You are on page 1of 156

DISSIPATION IN LIGHT - ATOM

INTERACTION
Concepts of Laser Cooling and Applications

Hanspeter Helm1

Lecture Notes SS 2012 July 19, 2012

m=-1 m=0 m=+1

40

m

1
http://frhewww.physik.uni-freiburg.de helm@uni-freiburg.de
Cover page:

Stern-Gerlach experiment with a Bose-Einstein


condensate of Rb F = 1 atoms. The top image shows
the line of sight along the optical dipole trap where
the condensate is prepared by forced evaporation.
The condensate contains ≈ 3 × 104 Rb atoms. The
conical coils serve as MOT coils in the preparation
stage. During free fall these coils establish the mag-
netic field gradient of the Stern-Gerlach separator.
m=-1 m=0 m=+1

The bottom image shows the three Zeeman compo-


nents which separate by 400 µm after a 10 ms free
fall in the earth’s gravitational field at a magnetic
field gradient of 7 G/cm. 40

m
C. Käfer et al., Phys. Rev. A 80 023409 (2009).

Boxes on margin numbered with integer n refer to solutions and drawings


Mn
obtained in the notebook DISS-MATHE-2012.nb (Mathematica-8).

This script is in support of 56 hours of lectures given in the summer semester


2012 at the University of Freiburg. Prof. Dr. Hanspeter Helm
Contents

1 A Short History 1

2 Mechanical Forces of Light 7


2.1 Spontaneous Force . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Dipole Force (Gradient Force) . . . . . . . . . . . . . . . . . . . . 11
2.3 Photon Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3 Optical Molasses 13
3.1 Classical Rate Model . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Radiation Force on Moving Atoms . . . . . . . . . . . . . . . . . 17
3.3 Friction in Optical Molasses . . . . . . . . . . . . . . . . . . . . . 17
3.4 Diffusion in Optical Molasses . . . . . . . . . . . . . . . . . . . . 18

4 Magneto-Optic Trap 21

5 Coupling between 2 States 27


5.1 Time-Independent Coupling . . . . . . . . . . . . . . . . . . . . . 27
5.2 2-Level Atom in the Field of a Single Mode . . . . . . . . . . . . 29
5.3 Spontaneous Emission . . . . . . . . . . . . . . . . . . . . . . . . 32

6 Dressed States 33
6.1 Uncoupled Atom-Laser States . . . . . . . . . . . . . . . . . . . . 34
6.2 Atom-Laser Coupling . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.3 Effects Of Spontaneous Emission . . . . . . . . . . . . . . . . . . 38

7 Density Matrix for a Two-Level Atom 41


7.1 Pure States and Density Operator . . . . . . . . . . . . . . . . . 41
7.2 Spontaneous Emission in Two-Level Atom . . . . . . . . . . . . . 45
7.3 Liouvillian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7.4 Solutions for Two-Level Atom . . . . . . . . . . . . . . . . . . . 46
7.5 Spontaneous Force . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.6 Susceptibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.7 Dipole Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

8 Dipole Force and Dissipation 51


8.1 Hamiltonian of Dressed States . . . . . . . . . . . . . . . . . . . . 52
8.2 Mollow Triplet . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
8.3 Mean Dipole Force . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8.4 Force on Stationary Atom . . . . . . . . . . . . . . . . . . . . . . 57

I
II CONTENTS

8.5 Dissipative Contributions . . . . . . . . . . . . . . . . . . . . . . 58


8.6 Atomic Motion in a Standing Wave . . . . . . . . . . . . . . . . . 61

9 Polarization Gradients 65

10 Lin-Lin Polarization Gradients 69


10.1 Sisyphus Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
10.2 Magnitude of Friction . . . . . . . . . . . . . . . . . . . . . . . . 72
10.3 Semiclassical Estimate . . . . . . . . . . . . . . . . . . . . . . . . 73

11 Orientational Cooling 75
11.1 Equilibrium for Stationary Atom . . . . . . . . . . . . . . . . . . 76
11.2 State of Moving Atom . . . . . . . . . . . . . . . . . . . . . . . . 77
11.3 Coupling in the Rotating System . . . . . . . . . . . . . . . . . . 79
11.4 Motion-Induced Orientation . . . . . . . . . . . . . . . . . . . . . 80
11.5 Light Pressure on an Oriented Sample . . . . . . . . . . . . . . . 80

12 VSCPT 83
12.1 Dark States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
12.2 Optical Pumping in Velocity Space . . . . . . . . . . . . . . . . . 85
12.3 Momentum-Families . . . . . . . . . . . . . . . . . . . . . . . . . 86
12.4 Motion Induced Atom-Laser Coupling . . . . . . . . . . . . . . . 87
12.5 Decay Due to Spontaneous Emission . . . . . . . . . . . . . . . . 88

13 Evaporative Cooling 93
13.1 Bose-Einstein Condensation . . . . . . . . . . . . . . . . . . . . . 93
13.2 Atom-Atom Interactions . . . . . . . . . . . . . . . . . . . . . . . 95
13.3 Collisions of Cold Atoms . . . . . . . . . . . . . . . . . . . . . . . 96
13.4 Mean-Field Approximation . . . . . . . . . . . . . . . . . . . . . 100
13.5 Magnetic Trapping . . . . . . . . . . . . . . . . . . . . . . . . . . 100
13.6 Evaporative Cooling . . . . . . . . . . . . . . . . . . . . . . . . . 103
13.7 Diagnostics on BE Condensates . . . . . . . . . . . . . . . . . . . 104

14 Sideband Cooling 105


14.1 Lamb-Dicke Parameter . . . . . . . . . . . . . . . . . . . . . . . . 106
14.2 Degenerate Raman Sideband Cooling . . . . . . . . . . . . . . . . 109
14.3 EIT Cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

15 Appendices 113
A-1 Ammonia Molecule and Maser . . . . . . . . . . . . . . . . . . . 113
A-2 Bloch Vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
A-3 Clebsch-Gordan Coefficients . . . . . . . . . . . . . . . . . . . . . 119
A-4 Partial Wave Expansion . . . . . . . . . . . . . . . . . . . . . . . 123
A-5 Rotation of Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
A-6 Slowly-Varying Amplitude Approximation . . . . . . . . . . . . . 129
A-7 Landau-Zener Crossing . . . . . . . . . . . . . . . . . . . . . . . . 131
A-8 Hyperfine Zeeman Structure of Rb . . . . . . . . . . . . . . . . . 133
A-9 Atom Interferometry . . . . . . . . . . . . . . . . . . . . . . . . . 137
A-10 Multiphoton Bragg Scattering . . . . . . . . . . . . . . . . . . . . 141
A-11 Feshbach Resonances . . . . . . . . . . . . . . . . . . . . . . . . . 143
A-12 Atomic Fountain Clock . . . . . . . . . . . . . . . . . . . . . . . . 146
CONTENTS III

Bibliography
Chapter 1

A Short History

Our daily experience tells us that neutral macroscopic objects can be moved,
slowed down, or held by physical contact with other bodies. This contact
appears to be entirely different from the way in which we control the motion
of electrically charged or magnetized objects with electromagnetic fields.
A closer look at these phenomena tells us however that there is little difference.

Whenever we hold a piece of paper between our fingers, our skin approaches
the surface of the paper to within atomic dimensions. Electrons at the surface
of the paper and our skin repel each other. This charge displacement gives rise
to dipoles and electric fields which are sufficiently strong to grab the paper.
Mechanical grip and friction, both of which which are required to move macro-
scopic objects have their origin in electromagnetic forces1 which appear when
objects approach each other to atomic distances [1].
Techniques were developed during the past 30 years which permit manipula-
tion of atoms and small objects, with astounding precision and ease, using forces
exerted by light. A main difference to the conventional mechanical manipula-
tion appears to be that neutral objects can be moved without the conventional
physical contact. Nevertheless contact is made by the electromagnetic field and
its interaction with atomic dipoles, quite analogous to the origin of mechanical
grip. Keys to the control of the external degrees of freedom of a neutral particle
by light are

• the momentum transfer in absorption and emission of photons, and

• the prudent choice of physical parameters such that light-atom interaction


gives rise to dissipative terms, that is friction.

A history of events which lead to this control shows the close-knit interac-
tion between basic research on one side and the development of new technologies
which pave the way to new applications. An original motivation was the de-
sire to reduce/eliminate uncertainties in precision spectroscopy, which originate
from thermal motion.

1 Van-der-Waals and Casimir-Polder forces are such examples.

1
2 CHAPTER 1. A SHORT HISTORY

Limitations in the process of measurement : The Doppler-effect is one of


the limitations to a precise determination of atomic transition frequencies. An
atom moving with the velocity ~v sees a stationary laser source at the frequency

ωA = ωL − ~k · ~v , (1.1)

where ωL is the frequency of the laser in the laboratory frame and ~k is the
wave vector.2 For a thermal speed distribution the Doppler effect gives rise to
a Gaussian distribution of the spectral absorption and emission profile.3

N a-atoms at 300 K have a Gaussian Doppler width of ∆ω = 2π ×1.7 GHz.


The natural linewidth of Na(2 P3/2 ) is by comparison Γ = 2π × 10 MHz.

Doppler-free experiments (e.g. two-photon absorption using counter propagat-


ing laser beams) reach high spectral resolution in that they select a signal which
originates primarily from atoms with no substantial speed component along ~k.
In principle, this technique permits to reach the natural linewidth. However
other limitations appear, among them the limit due to time-of-flight broaden-
ing: The duration of a measurement, ∆t, limits the frequency uncertainty ∆ν
of this measurement, ∆ν ∆t ≥ 1. One reason for attempting to cool atoms with
laser light was to slow down and trap atoms, undisturbed by neighboring atoms,
in order to observe them for a long time.

Images of individual atomic ions were first recorded in 1980 by Neuhauser


in Hamburg [2]. An atomic ion was isolated in a Paul trap and observed by
laser fluorescence. The author succeeded in trapping the atomic ion in vacuum,
nearly motionless, such that the same ion could be observed over many days,
even weeks. Laser cooling was a key issue here. In this experiment the kinetic
energy of the ion was reduced to about 10 mK.

Why should one try to cool atoms with laser light? Curiosity was often at
work, but there are numerous applications, some of which are realized already :

• Precise atomic clocks are based on high-resolution spectroscopy. Cur-


rent commercial clocks reach a spectral (time) resolution of 1 : 1013 . The
realization of atomic fountain clocks permits observation (measurement)
times which are much longer than those which can be achieved in conven-
tional atomic beams.4 Several laboratories work on increasing the preci-
sion of Rb and Cs frequency standards by several orders of magnitude, a
resolution of the order of 1 : 1018 appears possible. Why do we need atomic
clocks? High-resolution radio astronomy is based on proper synchroniza-
tion of signals recorded by different radio telescopes. Global Positioning
2 The expression ~
p k · ~v is the approximation for the relativistic Doppler-shift of frequency,
ω 0 − ω = ω 1 − 1 − β 2 /(1 + β cos θ) , in the limit of low speed, |~v |  c. Here β = |~v |/c and


θ is the angle between ~k und ~v . (Jackson, Classical Electrodynamics)


3 The reason is that the Maxwell-Boltzmann distribution in 1D (along the direction of

photon absorption/emission) is a Gaussian distribution and we make the assumption that the
Doppler-width is much larger than the natural line width
4 The observation of an atom resting on a surface is not an alternative, as the interaction

of the atom with the surface leads to a distortion of the atomic eigenfrequencies.
3

(GPS) only works because precise atomic clocks synchronize the signals
of different satellites. High-speed telecommunication is synchronized with
atomic clocks. Very likely a Strontium-based atomic clock, developed at
the PTB in Braunschweig, will form the new international time basis in
the near future.
• Atom optics: Lenses formed by light fields can be used to focus / col-
limate beams of slow atoms. Any optical element can be formed by ap-
propriately shaped light fields, one can built gratings, beam splitters and
mirrors for atoms from light. In general these elements only work for atoms
which are cooled to low translational energies. Reason is that the forces
exerted by light are typically small.
Light and glass, as we know it in conventional optics, can make the same
physics as atoms and light, a topic also known as atom optics.
Atomic-scale architecture appears in sight by controlled guiding of indi-
vidual atoms by tailored light fields.
In the same way as a photon can interfere with itself (and only with it-
self), atoms or molecules can interfere. In an atom interferometer an
atomic wavepacket passes along spatially different paths which are recom-
bined later. Path differences in the domain of a fraction of the atom’s
de-Broglie wavelength lead to noticeable shifts of the interference pattern.
Atom interferometry with freely falling atoms [4] permitted the most pre-
cise determination of the local gravitational acceleration. There are plans
for atomic gyroscopes for an improved inertial guiding system.
The de-Broglie wavelength of an atom moving with the recoil speed is
equivalent to the wavelength of the light causing the recoil. The recoil
speed, vR , can be calculated from h̄ k = M vR . For a Rb atom after
a recoil due to a photon at the resonance line at 780 nm de-Broglie
wavelength is λdB = h/p = h/(M vR ) = 780 nm.

• Bose-Einstein condensation of a dilute gas of weakly interacting atoms


was achieved after precooling an atomic cloud with laser light.

• Optical lattices permit the observation of the quantized motion of atoms


trapped in a periodic external potential. Based on predictions by Letokhov
and Minogin in 1977 such lattices with exactly known potential shape can
be made free of defects and the potential well depth can be changed by
varying the laser power without changing the lattice vectors [5].

• Recoil effects in photoionization or charge transfer can now be monitored


directly, a possible future application will be the precise determination of
the neutrino mass.5
• Antimatter : Only with electromagnetic fields can neutral antimatter
be manipulated and stored. First experiments succeeded in 2012 [3].

5 For example in the reaction 137 Cs →137 Ba + e− + ν the neutrino mass appears in the
55 56 e
energy/momentum balance. Provided one begins with a very cold Cs atom and measures the
recoil of the Ba atom, in coincidence with that of the electron, one might deduce the mass of
the neutrino from the conservation of momentum and energy.
4 CHAPTER 1. A SHORT HISTORY

Milestones

1619 Kepler postulates in his treatise ”De Cometis” that the pressure of light
is responsible for the direction of a comet’s tail (away from the sun).
1873 Crookes thinks that he observes the pressure of light in his radiometer.6
1873 Maxwell formulates the force exerted by an electromagnetic wave in
absorption and reflexion. This force is 5 orders of magnitude lower than
what Crookes observed in his radiometer (and of opposite sign). According
to Maxwell the force associated with the light pressure on a surface A is
W
F = (1 + r) = A · w · (1 + r) , (1.2)
c
where r is the coefficient of reflection of the surface.7 The power W
signifies the energy impinging on the surface per unit time ([Watt] =[J/s]),
c is the speed of light. The pressure acts in the direction of the propagation
of light. The light pressure is equal to the radiation energy per unit
volume (energy density of the electromagnetic field, w, with the dimension
[J/m3 ]).
In the photon picture the force (1.2) can be explained as follows. Each
photon transfers the momentum p = h̄ k (1 + r). The radiation intensity
(W/m2 ) is I = N h̄ω/A , where A is the area, h̄ω is the photon energy,
and N the number of photons hitting the area per second (photon flux).8
Thus we have

W = I A = N h̄ω . (1.3)

The radiation pressure acting on the surface A is equal to the momentum


transferred per second, F = N p = N h̄k (1 + r). With the expression for
N from (1.3) we obtain for the radiation pressure (using k = ω/c )
W W
F = h̄k (1 + r) = (1 + r) . (1.4)
h̄ω c

On a 1 m2 surface the bright sun exerts a radiation pressure correspond-


ing to 0.4 mg (black) and 0.8 mg (ideal mirror). Maxwell writes in a
textbook in 1883 : ”Concentrirtes elektrisches Licht wird wahrscheinlich
einen noch größeren Druck ausüben (als die Sonnenstrahlung), und es ist
nicht unmöglich, dass die Strahlen eines solchen Lichtes, wenn sie auf ein
dünnes metallisches Blättchen, das in einem Vacuum fein aufgehägt ist,
fallen, an diesem einen beobachtbaren mechanischen Effect ausüben.”[7]
1900 Lebedev in Moscow confirms [8] Maxwell’s prediction in an experiment.9
Lebedev illuminated small mirrors on a torsion balance with light from an
6 These radiometers (Lichtmühle) move in the wrong direction, as the pressure exerted on

the black surface by hot atoms (bad vacuum!) outweighs the radiation pressure.
7 The reflection is r = 0 for a black body and r = 1 for an ideal mirror.
8 The photon flux density is N/A = I/(h̄ω).
9 A Freiburg glassblower, C. Kramer, was instrumental in this experiment as he fabricated

the perfect glass valves which enabled Lebedev to produce very high vacuum, thus eliminating
Crookes’ problem.
5

arc discharge at a power of 70 mW. According to Maxwell the radiation


force should have been 2.3×10−10 N, Lebedev observed (3±0.2)·10−10 N.
1905: Poynting, then president of the Physical Society, commented this ex-
periment after Lebedev presented it in London [9]:
”A very short experience in attempting to measure these light forces is suf-
ficient to make one realize their extreme minuteness - a minuteness which
appears to put them beyond consideration in terrestrial affairs.”
1909: In his doctoral thesis ”On the radiation pressure” Debye searched for a
practical case of optical levitation (radiation pressure balances the gravi-
tational forces). He concluded that small particles in the vicinity of the
sun can fulfill this requirement.
1917 After Einstein introduced the concept of the photon, he showed that
momentum conservation is an essential aspect of the thermal equilibrium
between radiation and matter. A photon carries the energy E = h̄ω and
the momentum p~ = h̄~k.
1925 Compton and Simon, and later Bothe und Geiger carried out experi-
mental studies on the Compton effect (recoil experienced by an electron
in a collision with a photon).
1933 Otto Frisch in Hamburg demonstrates the momentum transfer of light
on atoms by deflecting a beam of sodium atoms.

Around 1970 such experiments became much simpler, as one then had access to
lasers and much higher photon flux. These opportunities lead to new thoughts:
Could one build an optical trap to hold atoms? First ideas to this topic came
from Lethokov and his group in Moscow [10] and from the Ashkin at Bell Labs.
1968 Letokhov suggested that forces associated with the light shift of atomic
levels in a standing wave may be sufficient to confine atoms. These forces
were demonstrated for the first time in an experiment in 1987 in Paris [6].
1978 Ashkin concluded: ”The forces exerted by a focused beam of laser light
are strong enough to push tiny particles around freely in various mediums.
Several applications based on this finding are proposed” [11].
1986 Stenholm reviewed the semiclassical theory of the mechanical action of
light in resonant interaction with atoms [12].

Many groups have since worked on laser cooling and atom trapping and many
novel technologies emerged from this research.
1997 Nobel Price awarded to
Steven Chu, William D. Phillips und Claude Cohen-Tannoudji
”for development of methods to cool and trap atoms with laser light.”
2001 Nobel Price awarded to
Eric A. Cornell, Carl E. Wieman and Wolfgang Ketterle ”for the
achievement of Bose-Einstein condensation in dilute gases of alkali atoms,
and for early fundamental studies of the properties of the condensates.”
6
Chapter 2

Mechanical Forces of Light

In 1925 Compton discovered the momentum transfer of an X-ray photon


to an electron in the change of wavelength of the scattered radiation and
explained it by considering the conservation of energy and momentum of
electron and photon, the photon momentum being p = h̄|k|. The momen-
tum transfer by the photon is the basis of all mechanical forces of light.
When photon of momentum h̄~k is scattered by a massive particle of mass M
momentum conservation requires

v
h̄~k + M~vi = h̄~k 0 + M~vf (2.1) a

where i and f signify the initial and final velocity of v - h k


m
the massive particle. b
Figures a) and b) show the case that ~k is antiparallel
to the atomic velocity ~vi , and a photon which is fully
absorbed in resonant excitation (~k 0 = 0). In this
c
event the atom of mass M gains the recoil

~vR = ~vi − ~vf = h̄~k/M . (2.2)

For sodium we have |k| = 2π/λ with λ = 586 nm, M = 23 amu and obtain

|~vR | = 3 cm/s . (2.3)

The momentum imparted by visible light is small, the temperature equivalent


for sodium being
2
M vR
T = ≈ 10−6 K . (2.4)
2kB

The Lorentz width for a natural lifetime τ = 16 ns corresponds to ∆ω =


2π × 10 MHz. By comparison the Doppler shift corresponding to a recoil speed
of 3 cm/s is

ωi − ωf = ∆ω = k vR ≈ 2π × 50 kHz . (2.5)

7
8 CHAPTER 2. MECHANICAL FORCES OF LIGHT

Velocity changes caused by the interaction of atoms with visible light are very
small. Nevertheless a proposal to use the photon momentum for cooling of
atoms was made in 1968 by Lethokov in Moskow [10]. In 1975 Wineland and
Dehmelt in Seattle [13] and Hänsch and Schawlow in Stanford [14] examined
schemes for cooling and by 1985 neutral atoms were cooled to ≈1 mK, using
a laser configuration now known as optical molasses. This paved the way for
trapping atoms. A variety of optical, magnetic, and magneto-optical traps in
use today. Soon temperatures of 100 µK, then 200 nK and lower were reached,
the lowest temperatures ever demonstrated in a laboratory.

In a simplified approach one may identify two origins for the force of light.
At high laser intensity and in spatially inhomogeneous light fields the distinction
between the two becomes blurred as we will see later.

