoionization

Properties of Conjugated Molecules, Cross Sections, A. L. Stewart
G. G. Hall and A. T. Amos Radiofrequency Spectroscopy of Stored Ions
Electron Affinities of Atoms and Molecules, I: Storage, H. G. Dehmelt
B. L. Moiseiwitsch Optical Pumping Methods in Atomic
Atomic Rearrangement Collisions, Spectroscopy, B. Budick
B. H. Bransden Energy Transfer in Organic Molecular
The Production of Rotational and Vibrational Crystals: A Survey of Experiments,
Transitions in Encounters between H. C. Wolf
Molecules, K. Takayanagi Atomic and Molecular Scattering from Solid
The Study of Intermolecular Potentials with Surfaces, Robert E. Stickney
Molecular Beams at Thermal Energies, Quantum, Mechanics in Gas Crystal-Surface
H. Pauly and J. P. Toennies van der Waals Scattering, E. Chanoch
Beder
High-Intensity and High-Energy Molecular
Beams, J. B. Anderson, R. P. Andres, and Reactive Collisions between Gas and Surface
J. B. Fen Atoms, Henry Wise and Bernard J. Wood
Volume 4
Volume 2
H. S. W. Massey—A Sixtieth Birthday
The Calculation of van der Waals Tribute, E. H. S. Burhop
Interactions, A. Dalgarno and
W. D. Davison Electronic Eigenenergies of the Hydrogen
Molecular Ion, D. R. Bates and
Thermal Diffusion in Gases, E. A. Mason, R. H. G. Reid
R. J. Munn, and Francis J. Smith
Applications of Quantum Theory to the
Spectroscopy in the Vacuum Ultraviolet, Viscosity of Dilute Gases,
W. R. S. Garton R. A. Buckingham and E. Gal
The Measurement of the Photoionization Positrons and Positronium in Gases,
Cross Sections of the Atomic Gases, P. A. Fraser
James A. R. Samson Classical Theory of Atomic Scattering,
The Theory of Electron–Atom Collisions, A. Burgess and I. C. Percival
R. Peterkop and V. Veldre Born Expansions, A. R. Holt and
Experimental Studies of Excitation in B. L. Moiselwitsch
Collisions between Atomic and Ionic Resonances in Electron Scattering by Atoms
Systems, F. J. de Heer and Molecules, P. G. Burke
Mass Spectrometry of Free Radicals, Relativistic Inner Shell Ionizations,
S. N. Foner C. B. O. Mohr
261
262 CONTENTS OF VOLUMES IN THIS SERIAL
Recent Measurements on Charge Transfer, The Diffusion of Atoms and Molecules,

J. B. Hasted E. A. Mason and T. R. Marrero
Measurements of Electron Excitation Theory and Application of Sturmian
Functions, D. W. O. Heddle and Functions, Manuel Rotenberg
R. G. W. Keesing Use of Classical mechanics in the Treatment
Some New Experimental Methods in of Collisions between Massive Systems,
Collision Physics, R. F. Stebbings D. R. Bates and A. E. Kingston
Atomic Collision Processes in Gaseous
Nebulae, M. J. Seaton
Volume 7
Collisions in the Ionosphere, A. Dalgarno
The Direct Study of Ionization in Space, Physics of the Hydrogen Master, C. Audoin,
R. L. F. Boyd J. P. Schermann, and P. Grivet
Molecular Wave Functions: Calculations and
Use in Atomic and Molecular Processes,
Volume 5 J. C. Browne
Localized Molecular Orbitals, Harel
Flowing Afterglow Measurements of
Weinstein, Ruben Pauncz, and
Ion-Neutral Reactions, E. E. Ferguson,
Maurice Cohen
F. C. Fehsenfeld, and A. L. Schmeltekopf
General Theory of Spin-Coupled Wave
Experiments with Merging Beams,
Functions for Atoms and Molecules,
Roy H. Neynaber
J. Gerratt
Radiofrequency Spectroscopy of Stored Ions
Diabatic States of Molecules—Quasi-
II: Spectroscopy, H. G. Dehmelt
Stationary Electronic States, Thomas
The Spectra of Molecular Solids, O. Schnepp F. O’Malley
The Meaning of Collision Broadening of Selection Rules within Atomic Shells,
Spectral Lines: The Classical Oscillator B. R. Judd
Analog, A. Ben-Reuven
Green’s Function Technique in Atomic and
The Calculation of Atomic Transition Molecular Physics, Gy. Csanak,
Probabilities, R. J. S. Crossley H. S. Taylor, and Robert Yaris
Tables of One- and Two-Particle Coefficients A Review of Pseudo-Potentials with
of Fractional Parentage for Configurations Emphasis on Their Application to Liquid
s λ s ′u pq , C. D. H. Chisholm, A. Dalgarno, Metals, Nathan Wiser and A. J. Greenfield
and F. R. Innes
Relativistic Z-Dependent Corrections to
Atomic Energy Levels, Holly Thomis Doyle Volume 8
Interstellar Molecules: Their Formation and