2.1 Spontaneous Force


The mean spontaneous force acts in the direction of the light beam and is
proportional to the product of the photon momentum h̄k, times the scattering
rate (how many spontaneous photons are emitted per second). We define the
detuning of the laser relative to the atomic frequency as

δ = ωL − ω0 − ~k · ~v = ωA − ω0 . (2.6)

When the laser is red detuned from resonance, ωL − ω0 < 0, the Doppler effect
leads to preferential absorption of photons by atoms drifting against the laser
beam (in which case ~k · ~v is negative). The atoms are effectively cooled be-
cause the direction of momentum transfer in spontaneous emission is in general
isotropic and hence the mean momentum transfer due to spontaneous emission
is zero.
This force fluctuates in time due to the random time at which absorption
happens and due to the random direction of spontaneous emission, leading to a
diffusion of atoms in momentum space. As we shall see this diffusion results in
a small heating term which limits the lowest temperature that can be achieved
by this method. The spontaneous force depends on detuning, the closer one is
to resonance the more frequently absorption can happen. However: stimulated
emission competes with spontaneous emission and stimulated emission compen-
sates the momentum transfer from absorption.1

To what magnitude can the spontaneous force grow ? For a classical


laser field, E(t) = E0 cos (ωL t), the laser intensity is

I = 0 cE02 /2 [Watt/cm2 ] . (2.7)

The Rabi frequency in a two level system (the rate at which population
oscillates between ground and excited state, see Chapter 5) is

Ω1 = |d| E0 /h̄ , (2.8)

where d is the dipole moment of the optical transition. The laser intensity can
in principle grow to arbitrarily high values, and equally the Rabi frequency.
1 This is true only if both processes occur with the same plane wave.
2.1. SPONTANEOUS FORCE 9

On the other hand, the rate of emission of spontaneous photons rises linearly
with intensity only at low values of intensity. The spontaneous rate eventually
saturates, as we need to give the excited state time to spontaneously emit.
We denote the rate of scattering photons2 by Γscatt . The spontaneous force
exerted by a beam of light which propagates along the vector ~k is

F~ = h̄~k Γscatt . (2.9)

In a two-level system the spontaneous force is (see Chapters 3 and 7)

Γ I/I0
F~ = h̄~k · , (2.10)
2 1 + I/I0 + (2δ/Γ)2

where I0 is called saturation intensity,

I 2Ω21
= . (2.11)
I0 Γ2

At the saturation intensity the squared Rabi frequency is equal to one half the
squared spontaneous rate. The saturation intensity is defined as the intensity
of a resonant laser (δ = 0) such that the scattering rate is equal to one quarter
of the spontaneous rate, Γscatt = Γ/4. The natural decay rate Γ is related to
the natural lifetime of the excited state by τ = 1/Γ. The maximal spontaneous
force (I → ∞) is

Γ
F~max = h̄~k (2.12)
2
in which case we have Γscatt = Γ/2.

We explore the magnitude of this force at the example of sodium (charac-


teristic quantities are : λ = 586 nm, Γ = 1/16 ns−1 = 2π × 10 MHz, M = 23 amu)

• the magnitude of the wave vector is

2π 2π 9
k= = 10 ≈ 107 m−1
λ 586
and the maximal acceleration due to light is

Fmax Γ (1/16) 109 2


amax = = h̄k = 10−34 · 107 ≈ 106 m/s
M 2M 2 · 23 · 1.6 · 10−27

By comparison the gravitational acceleration (g = 9.81 m/s2 ) is about 105


times smaller than the maximal spontaneous force

• the Coulomb force on a singly charged ion, F~ = q E,


~ at a field strength of
~ = 1 V/cm results in an acceleration of a Na+ -ion of
|E|

F qE 100 · 1.6 · 10−19


a= = = = 4 × 1010 m/s2
M M 23 · 1.6 · 10−27
2 The rate at which spontaneous photons are emitted by the atom.
10 CHAPTER 2. MECHANICAL FORCES OF LIGHT

We see that the spontaneous force can be much larger than the gravitational
force, but is typically much smaller than electrical forces on charged objects.

Next we consider the time required to slow down a thermal atom : The
mean thermal speed of sodium at 300 K is
r
8kB T
hvi = ≈ 500 m/s . (2.13)
πM
• the recoil speed of a sodium atom after absorption of a photon at 586 nm

vR = h̄k/M = 3 cm/s . (2.14)

• thus we need to absorb 500/0.03 ≈ 17 000 photons in order to completely


stop a thermal sodium atom
• in what time can we do this? According to (2.12) at least 2 times 17 000
spontaneous lifetimes are required, t = 2 × 17000/Γ ≈ 300 µs .
number of atoms

ta p e r
22869 atoms after b ia s
413497 scattering events
at a laser detuning of
a to m ic
∆ = -10 GHz b e a m
original
s o u rc e c o o lin g la s e r
distribution

B (z )
0 300 600 900 1200 1500
v HmsL z
M1
Problems appear if we direct a laser against an atomic beam and com-
pensate the Doppler shift by a suitable detuning ~k ·~v from resonance (for
v=500 m/s the optimal detuning would be ≈ −5 GHz). After cooling
for some time the atoms fall out of resonance.
To cool further we would need to tune the laser closer to resonance
again. This can be done by actively changing the laser frequency or by
spatially changing the atomic resonance frequency via the Zeeman effect
in a device named Zeeman-slower [15] (above, right).

Next we consider the deflection of an atomic beam. A resonant laser,


perpendicular to the atomic beam, accelerates the atom by

F h̄kΓ I/I0
a= = ·
M 2M 1 + I/I0
x
R
For a cylindrical lens and atoms guided along a 3
z
radius of curvature, R, we have
a to m ic b e a m
M vθ2 /R = F (~k · ~v = 0)
d e fle c tin g
In this way Ertmer [16] bent a beam of atoms la s e r b e a m

with v = 104 cm/s by an angle of 30o , along a


radius of curvature of R = 62 mm.
Atoms which do not fulfill the condition ~k · ~v = 0 fall out of resonance and
experience a friction which damps/accelerates the radial velocity.
2.2. DIPOLE FORCE (GRADIENT FORCE) 11

2.2 Dipole Force (Gradient Force)


Several semi-classical explanations exist for the dipole force. Intrinsic to all is
the polarizability of the atom.
1) An optical field (at any frequency) induces an oscillating electrical dipole
moment in the atom. This dipole moment in turn interacts with the oscillat-
ing electric field. In a homogeneous field a dipole experiences a torque, in an
inhomogeneous field it also experiences a force. The force is directed along the
gradient of the field strength. The phase of the dipole (its orientation along
the external field) has the typical resonance character, it is in phase with the
driving field at frequencies lower than resonance, but of opposite phase above
the resonance. Thus atoms are attracted to regions of higher intensity if the
driving frequency is below resonance. Above resonance the dipole is repelled
from regions of high intensity.
The induced dipole moment is d~ = α · E ~ where α is
the atomic polarizability. The interaction energy in
the field is Udip = −d~ · E
~ = α · |E|
~ 2 ∝ I.
U d ip The potential energy of the induced dipole in a Gaus-
ro t
sian laser beam is shown for red detuning. The spa-
k B T tial coordinate is horizontal. The minimum refers to
the position of the highest laser intensity.
For large detuning the dipole potential along some spatial co-ordinate r is

Udip (r) ≈ h̄ Ω21 (r)/(4 δ) , (2.15)

where δ = ωL − ω0 is the detuning, see Chapter 5.

2) The dipole force (at times referred to as stimulated force) can be seen as
a consequence of stimulated absorption / stimulated emission events involving
spatially different plane waves of identical frequency. This was formulated first
by Dalibard and Cohen-Tannoudji [17]. In the presence of plane waves
of different ~k, but equal |~k|, a redistribution of photons in the different waves
may occur, thus causing changes in the atom’s momentum, see Chapter 8. The
dipole force may also be viewed as a consequence of the light shift of atomic
energies, see Chapter 6.

3) An apparently different explanation goes back to Ashkin [11] who considers


a focussed laser beam in air and a sphere of refractive index ns > 1 (below, left)
and a noticeable gradient of light intensity across the cross section of the sphere.

F a
b F b
a

a
a
b
b
b
F b a
F a

If we neglect reflection at the interface we see that more light is refracted in the
direction of lower laser intensity. As a result a net acceleration of the sphere in
the direction of higher laser intensity occurs.
If on the other hand we consider a focussed laser beam in water (ns < nH2 O ),
the opposite holds, the sphere is driven out of the laser beam. This explanation
12 CHAPTER 2. MECHANICAL FORCES OF LIGHT

is consistent with that of the classical oscillator for which the refractive index
changes from a value n < 1 to n > 1 in the vicinity of the optical resonance.
An important application of this effect is realized so-called laser tweezers.
Using a focussed light beam, tuned far off resonance, one may steer and move
nearly macroscopic objects (cells), under the lens of a microscope. In this case
dissipation occurs due to friction of the particles in the liquid surroundings.

2.3 Photon Momentum


Feynman discussed the radiation pressure as a consequence of the magnetic
component of the electromagnetic wave [18]. A beam of light, linearly polarized
along x, propagates along the co-ordinate z and acts on a charge q. The electric
field leads to an oscillation of the charge along x with the velocity vx . Due to
the magnetic field, By , the charge experiences the Lorentz force

Fz = q · v x × B y , (2.16)

which acts in the direction ~k = {0, 0, kz }. Since |B| = |E|/c, the mean force is
q
hF i = hv Ei . (2.17)
c
The work done by a light wave per unit time, W (that is power) is equal to force
× velocity [Nm/s = J/s = W]. The velocity here being the speed of light, we
have hF i = W/c. The energy absorbed from the light wave per second is equal
to c times the force. With the definition of power in the photon picture (N is
the number of photons which impact on a surface A per second, and h̄ω is the
energy of a single photon)

W = I A = N h̄ω . (2.18)

Since ∆p/∆t = F , the momentum transferred by the light field is equal to the
energy absorbed in one second, divided by c. The momentum transferred by a
light wave of power W , which is fully absorbed, is therefore equal to
N h̄ω
∆p = , (2.19)
c
and the momentum of a single photon is therefore
h̄ω h̄2π
p= = = h̄k . (2.20)
c λ

According to Eq. (2.16) the momentum direction is that of ~k since vx ||E.


~

For the radiation pressure we have

P = F/A = w , (2.21)

the radiation pressure on a fully absorbing object being equal to w, the energy
density of the radiation field received by the object [Ws/m3 = J/m3 = N/m2 ].
Chapter 3

Optical Molasses

Atoms in the gas phase may be cooled using counter-propagating and red-
detuned laser beams. This configuration can be realized in 3 dimensions
and was named optical molasses (optische Melasse).
The first realization of molasses cooling was reported in 1986 by Chu et al. [19].
Origin for cooling is a friction term which appears in the spontaneous force
action of counter-propagating beams. We explore the force, the friction and the
associated diffusion in momentum space, at first using a classical rate model.

3.1 Classical Rate Model


The absorption profile of a stationary atom is given by a Lorentzian distribution

Γ2
L(ω) = L(ω0 ) (3.1)
4(ω − ω0 )2 + Γ2
where the angular frequency is ω = 2πν and L(ω0 ) is the amplitude of the pro-
file at the central frequency, ω0 . The full-width at half-maximum (FWHM) is
given by ∆ω = 2π∆ν = Γ, the natural linewidth.
1 If we introduce the detuning, δ,
δ = ω − ω0 , where ω is the frequency
of observation and normalize the maxi-
LH∆L

1
2
mum of the profile to one, we may write
M2

1
0 L(δ) = 2 . (3.2)
-3 -2 -1 0 1 2 3 1 + (2δ/Γ)
2∆G

Coefficient of absorption : If we irradiate a layer of gas atoms with thick-


ness x and atom density N [m−3 ] with a radiation intensity I, Beer’s law predicts
an attenuation to
I(x) = I(0)e−α(ω)x . (3.3)

13
14 CHAPTER 3. OPTICAL MOLASSES

The coefficient of absorption, α ( dimension [m−1 ] ), is related to the absorption


cross section ( dimension [m2 ] ) by
σ(ω) = α(ω)/N . (3.4)
For stationary atoms the spectral distribution of σ is given by the Lorentzian
1
σ(ω) = σ(ω0 ) 2 . (3.5)
1 + (2δ/Γ)
An approximate value for the resonant absorption cross section of a two-level
atom is given by1
3λ2
σ(ω0 ) ≈ = 6πc2 /ω02 . (3.6)

Optical Pumping : We consider a classical rate model for optical pumping


to evaluate the change in population density of a two-level atom. The pumping
rate ( dimension [s−1 ] ) is given by R = Bw(ω) = σ(ω)I(ω)/h̄ω, where B is the
Einstein B-coefficient. The population density in the ground state is N1 , that
in the excited state is N2 . The density of atoms is N = N1 + N2 . The temporal
change in densities
dN1 dN2
=− = −RN1 + RN2 + ΓN2 . (3.7)
dt dt
In stationary state we have dN1 /dt = 0,
0 = RN1 − N2 (Γ + R) .
Using N1 + N2 = N we obtain
Γ+R 1 + R/Γ
N1 = N =N . (3.8)
Γ + 2R 1 + 2R/Γ
The population difference between ground and excited state is
N N
N1 − N2 = = . (3.9)
1 + 2R/Γ 1+S
The populations (3.8) and (3.9) are shown below for the case N = 1.

1.0 1.0
0.9 0.8
N1 -N2

0.8 0.6
N1
0.7 0.4
0.6 0.2
0.5 0.0
0 1 2 3 4 5 0 1 2 3 4 5
S=2 R  G S=2 R  G

We have the following limiting cases


1 We use the Rb D-line (2 P 2 2
1/2 → S1/2 ) as example. The natural lifetime of the P1/2
state is τ = 28 ns, equivalent to a decay rate Γ = 1/τ = 2π × 5.7 MHz and a linewidth
∆ν = Γ/2π = 5.7 MHz. With the transition wavelength, λ = 795 nm, we obtain for the cross
section σ(ω0 ) = 3 × 10−9 cm2 .
3.1. CLASSICAL RATE MODEL 15

• R = 0 : all atoms are in the ground state, N1 = N .

• R  Γ : we have N1 = N/2.

• R = Γ/2 : we have N2 = N/4 and N1 = 3N/4.

The saturation parameter S is defined as the ratio between the pumping rate
and half the rate of spontaneous emission:

2R 2 I(ω)
S(ω) = = · σ(ω)
Γ h̄ω · Γ
2 I(ω) 1
= · σ(ω0 ) 2
h̄ω · Γ 1 + (2δ/Γ)
I(ω) 1
= · 2 , (3.10)
I0 1 + (2δ/Γ)

where we assumed ω ≈ ω0 . In the last expression we introduced the saturation


intensity

h̄ω0 · Γ
I0 = . (3.11)
2σ(ω0 )

For I = I0 , the pumping rate (at resonance δ = 0) is just one half the rate of
spontaneous emission.2

Saturation of absorption coefficient : Saturation modifies the absorption


coefficient at high intensity, as we no longer have all atoms in the ground state

N
α(ω) = σ(ω)Neff = σ(ω)(N1 − N2 ) = σ(ω) . (3.12)
1+S

Introducing the Lorentzian lineshape of the cross section we obtain

σ(ω0 ) N N σ(ω0 )
α(ω) = 2 · = 2 = σeff (δ, I) · N ,
1 + (2δ/Γ) 1+S 1 + I/I0 + (2δ/Γ)

were we have inserted for S the expression from Eq. (3.10). The spectral profile
of the absorption coefficient is again Lorentzian, but now with an increased
width and reduced height. The intensity dependence of the scattering rate by
a single atom (N = 1) is

I Γ
Γscatt = σeff = σeff I/I0
h̄ω 2σ(ω0 )
Γ I/I0
= · . (3.13)
2 1 + I/I0 + (2δ/Γ)2

2 For alkali atoms the saturation intensity is in the range of a few mW/cm2 .
16 CHAPTER 3. OPTICAL MOLASSES

0.4
II0 = 5
0.3

Gscat  G
1
0.2

0.1
M3 0.1
0.0
-4 -2 0 2 4
2∆G
The scattering rate reaches the value Γ/2 at high intensity and δ = 0.
Notice the power broadening with increasing intensity.
The profile approaches the natural linewidth at low intensity.

A more common approach : The result (3.13) is usually obtained in a


density matrix model of the driven two-level atom (Chapter 7). In this model
one obtains for the population in the excited state of a two level atom

Ω21
ρee = , (3.14)
Γ2 + 2Ω21 + 4δ 2

where Ω1 is the Rabi frequency Ω1 = |dge |E0 /h̄ . The relationship between
the Rabi frequency and the saturation intensity is

2Ω21 I
2
= . (3.15)
Γ I0
The rate of scattering of spontaneous photons is given by the product of
excited state population and the spontaneous rate
Γscatt = Γ · ρee
Γ 2Ω21
= · 2
2 Γ + 2Ω21 + 4δ 2
Γ 2Ω21 /Γ2
= ·
2 1 + 2Ω21 /Γ2 + (2δ/Γ)2
Γ I/I0
= · (3.16)
2 1 + I/I0 + (2δ/Γ)2
This expression is identical to the one which we derived from the classical rate
equation model, Equation (3.13).

In the following we explore the force experienced by an atom in a plane wave.


3.2. RADIATION FORCE ON MOVING ATOMS 17

3.2 Radiation Force on Moving Atoms


The mean force experienced by a stationary two-level atom is the product of
the momentum transfer per photo absorption, h̄~k, and the rate of scattering
spontaneous photons

F~ = h̄~k · Γscatt
Γ I/I0
= h̄~k . (3.17)
2 1 + I/I0 + (2δ/Γ)2

If our atom is moving with the velocity ~v , then the effective detuning needs to
account for the Doppler shift, δeff = δ − ~k · ~v = ωL − ω0 − ~k · ~v , hence

Γ I/I0
F~ (~v ) = h̄~k i2 . (3.18)
2
h
1 + I/I0 + 2(δ − ~k · ~v )/Γ

At low intensity (I  I0 ) this expression can be simplified to

Γ I/I0
F~ (~v ) ≈ h̄~k · i2 . (3.19)
2
h
~
1 + 2(δ − k · ~v )/Γ

3.3 Friction in Optical Molasses


We now consider two counter-propagating beams of equal intensity and equal
frequency. We assume that the intensity is homogeneous in space and we average
the effect of the two forces over dimensions longer than the wavelength of light,
by taking the total force on the atom to be the sum of the two contributions.
At low intensity, I/I0  1, we obtain

Γ I/I0 Γ I/I0
F (v) = h̄k · 2 − h̄k · . (3.20)
2 1 + [2(δ − kv)/Γ] 2 1 + [2(δ + kv)/Γ]2

Examples for two different detunings are given below.

1. 1.
G = 2Π‰10MHz G = 2Π‰10MHz
∆ = 60MHz ∆ = 25MHz
0.5 0.5
-1
k = 10 m7
k = 107 m -1

F 0. F 0.
M4

-0.5 -0.5

-1. -1.
-10 -5 0 5 10 -10 -5 0 5 10
v ms v ms
In the vicinity of v = 0, F (v) shows a linear dependence on v over a small
range of velocities, ∆v ≈ Γ/k. The slope of F (v) near v = 0 is maximal if
18 CHAPTER 3. OPTICAL MOLASSES

2|δ|/Γ = 1/ 3. For δ < 0 (red-detuned ) the force is always directed opposite to
the atomic motion. We may expand (3.20) in the vicinity of v = 0

1 1 8kδv/Γ2
2
− 2
= + O(v 3 ) + ...
1 + [2(δ − kv)/Γ] 1 + [2(δ + kv)/Γ] [1 + (2δ/Γ)2 ]2

Using the lowest order term we find in the vicinity of v = 0

2δ I/I0
F (v) ≈ 4h̄k 2 vh i2 = −α · v = −γ M v , (3.21)
Γ 2
1 + (2δ/Γ)

a force term proportional to v. The friction coefficient α = M γ has the dimen-


sion [ kg s−1 ], M being the atomic mass. For δ < 0 we have α > 0. This force
damps the atomic motion with the rate
α v̇
γ= = . (3.22)
M v
The characteristic time for damping the atomic motion is γ −1 .
At a given value of I/I0 this time is minimal for

2δ/Γ = −1/ 3 . (3.23)

This optimal detuning fixes an external time scale


   
1 h̄ I0
∝ · , (3.24)
γ max ER I

with ER being the recoil energy. Equation (3.24) tells us the average time after
which a photon recoil occurs. For sodium and I = I0 this time is 130 ns.

3.4 Diffusion in Optical Molasses


Up to now we have discussed the force averaged over time. In reality we have
a sequence of discrete momentum transfers whenever the atom absorbs or emits.
The graph shows schematically the random di-
! "
rection of the recoil momenta due to spontaneous
emission in 2D motion. This force action fluctu-
ates with discrete momentum steps. The fluctu-
ations lead to a small heating term. There are
two origins for the fluctuation. One due to the
number of photons absorbed per second and one
! #
due to the direction of spontaneous photons.
We assume that, on average, the force is zero, but consider the stochastic ab-
sorption and emission events. At low intensity both processes are random, hence
the atom undertakes a random walk with step size h̄~k.
For simplicity we assume two counter-propagating beams along x at low
intensity and that spontaneous emission is only permitted along the x-axis. The
spontaneously emitted photons are uncorrelated in their direction, hence they
induce a random walk with the step size h̄k. For each spontaneous scattering
3.4. DIFFUSION IN OPTICAL MOLASSES 19

process there are two steps, one due to absorption and one due to emisson. After
n spontaneous events the mean value is
hp2x i = 2nh̄2 k 2 = 2Γscatt t h̄2 k 2 (3.25)
where t gives the time over which the scattering rate is Γscatt . The factor
two originates from the two recoils per scattering event. The mean squared
momentum increases with time, hence the mean kinetic energy of the atom
rises. The temporal change of the mean square momentum is
dhp2x i/dt = 2Γscatt · h̄2 k 2 (3.26)

Xevents per atom\ = 10000 8


Xevents per atom\ = 10000
mean value = 9080
400 Xpx \ = 17 70 000
Xp2x \ = 9080 6

number of atoms
60 000
200
50 000
p2x  HÑkL2
px  Ñk

40 000 4
0
30 000
-200 20 000 2
10 000 9080 M5
-400
0
0 50 100 150 200 0
0 50 100 150 200 0 10 00020 00030 00040 00050 000
atom number
atom number p2x  HÑkL2

The result of a Monte-Carlo simulation of hpx i and hp2x i,


for 200 atoms after 104 spontaneous emission events with h̄k = 1.
The mean value obtained is hp2x i = 9080 ≈ Γscatt t.

We now equate the rate of increase of the kinetic energy due to momentum
diffusion with the rate of energy loss due to radiative friction. The latter is
derived from F = −αv, where α is taken from Eq. (3.21). The rate of cooling
by friction is obtained using the following argument: An atom travels per second
a distance of v meters against the friction force F . Hence the cooling rate is
equal to hF vi = −αhv 2 i. In stationary state we may write
d 2
hp i/2M = −hĖcooling i = αhv 2 i
hĖheating i = (3.27)
dt x
In this one-dimensional problem we have a single degree of freedom. For this
motion we define a temperature equivalent T ,
1 1
M hv 2 i = kB T . (3.28)
2 2
Inserting (3.27) and (3.21) we obtain a temperature which depends on detuning

2  
h̄Γ 1 + (2δ/Γ) h̄Γ Γ δ
kB T = − · =− + (3.29)
4 2δ/Γ 4 2δ 2Γ

The temperature is minimal for δ = −Γ/2 [20] 1.0


0.8
h̄Γ
(1 + O(I/I0 ) + . . .)
kB T  ÑG

kB T = 0.6
2 0.4 M6
The minimum, the Doppler cooling limit, 0.2

is 240 µK, corresponding to hvi = 30 cm/s for 0.0


0.0 0.5 1.0 1.5 2.0 2.5 3.0
Na and 125 µK (9 cm/s) for Cs. detuning H∆GL
20 CHAPTER 3. OPTICAL MOLASSES

In this model we have assumed that the radiation forces from the two laser
beams are additive. For I/I0 > 1 this model fails as it does not account for the
fact that the atom may absorb photons from one of the waves and stimulate it
into the other wave.
A typical arrangement of molasses cooling in
m o la s s e s 3D. After turning off the laser beams the
b e a m s
atoms fall due to gravity and are recorded
in time when they pass through the resonant
m o la s s e s
probe beam. From the temporal distribution
of the fluorescence of atoms in the probe beam
the temperature of the falling atoms can be es-
fa llin g timated. If an external force acts on the atom,
a to m s
the atom may reach a drift velocity such that
the friction just balances the external force

p ro b e
b e a m
vdrift = Fext /α (3.30)
c o lle c tio n
o p tic s
With the external force being gravity we have
d e te c to r Fext = −αv = −M g and obtain vdrif t = g/γ.

Under optimal conditions (3.23) we would expect a drift due to gravity of 0.4
mm/s for sodium and 5 mm/s for cesium. This would amount to a substantial
loss of atoms. Similar losses were expected in the case of unbalanced laser inten-
sities. Experiments however showed that molasses atoms are quite insensitive to
intensity imbalance. In addition molasses temperatures much lower than those
predicted by the Doppler limit were observed.
& ! !

% ! ! This graph shows the minimal temper-


) ( 4 5 ( 6 7 ) * 6 ( . / .! . . 8 3

ature measured as a function of laser


$ ! !
detuning. The full line gives the result
# ! ! of Eq. (3.29). The lowest temperature
" ! !
is reached at much larger values of the
detuning.
!
! " ! # ! $ ! % !
' ( )* + ,+ - ./ 0 1 2 3

The experiments also discovered that the minimal temperature depends on


the strength of magnetic fields and on the type of laser polarization [21, 22].
Origin for this discrepancy are additional cooling effects (Sisyphus-cooling,
polarization gradient cooling, see Chapters 9 to 11), as they appear for realistic
atomic systems which generally have multiple Zeeman ground state levels.
Chapter 4

Magneto-Optic Trap

The molasses concept does not provide a spatial confinement and concentration
of atoms in a trap. To achieve trapping a distinguished location in space must
be defined where cold atoms should accumulate. This can be achieved in a
magneto-optic trap as proposed by Dalibard and first realized in 1987 [23]. The
concept is most easily introduced in 1D, for an optical transition from a state
with total angular momentum F = 0 to a state with F = 1.
Opposing magnetic field coils (anti- F=1
E
Helmholtz configuration) form a spatial mF=+1
domain with a magnetic field zero at the
mF= 0
trap center. At this location the mF lev-
els in the excited state are degenerate, mF=-1
the degeneracy being lifted as we move hw0 s+ s-
away from the trap center. In sodium a hwL

magnetic-field gradient of 10 G/cm leads


to a splitting between the Zeeman levels F=0
of 2π × 14 MHz/cm. z

!#

!" !#

!$
!" !# M7
!
#

!"

!"

%&'()*+,-.+)/0-12+/3
.

21
22 CHAPTER 4. MAGNETO-OPTIC TRAP

We consider a red-detuned and left-hand circularily polarized laser (σ − ) of fre-


quency ωL , propagating along the negative z direction (in the figure above from
the right). An atom with positive velocity, vz > 0, moves away from the trap
center and is driven back to the trap center by the spontaneous force, as the Zee-
man effect brings the atom closer to resonance with the laser. We analyze this
situation in the case I/I0  1, ~k · ~v  Γ, gF µB B/h̄  Γ. Here mF gF µB B/h̄
stands for the Zeeman detuning of the excited state.

Near the trap center the magnetic field varies approximately linearly with
the distance z from the trap center. We write for the magnetic detuning due to
the Zeeman effect, ∆ωz = βz, where β = gµB h̄ dB/dz is the detuning gradient
in units of Hz/m. At low laser intensity the sum of forces of two counter-
propagating laser beams of equal frequency is therefore F (v) = Fσ+ + Fσ−
( )
h̄kΓ Iσ+ /I0 Iσ− /I0
F (v) =
2 2 − 2 (4.1)
1+4 [(δ−kv−βz)/Γ] 1+4 [(δ+kv +βz)/Γ]

Β  1, 2, 5 mTcm Β  1, 2, 5 mTcm
1 1
0.75 ∆ 2Π 10 MHz 0.75 ∆ 2Π 10 MHz
 2Π 10 MHz  2Π 10 MHz
0.5 v 0 ms 0.5 v 2 ms
0.25 0.25
F 0 F 0
0.25 5 0.25 5

0.5 2 0.5 2
0.75 1 0.75 1

10 8 6 4 2 0 2 4 6 8 10 10 8 6 4 2 0 2 4 6 8 10
z mm z mm

The spatial dependence of this force in units of 12 h̄kΓI/I0 is shown for


v = 0 and v = 2 ms−1 for three values of the magnetic field gradient.

When we set Iσ+ = Iσ− = I we obtain in the limit of small speed and small
M8 Zeeman energy shift from a series expansion in lowest order
8h̄kδ I/I0 α
F (v, z) ≈ 2 (kv + βz) = −αv + βz . (4.2)
Γ k

2
1 + (2δ/Γ)

an expression similar to that obtained for the optical molasses, but now with
the effective detuning kv + βz. A velocity dependent force term and a spatially
dependent force term appear. In this way the atoms feel where the center of
the trap is and undergoes a damped motion to this position.

Damped harmonic oscillator description The motion of the atoms under


the influence of the force (4.2) is that of a damped harmonic oscillator
2
z̈ + γ ż + ωtrap z = 0, (4.3)

where γ = α/M , see (3.21)

8h̄k 2 δ I/I0
γ= 2 . (4.4)


2
1 + (2δ/Γ)
23

From the spatial dependence in Eq. (4.2) one obtains for the trap frequency
2 8h̄kβδ I/I0
ωtrap = 2 . (4.5)


2
1 + (2δ/Γ)

The characteristic motion in the trap is governed by the relation


γ2 h̄k 3 2δ I/I0
2 = i2 . (4.6)
4ωtrap βM Γ
h
2
1 + (2δ/Γ)

The motion is overdamped, if this ratio is greater than one.


For 2δ/Γ = −1 and I/I0 = 1 we find
γ2 πER
2 = (4.7)
4ωtrap λβh̄
This corresponds to the ratio of the recoil energy over the change of the Zeeman
energy over a distance of one optical wavelength. For a typical experimental
value, β = 1 mT/cm (14 MHz/cm), we find for the ratio (4.7) the value of 25
for N a and 2.5 for Cs. For these parameters the oscillation frequency in the
2
trap ωtrap = βvR corresponds to about 2π×1 kHz in sodium and the static trap
depth corresponds to 2 K at a distance of 5 mm from the trap center. The
actual trap depth appears deeper for a moving atom, as the atom experiences
Doppler-cooling when it moves. This MOT scheme works great in 3D as well.

Limitations A limit in the peak atom density in the trap appears due to ra-
diation trapping [24] (Strahlungseinschluss). Light scattered by trapped atoms
may excite other atoms with incorrect polarization when the cloud of cold atoms
becomes opaque. One way to circumvent this problem is to use a dark spot in
the trap center [25]. An additional problem appears as a consequence of accel-
eration in atomic collisions of excited state atoms with ground state atoms. If
spontaneous emission occurs during the collision process, the collision partners
may emerge with high kinetic energy and leave the trap.

In realistic level schemes the situation is more complex. In the example


below the cooling laser operates on the F = 2 → F 0 = 3 transition. The red-
detuning of the cooling laser from the cooling transition is greatly exaggerated
in this picture. In order to achieve trapping one needs to prevent permanent
population transfer into dark states by optical pumping.
One cannot avoid off resonant pump-
F ' = 3 F ' = 3
2 2
ing on the F = 2 → F 0 = 2 transition
1 1 (this transition is separated by typi-
0 0
cally a few times 100 MHz from the
cooling laser frequency), the ground
c o o lin g re p u m p e r state level, F = 1, will be populated
as well, albeit at a low rate. Atoms in
this level are not pumped by the cool-
ing laser as the ground state hyper-
F = 2 F = 2 fine splitting is typically in the GHz
1 1 range. Hence a re-pump laser is re-
quired to recall atoms from F = 1.
24 CHAPTER 4. MAGNETO-OPTIC TRAP

Applications MOTs are the most commonly used atom traps today. Initially
they were fed from Zeeman slowers. Monroe [26] realized that one could directly
capture atoms from the room temperature gas phase into the MOT. As the MOT
is isotopically selective one may use it to trap rare isotopes from a mixture [27].
The density can be made so high that collisions and radiation trapping are
limiting factors [24], the maximal density achieved is typically 1011 cm−3 .
2D MOTs are used as atomic funnels to generate beams with near zero
longitudinal speed [28, 29], as slow atom sources [30] or as atomic fountains.

Vapor-cell MOT Monroe [26] showed that the slow component of a room
temperature Maxwellian distribution can be directly captured into the damped
HO potential of a MOT . In this way he generated a cloud of 107 Cs atoms
in a volume of ≈ 1 mm3 . The atom number results from the balance between
capture rate and loss rate.
The normalized velocity distribution of atoms in the gas phase is
r
4 v2
 2
v 2kB T
f (v) = √ 3 exp − 2 where u= (4.8)
πu u M
is the most probable speed. We define a maximal capture velocity vc . The rate
of capture is given by the number of atoms with speeds below vc which enter a
sphere with volume V , and radius r :

1 vc 1 v4 v6
Z
n v f (v) dv × 4πr2 = √ c3 n × 4πr2 + O( c5 ) . (4.9)
4 0 πu u
The number of atoms which enter this sphere per second is given by the rate
 3/2
M
R = 0.68 n V 2/3 vc4 , (4.10)
2kB T

where n is the density of atoms in the vapor at the temperature T (300 K). The
volume of the sphere is V ≈ 0.1 cm3 .
The loss rate of captured atoms due to collisions with hot Cs atoms is given by
p
1/τ = nσhvi = nσ 8kB T /πM (4.11)

The density of trapped atoms in the stationary state Ns is obtained from the
balance of both rates
dNs Ns
=R− =0 (4.12)
dt τ
Since initially N (t = 0) = 0, the cold atom density builds up according to

N (t) = Ns (1 − exp(−t/τ )) (4.13)

and leads to the stationary density


2
V 2/3 4

M
Ns = Rτ = 0.68 v . (4.14)
σ c 2kB T
25

which is independent of pressure. The time of loading the trap of course depends
on the vapor pressure.
Which parameters define the critical capture velocity vc ? Typically one uses a
detuning from resonance by about the natural linewidth. In this case we have

Γ≈δ ≈k·v → vc = 2Γλ . (4.15)

This is the speed which leads to a Doppler detuning of one linewidth. In Cs


where Γ = 5 MHz, we have vc = 1500 cm/s. If we now consider a Maxwellian
distribution at 300 K we find that about 1 in 104 atoms is slow enough to be
captured.

Parameters in Monroe’s experiment [26] :


Cooling of Cs on the F = 4 → F = 5 transition.
Repumping on F = 3 → F = 4 transition.
Three orthogonal beams with 5 mm diameter, each with 2 mW.
The vertical magnetic field gradient at the center is 10 G/cm.
Cs vapor pressure: 6 × 10−9 Torr (−20o C).
Radiation trapping limits the density of cold atoms to ≈ 5 × 1010 cm3 .
The lowest temperature reached was 30 µK.
Loading times: t = 0.06 s at 10−7 Torr and t = 4 s at 6 × 10−9 Torr.

The picture below shows a sketch of Monroe’s trap. The vacuum vessel is made
of glass with a glass-to-metal transition to the ion pump.
26 CHAPTER 4. MAGNETO-OPTIC TRAP
Chapter 5

Coupling between 2 States

5.1 Time-Independent Coupling


We discuss a two-level system with the eigenfunctions |ϕ1 i = |1i and |ϕ2 i = |2i.
In the unperturbed case the system is described by the Hamiltonian H0 ,

H0 |ϕn i = En |ϕn i . (5.1)

The spectrum has two discrete values E1 and E2 .


We set E1 = 0 and E2 = h̄ω0 . In presence of a time-
independent perturbation W the Hamiltonian is E2 2 

H = H0 + W , (5.2)
 Ω0
where the term W couples the states |1i and |2i while
hϕn |W|ϕn i = 0 ist. We look for solutions to
E1 1 
H |ψi = E |ψi . (5.3)

In this case the perturbed state is

|ψi = c1 |1i + c2 |2i (5.4)

where c1 and c2 are costants. We insert Eq. (5.4) into (5.3)

c1 (H − E)|1i + c2 (H − E)|2i = 0 , (5.5)

and multiply from the left with the bras h1| and h2| respectively. If we consider
the orthonormality hϕn |ϕm i = δnm and define Hnm = hϕn |H|ϕm i we obtain

c1 (H11 − E) + c2 H12 = 0
c1 H21 + c2 (H22 − E) = 0, (5.6)

Here H12 = H21 = W with W = hϕ1 |W|ϕ2 i . Non trivial solutions for (5.6)
appear when the determinant of the coefficients c1 and c2 vanishes:

H11 − E W
=0 (5.7)
W H22 − E

27
28 CHAPTER 5. COUPLING BETWEEN 2 STATES

With H11 = E1 = 0, H22 = E2 = h̄ω0 we obtain the perturbed eigenvalues,


E = E± , M9
r
1 1
E± = h̄ω0 ± (h̄ω0 )2 + W 2 (5.8)
2 4

Typical solutions for (5.8) are shown here:

E + E +
E 2
E +

E 2
E 2
h M 0 h M 0
2 W 2 W
E 1
E 1

E 1
E -
, E
E E -
-

In
a )
the right figure we considerb ) the case that the energy gapc ) between the
unperturbed states varies along the x-axis according to h̄ω0 = x. The
perturbation W leads to a repulsion of the states from the unperturbed
energies E1 , E2 . The repulsion is larger, the smaller the energy gap h̄ω0 .
The energy gap between the perturbed levels is
p
E+ − E− = (h̄ω0 )2 + 4W 2 . (5.9)

In the center figure we consider degenerate unperturbed states. We obtain

E+ − E− = 2W . (5.10)

In the left figure the perturbation is very small compared to the unper-
turbed energy gap. For W  h̄ω0 the energy shift is

1 W2
∆E = (E+ − E− − h̄ω0 ) ≈ , (5.11)
2 h̄ω0
which corresponds to the result from stationary perturbation theory in
second order, as we required that hϕn |W|ϕn i = 0.

We obtain the perturbed wavefunctions by calculating the coefficients c1 and


c2 from (5.6) with the solutions E± from (5.8)
|ψ+ i = sin β |2i + cos β |1i
|ψ− i = cos β |2i − sin β |1i . (5.12)
Here β is the mixing angle defined as
2W 8
tan 2β = ≈ 2β + β 3 + O(β 5 ) ... (5.13)
h̄ω0 3
When the perturbation W is small then tan 2β ≈ 2β, and we have sin β ≈ β and
cos β ≈ 1 and obtain the result known from second order perturbation theory
W W
|ψ− i ≈ |1i − |2i und |ψ+ i ≈ |2i + |1i . (5.14)
h̄ω0 h̄ω0
5.2. 2-LEVEL ATOM IN THE FIELD OF A SINGLE MODE 29

In case of degenerate states, tan 2β = ∞, β = π/4, we have


1 
|ψ± i = √ |2i ± |1i , (5.15)
2
the perturbed states are coherent superpositions of the unperturbed states.

→ Practical applications are discussed in Appendix A-1 and A-7.

5.2 2-Level Atom in the Field of a Single Mode


With mode one marks an electromagnetic field of fixed frequency ωL and fixed
linear polarization ˆx [31]. The state vector of the two level system with the
ground state |gi and the excited state |ei is

|ψ(t)i = ce (t)|ei + cg (t)|gi . (5.16)

The equation of motion for |ψ(t)i is the time-dependent Schrödinger equation


d i
|ψi = − H|ψi (5.17)
dt h̄
with the Hamiltonian

H = H0 + H1 (t) . (5.18)

H0 describes the unperturbed atom and H1 (t) the time-dependent atom-laser


interaction. If we use the completeness relation

|eihe| + |gihg| = 1 (5.19)

we have
   
H0 = |eihe| + |gihg| H0 |eihe| + |gihg|
= h̄ωe |eihe| + h̄ωg |gihg| (5.20)
with the eigenvalue equations H0 |ei = h̄ωe |ei and H0 |gi = h̄ωg |gi, where the
atomic frequency is equal to ω0 = ωe − ωg .
We define the dipole matrixelement d~eg = |q| he|ˆ
x x|gi, where q is the ele-
mentary charge. The field term now is
H1 (t) ~
= − |q| ˆx xE(t)
   
= − |eihe| + |gihg| |q|ˆ ~
x x |eihe| + |gihg| E(t)

= − d~eg E(t)
~ ( |eihg| + |gihe| ) (5.21)
with the classical laser field

~ 1
= ˆx E0 cos ωL t = ˆx E0 eiωL t + e−iωL t .

E(t) (5.22)
2
We define the Rabi-frequency, Ω1 , as
|deg |E0
Ω1 = . (5.23)

30 CHAPTER 5. COUPLING BETWEEN 2 STATES

We obtain the time dependent amplitudes by inserting (5.16) into (5.17). Using
(5.20) and (5.21) and applying from left the bras he| and hg| respectively on
(5.17) we obtain
i
ċe = −iωe ce + Ω1 cg eiωL t + e−iωL t

2
i
ċg = −iωg cg + Ω1 ce eiωL t + e−iωL t .

(5.24)
2
We introduce the time-dependent functions
ce = Ce e−iωe t and cg = Cg e−iωg t (5.25)
and obtain
i h i
Ċe = Ω1 Cg ei(ω0 −ωL )t + ei(ω0 +ωL )t (5.26)
2
i h i
Ċg = Ω1 Ce e−i(ω0 −ωL )t + e−i(ω0 +ωL )t (5.27)
2
Neglecting the rapidly varying terms e±i(ω0 +ωL )t (rotating wave approximation),
we obtain the simplified equations
i i
Ċe = Ω1 Cg e−iδt and Ċg = Ω1 Ce e+iδt , (5.28)
2 2
where we introduced the detuning δ = ωL − ω0 . To solve we differentiate (5.28)
once and insert Ċg and Ċe from (5.28)
Ω2
C̈e −iδ C˙e − 1 Ce ,
=
4
2

C̈g = +iδ C˙g − 1 Cg . (5.29)
4
With the generalized Rabi frequency
q
Ω = Ω21 + δ 2 (5.30)

the solutions for (5.29) are


 
Ce (t) = a1 eiΩt/2 + a2 e−iΩt/2 e−iδt/2 ,
 
Cg (t) = b1 eiΩt/2 + b2 e−iΩt/2 e+iδt/2 . (5.31)

The constants of integration result from the initial conditions


1
a1,2 = [(Ω ± δ)Ce (0) ± Ω1 Cg (0)] ,
2Ω
1
b1,2 = [(Ω ∓ δ)Cg (0) ± Ω1 Ce (0)] . (5.32)
2Ω
M 10 With the initial condition Ce (0) = 0; Cg (0) = 1 we obtain
 
Ω1 Ωt
Ce (t) = +i sin e−iδt/2
Ω 2
 
Ωt iδ Ωt
Cg (t) = cos − sin e+iδt/2 (5.33)
2 Ω 2
where |Ce (t)|2 + |Cg (t)|2 = 1.
5.2. 2-LEVEL ATOM IN THE FIELD OF A SINGLE MODE 31

Populations We first consider the initial condition Ce (0) = 0 und Cg (0) = 1


and obtain for the population difference

Ω21 − δ 2
     
2 2 2 Ωt 2 Ωt
|Ce (t)| − |Cg (t)| = sin − cos (5.34)
Ω2 2 2

At resonance (δ = 0) we have Ω = Ω1 . With sin2 α − cos2 α = − cos(2α) we


obtain the oscillating result

|Ce (t)|2 − |Cg (t)|2 = − cos (Ω1 t) . (5.35)

In the general case we have for the population in the excited state

Ω21 − δ 2
   
Ωt
|Ce (t)|2 = sin2 (5.36)
Ω2 2

The excited state population (5.36) is


1.0 ∆=0
shown for Ω1 = 1, at δ = 0 (black), and
0.8 ∆=0.25
at a detuning δ = 0.5 (blue). The pop-
2

∆=0.5
CeHtL

0.6
ulations return to the ground state af-
0.4 M 11
ter a period t = 2π/Ω. In the presence
0.2
of detuning the excited state population
0.0
0 Π 2Π 3Π never reaches the value 1 and oscillates
time more rapidly, with the generalized Rabi
frequency Ω.

Polarization : The oscillating electric field induces the atomic dipole moment

P(t) = qhψ(t)|x|ψ(t)i = c∗e (t)cg (t) |deg | + c.c. (5.37)

For the real part of the dipole moment we obtain


        
iΩ1 Ωt iδ Ωt Ωt iωL t
P(t) = 2<e |deg | cos − sin sin e (5.38)
Ω 2 Ω 2 2

1.0 È ce * ce È 1.0 È ce * ce È
∆=0 È ce * cg È ∆=0.2 È ce * cg È
Re@ ce * cg D Re@ ce * cg D
0.5 0.5

0.0 0.0

M 12
-0.5 -0.5
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
time time
Coherence and population at δ = 0 and δ = 0.2 for deg = 1, ωL = 100,
Ω1 = 1. Note that under realistic conditions the ratio ωL /Ω1 is orders
of magnitude larger, for alkalis at I = I0 we have ωL /Ω1 ≈ 107 s.

→ For more on this formalism see Bloch picture, Appendix A-2.


32 CHAPTER 5. COUPLING BETWEEN 2 STATES

5.3 Spontaneous Emission


The above considerations do not acount for the process of spontaneous emission.
For stimulated emission the excitation energy and the recoil momentum are
found in a photon of the laser field. By contrast the spontaneously emitted
photon is released into a bath which is, under normal circumstances, decoupled
from the two-level system. In general there is no further interaction of the atom
with the bath photons, once the spontaneous photon has been emitted (unless
our atom is inside a resonator, suitable for storing spontaneous photons).
As a consequence in spontaneous emission there is an irrversible change of
momentum and energy of the atom-laser system. A picture is appropriate in
which the driven two level atom is stochastically interrupted by the spontaneous
process. When this happens all information content contained in coherence is
lost. Irreversible are the changes in momentum, the loss of energy and the loss
of coherence (decoherence). After the spontaneous process (quantum jump) the
atom finds itself in the ground state and begins anew with Rabi oscillations.
This stochastic interrupt can be shown in a Monte-Carlo simulation
of the temporal dependence of excited state population. Eq. (5.36).