Volume 6 Destruction, D. McNally
Monte Carlo Trajectory Calculations of
Dissociative Recombination, J. N. Bardsley Atomic and Molecular Excitation in
and M. A. Biondi Thermal Systems, James C. Keck
Analysis of the Velocity Field in Plasmas Nonrelativistic Off-Shell Two-Body Coulomb
from the Doppler Broadening of Spectral Amplitudes, Joseph C. Y. Chen and
Emission Lines, A. S. Kaufman Augustine C. Chen
The Rotational Excitation of Molecules by Photoionization with Molecular Beams,
Slow Electrons, Kazuo Takayanagi and R. B. Cairns, Halstead Harrison, and
Yukikazu Itikawa R. I. Schoen
CONTENTS OF VOLUMES IN THIS SERIAL 263
The Auger Effect, E. H. S. Burhop and Role of Energy in Reactive Molecular

W. N. Asaad Scattering: An Information-Theoretic
Approach, R. B. Bernstein and R. D. Levine
Inner Shell Ionization by Incident Nuclei,
Volume 9 Johannes M. Hansteen
Stark Broadening, Hans R. Griem
Correlation in Excited States of Atoms,
A. W. Weiss Chemiluminescence in Gases, M. F. Golde
and B. A. Thrush
The Calculation of Electron–Atom Excitation
Cross Sections, M. R. H. Rudge
Collision-Induced Transitions between
Rotational Levels, Takeshi Oka Volume 12
The Differential Cross Section of
Low-Energy Electron–Atom Collisions, Nonadiabatic Transitions between Ionic and
D. Andrick Covalent States, R. K. Janev
Molecular Beam Electric Resonance Recent Progress in the Theory of Atomic
Spectroscopy, Jens C. Zorn and Thomas Isotope Shift, J. Bauche and
C. English R.-J. Champeau
Atomic and Molecular Processes in the Topics on Multiphoton Processes in Atoms,
Martian Atmosphere, Michael B. McElroy P. Lambropoulos
Optical Pumping of Molecules, M. Broyer,
G. Goudedard, J. C. Lehmann, and J. Vigué
Volume 10 Highly Ionized Ions, Ivan A. Sellin
Time-of-Flight Scattering Spectroscopy,
Relativistic Effects in the Many-Electron
Wilhelm Raith
Atom, Lloyd Armstrong, Jr. and Serge
Feneuille Ion Chemistry in the D Region, George
C. Reid
The First Born Approximation, K. L. Bell and
A. E. Kingston
Photoelectron Spectroscopy, W. C. Price
Dye Lasers in Atomic Spectroscopy, Volume 13
W. Lange, J. Luther, and A. Steudel
Atomic and Molecular Polarizabilities—A
Recent Progress in the Classification of the
Review of Recent Advances, Thomas
Spectra of Highly Ionized Atoms,
M. Miller and Benjamin Bederson
B. C. Fawcett
Study of Collisions by Laser Spectroscopy,
A Review of Jovian Ionospheric Chemistry,
Paul R. Berman
Wesley T. Huntress, Jr.
Collision Experiments with Laser-Excited
Atoms in Crossed Beams, I. V. Hertel and
Volume 11 W. Stoll
Scattering Studies of Rotational and
The Theory of Collisions between Charged Vibrational Excitation of Molecules,
Particles and Highly Excited Atoms, Manfred Faubel and J. Peter Toennies
I. C. Percival and D. Richards Low-Energy Electron Scattering by Complex
Electron Impact Excitation of Positive Ions, Atoms: Theory and Calculations,
M. J. Seaton R. K. Nesbet
The R-Matrix Theory of Atomic Process, Microwave Transitions of Interstellar Atoms
P. G. Burke and W. D. Robb and Molecules, W. B. Somerville
Volume 14 Inner-Shell Ionization, E. H. S. Burhop

Excitation of Atoms by Electron Impact,
Resonances in Electron Atom and Molecule D. W. O. Heddle
Scattering, D. E. Golden Coherence and Correlation in Atomic
The Accurate Calculation of Atomic Collisions, H. Kleinpoppen
Properties by Numerical Methods, Brian Theory of Low Energy Electron-Molecule
C. Webster, Michael J. Jamieson, and Collisions, P. G. Burke
Ronald F. Stewart
(e, 2e) Collisions, Erich Weigold and Ian
E. McCarthy Volume 16
Forbidden Transitions in One- and
Two-Electron Atoms, Richard Marrus and Atomic Hartree–Fock Theory, M. Cohen and
Peter J. Mohr R. P. McEachran
Semiclassical Effects in Heavy-Particle Experiments and Model Calculations to
Collisions, M. S. Child Determine Interatomic Potentials, R. Düren
Atomic Physics Tests of the Basic Concepts Sources of Polarized Electrons, R. J. Celotta
in Quantum Mechanics, Francis M. Pipkin and D. T. Pierce
Quasi-Molecular Interference Effects in Theory of Atomic Processes in Strong
Ion–Atom Collisions, S. V. Bobashev Resonant Electromagnetic Fields, S. Swain
Rydberg Atoms, S. A. Edelstein and Spectroscopy of Laser-Produced Plasmas,
T. F. Gallagher M. H. Key and R. J. Hutcheon
UV and X-Ray Spectroscopy in Astrophysics, Relativistic Effects in Atomic Collisions
A. K. Dupree Theory, B. L. Moiseiwitsch
Parity Nonconservation in Atoms: Status of
Theory and Experiment, E. N. Fortson and
Volume 15
L. Wilets
Negative Ions, H. S. W. Massey
Atomic Physics from Atmospheric and Volume 17
Astrophysical Studies, A. Dalgarno
Collisions of Highly Excited Atoms, Collective Effects in Photoionization of
R. F. Stebbings Atoms, M. Ya. Amusia
Theoretical Aspects of Positron Collisions in Nonadiabatic Charge Transfer,
Gases, J. W. Humberston D. S. F. Crothers
Experimental Aspects of Positron Collisions Atomic Rydberg States, Serge Feneuille and
in Gases, T. C. Griffith Pierre Jacquinot
Reactive Scattering: Recent Advances in Superfluorescence, M. F. H. Schuurmans,
Theory and Experiment, Richard Q. H. F. Vrehen, D. Polder, and
B. Bernstein H. M. Gibbs
Ion–Atom Charge Transfer Collisions at Low Applications of Resonance Ionization
Energies, J. B. Hasted Spectroscopy in Atomic and Molecular
Aspects of Recombination, D. R. Bates Physics, M. G. Payne, C. H. Chen,
The Theory of Fast Heavy Particle Collisions, G. S. Hurst, and G. W. Foltz
B. H. Bransden Inner-Shell Vacancy Production in Ion–Atom
Atomic Collision Processes in Controlled Collisions, C. D. Lin and Patrick Richard
Thermonuclear Fusion Research, Atomic Processes in the Sun, P. L. Dufton and
H. B. Gilbody A. E. Kingston
Volume 18 Spin Polarization of Atomic and Molecular

Photoelectrons, N. A. Cherepkov
Theory of Electron–Atom Scattering in a
Radiation Field, Leonard Rosenberg
Positron–Gas Scattering Experiments, Talbert Volume 20
S. Stein and Walter E. Kauppila
Ion–Ion Recombination in an Ambient Gas,
Nonresonant Multiphoton Ionization of D. R. Bates
Atoms, J. Morellec, D. Normand, and
G. Petite Atomic Charges within Molecules, G. G. Hall
Classical and Semiclassical Methods in Experimental Studies on Cluster Ions,
Inelastic Heavy-Particle Collisions, T. D. Mark and A. W. Castleman, Jr.
A. S. Dickinson and D. Richards Nuclear Reaction Effects on Atomic
Recent Computational Developments in the Inner-Shell Ionization, W. E. Meyerhof and
Use of Complex Scaling in Resonance J.-F. Chemin
Phenomena, B. R. Junker Numerical Calculations on Electron-Impact
Direct Excitation in Atomic Collisions: Ionization, Christopher Bottcher
Studies of Quasi-One-Electron Systems, Electron and Ion Mobilities, Gordon
N. Anderson and S. E. Nielsen R. Freeman and David A. Armstrong
Model Potentials in Atomic Structure, On the Problem of Extreme UV and X-Ray
A. Hibbert Lasers, I. I. Sobel’man and
Recent Developments in the Theory of A. V. Vinogradov
Electron Scattering by Highly Polar Radiative Properties of Rydberg States in
Molecules, D. W. Norcross and Resonant Cavities, S. Haroche and
L. A. Collins J. M. Ralmond
Quantum Electrodynamic Effects in Few- Rydberg Atoms: High-Resolution
Electron Atomic Systems, G. W. F. Drake Spectroscopy and Radiation Interaction—
Rydberg Molecules, J. A. C. Gallas,
G. Leuchs, H. Walther, and H. Figger
Volume 19
Electron Capture in Collisions of Hydrogen Volume 21

Atoms with Fully Stripped Ions,
B. H. Bransden and R. K. Janev Subnatural Linewidths in Atomic
Interactions of Simple Ion–Atom Systems, Spectroscopy, Dennis P. O’Brien, Pierre
J. T. Park Meystre, and Herbert Walther
High-Resolution Spectroscopy of Stored Molecular Applications of Quantum
Ions, D. J. Wineland, Wayne M. Itano, and Defect Theory, Chris H. Greene and
R. S. Van Dyck, Jr. Ch. Jungen
Spin-Dependent Phenomena in Inelastic Theory of Dielectronic Recombination,
Electron–Atom Collisions, K. Blum and Yukap Hahn
H. Kleinpoppen Recent Developments in Semiclassical
The Reduced Potential Curve Method for Floquet Theories for Intense-Field
Diatomic Molecules and Its Applications, Multiphoton Processes, Shih-I Chu
F. Jenč Scattering in Strong Magnetic Fields,
The Vibrational Excitation of Molecules by M. R. C. McDowell and M. Zarcone
Electron Impact, D. G. Thompson Pressure Ionization, Resonances, and the
Vibrational and Rotational Excitation in Continuity of Bound and Free States,
Molecular Collisions, Manfred Faubel R. M. More
Volume 22 Angular Correlation in Multiphoton