Here we assume that the probability for


1.0
spontaneous emission is proportional to 0.8 1 atom
|Ce (t)|2 . We also assume that the two level 0.6
0.4
atom begins with Cg = 1, Ce = 0 after a 0.2
spontaneous event. 0.0
0 2 4 6 8
For a single atom we see the interrupt from
1.0
spontaneous emission at the times marked 0.8 10 atoms
by an arrow. 0.6
M 13 0.4
For several atoms we see that the oscilla- 0.2
tions are damped, the excitation probabil- 0.0
0 2 4 6 8
ity approaches a stationary state.
1.0
At resonance (δ = 0) and for high intensity 0.8 500 atoms
the stationary state approaches the value 0.6
1 1 0.4
2 , otherwise it is smaller than 2 . The pe- 0.2
riodic change of |Ce (t)|2 after turn-on is 0.0
0 2 4 6 8
known as optical nutation.
Chapter 6

Dressed States

In 1977 Cohen-Tannoudji and Reynaud [32] presented a framework for light-


atom interaction which has come to be known as dressed atom.1 The dressed
state approach provides an intuitive picture of light-atom interaction and allows
the quantitative treatment of stimulated emission and absorption.

The dressed atom approach is specifically suited in the limit of high laser
intensity. This can be understood from the following argument. For Rabi fre-
quencies Ω1 , which are large compared to the damping rate by spontaneous
decay, it is a good approximation to first consider energy levels of the combined
system: atom + laser photons together (the dressed states) and only later in-
troduce the spontaneous channel. In this approach one first defines uncoupled
atom-laser states which describe the total energy of atom + laser photons.
In a second step the interaction Hamiltonian H1 is considered. Its action is to
shift the energies of the dressed states. These new states (they are diagonal in
H1 ) are termed coupled dressed states. These are coherent superpositions
of ground and excited state of our atom, in a firm phase relationship with the
driving laser field(s). In a third step spontaneous emission (coupling with
the other modes of the radiation field, the bath) is taken into account. This
process appears as population transfer between the coupled dressed states.
In the dressed states scheme one can easily account for effects of the spatial
variation of the properties of the laser field. This is an important bridge to
realistic environments encountered in experiments:

• A focussed laser beam requires a description which accounts for the spatial
dependence of the Rabi-frequency.

• Superimposing two or more light fields typically means that the polariza-
tion properties of the resultant field show a spatial dependence.

The spatial variation of intensity and polarization are readily accounted for in
the dressed atom approach, even for multi-level atoms.

33
34 CHAPTER 6. DRESSED STATES

6.1 Uncoupled Atom-Laser States


We consider an atom with two stationary states, |gi and |ei at energies 0 und
h̄ω0 . The atomic Hamiltonian is

HA = h̄ω0 |eihe| + 0 |gihg| .

The atomic electric dipole operator is odd (ungerade). It couples only states of
opposite parity. We assume that the states |ei and |gi are of opposite parity
and that our two-level atom is embedded in a laser field with frequency ωL with
a detuning from resonance by δ = ωL − ω0 .
The electromagnetic field is quantized on to a complete set of orthonormal
~ λ (~r), one of which is the laser field, E
field distributions, E ~ L (~r). Here ~r is
the position vector, introduced to indicate the spatial dependence of the field
properties.2 The Hamiltonian of the free radiation field is
 
h̄ωλ a†λ aλ + 1/2
X
Hλ = (6.1)
λ

where a†λ and aλ are the creation- und annihilation-operators of a photon in the
mode λ. For the laser Hamiltonian we write
 
HL = h̄ωL a†L aL + 1/2 , (6.2)

where a†L and aL are the creation- und annihilation-operators for photons in
the laser mode. The number operator a†L aL applied to the eigenfunctions of HL
gives us the number of photons N in the laser mode.

ÈN+1\
We first define uncoupled atom-laser
states: We turn off the photon-atom Èe\ ΩL
interaction and dress the atomic basis
M 14 states with a given number of photons, Ω0 ÈN\
N . This is equivalent to trapping an
atom and N photons in a fictitious box Èg\ ΩL
and calculating the total energy con-
tained in this box. ÈN-1\

These dressed states are eigenstates of


HA +HL with two quantum numbers, |gi Èg,N+3\
EHN+2L
or |ei, and the number of laser photons Èe,N+2\
N . When the detuning is very small, δ = ΩL
ωL − ω0  ω0 , the states |g, N + 1i and ∆
|e, N i are close to each other (degenerate EHN+1L
Èg,N+2\ ¯
M 15
Èe,N+1\
for δ = 0). Along the energy axis these ­
two states form a pair in the manifold ΩL
of possible states, separated from each Èg,N+1\
EHNL
other by h̄δ, each pair separated from its Èe,N\
neighbors by the photon energy h̄ωL .
n o
We characterize each pair (family) by E(N ) = |g, N +1i, |e, N i .
2 In the following we suppress this explicit dependence on ~
r.
6.2. ATOM-LASER COUPLING 35

Note that the dressed states in each family have equal parity.3 However, they
differ in parity from that of the next neighbor pairs, E(N ± 1). The infinite
number of possible dressed states thus groups into families of alternating parity.

6.2 Atom-Laser Coupling


In the electric dipole approximation (atomic dipole moment d~ = q · ~x) we derive
a coupling from the energy of the atomic dipole in the external electric field of
the laser

VAL = −d~ · E
~L (6.3)

The dipole operator connects states of opposite parity, see Eq. (5.21),

d~ = d~eg (|eihg| + |gihe|) = d~eg b† + b



(6.4)

where d~eg = hg|d|ei


~ ~
= he|d|gi is assumed to be real and the b’s are ladder
operators for the atomic internal state. As we count the photon number in the
laser field, we introduce the quantized field expression [35]
r
h̄ωL  
~L =
E ~L aL + a†L , (6.5)
20 V

where V p
is the mode volume and ~L is the laser polarization. Note that the ex-
pression h̄ωL /20 V has the dimensions of an electric field. With the definition

r
h̄ωL
v=− ~L · d~eg (6.6)
20 V

we may write for the coupling term (6.3)


 
VAL = v b + b† aL + a†L . (6.7)

The operator aL destroys a photon in the laser mode, the operator a†L creates a
photon. The operator b takes our atom from the excited state into the ground
state, the operator b† takes the atom from the ground state into the excited
state. The term VAL in (6.7) contains four contributions,

• a†L b describes the transition from the excited into the ground state, con-
current with creation of a photon in the laser mode (stimulated emission).

• aL b† describes the transition from ground to excited state, concurrent


with the loss of one photon from the laser mode (absorption).

• The terms aL b und a†L b† do not obey energy conservation and are not
considered further.
3 This is because in each pair the states of opposite parity, |ei and |gi, are dressed with a

number of photons that differ by One.


36 CHAPTER 6. DRESSED STATES


In the HO description
√ of the laser mode we have a†L |N i = N + 1|N + 1i and
aL |N + 1i = N + 1|N i. Hence we can write for the coupling matrix element
for stimulated absorption between the states in each family, E(N ),

he, N |VAL |g, N +1i = v he, N | b† aL |g, N +1i




= v N +1
p
≈ v hN i . (6.8)

An identical expression appears for stimulated emission, hg, N + 1|VAL |e, N i.


Here we have used the approximation that many photons are present in the
~0
laser mode, N + 1 ≈ hN i. In this case we can write for the field amplitude E
~
and for the intensity I of the classical laser field, E(t) ~ 0 cos ωL t,
=E
p √
|E0 | ∝ hN i/V ∝ I

and relate these quantities with those of the quantized field [36]
r
~ h̄ωL p
E0 = 2~L hN i . (6.9)
20 V
With the definition of the Rabi-frequency

|d~ge E~ 0|
Ω1 = (6.10)

we obtain for the interaction terms (6.8)
1
he, N |VAL |g, N +1i = h̄Ω1
2
1
hg, N +1|VAL |e, N i = h̄Ω1 . (6.11)
2
Next we consider the consequences of the interaction terms (6.11). In the
absence of interaction, the states of each pair in a family are spaced by the laser
detuning from resonance δ = ωL − ω0 . If we introduce an energy scale relative
to the energy (N+1) h̄ωL − h̄δ/2 we can write for the Hamiltonian of the dressed
states of each family
 
+δ Ω1 (r)
h̄ 
HDA = . (6.12)
2 Ω1 (r) −δ

Diagonalizing this basis of uncoupled states we obtain the family of coupled


dressed states which we label |1, N i und |2, N i. Their energies are at
1
U|1,N i = + h̄Ω (6.13)
2
1
U|2,N i = − h̄Ω . (6.14)
2
Note that this result is identical to that which we had discussed in Eq. (5.9).
Turning on the interaction between the states of each pair, their energy spacing
increases to
6.2. ATOM-LASER COUPLING 37

q
h̄Ω = h̄ Ω21 + δ 2 . (6.15) È1,N\

The magnitude of this splitting corre- Èg,N+1\

sponds to the generalized Rabi-frequency M 16


∆ W
of Equation (5.30). This energy shift,
(Ω − δ)/2, is often termed ac-Stark Èe,N\
shift (light shift, Wechselfeld-Stark Ver-
È2,N\
schiebung).

It is important to realize that the coupled dressed states, |1, N i und |2, N i, are
now superpositions of the uncoupled dressed states, see (5.12),

|1, N i = sin β |g, N +1i + cos β |e, N i


|2, N i = cos β |g, N +1i − sin β |e, N i , (6.16)

where the mixing angle β is given as


tan 2β = Ω1 /δ , È1,N\ Èg,N+1\

sin 2β = Ω1 /Ω , or Èe,N\
energy

cos 2β = −δ/Ω . Ωo
W1
M 17
The figure shows the perturbed energy È2,N\
of the two states of a family by the full
lines as a function of the detuning. The 0
noninteracting dressed states are shown detuning ∆
by the dashed lines.
For a detuning of δ = 0 we have β = π/4 (complete mixing), for Ω1 = 0 we
obtain β = π/2 (unperturbed states). The light-shifting of dressed states may
be projected back into the conventional energy representation of the free atom.
From this we deduce the following light shift of atomic energies:

∆<0 ∆>0

Èe\ Èe\
: Ω0 ΩL >: Ω0 ΩL > M 18
Èg\ Èg\
VAL = 0 VAL ¹ 0 VAL = 0 VAL ¹ 0

• For red detuning, δ < 0, the states |gi und |ei are energetically pushed
apart, the potential energy of the ground state decreases whereas the
potential energy of the excited state increases.

• The opposite effect occurs in the case of blue detuning: The interaction
pushes the states |gi und |ei closer to each other when δ > 0. In other
words the potential energy of the ground state increases whereas the po-
tential energy of the excited state decreases.
38 CHAPTER 6. DRESSED STATES

Note that these new states are superpositions of |gi und |ei, they each contain
character of both, |gi und |ei, to a degree given by the mixing angle.

In general the light shift is tiny! To see its magnitude we choose a strong optical
transition with a dipole moment of magnitude of 1 debye = 3.33 × 10−30 C m.
(A dipole moment of 1 debye corresponds to two elementary charges separated
by 20 pm. This definition was made when CGS units were in use.) At an
electric field strength of 1 V/cm (corresponds to I ≈ 1.3 mW/cm2 ) we have a
Rabi-frequency, Ω1 = 3 MHz, corresponding to an energy shift at resonance
of h̄Ω1 = 2 neV ≈ 20 µK . In the context of cold atoms, when translational
energies are in the µK range and below, these small energy shifts translate into
Himalayan-like potential energy landscapes an atom has to navigate through,
on length scales of the size of a laser focus or the wavelength of light.

6.3 Effects Of Spontaneous Emission


We have not yet accounted for effects of spontaneous emission of photons into
the empty modes of the radiation field. This coupling is essential for various
sub-Doppler cooling effects, which enter via an imbalance in the intensity of the
side bands of the so-called Mollow triplet. Before introducing the Mollow triplet
we consider the effect of spontaneous emission in the dressed state picture.
We show the dressed-state energy ladders by separating each family pair
such as shown below. Here all states built by dressing the ground state are col-
lected on the left and all states built by dressing the excited state are collected
on the right hand side. As the laser frequency is much larger than the detuning,
each family appears practically degenerate on this scale.

The interactions (6.11) describe stim-


ulated absorption and emission. They Èg,N+1\ Èe,N\
are indicated by blue horizontal arrows.
These connect the states of each family.
Èg,N\ Èe,N-1\
M 19
In the uncoupled case the only spon-
taneous transitions permitted connect
|e, N i to |g, N i, |e, N −1i to |g, N −1i, Èg,N-1\ Èe,N-2\
etc. as indicated by the wiggly lines.

When we switch from the uncoupled dressed states to the coupled basis
each state contains both |gi as well as |ei character.

As a consequence there will be tran- Èg,N+1\ È1,N\


sitions from each state of a each fam-
ily E(N )-pair into each state of the Èe,N\ È2,N\
lower-lying family. Two of the four
M 20 ΩL
transitions are degenerate. The re-
sulting fluorescence spectrum now com-
prises three individual transitions, the Èg,N\ È1,N-1\
so called Mollow Triplet [33]. ∆ W
Èe,N-1\ È2,N-1\
6.3. EFFECTS OF SPONTANEOUS EMISSION 39

A quantitative estimate of the intensity distribution in the Mollow-Triplet


may be deduced from the following consideration. According to Equation (6.16)
the coupled states |1, N i and |2, N i are mixtures of |g, N i and |e, N − 1i respec-
tively. This is also true for the members of the lower-lying pair of coupled states
|1, N − 1i and |2, N − 1i.
The strength of the transitions in the Mollow-triplet is obtained by weighting
the state amplitudes +cos β |e, Ni and −sin β |e, Ni of the excited state of family
E(N ) with the amplitudes +sin β |g, Ni and cos β |g, Ni of the ground state of
family E(N − 1) and finally also with the steady state populations in the two
states. The transition energies and relative line strengths of the four optical
transitions are given by
2
+Ω + ωL with (cos β cos β) (a)
2
ωL with (cos β sin β) (b)
2
−Ω + ωL with (sin β sin β) (c)
2
ωL with (sin β cos β) (d)
In steady state the sidebands (a) and (c) appear with equal intensity at ωL ± Ω,
in addition to the strong central line at ωL , see Equ. (8.24) and (8.25). The
magnitude of the side bands depends on the Rabi-frequency which controls the
mixing angle β. At low laser intensity the transitions (a) and (c) are weak. In
the strong mixing case (β = π/4) the sidebands are each half of the height of
the central line.

Two so-called Sisyphus-cooling mechanisms (Chapters 8 und 11) derive


from an imbalance of the two sidebands of the Mollow triplet. Such an imbalance
may occur for slowly moving atoms. When the side bands in the the Mollow
triplet are balanced in intensity, stimulated emission and stimulated absorption
occur with equal rates and no net momentum transfer occurs.

Autler-Townes profile :

Èp,N\
Èp,N\
When a strong laser field couples the
transition |gi to |ei, a weak laser
probing the transition from some Ωp
other state, |p, N i, at photon energies M 21
near ωp , will experience absorption at
the energies of both coupled states of Èe,N\ È1,N\
the family, {|1, N i, |2, N i}. ∆ W
Èg,N+1\ È2,N\

The Rabi-splitting can be observed in transitions from or to state |p, N i.


Note that this state may have the parity of either state, |ei or |gi.
The splitting Ω of the spectral peak is termed Autler-Townes doublet.
40 CHAPTER 6. DRESSED STATES
Chapter 7

Density Matrix for a


Two-Level Atom

Dissipation and decoherence describe effects which stochastically change


a quantum state. Examples are spontaneous emission or collisions which
change populations, momentum or merely the phase of an atom’s state
vector. These effects lead to loss of information about the system. They can
be accounted for in a quantitative manner in the density matrix formalism.
Dissipation usually refers to the case that energy of the quantum system is
irreversibly lost to an environment (friction). An analogous process is thermal
excitation of the quantum system by contact with fluctuating forces from an
incoherent bath. On the other hand coherent effects displayed by the quan-
tum system, such as interference in space and time, are gradually destroyed
by contact of the quantum system with an environment. This effect is known
as decoherence or dephasing. It may occur without dissipation, however when
dissipation occurs in a coherently driven system, decoherence is always present.
In Chapter 5.2 we considered the coherent evolution of a two-level system
in an external driving field in the absence of dissipation or decoherence. This
example intuitively conveys the meaning of the elements of the density matrix
and its expansion to include effects of spontaneous emission.

7.1 Pure States and Density Operator


If we possess complete information about a quantum state we describe it with
a state vector |Ψi. In this case we talk about a pure state. We first discuss the
pure state of a driven two-level system. Our atom has two stationary states, an
excited state at energy Ee = h̄ωe and a ground state Eg = h̄ωg . The respective
wave functions, |ei and |gi, are normalized and orthogonal
he|ei = hg|gi = 1 and he|gi = hg|ei = 0 . (7.1)
The closure relation for this basis is |eihe|+|gihg| = 1 . Any possible state vector
of this system can be written as
|Ψi = cg |gi + ce |ei . (7.2)

41
42 CHAPTER 7. DENSITY MATRIX FOR A TWO-LEVEL ATOM

The coefficients give the probability amplitudes for finding the system in state
|ei and |gi respectively. The expectation value of an observable A is
hAi = hΨ|A|Ψi = |ce |2 Aee + |cg |2 Agg + c∗e cg Aeg + c∗g ce Age . (7.3)
The total Hamiltonian contains one part for the free atom and a second part
describing the interaction with an external field
H = HA + H1 . (7.4)
The free atom Hamiltonian is1
HA = h̄ωe |eihe| + h̄ωg |gihg| (7.5)

Free two-level atom


The temporal development of the free atom is given as
d i i  
|Ψi = − HA |Ψi = − h̄ωe |eihe| + h̄ωg |gihg| |ψi . (7.6)
dt h̄ h̄
Inserting the state vector (7.2) in (7.6) and multiplying from the left, once with
he| and once with hg|, we obtain
ċe = −i ωe ce and ċg = −i ωg cg . (7.7)
Solutions to (7.7) are, apart from a global phase eiφ
ce (t) = e−i ωe t ce (0) and cg (t) = e−i ωg t cg (0) . (7.8)
Hence the temporal development of the state vector of the free atom is
|Ψ(t)i = e−i ωg t cg (0) |gi + e−i ωe t ce (0) |ei . (7.9)

Coherently driven two-level atom


1 −iωL t
Including the action of the external laser field E = 2 E0 (e + e+iωL t ) and
defining the Rabi-frequency
Ω1 = |deg |E0 /h̄ (7.10)
we have for the interaction term
h̄Ω1 −iωL t  
H1 = − (e + e+iωL t ) |eihg| + |gihe| . (7.11)
2
As we discussed in chapter 5 solutions may be found in terms of the rotating
wave approximation. With the initial conditions ce (0) = 0 and cg (0) = 1 solutions
for the time evolution are apart from a global phase eiφ , see Eq. (5.33)
 
−i ωe t Ω1 Ωt −iδt/2
ce (t) = e +i sin e
Ω 2
 
−i ωg t Ωt iδ Ωt +iδt/2
cg (t) = e cos − sin e . (7.12)
2 Ω 2
p
Here δ = ωL − ω0 , ω0 = ωe − ωg , and Ω = Ω21 + δ 2 . Eq. (7.12) describe
coherent Rabi oscillations of population and polarization of the two-level atom
(see page 31).
1 How do we get this form? We multiply H from left and right with the closure relation
A
: (|eihe| + |gihg|) HA (|eihe| + |gihg|) = h̄ωe |eihe| + h̄ωg |gihg| .
7.1. PURE STATES AND DENSITY OPERATOR 43

Description by the density operator


The relationship (7.3) for the expectation value hAi = hΨ|A|Ψi = nm c∗n cm Anm
P
contains the state amplitudes in quadratic form c∗n cm . They are the matrix el-
ements of the operator |ΨihΨ|, which is the projector on the state vector |Ψi
which we represented in some basis {|un i},
hum |ΨihΨ|un i = c∗n cm . (7.13)
For this reason it appears natural to introduce the density operator
ρ = |ΨihΨ| . (7.14)
The density operator (the density matrix) is built from the elements
ρmn = hum |ρ|un i = c∗n cm . (7.15)

Properties of the density operator


• The density matrix is Hermitian2 ρ∗mn = c∗m cn = ρnm .
• The specification of ρ is sufficient to characterize all physical predictions
of our quantum system, for example the conservation of probability
X 2
X
|cn | = ρnn = Tr ρ = 1 . (7.16)
n n

• The expectation value of an observable A is M 22


X X
hAi = hΨ|A|Ψi = c∗n cm Anm = ρnm Anm = Tr{ρA} (7.17)
nm nm

• For the equation of motion of the density operator we obtain the general-
ized Schrödinger equation (von-Neumann equation)
   
d d d d
ρ= (|ΨihΨ|) = |ψi hψ| + |ψi hψ|
dt dt dt dt
i i
= − H |ψihψ| + |ψihψ| H
h̄ h̄
i h i
= − H, ρ (7.18)

• Two additional properties appear for a pure state : M 23

ρ2 = ρ Tr (ρ2 ) = 1 .

• For a mixed state we have Tr (ρ2 ) < 1 .


While the description of a pure state in terms of a state vector and in terms
of the density operator are equivalent there are certain advantages when using
the density operator. For one, the (unobservable) global phase is absent in the
density matrix approach and secondly the calculation of an expectation value is
linear in case of the density operator but quadratic in case of the state vector.
2 The diagonal terms are real, the real parts of the matrix are symmetric: Re(ρ ) = Re(ρ ),
ij ji
and the imaginary parts of the matrix are antisymmetric: Im(ρij ) = −Im(ρji ).
44 CHAPTER 7. DENSITY MATRIX FOR A TWO-LEVEL ATOM

Explicit density operator for the pure state of a two-level atom


In the absence of stochastic effects the state vector of the two-level atom

|Ψ(t)i = ce (t)|ei + cg (t)|gi (7.19)

represents a pure state. Its density operator


  
ρ = |ΨihΨ| = ce |ei + cg |gi c∗e he| + c∗g hg| (7.20)

has the matrix elements

ρee = he|ρ|ei = he|ψihψ|ei = |ce |2


ρgg = hg|ρ|gi = hg|ψihψ|gi = |cg |2
ρeg = he|ρ|gi = he|ψihψ|gi = ce c∗g = ρ∗ge (7.21)

In matrix form it is written as


 
ρee ρeg
ρ= . (7.22)
ρge ρgg

The diagonal elements describe the probability (population) of the atom to be in


the excited or ground state, respectively. The off-diagonal elements, ρeg and ρge ,
describe the atomic coherence. Using the dipole matrix element deg = he|x|gi,
we may write for the atomic polarization3

pA = hq xi = q hΨ|x|Ψi = ρge deg + ρeg deg . (7.23)

The atomic Hamiltonian (7.5) is in matrix form


 
h̄ωe 0
HA = . (7.24)
0 h̄ωg

With (7.22) we obtain for the commutator in (7.18)

1 h i  0 −i(ωe − ωg )ρeg

HA , ρ = (7.25)
ih̄ i(ωe − ωg )ρge 0

and hence for the evolution of the density matrix elements

ρ̇ee = ρ̇gg = 0
ρ̇eg = −i (ωe − ωg ) ρeg . (7.26)

The solutions are equivalent to what we found in Equation (7.8)

ρee (t) = const. ρgg = const.


−i (ωe −ωg )t
ρeg (t) = e ρeg (0) . (7.27)

When including the interaction Hamiltonian, solutions equivalent to (7.12) are


obtained which we discuss in the context of the Bloch vector (Appendix A-2).
3 Here we assume that the states |ei and |gi are of opposite parity.
7.2. SPONTANEOUS EMISSION IN TWO-LEVEL ATOM 45

7.2 Spontaneous Emission in Two-Level Atom


Inclusion of spontaneous emission transforms the coherently driven atom into a
dissipative quantum system with decoherence. This quantum system is now
open, it exchanges energy, polarization, and momentum with its environment
and coherence imprinted in the system by the coherent driving term H1 suffers
decoherence when stochastic emission of a spontaneous photons appears. The
process of spontaneous emission is usually irreversible.4
This situation is handled by introducing an
unobserved reservoir R which couples to the
atom via an interaction VAR . The reservoir
quantum VAR
is assumed to have such a large number of system
reservoir M 24
degrees of freedom that no back action on the
atom occurs and that the reservoir effectively
remains in equilibrium.
By contrast the atom has only a limited number of degrees of freedom. Also
present is a coherent radiation field which interacts with the atom via H1 .
The laser field itself is described by the Hamiltonian HL , the reservoir by the
Hamiltonian HR , the exact Hamiltonian being

Hex = HA + HL + H1 + HR + VAR . (7.28)

If the reservoir remains unobserved and if (due to the huge size of the reservoir)
there is no retroactive effect from the reservoir on to the atom, we may neglect
the term HR and treat the spontaneous emission VAR in a phenomenological
relaxation matrix. We also assume that the photon number in the laser
mode is so large that stimulated absorption and stimulated emission leave the
photon number in the laser mode unchanged. In this case we stay with the
approximation to the Hamiltonian used so far

H ≈ H A + H1 . (7.29)

Phenomenological approach to dissipation


In the Schrödinger picture we can account for dissipative effects in a phenomeno-
logical manner by introducing an additional term in Equation (5.28)

i Γ
Ċe = Ω1 Cg e−iδt − Ce . (7.30)
2 2
Without coupling to a laser field (Ω1 = 0) the relaxation term describes the
exponential decay of the excited state

|Ce (t)|2 = |Ce (0)|2 e−Γt . (7.31)

In the Master Equation (7.18) this relaxation brings about the additional term

ρ̇ij = (ρ̇ij )no dissipation + (ρ̇ij )dissipation (7.32)


4 De facto the spontaneous emission process is reversible: a photon emitted from an excited

atom into the empty reservoir of vacuum modes might be reabsorbed. However this can only
happen with significant probability if the atom is embedded in an appropriate resonator.
46 CHAPTER 7. DENSITY MATRIX FOR A TWO-LEVEL ATOM

For population elements this expression gives the exponential decay ρii ∝ exp(−Γt).
If Γ describes the spontaneous decay rate, then ρii Γ is equal to the rate of emis-
sion of spontaneous photons from state i. In our two-level atom the excited
state will loose population

ρ̇ee = −Γρee , (7.33)

which is gained by the ground state:

ρ̇gg = +Γρee . (7.34)

Due to this decay the off-diagonal terms decay with a rate [31]

Γ
ρ̇eg = − ρeg . (7.35)
2
These loss and gain terms result automatically when introducing the Liouvillian.

7.3 Liouvillian
Dissipation and decoherence is formally introduced in the Master Equation in
the so-called Liouvillian, L, in Lindblad form [37] using the atomic transition
operators b = |gihe| (transition of atom from excited to ground state) and
b† = |eihg| (transition from ground state to excited state)

d i i Γ †
ρ b b + b† b ρ − 2b ρ b† .

ρ = − [H, ρ] + Lρ = − [H, ρ] − (7.36)
dt h̄ h̄ 2
We abbreviate (7.36) by defining an energy-conserving and a dissipative part
   
d d d
ρ= ρ + ρ . (7.37)
dt dt 0 dt dis

We use this formalism in the following to derive solutions to the Master Equation
of the coherently driven two-level atom under inclusion of spontaneous emission.
The resulting equations (Optical Bloch Equations) provide a realistic description
of the two-level atom and serve to explain forces experienced by the two-level
atom in an external laser field. They also provide a microscopic explanation of
the susceptibility of a dielectric medium.

7.4 Solutions for Two-Level Atom


For an external laser field E = 12 E0 (e−iωL t + e+iωL t ) with the Rabi-frequency
M 25 Ω1 = |deg |E0 /h̄ we have the interaction Hamiltonian

h̄Ω1 −iωL t  
H1 = − (e + e+iωL t ) |eihg| + |gihe| . (7.38)
2
If we set the energy reference at the ground state level (h̄ωg = 0), the atomic
Hamiltonian is (ω0 = ωe − ωg )

HA = h̄ω0 |eihe| (7.39)


7.4. SOLUTIONS FOR TWO-LEVEL ATOM 47

Using H = HA + H1 we obtain
˙ 0 = +(i/2)Ω1 (e−iωL t +e+iωL t ) (ρeg −ρge ) ,

he|ρ̇0 |ei = ρee
˙ 0 = −(i/2)Ω1 (e−iωL t +e+iωL t ) (ρeg −ρge ) ,

hg|ρ̇0 |gi = ρgg
˙ 0 = +(i/2)Ω1 (e−iωL t +e+iωL t ) (ρee −ρgg ) − iω0 ρeg ,

he|ρ̇0 |gi = ρeg
˙ 0 = −(i/2)Ω1 (e−iωL t +e+iωL t ) (ρee −ρgg ) + iω0 ρge .

hg|ρ̇0 |ei = ρge
Accounting for spontaneous emission of the excited state with a rate5 Γ we
obtain the dissipative and decoherence terms

he|ρ̇dis |ei = ρ̇ee dis = −Γ ρee ,

hg|ρ̇dis |gi = ρ̇gg dis = +Γ ρee ,

he|ρ̇dis |gi = ρ̇eg dis = −(Γ/2) ρeg ,

hg|ρ̇dis |ei = ρ̇ge dis = −(Γ/2) ρge .
We may suppress the explicit time dependence by introducing the rotating basis
(rotating wave approximation, RWA):
ρ̃eg = ρeg e+iωL t ρ̃ge = ρge e−iωL t
ρ̃gg = ρgg ρ̃ee = ρee (7.40)
Neglecting the high frequency terms e±2iωL t we obtain with δ = ωL −ω0
i  
ρ̃˙ ee = + Ω1 ρ̃eg − ρ̃ge − Γ ρ̃ee
2
i  
ρ̃˙ gg = − Ω1 ρ̃eg − ρ̃ge + Γ ρ̃ee
2
i  
ρ̃˙ eg = + Ω1 ρ̃ee − ρ̃gg + iδ ρ̃eg − (Γ/2) ρ̃eg
2
i  
ρ̃˙ ge = − Ω1 ρ̃ee − ρ̃gg − iδ ρ̃ge − (Γ/2) ρ̃ge (7.41)
2
These equation are also known under the term optical Bloch-Equations. They
describe the coherently driven two-level atom in analogy to the treatment in
Chapter 5 but include dissipation. The so called Bloch-vector graphically il-
lustrates the evolution of the coherently driven two-level atom. This topic is
discussed in Appendix A-2.

Stationary state solutions


In the stationary case (ρ̇ij = 0) we obtain from (7.41) for the populations
(ρ̃ee = ρee , ρ̃gg = ρgg )
Ω21
ρee = ρgg = 1 − ρee (7.42)
4δ 2 + Γ2 + 2Ω21
and the coherences (ρ̃eg = ρeg e+iωL t , ρ̃ge = ρge e−iωL t )
2Ω1 (δ − iΓ/2) ∗
ρ̃eg = − ρ̃ge = ρ̃eg (7.43)
4δ 2 + Γ2 + 2Ω21
These results permit the derivation of important equations for light forces and
for the susceptibility of a two level atom.
5Γ = ω03 |deg |2 /(3π0 h̄c3 ), also termed Einstein-A coefficient.
48 CHAPTER 7. DENSITY MATRIX FOR A TWO-LEVEL ATOM

7.5 Spontaneous Force


With the definition of the saturation intensity I0 (3.11)

2Ω21 I
= (7.44)
Γ2 I0
we rewrite the population of the excited state as
1 I/I0
ρee = · . (7.45)
2 1 + I/I0 + 4δ 2 /Γ2
The rate of scattering of spontaneous photons of an atom at rest is
Γ I/I0
Γscat = Γ ρee = · . (7.46)
2 1 + I/I0 + 4δ 2 /Γ2
This is equal to the expression which we had derived from the classical rate-
equation model in section (3). The spontaneous force is obtained by multiplying
the scattering rate with the photon momentum
Γ I/I0
F~spontan = h̄~k Γ ρee = h̄~k · . (7.47)
2 1 + I/I0 + 4δ 2 /Γ2
Note that this so-called spontaneous force does not describe the momentum
transfer due to spontaneous emission, but rather the momentum transfer which
occurs prior to spontaneous emission, when the atom undergoes stimulated ab-
sorption from the directed laser beam with wave vector ~k. The momentum
vector in spontaneous emission is taken to be zero, on average, based on the
assumption that spontaneous photons are emitted isotropically.

7.6 Susceptibility
In Appendix A-6 we consider the propagation of an electromagnetic wave in a
medium and interpret the relationship between the real and imaginary part of
the susceptibility (χ = χ0 + i χ00 ) with the refractive index and the coefficient of
absorption of the medium
1 ω 00
nr = 1 + χ0 and α= χ . (7.48)
2 c
In the following we attempt a microscopic description of the medium by relating
refractive index and absorption coefficient to properties of the two-level atom.

Atomic basis for the susceptibility


The polarization of an atomic gas is taken to be the product of the atom density
N and the atomic polarization P0 = N pA . For the two-level atom the atomic
polarization is the product of dipole moment deg and the coherence ρ̃eg

p̃A = |deg | (ρ̃eg + ρ̃ge ) = |deg | (ρeg e+iωL t + ρ∗eg e−iωL t ) . (7.49)

The envelope of the polarization density of the atom is given as

P0 = N |deg | ρeg .
7.7. DIPOLE FORCE 49

In Appendix A-6, Eq. (A-6.8), we have defined P0 = 0 χ E0 . Thus we have


a relationship between susceptibility and coherence

|deg | |deg |2
χ = 2N ρeg = 2N ρeg . (7.50)
0 E0 0 h̄ Ω1
The complex coherence (7.43) implies a complex susceptibility

χ = χ0 + i χ00 (7.51)

where
|deg |2 2δ
χ0 = −N (7.52)
0 h̄ 4δ 2 + Γ2 + 2Ω21
|deg |2 Γ
χ00 = +N . (7.53)
0 h̄ 4δ 2 + Γ2 + 2Ω21
For small Rabi frequency Ω1  Γ, the last term in the denominator can be
neglected and we obtain a susceptibility which is independent of the intensity of
the pump laser. The term 2Ω21 in the denominator is origin for power broadening.

0.2
Im! Χ"$ Χ"
Re! Χ"$ Χ'

0.2
0

!.2
0
!3 !2 !1 0 1 2 3 !3 !2 !1 0 1 2 3
∆ ∆

7.7 Dipole Force


The potential energy of an atomic dipole in an external field can be written as
~ r) .
Udip (~r) = − p~A · E(~ (7.54)

If we equate the atomic dipole moment p~A to the polarization induced by the
laser field (7.49) we obtain with the definition of the Rabi frequency, Ω1 =
|deg |E0 /h̄,
   
h̄ Ω1 (~r)
Udip (~r) = −|deg | ρ̃eg + ρ̃ge E0 (~r) = − ρ̃eg + ρ̃ge . (7.55)
2

Using the expression for the coherences (7.43)

2Ω1 (δ − iΓ/2) ∗
ρ̃eg = − ρ̃eg = ρ̃ge (7.56)
4δ 2 + Γ2 + 2Ω21

we obtain for the dipole potential

h̄ δ Ω21
Udip (~r) = . (7.57)
4δ 2 + Γ2 + 2Ω21
50 CHAPTER 7. DENSITY MATRIX FOR A TWO-LEVEL ATOM

The spatial dependence of Udip (~r) typically arises usually from the spatial distri-
bution of laser intensity, I(~r) ∝ Ω21 (~r). For large detuning the dipole potential
(7.57) is equivalent to the results in Eq. (6.15) and Eq. (5.11)

h̄ Ω21 (~r)
Udip (~r) ≈ . (7.58)

With the saturation intensity from Equation (3.11) we may write

h̄ Γ2 I(~r)
Udip (~r) ≈ . (7.59)
8δ I0
Rather different looking equations are found in the literature. One of them
appears as Eq.(1) in [6]

2Ω2
 
h̄ δ
Udip = ln 1 + 2 1 2 . (7.60)
2 4δ + Γ

Various expressions for the dipole potential are compared in the figure below.

0.2 Ñ∆ W1 2
ln@1+ D
0.1 4 ∆2

Udip 0.0

-0.1
ÑW21
-0.2 4∆
M 26 -3 -2 -1 0 1 2 3
detuning ∆  W1
Predictions of Eq. (7.57) (red) and Eq. (7.60) (green) for the dipole po-
tential for Ω1 = 1, Γ = 1. The blue curve is an alternate expression
sometimes found in the literature. The dashed line is the energy of the
coupled dressed state |1, N i in the limit Ω1 /δ  1 , Eq. (7.58).
For the red, blue and green curves, starting from δ = 0, the potential
initially grows with detuning, δ, reaches a maximum and then falls off for
larger detuning, the overall sign of the potential depending on the sign
of δ. The four equations agree in their prediction at large detuning.

For Ω1  δ or Ω1  Γ, the force acting on the atom in the dipole potential is


h̄δ
F~ = −∇U
~ dip ≈ − ~ 21 .
∇Ω (7.61)
4δ 2 + Γ2
Chapter 8

Dipole Force and


Dissipation

In 1985 Dalibard und Cohen-Tannoudji [17] used the dressed-atom ap-


proach to analyze the dipole force. In this framework atomic energy levels
experience a light shift which is proportional to the square of the Rabi-
frequency if δ  Ω1 . In spatially inhomogeneous fields, as they appear
in focused or standing-wave laser beams, the Rabi frequency depends on
the spatial position of the atom. The spatial gradient of the light-shifted
energy can then be related to the dipole force.
This treatment uncovered dissipative terms which arise from an imbalance
of the sidebands of the Mollow triplet. A slowly-moving atom may dissi-
pate translational energy in the form blue detuned spontaneous photons
when moving in a spatially inhomogeneous laser field.
Dalibard und Cohen-Tannoudji, in the following referred to DCT, make the as-
sumption that the atomic wavepacket is small compared to the laser wavelength,
∆r  λ, and that the Doppler dispersion is small in comparison to the natural
width of the excited state,

k ∆v  Γ . (8.1)

The uncertainty relationship requires M ∆r ∆v ≥ h̄. These assumptions are


fulfilled if the recoil energy is small compared to the natural width of the excited
state,

h̄2 k 2
ER =  h̄Γ . (8.2)
2M

from (8.1) ∆v  1 m/s k 8 × 106 m−1


h̄/M ∆v ∆r < 1 nm λ 780 nm
ER 3.5 kHz Γ 2 π×6 MHz

Under these conditions DCT consider a two-level atom with momentum P at


the position r (which in general should be written as vectors). In the absence

51
52 CHAPTER 8. DIPOLE FORCE AND DISSIPATION

of spontaneous emission the total Hamiltonian is


H(r) = HA + HL + VAL . (8.3)
With the projectors b† = |eihg|, b = |gihe| the atomic Hamiltonian is
P2 P2
HA = + h̄ω0 b† b = + h̄ω0 |eihe| . (8.4)
2M 2M
The laser Hamiltonian is
 
HL = h̄ωL a†L aL + 1/2 , (8.5)

where a†L , aL are the creation and annihilation operators for a photon in the
laser mode. The expectation value a†L aL gives us the number of photons in the
laser mode (N ). The atom couples to the quantized electromagnetic field EL (r)
in stimulated one-photon absorption and emission (d = dipole matrix element)
 
VAL (r) = −d EL (r) b† aL + b a†L . (8.6)

8.1 Hamiltonian of Dressed States


If the atom-laser coupling is much stronger than the atom-vacuum coupling,
we may neglect spontaneous emission in a first step of our treatment. Initially
we consider a stationary atom (v = 0). The sum of internal energy, the energy
of the laser field and the atom-laser coupling can be written in a dressed-state
Hamiltonian as
HDA (r) = h̄ (ωL −δ) b† b + h̄ωL a†L aL + VAL (8.7)
with the detuning δ = ωL − ω0 . We have suppressed the zero-point energy of
the laser field. At first we turn off the coupling (8.6).
In this dressed basis an infinite num-
ber of families of pairs of states Èg,N+3\
EHN+2L
E(N ) = {|e, N i, |g, N +1i} appears. Èe,N+2\
The states in each pair are separated ΩL

by the detuning h̄δ, neigbouring fam-
EHN+1L
Èg,N+2\ ¯
ilies are separated by the photon en- Èe,N+1\
­
ergy h̄ωL . ΩL
The coupling, VAL , connects only Èg,N+1\
EHNL
states of one family, E(N ), in stim- Èe,N\
ulated emission and absorption,
1
he, N |VAL |g, N +1i = h̄Ω1 (r) (8.8)
2
1
hg, N +1|VAL |e, N i = h̄Ω1 (r) (8.9)
2
with Ω1 = d · E0 /h̄ being the Rabi frequency. On an energy scale relative to the
energy (N +1) h̄ωL − h̄δ/2 the matrix of coupled dressed states is
 
+δ Ω1 (r)
h̄ 
HDA =  (8.10)
2 Ω1 (r) −δ
8.1. HAMILTONIAN OF DRESSED STATES 53

After diagonalization we obtain the coupled dressed states of family E(N ) which
we label by |1, N ; ri und |2, N ; ri. These states lie at the energy
1
U1,N (r) = + h̄Ω(r) (8.11)
2
1
U2,N (r) = − h̄Ω(r) . (8.12)
2
They are spaced by h̄ times the effective Rabi frequency
q
Ω(r) = Ω21 + δ 2 . (8.13)

These new states are mixtures of the basis states |g, N +1i und |e, N i with the
state vectors
|1, N ; ri = + cos β |e, N i + sin β |g, N +1i (8.14)
|2, N ; ri = − sin β |e, N i + cos β |g, N +1i . (8.15)
The mixing angle, β = β(r), is defined by

δ Ω1 (r)
cos 2β(r) = − and sin 2β(r) = . (8.16)
Ω(r) Ω(r)

• For Ω1 = 0 (no laser field) we have sin 2β = 0 and a mixing angle β = 90o ,
hence cos β = 0 and the states |1, N ; ri ≡ |g, N + 1i, |2, N ; ri ≡ |e, N i
are pure basis states. For δ = 0 we have cos 2β = √ 0 and a mixing angle
β = 45o . In this case the state |1, N ; ri = (1/ 2)(|g, N + 1i + |e, N i)
aquires a 50 % admixture of excited state character.
The origin for the dipole force is that in an inhomogeneous laser field the energies
and eigenvectors depend on the position of the atom, as intensity and therefore
Rabi frequency vary in space. This is shown schematically in a cross section
through a laser beam profile.
Outside the laser field the !) *% +&
dressed states coincide with !" #$% $- $) & !" #% $- $) &
the uncoupled bare states,
! " ! ! *% +
the detuning of each family
pair being δ. Inside the laser !' #$% & !' #$% &
field the states are superposi- !, *% +&
tions of |g, N + 1i and |e, N i . / 01 23 ' $04 ' :; 1 23 ' $04 ' . / 01 23 ' $04 '
and their separation is larger, 56 1 ' 7 $8 ' 6 9 56 1 ' 7 $8 ' 6 9 56 1 ' 7 $8 ' 6 9
h̄Ω.
The direction of the light shift < = 1 202= ;
of the states |gi und |ei de- !) *% $($) +&
pends on the sign of the de- !" #$% & !" #% &
tuning. The light shift is ! ! ! *% $($) +
"
greatly exaggerated in this
drawing. In this example the !' #$% $($) & !' #$% $($) &
detuning is positive. !, *% $($) +&

The state |g, N +1i and hence |gi shifts up in energy when δ is positive, but it
shifts down in energy when δ is negative. This sign agrees with the prediction
of the dipole potential which we had derived in Eq. (7.57).
54 CHAPTER 8. DIPOLE FORCE AND DISSIPATION

8.2 Mollow Triplet


We consider the coupling of the dressed atom to the empty modes of the reser-
voir into which spontaneous emission may occur. In the bare basis the sole
allowed transitions are those from |e, N i to |g, N i, see page 38.

Èg,N+1\ È1,N\
In the coupled basis, both states
EHNL in each family E(N ) share |gi
Èe,N\ È2,N\ and |ei character, hence both con-
tribute in spontaneous transitions
ΩL to the members of the lower lying
family, E(N −1). These four tran-
È1,N-1\
sitions separate into the Mollow
Èg,N\
triplet [33]. Two side bands ap-
∆ W EHN-1L
pear at ωL ± Ω, symmetric to the
Èe,N-1\ È2,N-1\ central line at frequency ωL .