Ionization of Atoms, S. J. Smith and
Positronium—Its Formation and Interaction G. Leuchs
with Simple Systems, J. W. Humberston Optical Pumping and Spin Exchange in Gas
Experimental Aspects of Positron and Cells, R. J. Knize, Z. Wu, and W. Happer
Positronium Physics, T. C. Griffith Correlations in Electron–Atom Scattering,
Doubly Excited States, Including New A. Crowe
Classification Schemes, C. D. Lin
Measurements of Charge Transfer and Volume 25
Ionization in Collisions Involving
Hydrogen Atoms, H. B. Gilbody Alexander Dalgarno: Life and Personality,
Electron–Ion and Ion–Ion Collisions with David R. Bates and George A. Victor
Intersecting Beams, K. Dolder and B. Pearl Alexander Dalgarno: Contributions to
Electron Capture by Simple Ions, Edward Atomic and Molecular Physics, Neal Lane
Pollack and Yukap Hahn Alexander Dalgarno: Contributions to
Relativistic Heavy-Ion–Atom Collisions, Aeronomy, Michael B. McElroy
R. Anholt and Harvey Gould Alexander Dalgarno: Contributions to
Continued-Fraction Methods in Atomic Astrophysics, David A. Williams
Physics, S. Swain Dipole Polarizability Measurements, Thomas
M. Miller and Benjamin Bederson
Flow Tube Studies of Ion–Molecule
Volume 23 Reactions, Eldon Ferguson
Differential Scattering in He He and
Vacuum Ultraviolet Laser Spectroscopy of He+ He Collisions at KeV Energies,
Small Molecules, C. R. Vidal R. F. Stebbings
Foundations of the Relativistic Theory of Atomic Excitation in Dense Plasmas,
Atomic and Molecular Structure, Jon C. Weisheit
Ian P. Grant and Harry M. Quiney
Pressure Broadening and Laser-Induced
Point-Charge Models for Molecules Derived Spectral Line Shapes, Kenneth M. Sando
from Least-Squares Fitting of the Electric and Shih-I Chu
Potential, D. E. Williams and Ji-Min Yan
Model-Potential Methods, G. Laughlin and
Transition Arrays in the Spectra of Ionized G. A. Victor
Atoms, J. Bauche, C. Bauche-Arnoult, and
M. Klapisch Z-Expansion Methods, M. Cohen
Photoionization and Collisional Ionization of Schwinger Variational Methods, Deborah
Excited Atoms Using Synchroton and Kay Watson
Laser Radiation, F. J. Wuilleumier, Fine-Structure Transitions in Proton-Ion
D. L. Ederer, and J. L. Picqué Collisions, R. H. G. Reid
Electron Impact Excitation, R. J. W. Henry
and A. E. Kingston
Volume 24 Recent Advances in the Numerical
Calculation of Ionization Amplitudes,
The Selected Ion Flow Tube (SIDT): Studies Christopher Bottcher
of Ion–Neutral Reactions, D. Smith and The Numerical Solution of the Equations of
N. G. Adams Molecular Scattering, A. C. Allison
Near-Threshold Electron–Molecule High Energy Charge Transfer, B. H. Bransden
Scattering, Michael A. Morrison and D. P. Dewangan
Relativistic Random-Phase Approximation, Electron–Atom Ionization, I. E. McCarthy

W. R. Johnson and E. Weigold
Relativistic Sturmian and Finite Basis Set Role of Autoionizing States in Multiphoton
Methods in Atomic Physics, G. W. F. Drake Ionization of Complex Atoms, V. I. Lengyel
and S. P. Goldman and M. I. Haysak
Dissociation Dynamics of Polyatomic Multiphoton Ionization of Atomic Hydrogen
Molecules, T. Uzer Using Perturbation Theory, E. Karule
Photodissociation Processes in Diatomic
Molecules of Astrophysical Interest,
Kate P. Kirby and Ewine F. van Dishoeck Volume 28
The Abundances and Excitation of
Interstellar Molecules, John H. Black The Theory of Fast Ion–Atom Collisions,
J. S. Briggs and J. H. Macek
Some Recent Developments in the
Volume 26 Fundamental Theory of Light,
Peter W. Milonni and Surendra Singh
Comparisons of Positrons and Electron Squeezed States of the Radiation Field,
Scattering by Gases, Walter E. Kauppila Khalid Zaheer and M. Suhail Zubairy
and Talbert S. Stein Cavity Quantum Electrodynamics,
Electron Capture at Relativistic Energies, E. A. Hinds
B. L. Moiseiwitsch
The Low-Energy, Heavy Particle
Collisions—A Close-Coupling Treatment, Volume 29
Mineo Kimura and Neal F. Lane
Vibronic Phenomena in Collisions of Atomic Studies of Electron Excitation of Rare-Gas
and Molecular Species, V. Sidis Atoms into and out of Metastable Levels
Associative Ionization: Experiments, Using Optical and Laser Techniques,
Potentials, and Dynamics, John Weiner, Chun C. Lin and L. W. Anderson
Françoise Masnou-Sweeuws, and Cross Sections for Direct Multiphoton
Annick Giusti-Suzor Ionionization of Atoms, M. V. Ammosov,
On the β Decay of 187 Re: An Interface of N. B. Delone, M. Yu. Ivanov, I. I. Bondar,
Atomic and Nuclear Physics and and A. V. Masalov
Cosmochronology, Zonghau Chen, Collision-Induced Coherences in Optical
Leonard Rosenberg, and Larry Spruch Physics, G. S. Agarwal
Progress in Low Pressure Mercury-Rare Gas Muon-Catalyzed Fusion, Johann Rafelski and
Discharge Research, J. Maya and Helga E. Rafelski
R. Lagushenko Cooperative Effects in Atomic Physics,
J. P. Connerade
Multiple Electron Excitation, Ionization, and
Volume 27 Transfer in High-Velocity Atomic and
Molecular Collisions, J. H. McGuire
Negative Ions: Structure and Spectra,
David R. Bates
Electron Polarization Phenomena in Volume 30
Electron–Atom Collisions,
Joachim Kessler Differential Cross Sections for Excitation of
Electron–Atom Scattering, I. E. McCarthy Helium Atoms and Helium-Like Ions by
and E. Weigold Electron Impact, Shinobu Nakazaki
Cross-Section Measurements for Electron Electron–Ion and Ion–Ion Recombination

Impact on Excited Atomic Species, Processes, M. R. Flannery
S. Trajmar and J. C. Nickel Studies of State-Selective Electron Capture in
The Dissociative Ionization of Simple, Atomic Hydrogen by Translational Energy
Molecules by Fast Ions, Colin J. Latimer Spectroscopy, H. B. Gilbody
Theory of Collisions between Laser Cooled Relativistic Electronic Structure of Atoms
Atoms, P. S. Julienne, A. M. Smith, and and Molecules, I. P. Grant
K. Burnett The Chemistry of Stellar Environments,
Light-Induced Drift, E. R. Eliel D. A. Howe, J. M. C. Rawlings, and
Continuum Distorted Wave Methods in D. A. Williams
Ion-Atom Collisions, Derrick Positron and Positronium Scattering at Low
S. F. Crothers and Louis J. Dubé Energies, J. W. Humberston
How Perfect are Complete Atomic Collision
Experiments?, H. Kleinpoppen and
Volume 31 H. Handy
Adiabatic Expansions and Nonadiabatic
Energies and Asymptotic Analysis Effects, R. McCarroll and D. S. F. Crothers
for Helium Rydberg States, Electron Capture to the Continuum,
G. W. F. Drake B. L. Moiseiwitsch
Spectroscopy of Trapped Ions, How Opaque Is a Star? M. J. Seaton
R. C. Thompson
Studies of Electron Attachment at
Phase Transitions of Stored Laser-Cooled Thermal Energies Using the Flowing
Ions, H. Walther Afterglow–Langmuir Technique,
Selection of Electronic States in Atomic David Smith and Patrik Španěl
Beams with Lasers, Jacques Baudon, Exact and Approximate Rate Equations in
Rudolf Düren, and Jacques Robert Atom-Field Interactions, S. Swain
Atomic Physics and Non-Maxwellian Atoms in Cavities and Traps, H. Walther
Plasmas, Michèle Lamoureux
Some Recent Advances in Electron-Impact
Excitation of n = 3 States of Atomic
Hydrogen and Helium, J. F. Williams and
Volume 32 J. B. Wang
Photoionization of Atomic Oxygen and

Atomic Nitrogen, K. L. Bell and
A. E. Kingston Volume 33
Positronium Formation by Positron Impact on Principles and Methods for Measurement of
Atoms at Intermediate Energies, Electron Impact Excitation Cross Sections
B. H. Bransden and C. J. Noble for Atoms and Molecules by Optical
Electron–Atom Scattering Theory and Techniques, A. R. Filippelli, Chun C. Lin,
Calculations, P. G. Burke L. W. Andersen, and J. W. McConkey
Terrestrial and Extraterrestrial H+
3 , Alexander Benchmark Measurements of Cross Sections
Dalgarno for Electron Collisions: Analysis of
Indirect Ionization of Positive Atomic Ions, Scattered Electrons, S. Trajmar and
K. Dolder J. W. McConkey
Quantum Defect Theory and Analysis of Benchmark Measurements of Cross Sections
High-Precision Helium Term Energies, for Electron Collisions: Electron Swarm
G. W. F. Drake Methods, R. W. Crompton
Some Benchmark Measurements of Cross Indirect Processes in Electron Impact