The strength of the transition moment in each line is defined by the dipole
matrix elements, now defined in terms of the coupled states. The actual line
strength observed in the Mollow triplet also reflects the population of the re-
spective coupled state, as will be discussed later. We derive the transition
moment for each line from the dipole matrix elements
dij = d hi, N −1| b |j, N i (8.17)
where the operator b projects the |ei-admixture in the members of family |j, N i
by spontaneous emission into the |gi-admixture in the members of family |i, N−
1i. The dipole transition matrix element between the bare atomic states is
d = hg|x|ei. With i = 1, 2 we characterize the two states in the family, the
second index refers to the emitting member. Inserting the state vectors from
Eq. (8.14) and (8.15) we obtain
d11 = −d22 = +d cos β sin β
d12 = −d sin2 β (8.18)
2
d21 = +d cos β
The dipole matrix elements dij depend on Ω via the mixing angle and therefore
also on the position of the atom, r. One may now define a Master equation for
a density matrix ρ(r), for the dressed atom at position r which describes the
time dependence of the two-level properties in the presence of the laser field,
including the coupling to the empty modes of the vacuum. DCT show that an
explicit solution of the Master equation is not actually required for gaining a
deeper understanding of the dynamics of the moving atom. Rather it is sufficient
to evaluate the reduced populations, Πi , by summing over the states
X
Πi (r) = ρii = hi, N ; r|ρ(r)|i, N ; ri . (8.19)
N

and, in the case that higher order contributions to the dipole force are consid-
ered, also information about the reduced coherences, ρij (r),
X
ρij (r) = hi, N ; r|ρ(r)|j, N ; ri . (8.20)
N
8.2. MOLLOW TRIPLET 55

In the following the assumption is made that the three lines in the Mollow triplet
are well separated in energy, Ω1 (r)  Γ, the large intensity limit.

We first consider a stationary atom at the position r, seeing a fixed laser


intensity and consider the change in population with time. Note that transitions
at the central line of the Mollow triplet, ωL do not change the populations Πi , as
these are defined as the sum over all families, see Eq. (8.19). However changes
in the populations occur in side-band transitions,

dΠ1 /dt = −Γ21 Π1 + Γ12 Π2 ,


dΠ2 /dt = +Γ21 Π1 − Γ12 Π2 . (8.21)

The transfer rates 1 ↔ 2 are proportional to the squared matrix elements of


Eq. (8.18)

Γ12 = Γ sin4 β
Γ21 = Γ cos4 β (8.22)

with Γ being the spontaneous rate of the pure atomic state, |ei. In the stationary
case the derivatives on the left hand side of Equation (8.21) are set to zero. As
the sum of the reduced populations has to equal to one, Π1 + Π2 = 1, we can
predict the stationary populations from (8.21)

Γ12 sin4 β
Πst
1 = = 4
Γ21 + Γ12 sin β + cos4 β
Γ21 cos4 β
Πst
2 = = 4 . (8.23)
Γ21 + Γ12 sin β + cos4 β
With this result we can now predict the intensity distribution in the Mollow
triplet for a stationary atom. The number of spontaneous photons per second
is given by the product of transition rate and population, Πst
j Γij .

1.0 90
85
P1 st
steady state population

0.8 80
mixing angle Β

75
0.6
70
0.4 65
60
0.2 P2 st 55
50
0.0 45
45 60 75 90 0 2 4 6 8 10 12
mixing angle Β W1  ∆

In steady state we have for the two contributions to the central line

sin4 β
Πst
1 Γ11 = cos β 2 sin β 2
sin4 β + cos4 β
cos4 β
Πst
2 Γ22 = cos β 2 sin β 2 . (8.24)
sin4 β + cos4 β

They differ greatly from each other unless δ = 0, e.g. β = π/4. For the two side
56 CHAPTER 8. DIPOLE FORCE AND DISSIPATION

bands we obtain
sin4 β
Πst
1 Γ21 = 4 cos β 4
sin β + cos4 β
Πst
2 Γ12 = Πst
1 Γ21 . (8.25)
We see that, under stationary conditions and for an atom at rest, an equal
number of spontaneous photons appears in the two side bands. In the special
case that δ = 0 (or Ω1 → ∞), the sidebands are each half the height of the
central peak.
In general the intensity varies with spatial position and hence Π and β
depend on the position of the atom. A key question now is how much time does
an atom require to reach the stationary condition if we move it from one point
in space to another? A rough estimate of this time is given by the inverse of
the sum of the two side-band rates
Γpop = Γ12 + Γ21 = Γ sin4 β + cos4 β

(8.26)
This internal time scale of the dressed atom will be used in an estimate of the
magnitude of dissipation, see Section 8.5.

8.3 Mean Dipole Force


A mean dipole force may be derived from the spatial variation of the relative
potential energy of the dressed atom in a spatially inhomogeneous laser field
1
U1 (r) = + h̄Ω(r)
2
1
U2 (r) = − h̄Ω(r) = −U1 (r) (8.27)
2
For large detuning, (Ω1 /δ  1), we may write
q 
1 1
U1 (r) − h̄δ/2 = h̄ {Ω(r) − δ} = h̄ Ω21 + δ 2 − δ
2 2
q 
1
= h̄δ 1 + Ω21 /δ 2 − 1
2
h̄Ω21
 
1 1 2 2
= h̄δ Ω1 /δ − . . . ≈ . (8.28)
2 2 4δ
We now consider the spatial variation of Eq. (8.27) along the cross section of a
Gaussian laser beam for two cases, positive detuning (left figure) and negative
detuning (right figure). The sizes of the black circles represents a measure of
the stationary populations at a selected position, r.

!' #% ( !' #% ,
!" #$% $& $' ( !) #$% (

!) #$% ( !" #$% $& $' (


!* #% ( !* #% (
! ( + ! - +
8.4. FORCE ON STATIONARY ATOM 57

In order to derive a value for the force acting on an atom in the light field, we
have to take into account that the population is distributed over both members
of a family pair. The force acting on the atom is proportional to the spatial
gradient of the light-shifted potential. Owing to the symmetry of Eq. (8.27)
the force is equal but of opposite sign for the two dressed-atom states. The
overall force results from the sum of forces on both dressed-atom states, the
contribution of each being weighted by the respective population. Obviously we
end up with a repulsive light potential in the case of blue detuning, when the
dominant population resides in state |1, N i, but an overall attractive potential
for red detuning.
If we assume that the stationary population (8.23) has been established by some
means and evaluate the dipole force from the mean of energy gradients, weighted
by the respective populations, we obtain

F~dip
st
(r) = −Πst ~ st ~
1 (r) ∇U1 (r) − Π2 (r) ∇U2 (r) (8.29)

8.4 Force on Stationary Atom


To simplify the notation we suppress the explicit spatial dependence in the
following. Inserting from Equations (8.23) and (8.27) into (8.29) we obtain
sin4 β − cos4 β
   
~ st 1 ~ 1 ~ cos 2β
Fdip = − h̄∇Ω = − h̄∇Ω
2 sin4 β + cos4 β 2 1 − 12 (sin 2β)2
and with Eq. (8.16)
Ω2
F~dip
st
= − h̄δ ~ ,
∇Ω (8.30)
Ω21 + 2δ 2
which may also be written as

h̄δ ~ (I/I0 )

F~dip
st
=− . (8.31)
4 1 + I/I0 + (2δ/Γ)2

Without assumping a large detuning in Eq. (8.29) the dipole force appears as
2
  
F~dip
st
= −∇~ h̄δ ln 1 + Ω1 (8.32)
2 2δ 2
with the dipole potential

Ω21
 
h̄δ
Udip = ln 1 + 2 . (8.33)
4 2δ

We compared on page 50 the dipole potentials from the approximation (8.28)


with that from (8.33). The dashed lines give the potential energy of the inter-
acting dressed state |1, N i in the limit Ω1 /δ  1.

Important conclusion at this point: There is no dissipative force for


an atom for which the stationary population of the coupled dressed states has
somehow been established. For such an atom the dipole force is conservative.
Dissipative terms appear only for a moving atom as we see next.
58 CHAPTER 8. DIPOLE FORCE AND DISSIPATION

8.5 Dissipative Contributions


We consider the energy balance of an atom moving in the inhomogeneous laser
field. We may ask what work is required to move the dressed atom by a small
distance dr? Two contributions appear, one describing the change in potential
energy of the atom, the second describing the change in the energy content of
the electromagnetic field,

dW = −Fdip dr = dUA + dUF . (8.34)

The change in potential energy of the atom is

dUA = Π1 ∆U1 + Π2 ∆U2 . (8.35)

The change of the energy of the electromagnetic field (laser photons + fluores-
cence photons) during the time dt, which we need to move the atom by the
distance dr, is

dUF = (Γ21 Π1 h̄Ω − Γ12 Π2 h̄Ω) dt . (8.36)

During the time dt, dN laser photons with energy h̄ωL disappear and dN flu-
orescence photons are emitted. Photons, which are spontaneously emitted in
transitions of the central lines of the Mollow triplet,

|i, N ; ri → |i, N −1; ri (8.37)

need not be counted, as they have the same energy as the laser photons. However
each emission in the side bands
|1, N ; ri → |2, N −1; ri
|2, N ; ri → |1, N −1; ri (8.38)
changes the energy of the electromagnetic field1 by the amount +h̄Ω or −h̄Ω.
In case that dUF is different from zero the intensity of the side bands in the
Mollow triplet is not balanced. From this we may derive an expression for the
actual force expression on a slowly moving atom.

Mean dipole force for a very slow atom : We use the force expression

F~dip = −Π1 ∇U
~ 1 − Π2 ∇U
~ 2 (8.39)

and assume that the Doppler detuning is much smaller than the spontaneous
rate, kv  Γ. This corresponds to a situtation when

v  λΓ. (8.40)

Under this condition the atom moves only a very small portion of an optical
wavelength during one spontaneous relaxation time. As a consequence the pop-
ulations in the moving atom will be very close to that given by the stationary
1 Actually because of the Doppler effect, the mean frequencies of the three components

of the fluorescence spectrum are slightly shifted from the values ωL , ωL + Ω, ωL − Ω. The
spectrum is shifted to a center at ωL − ~k · ~v . If n fluorescence cycles occur per second, the
energy balance dUF of the field must be corrected by an amount −n dt h̄~k ·~v = −n h̄~k ·d~
r. This
corresponds to the work done against the radiation pressure force nh̄~k. As we are interested
in the dipole force, we neglect this contribution here.
8.5. DISSIPATIVE CONTRIBUTIONS 59

conditions in Equation (8.23). We now express the temporal variation of the


~ i (r)
population dΠi (r)/dt in terms of the product ~v · ∇Π
~ i (r) .
dΠi (r)/dt = ~v · ∇Π (8.41)

and remember that Γpop signifies the rate at which the stationary state popu-
lation is reached. The characteristic time for this is τpop = 1/Γpop . Hence we
may approximate the gradient in (8.41) using the stationary populations
~ st
τpop~v · ∇Π i (r) ≈ −Πi (r) + Πst
i (r) (8.42)
Πi (r) ≈ Πst
i (r − v · τpop ) (8.43)
and obtain

dΠi (r)/dt ≈ −Γpop Πi (r) − Πst


 
i (r) . (8.44)

DCT’s argument: The relaxation between the states |1, N i und |2, N i requires
a characteristic time, τpop . A moving atom cannot instantaneously respond
to the changing Rabi-frequency. As a result there will be a delay in reaching
equilibrium, with the consequence that (for an atom moving with speed v) the
population at a spatial position r will approximately correspond to the popula-
tion which was present at some earlier time, when the atom was at r − v · τpop .
We now consider the consequence of this retardation for the case of blue de-
tuning. As we move with the atom out of the laser field, the state |1, N ; ri
progressively acquires more |g, N + 1i character. For δ > 0 the state |1, N ; ri
always carries a higher admixture of |g, N + 1i than does the state |2, N ; ri.
As the inversion always stays below 50 %, the state |1, N ; ri carries a higher
population than the state |2, N ; ri. We therefore have

Πst st
1 (r) > Π2 (r) . (8.45)

At the same time the contamination of state |1, N ; ri by |e, N i increases when
the Rabi frequency increases. From this it follows that Πst1 (r) decreases when
we move the atom in the direction of higher laser intensity, that is

Πst st
1 (r−dr) > Π1 (r) . (8.46)

The figure below, on the left, depicts the situation of the stationary atom from
(8.45) and (8.46) at two positions, r and r − dr.

( )
! & *+. ! *+.#1 ! ( )
*+,- +.
& &

! ( )
*+,- +. & "$ 0+ & "$ 0+
& - + # 1 #"# # # # 3 4 3

! "#$ % & ! "#$ % &


"
" #1 #2

' "$ ' "$


( )
! / *+,- +.
/ "$ 0+ ( ) / "$ 0+
! / *+.#1 ! / *+,- +.
( )
! / *+.

+,- + + +
60 CHAPTER 8. DIPOLE FORCE AND DISSIPATION

We now consider an atom which moves with the speed v in the direction
of higher laser intensity (see the figure above, on the right). At the position r
the population in the moving atom Π1 (r) is not equal to the population Πst 1 (r),
but rather it corresponds approximately to the population Πst 1 (r − dr), where
dr = vτpop . From this we conclude

Π1 (r) = Πst st
1 (r−vτpop ) > Π1 (r) , (8.47)

Π2 (r) = Πst st
2 (r−vτpop ) < Π2 (r) . (8.48)

The force at which a moving atom is pushed out of the blue-detuned field,

~
h̄∇Ω
F~dip = − [Π1 (r) − Π2 (r)] , (8.49)
2
is larger than the force experienced by a nonmoving atom in the stationary
case, as given in Equation (8.29). This additional force depends on the speed
of the atom and acts like a friction force. The higher the speed of the atom, the
larger is the deviation of the population at the position r from the stationary
population. From this (so far qualitative) argument we conclude that the dipole
force contains a friction term when δ is positive.2

A first quantitative estimate of the speed dependence may be obtained by in-


serting the approximation

Πi (r) = Πst
i (r−vτpop ) (8.50)

into Equation (8.49) and use for the stationary populations the expression (8.23)

3
Ω21 (r)

2h̄δ
F~dip (r, v) = F~dip
st
(r) − 2 α · ~v ) α
(~ ~ (8.51)
Γ Ω1 (r) + 2δ 2

the abbreviation α
~ is
~ 1 (r)
∇Ω Ω ~
α
~= = 2 ∇Ω . (8.52)
Ω1 (r) Ω1

In the following we apply this model in a quantiative description of dipole


force effects in a standing-wave laser beam. In this context Cohen-Tannoudji
and Dalibard have first coined the expression Sisyphus cooling.3 Note that
other sisyphus cooling schemes for red detuned laser beams at very low laser in-
tensity have since been discovered. They are based on a fundamentally different
sequence of events, but they also rely on the motion of dressed-atom population
over the hills and valleys of spatially modulated light shifts.

2 When the detuning is negative, δ < 0, the reverse holds, the speed-dependent term to the

dipole force leads to heating.


3 Greek mythology has it that Sisyphus must endlessly roll a stone up a hill in the under-

world. As soon as he reaches the top the stone rolls down again.
8.6. ATOMIC MOTION IN A STANDING WAVE 61

8.6 Atomic Motion in a Standing Wave


In a standing wave the Rabi frequency varies 1

periodically with the position z along the E 0


standing wave,
"1
0 Λ!2 Λ 3Λ!2
Ω1 (z) = Ωmax
1 · | cos kz| (8.53)
#1

where Ωmax
1 is the Rabi frequency at the
$1
antinodes of the standing wave. The figures
show as a function of z, for the parameters 0
Λ!2 Λ
Ωmax
1 = 1 and δ = +0.2, 0 3Λ!2

90

• the electric field strength E with its nodes 75


at the positions z = λ4 , 3λ
4 , ..,
Β
60

45
0 Λ!2 Λ 3Λ!2
• the Rabi frequency, Ω1 (z),
#1

• the mixing angle β. It shows pure atomic &1


states, β = 90o , at the nodes, &2

• the stationary populations, Πst st


1 and Π2 ,
0 Λ!2 Λ 3Λ!2

1
• the light shifted energies U1 und U2 . At the U1
0 $ ∆
nodes the energy gap is small, ∆U = δ, at the U2

antinodes it is maximal, ∆U = Ω) "1


0 Λ!2 Λ 3Λ!2

"g(
• the absolute amplitudes of |gi and |ei of #1

state |1i. State |1i will preferentially emit


"1( "e(
near z = λ2 , 2λ
2 , .., (blue arrows), the position
where its contamination by |ei character 0
0 Λ!2 Λ 3Λ!2
maximizes.
"e(
#1

• the absolute amplitudes of |gi and |ei of


"2( "g(
state |2i. State |2i will preferentially emit near
z = λ4 , 3λ
4 , .., (blue arrows), the position where 0
0 0.5Λ Λ 1.5Λ
its contamination by |ei character maximizes.

The preferential spontaneous emission from state |1i near z = λ2 , 2λ 2 , .. and


from state |2i near z = λ4 , 3λ
4 , .. does not change the potential energy of the
dressed state if the transition occurs in the central line of the Mollow triplet.
By contrast, in the case of blue detuning, transitions in the side bands always
reduce the potential energy of the dressed states. This is because the side band
emitted from state |1i near z = λ2 , 2λ
2 , .. terminates in the potential minimum of
state |2i and because the the side band emitted from state |2i near z = λ4 , 3λ 4 , ..
terminates in the potential minimum of state |1i.
Spontaneous transitions may of course occur at any position along the stand-
ing wave, however their probability is highest at the positions indicated by the
blue arrows in the two figures at the bottom of page 61. This preference can be
condensed into a graphic representation of events which helps to visualize the
62 CHAPTER 8. DIPOLE FORCE AND DISSIPATION

dissipative nature of the dipole force for blue detuning. In order to do this we
show the potential energy landscape of the dressed states of three families along
the standing wave. The periodic splitting in potential energy for each family
serves to indicate the continuous exchange between potential and kinetic energy
of the atom as it moves along the standing wave. What follows is valid for blue
detuning, δ > 0, the reverse conclusion applies in case of red detuning.
One side band is preferentially emitted
near the antinodes. At the antinodes the
1 ,N + 1 ;r state |1, N + 1; ri (purely ground state
- n + 1 in the absence of the laser field) has the
2 ,N + 1 ;r highest contamination by |e, N i. At these
positions a spontaneous transition into
state |2, N ; ri of the lower family has the
highest probability. The latter is purely
1 ,N ;r excited state in the absence of the laser
- n field, but has its highest contamination by
2 ,N ;r
|gi character at the antinode.

The other side band preferentially appears


at the nodes. At these positions state
1 ,N -1 ;r |2, N ; ri is purely excited state. Emis-
- n -1 sion from this state at the nodes prefer-
2 ,N -1 ;r entially populates state |1, N −1; ri, which
has purely |gi character at the node.

In between these two points of preferential emission in the side bands the atom
always experiences more uphill than downhill paths. As a result this process con-
tinuously extracts kinetic energy from the atom and redistributes it into energy
of the spontaneous photons. This is the reason for the name Sisyphus cooling.

To quantify this effect, we explore the change in the kinetic energy of an atom
which moves over a distance equal to one wavelength of the standing wave laser:

Z λ
∆W = Fdip dz = λ hF i (8.54)
0

According to Equation (8.34) the mean change in energy derives from

• the change in the potential energy of the atom and

• the change in energy contained in the electromagnetic field.


Z λ Z λ
1 1
hF i = − dUA − dUF (8.55)
λ 0 λ 0

As Πi and Ui are periodic functions of z, the first contribution vanishes. The


second contribution however is unequal to zero when the Mollow triplet exhibits
asymmetric side bands. It is this change in the energy contained in the elec-
tromagnetic field which dictates a change in the kinetic energy of the moving
atom. We consider this case at various speed conditions:
8.6. ATOMIC MOTION IN A STANDING WAVE 63

At very low atomic speed : Here the time the atom takes to move over the
distance of one wavelength is very much larger then the spontaneous emission
time. In other words the atom undergoes many spontaneous emission events
during the time it travels over one wavelength. In this situation the speed de-
pendence of the dipole force derives from the small time delay required to attain
the stationary population.

From Equation (8.51) we obtain

1 λ
Z
hF i = Fdip (z, v)dz
λ 0
3
1 λ st 1 λ 2h̄δ Ω21 (r)
Z Z 
= F (z)dz − α2 v dz (8.56)
λ 0 dip λ 0 Γ Ω21 (r) + 2δ 2
1 λ
Z
= − γ v dz (8.57)
λ 0
The first contribution on the right hand side of Equation (8.56) vanishes, the
second term in (8.57) may be written (with M being the mass of the atom)

Mv
hF i = − . (8.58)
τ
The characteristic time τ is positive for δ > 0 and negative for δ < 0. The time
τ is proportional to 1/δ and it can be much shorter than the characteristic time
for cooling by the spontaneous force, see Equation (3.24),

M
τspon = = h̄/ER (8.59)
h̄k 2
We conclude that in a standing wave of high laser intensity the friction in the
dipole force can, for low atomic speed, be much larger than the friction term of
the spontaneous force.
-1 0 0 F o rc e f
An explicit calculation of the dependence ( u n its h k / /2 )
of the dipole force on atomic speed (in -5 0
-7 5
units of Γ/k) is shown here. The mean
dipole force is given in units of the maxi-
F (h k / /2 )

-5 0
mal spontaneous force, h̄kΓ/2, see Equa- 0 0 .1 0 .2
tion (3.17). The insert gives the depen-
dence at low speed, the linear result which -2 5
V e lo c ity v
we found in Equation (8.58) is given by ( u n its / /k )

the dashed straight line. 0


0 2 .5 5 7 .5 1 0

The dependence of the force on the atomic speed in the figure above
may be understood from the following arguments

• The maximum in the force occurs at a speed v ≈ 0.5 Γ/k. At this speed
the atom travels over a distance of half a wavelength during one spon-
taneous emission time, optimal for the Sisyphus effect. Without detailed
calculation we understand the effective exchange between kinetic energy
64 CHAPTER 8. DIPOLE FORCE AND DISSIPATION

and the energy of the electromagnetic field when the emission in the Mol-
low side bands occurs from near a potential maximum of the light-shifted
states.

• At very low speed the force increases linearly with speed. This is because
the non-stationary population is carried over ever increasing distances of
the potential energy difference and the asymmetry in the Mollow side band
increases with increasing v, until the optimal transport of population over
a mean distance of λ/2 occurs, the position of the force maximum.

• This Sisyphus-effect leads to a continuous energy loss. In characterizing


the motion along the light-shifted states we had assumed that the motion
follows these states in an adiabatic manner. This is a good approximation
for slow speeds but at higher speed the so-called kinetic, nonadiabatic
coupling also needs to be taken into account.