Sections for Collisions of Simple Heavy Ionization of Positive Ions, D. L. Moores
Particles, H. B. Gilbody and K. J. Reed
The Role of Theory in the Evaluation and Dissociative Recombination: Crossing and
Interpretation of Cross-Section Data, Tunneling Modes, David R. Bates
Barry I. Schneider
Analytic Representation of Cross-Section
Data, Mitio Inokuti, Mineo Kimura, Volume 35
M. A. Dillon, Isao Shimamura
Electron Collisions with N2 , O2 and O: What Laser Manipulation of Atoms, K. Sengstock
We Do and Do Not Know, Yukikazu and W. Ertmer
Itikawa Advances in Ultracold Collisions:
Need for Cross Sections in Fusion Plasma Experiment and Theory, J. Weiner
Research, Hugh P. Summers Ionization Dynamics in Strong Laser Fields,
Need for Cross Sections in Plasma Chemistry, L. F. DiMauro and P. Agostini
M. Capitelli, R. Celiberto, and Infrared Spectroscopy of Size Selected
M. Cacciatore Molecular Clusters, U. Buck
Guide for Users of Data Resources, Jean Femtosecond Spectroscopy of Molecules and
W. Gallagher Clusters, T. Baumer and G. Gerber
Guide to Bibliographies, Books, Reviews, Calculation of Electron Scattering on
and Compendia of Data on Atomic Hydrogenic Targets, I. Bray and
Collisions, E. W. McDaniel and A. T. Stelbovics
E. J. Mansky Relativistic Calculations of Transition
Amplitudes in the Helium Isoelectronic
Sequence, W. R. Johnson, D. R. Plante, and
Volume 34 J. Sapirstein
Rotational Energy Transfer in Small
Atom Interferometry, C. S. Adams, O. Carnal, Polyatomic Molecules, H. O. Everitt and
and J. Mlynek F. C. De Lucia
Optical Tests of Quantum Mechanics, R. Y.
Chiao, P. G. Kwiat, and A. M. Steinberg
Classical and Quantum Chaos in Atomic Volume 36
Systems, Dominique Delande and
Andreas Buchleitner Complete Experiments in Electron–Atom
Measurements of Collisions between Collisions, Nils Overgaard Andersen, and
Laser-Cooled Atoms, Thad Walker and Klaus Bartschat
Paul Feng Stimulated Rayleigh Resonances and
The Measurement and Analysis of Electric Recoil-Induced Effects, J.-Y. Courtois and
Fields in Glow Discharge Plasmas, G. Grynberg
J. E. Lawler and D. A. Doughty Precision Laser Spectroscopy Using
Polarization and Orientation Phenomena in Acousto-Optic Modulators, W. A. van
Photoionization of Molecules, Wijngaarden
N. A. Cherepkov Highly Parallel Computational Techniques for
Role of Two-Center Electron–Electron Electron–Molecule Collisions,
Interaction in Projectile Electron Carl Winstead and Vincent McKoy
Excitation and Loss, E. C. Montenegro, Quantum Field Theory of Atoms and
W. E. Meyerhof, and J. H. McGuire Photons, Maciej Lewenstein and Li You
Volume 37 Volume 40
Evanescent Light-Wave Atom Mirrors, Electric Dipole Moments of Leptons, Eugene

Resonators, Waveguides, and Traps, D. Commins
Jonathan P. Dowling and Julio High-Precision Calculations for the Ground
Gea-Banacloche and Excited States of the Lithium Atom,
Optical Lattices, P. S. Jessen and Frederick W. King
I. H. Deutsch Storage Ring Laser Spectroscopy, Thomas
Channeling Heavy Ions through Crystalline U. Kühl
Lattices, Herbert F. Krause and Sheldon Laser Cooling of Solids, Carl E. Mungan and
Datz Timothy R. Gosnell
Evaporative Cooling of Trapped Atoms, Optical Pattern Formation, L. A. Lugiato,
Wolfgang Ketterle and N. J. van Druten M. Brambilla, and A. Gatti
Nonclassical States of Motion in Ion Traps,
J. I. Cirac, A. S. Parkins, R. Blatt, and
P. Zoller Volume 41
The Physics of Highly-Charged Heavy Ions
Revealed by Storage/Cooler Rings, Two-Photon Entanglement and Quantum
P. H. Mokler and Th. Stöhlker Reality, Yanhua Shih
Quantum Chaos with Cold Atoms,
Mark G. Raizen
Volume 38 Study of the Spatial and Temporal Coherence
of High-Order Harmonics, Pascal Salières,
Electronic Wavepackets, Robert R. Jones and Ann L’Huiller Philippe Antoine, and
L. D. Noordam Maciej Lewenstein
Chiral Effects in Electron Scattering by Atom Optics in Quantized Light Fields,
Molecules, K. Blum and D. G. Thompson Matthias Freyburger, Alois M. Herkommer,
Optical and Magneto-Optical Spectroscopy of Daniel S. Krähmer, Erwin Mayr, and
Point Defects in Condensed Helium, Wolfgang P. Schleich
Serguei I. Kanorsky and Antoine Weis Atom Waveguides, Victor I. Balykin
Rydberg Ionization: From Field to Photon, Atomic Matter Wave Amplification by
G. M. Lankhuijzen and L. D. Noordam Optical Pumping, Ulf Janicke and
Studies of Negative Ions in Storage Rings, Martin Wilkens
L. H. Andersen, T. Andersen, and
P. Hvelplund
Single-Molecule Spectroscopy and Quantum Volume 42
Optics in Solids, W. E. Moerner,
R. M. Dickson, and D. J. Norris Fundamental Tests of Quantum Mechanics,
Edward S. Fry and Thomas Walther
Wave-Particle Duality in an Atom
Volume 39 Interferometer, Stephan Dürr and
Gerhard Rempe
Author and Subject Cumulative Index Atom Holography, Fujio Shimizu
Volumes 1–38 Optical Dipole Traps for Neutral Atoms,
Author Index Rudolf Grimm, Matthias Weidemüller, and
Subject Index Yurii B. Ovchinnikov
Appendix: Tables of Contents of Volumes Formation of Cold (T 1K) Molecules,
1–38 and Supplements J. T. Bahns, P. L. Gould, and W. C. Stwalley
High-Intensity Laser-Atom Physics, Fundamental Processes of Plasma–Surface

C. J. Joachain, M. Dorr, and N. J. Kylstra Interactions, Rainer Hippler
Coherent Control of Atomic, Molecular, and Recent Applications of Gaseous Discharges:
Electronic Processes, Moshe Shapiro and Dusty Plasmas and Upward-Directed
Paul Brumer Lightning, Ara Chutjian
Resonant Nonlinear Optics in Phase Coherent Opportunities and Challenges for Atomic,
Media, M. D. Lukin, P. Hemmer, and Molecular, and Optical Physics in Plasma
M. O. Scully Chemistry, Kurl Becker, Hans Deutsch, and
The Characterization of Liquid and Solid Mitio Inokuti
Surfaces with Metastable Helium Atoms,
H. Morgner
Quantum Communication with Entangled Volume 44
Photons, Harald Weinfurter
Mechanisms of Electron Transport in
Electrical Discharges and Electron
Collision Cross Sections, Hiroshi Tanaka
Volume 43 and Osamu Sueoka
Theoretical Consideration of Plasma-
Plasma Processing of Materials and Atomic, Processing Processes, Mineo Kimura
Molecular, and Optical Physics: An Electron Collision Data for Plasma-
Introduction, Hiroshi Tanaka and Processing Gases, Loucas
Mitio Inokuti G. Christophorou and James K. Olthoff
The Boltzmann Equation and Transport Radical Measurements in Plasma Processing,
Coefficients of Electrons in Weakly Ionized Toshio Goto
Plasmas, R. Winkler Radio-Frequency Plasma Modeling for
Electron Collision Data for Plasma Chemistry Low-Temperature Processing, Toshiaki
Modeling, W. L. Morgan Makabe
Electron–Molecule Collisions in Electron Interactions with Excited Atoms and
Low-Temperature Plasmas: The role Molecules, Loucas G. Christophorou and
of Theory, Carl Winstead and James K. Olthoff
Vincent McKoy
Electron Impact Ionization of Organic
Silicon Compounds, Ralf Basner, Volume 45
Kurt Becker, Hans Deutsch, and
Martin Schmidt Comparing the Antiproton and Proton, and
Kinetic Energy Dependence of Ion–Molecule Opening the Way to Cold Antihydrogen,
Reactions Related to Plasma Chemistry, G. Gabrielse
P. B. Armentrout Medical Imaging with Laser-Polarized Noble
Physicochemical Aspects of Atomic and Gases, Timothy Chupp and Scott Swanson
Molecular Processes in Reactive Plasmas, Polarization and Coherence Analysis of the
Yoshihiko Hatano Optical Two-Photon Radiation from the
Ion–Molecule Reactions, Werner Lindinger, Metastable 22 S1/2 State of Atomic
Armin Hansel, and Zdenek Herman Hydrogen, Alan J. Duncan, Hans
Uses of High-Sensitivity White-Light Kleinpoppen, and Marlan O. Scully
Absorption Spectroscopy in Chemical Laser Spectroscopy of Small Molecules,
Vapor Deposition and Plasma Processing, W. Demtröder, M. Keil, and H. Wenz
L. W. Anderson, A. N. Goyette, and Coulomb Explosion Imaging of Molecules,
J. E. Lawler Z. Vager
Volume 46 Formation of Ultracold Molecules

(T ≤ 200 μK) via Photoassociation in a
Femtosecond Quantum Control, T. Brixner, Gas of Laser-Cooled Atoms, Francoise
N. H. Damrauer and G. Gerber Masnou-Seeuws and Pierre Pillet
Coherent Manipulation of Atoms and Molecular Emissions from the Atmospheres
Molecules by Sequential Laser Pulses, of Giant Planets and Comets: Needs for
N. V. Vitanov, M. Fleischhauer, B. W. Shore, Spectroscopic and Collision Data,
and K. Bergmann Yukikazu Itikawa, Sang Joon Kim, Yong Ha
Slow, Ultraslow, Stored, and Frozen Light, Kim, and Y. C. Minh
Andrey B. Matsko, Olga Kocharovskaya, Studies of Electron-Excited Targets Using
Yuri Rostovtsev, George R. Welch, Recoil Momentum Spectroscopy with
Alexander S. Zibrov, and Marlan Laser Probing of the Excited State, Andrew
O. Scully James Murray and Peter Hammond
Longitudinal Interferometry with Atomic Quantum Noise of Small Lasers,
Beams, S. Gupta, D. A. Kokorowski, J. P. Woerdman, N. J. van Druten,
R. A. Rubenstein, and W. W. Smith and M. P. van Exter
Volume 47
Nonlinear Optics of de Broglie Waves,