• The nonadiabatic coupling permits transitions between |1, N ; ri and |2, N ; ri


states.4 The derivation5 shows that above a critical speed
1/3
vcr = 2πΓδ 4 /Ω21 (8.60)

the dipole force falls off proportional to 1/v.

Under realistic conditions the dipole force can greatly exceed the sponta-
neous force. A 1 W laser, focussed to a spot of 100 µm diameter can generate
a Rabi frequency Ω1 = 1000 Γ. The figure on page 63 gives the dipole force in
this case for a detuning of δ = +200 Γ.
Aspect et al. [38] have used a blue-detuned standing wave to demonstrate the
cooling aspect of the dipole force. Directing an atomic beam perpendicular to
the standing wave laser beam, they were able to show the collimating force on
the atomic beam. Since this original demonstration of the cooling aspect of the
dipole force only few applications for trapping and cooling by the dipole force
have appeared. This is largely related to the difficulty of trapping atoms using
repulsive light-shifts of Gaussian beams. Nevertheless the ability of generating
hollow light beams promises future applications of the dissipative dipole force
for damping the radial motion of atoms in tubes of blue-detuned light.

The laudation by the Nobel committeea shows pic-


tures describing Sisyphus’ work and a cooling scheme
which appears when the atom experiences polariza-
tion gradients in space. This was also discovered by
DCT in a later paper in 1989, which we discuss next.

a http://nobelprize.org/nobel prizes/physics/laureates/1997/illpres/doppler.html

4 See appendix A-7, the Landau-Zener model.


5 Equations (5.10) - (5.18) in [17]
Chapter 9

Polarization Gradients

In 1989 Dalibard und Cohen-Tannoudji [39] discussed two cooling schemes


which explain how temperatures well below the Doppler limit are reached
in red detuned molasses cooling. These cooling schemes require atoms
with Jg > 0, where Jg is the total angular momentum of the ground state.
The cooling schemes work efficiently at low laser intensity and large red
detuning, much larger than recommended by the theory of Doppler cooling.
In Doppler cooling we added the forces exerted by the counter-propagating red-
detuned laser beams. The key to DCT’s scheme is that one first adds the electric
fields of the two counter-propagating laser beams and then considers the resul-
tant modulation of the electromagnetic field in space. Polarization gradients
appear in most such superpositions. Adding fields before considering forces is
required when the internal clock of the atom ticks faster than or on the order
of the time required by the atom to travel the distance of a wavelength.

Two fundamentally different effects from polarization gradients were identified:

Lin⊥Lin Two counterpropagating but linearly polarized beams, their


polarization being orthogonal.
In this case the light shift of Zeeman sublevels is modulated periodically
in space and a Sisyphus effect emerges, similar to that discussed in the
previous chapter, but now for red-detuned light at low laser intensity.

σ+ − σ− Two counterpropagating circularly polarized beams, their


circular polarization being orthogonal.
The resulting polarization ˆ is then linear, but it rotates in space along
the axis of propagation. In this situation a radically different mechanism
for dissipation emerges. In equilibrium, the alignment of the angular mo-
mentum of the ground state |gi is parallel to ˆ. However, when the atom
moves, it sees this axis rotating in space. The work required to rotate the
atomic alignment is the fundamental origin for this cooling process. The
reason is that even for very slowly moving atoms a substantial population
difference among the magnetic substates can build up and the radiation
pressure from left and right-directed laser beams is unbalanced.

65
66 CHAPTER 9. POLARIZATION GRADIENTS

Both schemes require low intensity, low enough that the time for redistribution
of population among Zeeman-sublevels in the ground state is of the order of the
time the atom takes to travel the distance of about a quarter of the wavelength
of the laser light. Under such conditions very strong cooling forces appear.

In the following two chapters we derive semi-classical expressions for the friction
coefficient for both cases. We will see that under proper conditions the temper-
ature of atoms can approach the recoil energy. At this point one enters a regime
where the de-Broglie wavelength of the atom is of the order of the wavelength
of the light. At this point two wave descriptions meet and the semi-classical
treatment, in which the atomic position is precisely identified, becomes invalid.

Time Scales
The initial proposal for laser-cooling was based on the Doppler-effect. An
atom which moves at red-detuning (ωL − ω0 = δ < 0) in a standing wave
is closer to resonance with the counterpropagating wave. This wave exerts a
stronger radiation pressure and damps the atomic motion in a way as if the
atom moves in a viscous medium (optical molasses). The range of velocities
which can be captured in this process is governed by the natural width of the
excited state1

k ∆v ≈ Γ . (9.1)

Spontaneous emission which is associated with the spontaneous force leads to a


diffusive heating of the sample, limiting the lowest possible temperature to the
Doppler limit,2 (see page ??):

Γ
kB TD = h̄ . (9.2)
2
In experiments around 1988 evidence was found that experimental temperatures
could fall well below the Doppler limit [21, 22]. For Rb, Na and Cs temperatures
were observed near the recoil energy3

h̄2 k 2
kB TR = . (9.3)
2M
In reaction to these findings two groups [39, 40] independently presented models
for new cooling mechnisms. Arguments why such low temperatures can be
reached are:

1) Friction arises for a moving atom, if the internal atomic equilibrium popu-
lation can not be established rapidly enough. With internal atomic equi-
librium we refer to the equilibrium of population in the Zeeman-sublevels,
as appropriate for the local radiation field. This process of approaching
equilibrium can be characterized by an adiabaticity-parameter, ζ, which is
1 For Rb we have ∆v = Γ/k ≈ 6·106 · 780·10−9 ≈ 5 m/s.
2 For Rb we have TD = h̄Γ/(2kB ) ≈ 2π · 10−34 · 6·106 /(2 · 1.38·10−23 ) ≈ 140 µK.
2
3 For Rb TR ≈ 2π 10−34 /(780·10−9 ) /(2 · 87·1.6·10−27 · 1.38·10−23 ) ≈ 170 nK .

67

given by ratio of the distance v τ , which the atom covers during the char-
acteristic internal relaxation time τ , and the laser wavelength, λ = 2π/k,


ζ = 2π = kvτ . (9.4)
λ
2) For a two-level system we have a single internal characteristic time, the
time for spontaneous relaxation, τ = 1/Γ. The corresponding adiabaticity
parameter is4
ζS = k v τ = k v/Γ . (9.5)
However alkali-atoms have several Zeeman substates in their ground elec-
tronic state. This means that there is an additional internal time, the time
for optical pumping from one Zeeman-sublevel, gM , into another Zeeman
sublevel, for example g−M . We call this time for optical pumping
−1
τZ = Γ0 . (9.6)
where Γ0 is the mean scattering rate of the ground state5 . The scattering
rate Γ0 depends on laser intensity and can be made very small. For an
atom with more than one ground state level we may therefore introduce
a second adiabaticity parameter,
ζZ = k v τZ = kv/ Γ0 . (9.7)
At low laser intensity, the Rabi frequency Ω1 is small compared to Γ. Then
we have an optical pumping rate, Γ0 , which can be many times smaller
than the spontaneous emission rate Γ:
Ω1  Γ → Γ0  Γ → ζZ  ζS (9.8)
From this it follows that strong nonequilibrium effects may occur at speeds,
kv ≈ Γ0 , much smaller than the typical speed in the Doppler scheme,
kv ≈ Γ. We show below that the resulting non-equilibrium effects give
rise to strong friction for slow atoms.
3) The most important ingredient for this cooling scheme concerns the spa-
tial polarization gradients. Strong dissipation only appears when the
anisotropy of the ground state equilibrium configuration depends strongly
on the atom’s spatial position. The word anisotropy refers to an orienta-
tion of the ground state wavefunction.
Cooling by polarization gradients requires the presence of several Zeemann
sublevels in the ground state of the atom. In this case one needs to consider that
the Rabi frequency depends on the laser polarization and the magnetic quantum
number of the atomic state. The latter dependence is brought about by the de-
pendence of the line strength on magnetic quantum number (which is expressed
by the Clebsch-Gordan coefficients, see appendix A-3). As a consequence a spa-
tial modulation of light-shifted energies and of the atomic wavefunction results
which can lead to a sisyphus effect of dissipation.
4 For Rb the value of ζ = k v/Γ ≈ v/(5 m/s). This means that for a Rb-atom with a speed

of v=5 m/s one spontaneous emission occurs during the time the atom takes to travel one
wavelength.
5 h̄Γ0 describes the energy width of the ground state.
68 CHAPTER 9. POLARIZATION GRADIENTS

What is Orientation?
Orientation occurs when Zeeman substates with M -values of opposite sign are
not equally populated.

What is Alignment?
Alignment indicates that the Zeeman sublevels are not in equilibrium, but states
of equal values of |M | are equally populated.
M 27 .
Chapter 10

Lin-Lin Polarization
Gradients

We consider a one-dimensional molasses from two plane wave lasers at the same
frequency, ωL , with orthogonal linear polarization (lin⊥lin). The laser beams
propagate in opposite directions along the z-axis. In this case the resultant
polarization varies in space periodically from σ + to σ − to σ + . As a result we
have a spatially periodic situation where the equilibrium ground state popula-
tion changes from one where all population is pumped into the ground state
level g+M , to one where all population is transferred to g−M . The periodic
transfer of population among the M -states can lead to cooling if the differ-
ent M -state energies experience different light-shifts, in which case the spatial
gradients of the light shift exert a dipole force on the atom.
The two laser beams are assumed to have the same real amplitude, E0 , which
gives us for the superposition field of two counterpropagating waves1

~ t) = E + (z) E0 e−iωL t ,
E(z, E + (z) = ˆx eikz + ˆy e−ikz . (10.1)

giving

 
+ ˆx + ˆy ˆy − ˆx
E (z) = 2 √ cos kz − i √ sin kz (10.2)
2 2

M 28

1 A change of polarization in space will occur even if the amplitudes of the two waves are

not taken as equal. A change in polarization type in space in general always occurs when
counterpropagating light fields meet, the only exceptions being two parallel, linearly polarized
waves (lin || lin) or two equal circularly polarized waves ( σ ± ||σ ± ).

69
70 CHAPTER 10. LIN-LIN POLARIZATION GRADIENTS

which we may abbreviate as



E + (z) = 2 (ˆ
X cos kz − i ˆY sin kz) (10.3)

This corresponds to √ two fields e e


x 1 s -
s +
with √ amplitudes E 0 2 cos kz and
E0 2 sin kz, polarized along two
orthogonal vectors e 2
-e 1
e y
x

ˆX = (ˆ x + ˆy ) / 2
√ 0
ˆY = (ˆ y − ˆx ) / 2 . l /8 l /4 3 l /8 l /2 z
y
The resulting field has an ellip-
e -3 /2 e -1 /2 e + 1 /2 e + 3 /2
ticity which varies along the z-
axis. Over a distance of λ/8 the
field changes from linear to cir- g -1 /2 g + 1 /2

cular and back to linear over an-


other distance of λ/8.

Position kz Polarization type Orientation of polarization


z =0 0 linear, % X √
z = λ/8 π/4 circular, σ− + (X − iY ) / 2
z = λ/4 π/2 linear, & Y √
z = 3λ/8 3π/4 circular, σ+ − (X + iY ) / 2
z = λ/2 π linear, . −X √
z = 5λ/8 5π/4 circular, σ− − (X − iY ) / 2
... ... ... ...
We now explore the optical transition Jg = 1/2 → Je = 3/2 in the laser
field (10.3). The Zeeman sublevels which participate in optical transitions,
together with the respective Clebsch-Gordan coefficients (see App. A-3) are
shown below for transitions with linearly and with circularly polarized light. The
Rabi frequency is proportional to the Clebsch-Gordan coefficient. The square
of the Rabi frequency and hence the square of the Clebsch-Gordan coefficients
is a measure for the strength of the optical transition. The σ + transition from
g+1/2 is three times stronger than the σ − transition from g+1/2 . The opposite
is true for the state g−1/2 . This will give rise to E an e spatially
rg y dependent light shift
for the different Zeeman components.

! "# $% ! "& $% ! ' & $% ! ' # $%

0 l /8 l /4 3 l /8 l /2 5 l /8 z

% & & % L in s - L in s + L in s -

& # # &
# #

( ( g -1 /2 g + 1 /2 g -1 /2
"& $% ' & $%

• Linear polarization (at z = 0, λ/4, λ/2, . . .): The transitions from g+1/2 and
from g−1/2 have equal strength. According to Eq. (8.28) the energy shift of the
two Zeeman substates is proportional to 2/3.
10.1. SISYPHUS EFFECT 71

• Circular, σ − , (at z = λ/8, 5λ/8, . . .): The transition from g−1/2 is three
times stronger than g+1/2 . At the position of σ − -polarization the g−1/2 state is
light shifted three times more than the state g+1/2 .

• Circular, σ + , (at z = 3λ/8, 7λ/8, . . .): The transition from g+1/2 is three
times stronger than g−1/2 . At the position of σ + -polarization the light shift of
the g+1/2 state is three times stronger than that of g−1/2 .

10.1 Sisyphus Effect


Equilibrium state of a stationary atom : We assume δ = ωL − ω0 < 0,
that is red detuning. In this case the ground state levels are pushed to lower
potential energy in the presence of the field (Figure on page 70).
• At the position z = λ/8 the equilibrium populations are Π(g−1/2 ) = 1
and Π(g+1/2 ) = 0. The light shift of the |g−1/2 i-state, ∆− , is three times
larger than the light shift of the |g+1/2 i-state, ∆+ . At the position of
σ − -polarization we have ∆− = 3∆+ .
• For an atom at z = 3λ/8 (a position of σ + -polarization) the opposite
holds, ∆+ = 3∆− and the equilibrium populations are Π(g−1/2 ) = 0 and
Π(g+1/2 ) = 1.
• At the positions z = 0, λ/4, λ/2, . . . we have linear polarization and both
sublevels have equal equilibrium populations.
In the figure on page 70 we have marked the equilibrium populations by the
thickness of the black dots.

State of the moving atom : A Sisyphus effect appears when the time
required for repumping between the two Zeeman substates, τZ , is of the order of
the time that the atom takes to move over a distance of λ/4. In this case the
atom stays in state g−1/2 and climbs up the potential hill prior to absorption of
a σ + -photon, which pumps it back into the state g+1/2 .
E n e rg y This is similar to the case in stimulated
molasses (see page 62), but here the laser
intensity is much lower and the laser is red
detuned. When cooling with polarization
gradients, the light shift ∆ is made much
smaller than Γ. Kinetic energy of the atom
is effectively transferred to potential en-
ergy of the atom, given that the time re-
quired for repumping between the Zeeman
0 l /8 l /4 3 l /8 l /2 5 l /8 z levels is chosen properly.
The atomic momentum can decrease when climbing up the potential hill be-
cause of a coherent redistribution of photons between the two counterpropagat-
ing waves. The atom absorbs photons from one wave and stimulates it into the
opposite wave. In this fashion arbitrary changes in the atomic momentum in
72 CHAPTER 10. LIN-LIN POLARIZATION GRADIENTS

units of the recoil momentum can be achieved. This by itself is a conservative


process, free of friction. However, an irreversible energy loss will occur when
optical pumping leads to a redistribution of population in the Zeeman substates
in the event of spontaneous emission.

10.2 Magnitude of Friction


Suppose the atom starts in state g−1/2 at z = λ/8 and climbs up the potential
hill towards z = 3λ/8. There the probability that the atom is pumped by σ + -
light into e+1/2 is large. When this happens the probability for emission into the
state g+1/2 is larger by a factor of two than returning by spontaneous emission
back to g−1/2 . The blue-detuned spontaneous photon which is emitted when
the transition goes into g+1/2 will carry away the potential energy that the atom
had gained in climbing (and had lost as translational energy).
In order for this to occur efficiently the laser intensity has to be made lower,
the slower the atom is. Obviously this scheme can only work if more than one
ground state is present (Jg > 0).
For the friction force we choose the ansatz

F = −α · v (10.4)

We expect it to be maximal, when v · τZ ≈ λ/4. This is the case for

k · v ≈ Γ0 = 1/τZ (10.5)

At this speed the energy dissipated in a time span τZ is of the order of −h · ∆


(negative, because of red detuning, δ < 0 ). Then we have for the energy loss
per second:
dW ∆
= −h̄ · = −h̄ · ∆ · Γ0 (10.6)
dt τz
With the assumption that F increases linearly with v (10.4) we may write
dW
= −F · v = −α · v 2 (10.7)
dt
If we equate (10.6) and (10.7) we obtain

α ≈ −h̄ · k 2 . (10.8)
Γ0
Since the light shift, ∆, is proportional to the laser intensity and the repumping
time between Zeeman sublevels, Γ0 , is also proportional to the laser intensity
we get a friction coefficient α which is independent of laser intensity.
By contrast, the friction coefficient for Doppler-cooling is proportional to the
laser intensity, see Eq. (3.21). Considering the effect of momentum diffusion,
equivalently to that which we made for the Doppler cooling limit, see Eq. (3.27),
one obtains the result that the final temperature that can be achieved by this
polarization gradient cooling process scales with the laser intensity. As it can
be made arbitrarily small, there is no lower limit in temperature - at least in
this semiclassical approach.
10.3. SEMICLASSICAL ESTIMATE 73

An explicit calculation [39] gives the dependence of the force on velocity as


shown below for the parameters Ω1 = 0.3Γ , δ = −Γ.

! #$ 2 5 4 2 !2 2 3 $
! ! #! %
!

"! #! $

& ' ( ) * +, -
. / 0 1 2!2 2 3 4
"$ $

"! #$

The dashed line represents the radiation pressure from the sum of the two
Doppler-shifted counterpropagating waves, Eq. (3.18). The force resulting from
the polarization gradient is given by the full line, shown in closer detail in the
inset. It is obvious that at very low speed the force increases linearly, much
steeper than the conventional Doppler-cooling (dashed line).
We may rewrite Eq. (10.8) for conditions appropriate at low laser intensity,
(Ω1  Γ). If we also demand |δ|  Γ we obtain from Eq. (3.16)
Γ0 ∝ Ω21 Γ/δ 2
∆ ∝ Ω21 /δ
αZ ∝ −hk 2 · δ/Γ (10.9)
At large detuning, δ, the friction coefficient αZ is much larger than the optimal
friction coefficient for Doppler-cooling αD (see Eq. 3.29 for 2δ/Γ = −1)

αD ≈ −hk 2 . (10.10)

Caveat : the range of velocities which can be captured by polarization-gradient


cooling is very much smaller than in the case of Doppler cooling, k∆v ≈ Γ. In
a typical experiment both schemes are active. One first captures atoms with
Doppler-cooling at low red detuning and high laser intensity. Once the atoms
are pre-cooled one may detune the laser further and decrease its intensity – and
exploit cooling effects by polarization gradients.

10.3 Semiclassical Estimate


For convenience we shift the origin of the coordinate
√ system of the z-axis by
λ/8. With
√ cos(kz + π/4) = [cos kz − sin kz]/ 2 and sin(kz + π/4) = [cos kz +
sin kz]/ 2 we may rewrite the field polarization (10.3) as

E + (z) = [(ˆ
X − i ˆY ) cos kz − (ˆ
Y + i ˆX ) sin kz] (10.11)

(10.11) represents any field configuration along the z-axis in terms of a su-
perposition of σ + and σ − light, each weighted by the respective trigonomet-
74 CHAPTER 10. LIN-LIN POLARIZATION GRADIENTS

ric function. So we have σ − light at z = 0, π, 2π, ..., we have σ − light at


z = π/2, 3π/2, ..., and we have mixtures of of σ + and σ − in between.
Next we consider the Clebsch-Gordan coefficients and calculate the light
shift for the g+1/2 state. It experiences a Rabi frequency proportional to 1 at

positions of σ + light and proportional to 1/ 3 at positions of σ − light, and
weighted mixtures of both in between.
 
1
Ω+1/2 (z) = Ω1 √ (ˆ X − i ˆY ) cos kz − 1 (ˆ
Y + i ˆX ) sin kz , (10.12)
3
which gives for the squared Rabi frequencies
 
2 1
|Ω±1/2 (z)|2 = 2Ω21 ± cos(2kz) . (10.13)
3 3
The spatial modulation of the potential energy of the Zeeman components is
obtained from the dipole potential Eq. (7.57) in the limit Γ  Ω1 ,

2Ω21 /Γ2
 
2 1
U±1/2 = h̄∆± = h̄δ ± cos 2kz . (10.14)
1 + (2δ/Γ)2 3 3
With this potential energy we define the spatially dependent force for each
Zeeman component
d 2 2Ω21 /Γ2
f±1/2 = − U±1/2 = ∓ h̄kδ sin 2kz , (10.15)
dz 3 1 + (2δ/Γ)2
and a mean force

hF i = f+1/2 Π+1/2 + f−1/2 Π−1/2 . (10.16)

The equilibrium populations for a stationary atom are easily determined at the
special points where the polarization is either circular or linear. D&CT use opti-
cal Bloch equations to predict the z-dependence of the equilibrium populations
for a stationary atom and obtain2
2
Πstat
+1/2 = sin kz , Πstat 2
−1/2 = cos kz . (10.17)

With these populations the mean force on a stationary atom is


2 2Ω21 /Γ2
hF (z, v = 0)i = h̄kδ sin 2kz cos 2kz . (10.18)
3 1 + (2δ/Γ)2
For calculating the force on a moving atom a retarded population must be
chosen, analogous to the considerations we had made in Eq. (8.41-8.43). With
this approach D&CT[39] obtain the dependence of the force on speed shown on
page 73 by the full line. For the equilibrium temperature they obtain

hΩ21
kB T = , (10.19)
8|δ|

a value much smaller than the Doppler limit Eq. (3.29). For the limiting mean
|δ|
speed they predict vrms > h̄k
M Γ which is larger than the recoil speed.
2 See Eq. (4.5) to (4.14) in [39].
Chapter 11

Orientational Cooling

Dalibard und Cohen-Tannoudji [39] also explored the cooling mechanism


which appears when plane waves of σ + −σ − light counterpropagate. For
two beams of equal intensity the resulting field has always linear polariza-
tion, but the direction of polarization rotates in space, the laser intensity
being constant. In this case no dipole force appears but a different cooling
effect emerges : At very low atomic speeds a population difference among
the Zeeman sublevels builds up. The field attempts to align the atom, but
equilibrium is not reached for a moving atom. Rather, the atom’s angular
momentum becomes partially oriented along the direction of atomic mo-
tion. This orientation leads to an imbalance of forces exerted by the two
counterpropagating laser beams and gives rise to strong dissipation.
We consider two counter-propagating plane waves of frequency ωL with equal
amplitudes E0 ,
ˆx + iˆ
 ˆ − iˆ

E(z, t) = E0 e−iωL t − √ y eikz + x √ y e−ikz .