P. Meystre
90018
9 780120 038473
ISBN 0-12-003847-1

(Meng) Advances in Atomic, Molecular, and Optical Physics Volume 47

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

(Meng) Advances in Atomic, Molecular, and Optical Physics Volume 47

Caricato da

Copyright:

Formati disponibili

Advances in

ATOMIC, MOLECULAR, AND OPTICAL PHYSICS

San Diego San Francisco New York

All Rights Reserved.

Explicit permission from Academic Press is not required to reproduce a maximum of

International Standard Serial Number: 1049-250X

International Standard Book Number: 0-12-003847-1

PRINTED IN THE UNITED STATES OF AMERICA

Nonlinear Optics of de Broglie Waves

Formation of Ultracold Molecules (T ≤ 200 μK) via

Molecular Emissions from the Atmospheres of Giant Planets

Studies of Electron-Excited Targets Using Recoil Momentum

Quantum Noise of Small Lasers

PETER HAMMOND (163), Department of Physics, University of Western Australia,

ATOMIC, MOLECULAR, AND OPTICAL PHYSICS

1 Copyright C 2001 by Academic Press

II. s-Wave Scattering

Collisions play in atom optics a role similar to nonlinear susceptibilities in optics.

in a potential U(r), which we assume to be central in the following,

where f k (θ, φ) is the scattering amplitude. In addition to the momentum k of the

for the incident particles and is approximately given by

In the case of a central potential, it is convenient to expand the wave function

where u k,l (r ) are the solutions of the radial Schrödinger equation,

h2 d 2 l(l + 1)h 2 h2k2

The phase shift δl appearing in this expression is determined by the requirement

eikr e−i(lπ/2−δl ) − c.c. 0

result over the 4π solid angle. With the orthonormality relation

this readily gives the familiar result

practical purposes as an spherical wave, so that by conservation of angular momen-

δl ∝ (kr0 )2l+1 , (24)

and from the differential cross section (7), we find

where we have introduced the scattering length,

wave functions well-behaved at the origin, it takes the form

III. A Microreview of Manybody Theory

where Hi is the Hamiltonian of particle i. The N-particle wave function must be

and for fermionic particles

In this last expression, [. . .]+ is an anticommutator, [A, B]+ = AB + B A. As we

In terms of the Schrödinger field creation operator  ˆ † , the N-particle wave

where |0 is the vacuum state (absence of particles),

Introducing the number operator

and using the commutator identity

it is easily shown that

Similarly, one finds

The equivalence of the formalism of second quantization with the conven-

is equivalent to the evolution given by the Fock space Schrödinger equation

yields the continuity equation for the quantum field operators,

where the flux

compare with Eq. (4).

where we have made use of the orthonormality relation

which we recognize from the quantization of the electromagnetic field. For

Va− f = hg(a ĉn† ĉm + h.c.),

whereby the absorption of a photon is accompanied by a particle being “annihi-

finds readily that

N̂ n ĉn† |0 = ĉn† |0 , (57)

IV. Mean-Field Theory of Bose–Einstein Condensates

where the effective single-particle states ϕ N (r) are assumed to be normalized. In

For a manybody Hamiltonian including a two-body interaction of the form (58),

(r, t) = (r, t) . (64)

In this approximation, it is assumed that the condensate possesses a well-defined

We observe that this derivation implicitly makes the factorization ansatz,

compare with Eq. (61), where E is the Gross–Pitaevskii energy functional,

The condensate ground state may be obtained by expressing the condensate

(r, t) = (r)e−iμt/h , (73)

In terms of the Schrödinger field creation operator ˆ † , the N-particle wave

(r, t) = (r, t) . (64)

(r, t) = (r)e−iμt/h , (73)

μ(r) = [Vtrap (r) + g|(r, t)|2 ](r) (77)

where (r) is the condensate wave function, corresponding to a macroscopic pop-

(r, t) = e−iμt/h [(r) + u(r)e−iωt + v ∗ (r)eiωt ], (81)

hωu j (r) = [H0 − μ + 2g2 (r)]u j (r) + g2 (r)v j (r),

n c (r) ≡ |(r)|2 , (87)

p(t) = −i h(t)|∇|(t) . (104)

are automatically satisfied if either i = j = m = n or i = j = m = n. The first of

(2) (2) (2) 1/2

(2) (2) (1) (2) (1) 1/2

|gkλ |2 b̂† b̂kλ

These two-mode correlation functions (i = j) arise, e.g., if we consider a mea-

(Ni j )2 = (Ni )2 + (N j )2 . (164)

(2) (2) 1/2 (2) 1/2