(11.1)
2 2
The resulting field has linear polarization along the unit vector ˆY ,
√ √
E + (z) = 2 (ˆ y cos kz + ˆx sin kz) = 2 ˆY . (11.2)

M 29

At z = 0 the direction of linear polarization ˆY , coincides with ˆy . This


linear polarization rotates in space by an angle ϕ(z) = kz, the intensity be-
ing constant.1 Under these conditions a new mechanism, orientational cooling,
appears in the case that the total angular momentum of the ground state is
1 For unequal amplitudes the resulting field is elliptical, but the axes of the ellipse rotate

in space, the length of the axes staying constant.

75
76 CHAPTER 11. ORIENTATIONAL COOLING

Jg > 1/2 and the lasers are red-detuned. This cooling mechanism is active if a
common excited state exists which is coupled by σ + −σ − light with two Zee-
man-components of the ground state. In the following example we consider the
Clebsch-Gordan coefficients for a transition Jg = 1 ↔ Je = 2. In this scheme
the common excited state is |e0 i.

I +
I -
e -2 e -1 e 0 e + 1 e + 2
1 1
A y 6 6
1 1 1
x 1 2 2 2
1 2 1
2 3

0
z
y g -1 g 0 g + 1

This state is coupled via σ + and σ − light with |g−1 i and |g+1 i. The thermal
N
(statistical) population of the Zeeman ground state levels is NM = 2J+1 , where
N is the total number of atoms and NM is the number of atoms in the level with
magnetic quantum number M . At equilibrium and without external radiation
field, the three ground-state sublevels each carry one-third of the population.
This equilibrium is modified by optical pumping.

11.1 Equilibrium for Stationary Atom


At z = 0 the linear laser polarization is oriented along y . We label the eigen-
states of Jy by |g−1 iy , |g0 iy and |g+1 iy . The linear laser polarization con-
centrates population in the state |g0 iy , because
pthe poptical pump rate for the
2
transition |g−1 iy → |g0 iy is proportional
p p to (2 1/2 1/2) = 1/4, the pump
rate from |g0 iy → |g−1 iy is only ( 2/3 1/6) = 1/9.
To deduce the Zeeman populations for a stationary atom we set up rate
equations for their temporal change. Using the Clebsch-Gordan coefficients for
pumping with linear polarization and accounting for spontaneous decay we have
2 2 2 1 1 1 1
Π̇0 = − Π0 + · Π0 + · Π1 + · Π−1
3 3 3 2 2 2 2
1 2 1 1 1
Π̇±1 = − Π±1 + · Π0 + · Π±1
2 3 6 2 2
In steady state (Π̇i = 0) we obtain from the second equation 9Π±1 = 4Π0 .
With Π−1 + Π0 + Π+1 = 1 we obtain the stationary populations the values
4/17, 9/17, and 4/17. In the presence of the laser fields the angular momentum
of a stationary atom in the ground state shows some alignment along the local
polarization vector ˆy = ˆY (z = 0).An important feature is that the π-transition
(∆M = 0) from |g0 iy is stronger by a factor of 4/3 than the π-transitions from
|g−1 iy and |g+1 iy . As a consequence the light shift of the |g0 iy state is 4/3
times larger than that of |g−1 iy and |g+1 iy :
3
∆1 = ∆−1 = ∆0 .
4
When the laser is red-detuned we have ∆0 < 0, the Zeeman-substates of the
ground state are pushed down in energy.
11.2. STATE OF MOVING ATOM 77

Next we explore the fluorescence spectrum under the condition that the
laser intensity is very weak, Ω1  Γ and the detuning is large, |δ|  Γ. We
expect a Raleigh line at the laser frequency ωL , corresponding to absorption and
emission back into the same Zeeman ground state. In addition we also expect a
Stokes-line at ωL + ∆0 /4 (when the atom starts in |g0 iy and emits into |g−1 iy
or |g+1 iy ) and an anti-Stokes line at ωL − ∆0 /4 (when the atom starts in |g−1 iy
or |g+1 iy and emits into |g0 iy ). At equilibrium the two side bands are of equal
strength, each side band being three times weaker than the central line.2
When the atom is at some other po-
sition along z, an identical equilib-
rium builds up, a stationary atom in
h , 1 = 34 h , 0 h , 1 = 34 h , 0
the ground state showing some align- h , 0
ment along the local polarization vec-
tor ˆY (z). As the laser intensity is con-
stant everywhere, the times it takes to g -1
y
g + 1
y
reach this equilibrium is independent
of z. What is different is that the laser g 0
y

polarization direction depends on z.


Hence the alignment vector of the ground-state angular momentum of our atom
is along the local polarization axis ˆY which rotates in space as the atom moves
along z. An atom moving along z is constantly asked to realign its angular
momentum vector to the local polarization axis ˆY . Work has to be done in
order to realign the atomic flywheel associated with the angular momentum,
with the consequence that dissipation appears.

11.2 State of Moving Atom


An atom moving along z with speed v covers a distance z = v t in a time span t.
The atom sees the linearly polarized em-field, the polarization vector rotating
in the x-y plane, oriented at an angle ϕ with respect to the y-axis,

ϕ(z) = −k z = −k v t . (11.3)

To treat this case D&CT transform into a frame centered at the atom and
rotating, such that the laser polarization in this new frame keeps the fixed
direction Y . This transformation introduces an inertial force (a pseudo force
which acts like a magnetic field along the z-axis) in form of an additional term in
the Hamiltonian, Vrot = kvJz , which couples the eigenstates |g±1 iy with |g0 iy .
This coupling is indicated in the Figure above by the thick arrows. It is pro-
portional to the atomic velocity v and involves a fictitious angular momentum,
Jz , pointing along the direction of translational motion.
When taking into account the coupling a set of new perturbed eigenstates |Gm iy
appears, which exhibit some orientation parallel to the z-axis. This oriented
ground state experiences the radiation pressure from the two laser beams dif-
2 We predict the relative strength in the spontaneous spectrum from a balance of events:
we multiply the equilibrium ground state population with the rate of excitation and multiply
it with the rate of spontaneous decay. In this way we get for the contributions to the central
9 · 2 · 2 + 4 · 1 · 1 + 4 · 1 · 1 = 6 , and for the Stokes line at ω − ∆ /4 (equivalently
line at ωL , 17 3 3 17 2 2 17 2 2 17 L 0
for anti-Stokes) 17 9 · 2 · 1+ 9 · 2 · 1 = 2 .
3 6 17 3 6 17
78 CHAPTER 11. ORIENTATIONAL COOLING

ferently. This is the actual origin for strong cooling at red detuning.

Transformation into the rotating frame:


We consider the wave function Ψ0 , which appears after transformation of the
laboratory-based wave function Ψ into the rotating frame. For the laboratory
based coordinate ~r we choose the unit vectors {êx , êy , êz }. In the moving and
rotating frame we choose the unit vectors {êX , êY , êZ = êz } for the coordinate
~ At time t = 0 both coordinate systems coincide at the position z = 0 where
R.
the polarization axis points along êy . The unitary transformation3 into the
rotating frame is accomplished by the rotation operator

T (t) = exp(−ikv t Jz /h̄) . (11.4)

The properties of the unitary transformation operator T and its adjoint T + are
T ~
Ψ(~r) −→ Ψ0 (R)
↓ H ↓ T HT +
T ~
HΨ(~r) −→ T HT + Ψ0 (R)

From this we read the relationships T Ψ = Ψ0 and Ψ = T + Ψ0 . We now transform


the time-dependent Schrödinger equation

∂Ψ
ih̄ = HΨ , (11.5)
∂t
by replacing Ψ with T + Ψ0

∂T + 0 ∂Ψ0 +
ih̄ Ψ + ih̄ T = HT + Ψ0 (11.6)
∂t ∂t
and applying the transformation T

∂T + ∂Ψ0
ih̄ T Ψ0 + ih̄ T T + = T HT + Ψ0 . (11.7)
∂t ∂t
Since T T + = 1 we obtain for the Schrödinger equation in the moving frame
∂Ψ0 ∂T +
T + T HT + Ψ0

ih̄ = − ih̄
∂t ∂t
Vrot + H0 Ψ0

= (11.8)

where Vrot + H0 is the Hamiltonian in the rotating frame.

Hamiltonian in the rotating frame :


The part of the Hamiltonian which is affected by this transformation is VAL . The
laser-atom interaction is proportional to the component of the dipole moment
which appears along the polarization vector of the field, ˆY , which is present at
3 Unitary transformations do not change the physics. An operator T is unitary if its inverse
+
T −1 is equal to its adjoint T + . As Jz is Hermitean, we have T + = eiJz /h̄ = eiJz /h̄ and
+
T T = 1, see [41]
11.3. COUPLING IN THE ROTATING SYSTEM 79

the position z. If we replace the positional coordinate z by vt we obtain for the


interaction in the moving frame
~ · ˆY = Dx sin kvt + Dy cos kvt .
VAL ∝ D (11.9)

The unitary transformation of VAL into the rotating frame is

T (t) [Dx sin kvt + Dy cos kvt] T + (t) = Dy . (11.10)

This simple result follows from an elaborate calculation [42] outlined in the box
on page 82.4 We retain only the dipole matrix element Dy . The price for this
simplicity is the additional term which appears as the first term on the right
hand side of Eq. (11.8), which D&CT refer to as inertial term. It arises because
T explicitly depends on time,
∂T + (t)
−ih̄ T (t) = kvJz = Vrot . (11.11)
∂t
In the following we consider the moving atom under quasi-stationary condi-
tions and account for the inertial term in the Hamiltonian by first-order pertur-
bation theory.

11.3 Coupling in the Rotating System


If we neglect the inertial term (11.11), the ground state levels {|gm iy } are eigen-
states of Jy . We may also describe these states in the basis of eigenstates of Jz ,
{|gm iz }. This we accomplish by a rotation of the basis by an angle π/2 along
the old x-axis (see Appendix A-5)
1 √ 
|g+1 iy = + |g+1 iz − |g−1 iz + i 2 |g0 iz , (11.12)
2
1  √ 
|g−1 iy = − |g+1 iz + |g−1 iz + i 2 |g0 iz , (11.13)
2
1  
|g0 iy = √ |g+1 iz + |g−1 iz . (11.14)
2
The eigenvalue of the Jz operator in the z-basis is Jz |gm iz = mh̄|gm iz . Hence
the matrix elements of the inertial term in the z-basis are

z hgm0 |kvJz |gm iz = h̄k v m δm0 m .

With this relationship we predict the matrix elements in the y-basis using the
expansion (11.12)-(11.14)

y hg+1 |Vrot |g0 iy = y hg0 |Vrot |g+1 iy = +h̄kv/ 2 (11.15)

y hg−1 |Vrot |g0 iy = y hg0 |Vrot |g−1 iy = −h̄kv/ 2 (11.16)
y hgm |Vrot |gm iy = 0 (11.17)
No diagonal elements of Vrot appear, hence the energy diagram of the ground
states on page 77 is not modified. Next we account for the couplings (11.15)
and (11.16) in first order perturbation theory.
4 The transformation (11.4) was chosen to accomplish such a simple result, the polarization

and interaction are constants in the moving coordinate system.


80 CHAPTER 11. ORIENTATIONAL COOLING

11.4 Motion-Induced Orientation


The goal is to find the effect of the inertial term on the energy-exchange balance
between the laser fields and the moving atom. The couplings (11.15)-(11.16)
among the ground-state levels {|gm iy } lead to perturbed states which we call
{|Gm iy }. From first order perturbation theory we obtain

kv/ 2
|G+1 iy = |g+1 iy + |g0 iy (11.18)
∆ 1 − ∆0

kv/ 2
|G−1 iy = |g−1 iy − |g0 iy (11.19)
∆1 − ∆0
√ √
kv/ 2 kv/ 2
|G0 iy = |g0 iy + |g+1 iy − |g−1 iy . (11.20)
∆0 − ∆ 1 ∆0 − ∆1
In this perturbed basis the expectation values of Jz are
1 h̄kv
y hG±1 |Jz |G±1 iy =
2 ∆1 − ∆0
h̄kv
y hG0 |Jz |G0 iy = (11.21)
∆0 − ∆1
The implication of these expectation values is that the amplitudes of the eigen-
states of Jz , |g−1 iz und |g+1 iz , are not equal but orientation along z is present.
If we assume for simplicity that the stationary population in the perturbed
states {|Gm iy } is about equal to that in the unperturbed states {|gm iy }, we can
weigh the terms in Equation (11.21) with the stationary populations 4/17, 9/17
und 4/17 and sum over all m-values to obtain an expectation value for Jz
 
stat h̄kv 9 2 2 20 h̄kv
hJz i = − − = . (11.22)
∆0 − ∆1 17 17 17 17 ∆0
This expectation value of Jz describes an orientation of the ground state pop-
ulation, induced by the atomic motion with velocity v. We may interpret this
result further by saying that the population in the states |g±1 iz differ by
20 kv
Πz+1 − Πz−1 = hJz istat /h̄ = . (11.23)
17 ∆0
This population difference increases with velocity and is inversely proportional
to the light shift, and this is the origin for the new cooling mechanism. It
is superior to Doppler cooling at low speeds and is most prominent when the
internal Zeeman re-pumping clock of the atom ticks about once in the time span
required for the atom to travel one wavelength.

11.5 Light Pressure on an Oriented Sample


For red detuning (δ < 0) the ground states are shifted down in energy, see
page 77. We consider an atom which moves into the direction z > 0, that is
v > 0. We conclude from Equation (11.23) that in this situation the state |g−1 iz
is more strongly populated than state |g+1 iz .
To estimate the consequence of this unequal population we consider the
Clebsch-Gordan coefficients for absorption of circularly polarized light for the
11.5. LIGHT PRESSURE ON AN ORIENTED SAMPLE 81

set-up of beams shown on page 76: σ + -light propagates along the z-axis while
the σ − -light propagates in the opposing direction. The probability that an atom
in state |g−1 iz absorbs a σ − photon is six times larger than the probability for
absorbing a σ + photon. The opposite holds for the |g+1 iz state. Since the
|g−1 iz is more strongly populated than |g+1 iz we conclude that the radiation
pressure from both beams is unbalanced: The atom scatters more photons from
the counterpropagating σ − beam than from the copropagating σ + beam. This
difference is not due to the Doppler effect but has its origin in the uneven
population of Zeeman levels of equal |m|, that is orientation. This orientation is
carried into regions of new alignment before spontaneous relaxation can achieve
equilibrium. The number unbalanced photons scattered per unit time is

Γ0
Γ0 Πz+1 − Πz−1 ∝

kv (11.24)
∆0

where Γ0 is the mean scattering rate of a ground state atom. Every σ − photon
transfers to the atom −h̄k and each σ + photon transfers +h̄k. The mean rate
of momentum transfer appears as independent of laser intensity

Γ0
F ∝ −h̄k 2 v. (11.25)
∆0

A detailed consideration [39] of the 0 .1 5 F O R C E ( U n it h k / /2 )


0 0 .0 4
speed dependence of the force gives 0

the result shown in the figure here for


the conditions Ω1 = Γ/4 and δ = −Γ/2. -0 .0 1 5

V E L O C IT Y
Near v = 0 the friction coefficient is u n it / /k
-1 1
very large. This cooling method op-
erates efficiently only at low atomic
speeds, k∆v ≈ ∆. As the speed
increases the result evolves into the
-0 .1 5
Doppler force result, Eq. (3.20), which
is shown by the dashed line.
The kinetic energy is not dissipated from the atom through a sisyphus-type ef-
fect, but rather because the atom scatters more photons from the beam counter-
propagating its motion than from the one co-propagating. At low speed it does
this much more frequently than expected from the Doppler-shift.

In conclusion the following dependence on laser intensity can be given :


friction capture range final temperature
Doppler cooling proportional to I independent of I independent of I
Pol.-gradients independent of I proportional to I proportional to I

Considerations so far were based on a continuous energy loss, averaged over the
distance of the order of a wavelength. At temperatures well below the Doppler limit one
enters a realm where the de-Broglie wavelength is of the order of the laser wavelength.
In this case the mean energy of the atom is of the order of the recoil energy (Rb at
780 nm has TR ≈ 700 nK).
To account for the discrete momentum transfer DCT [39] carried out a full quantum
treatment at low speeds using a simplified five level W -scheme, see below. Beginning
82 CHAPTER 11. ORIENTATIONAL COOLING

with an initial Gaussian distribution with a width of 15 recoil energies, a sample


peaking with a width of only two recoil speeds develops in a time of only 1000 Γ−1 .
This time corresponds to ≈ 10−4 s.

e -2 e 0 e 2
Parameters:

1 1
t = 2 0 0 0 / -1
σ+ − σ−
1 1
2 2 t = 1 0 0 0 / -1
δ = −Γ
-1
t = 3 0 0 /
Ω1 = 0.2Γ
t = 0
g -1 g + 1 -5 0 h k 0 5 0 h k

Proof of Eq. (11.10) by M. Gessner [42].


~ = (Dx , Dy , Dz ) is a vector operator, its components satisfy the
Since D
commutation relations with Jz (see e.g. Eq. (A3) in [39] or Chapter 6.6.5 in
Quantenmechanik (volume I) by C. Cohen-Tannoudji, D. Diu and F. Lalo)
[Jz , Dx ] = ih̄Dy , [Jz , Dy ] = −ih̄Dx . (11.26)
Defining the functions
fx (t) : = T (t)Dx T † (t),
fy (t) : = T (t)Dy T † (t), (11.27)
with T (t) = exp(−ikvtJz /h̄), we obtain using Eq. (11.26)
d
fx (t) = −ikv/h̄T (t)[Jz , Dx ]T † (t) = kvfy (t),
dt
d
fy (t) = −ikv/h̄T (t)[Jz , Dy ]T † (t) = −kvfx (t), (11.28)
dt
and
d2
fx (t) = −(kv)2 fx (t),
dt2
d2
fy (t) = −(kv)2 fy (t). (11.29)
dt2
These differential equations are solved by the ansatz A cos(kvt) + B sin(kvt).
Using Eqs. (11.27) and (11.28) we obtain the initial conditions
fx (0) = Dx , fy (0) = Dy , (11.30)

d d
fx (t) = kvDy , fy (t) = −kvDx , (11.31)
dt t=0 dt t=0

which yield the solutions


fx (t) = Dx cos(kvt) + Dy sin(kvt),
fy (t) = Dy cos(kvt) − Dx sin(kvt). (11.32)
Inserting this into Eq. (11.12) yields
T (t)[Dx sin(kvt) + Dy cos(kvt)]T † (t) = fx (t) sin(kvt) + fy (t) cos(kvt)
= Dy (cos2 (kvt) + sin2 (kvt))
= Dy . (11.33)
Chapter 12

VSCPT
Velocity Selective Coherent Population Trapping
All cooling mechanism discussed so far show under certain external condi-
tions (detuning, intensity, polarization) a friction term which slows the motion
of the atom. Spontaneous emission is always the reason that kinetic energy of
the atom can be removed from the atomic system. For example in the case of
sisyphus cooling: in the case of blue detuning and high intensity as well as in the
case of low intensity and red detuning it is spontaneous emission which extracts
potential energy which the atom previously gained from kinetic energy. Also in
Doppler cooling and in polarization gradient cooling with σ +−σ − light the blue
detuned spontaneous emission (relative to the absorbed photon energy) is the
origin of dissipation.
In all these laser cooling processes fluorescence never stops and as a conse-
quence the recoil energy ER should appear as an ultimate limit for cooling.
At this limiting temperature TR the
ER = kB TR = h̄2 k 2 /(2M ) (12.1) spatial extent of the atomic wavepacket
(deBroglie wavelength) is equal to the
Atom λL ER vR λdB a photon wavelength. In this situation
He 1082 ≈4 90 1082 the atom will average over effects which
Na 589 ≈1 30 589 change over the distance of a wave-
Cs 894 ≈ 0.1 3 894 length. In this case it is necessary to
nm µK mm/s nm also describe the motion of the atomic
a at v = vR
wave packet quantum mechanically.

There are however possibilities to cool below the recoil limit. In order to do
so spontaneous emission must cease. This can be achieved for certain laser-atom
combinations, for example für σ + −σ − –light and transitions between Jg = 1 →
Je = 1. In this case the rate of fluorescence of the atom can become zero, if
the atom is motionless and if the atom is given sufficient time to evolve into a
so called dark state. A dark state develops if the atom is pumped by two laser
beams into a non-absorbing superposition state of two ground state levels. In
case of VSCPT the superposition state is a combination of two ground-state
Zeeman levels with opposing linear (mechanical) momentum. This combination
is a perfect trap for atoms with v = 0, and less good so for |v| > 0 [43, 44, 45].
The VSCPT-scheme selects by itself (stochastically) atoms with velocities in the
vicinity of v = 0 and shields these from the unwelcome spontaneous emission,
for ever longer times the closer v is to zero.

83
84 CHAPTER 12. VSCPT

The enrichment of atoms in the velocity range near v = 0 has been termed
Velocity Selective Coherent Population Trapping . This enrichment de-
velops by itself in the process of the statistical distribution of recoil momenta
in spontaneous emission.

12.1 Dark States


A non-absorbing dark state appears in a three level atom (a so called Λ-system)
when pumped with two lasers as shown below. The lasers are detuned from
two-photon resonance by δ. The rate of fluorescence from the excited state |ei
and the absorption rate from the ground state shows a mininum near δ = 0.
In this two-photon resonant case the atoms develops into a quantum mechan-
ical superposition of |g1 i and |g2 i, such that the absorption amplitudes from
|g1 i to |ei and from |g2 i to |ei interfere destructively. The width in the dip in
the absorption profile is a measure of the inverse lifetime of the dark state, Γ0 ,
which may be interpreted as the energetic width of the ground state. It is much
smaller than the Rabi frequency and in the absence of collisions much smaller
than the width of the excited state, Γ. In the absence of dephasing collisions of
the superposition state the minimum on the absorption scale reaches zero.

e ) > I

M L 1
M L 2 /  /

@
g 2
g 1  @

As an example for such a Λ-system we consider the transition from metastable


helium 23 S1 ↔ 23 P1 . For this transition the Clebsch-Gordan coefficients are
shown in the figure below.
4
s H e s z
+ -

e e e e - e 0 e + e = 2 3 P
-1 0 + 1 1

s + s -
1 1 1 1
2 2 2 2
1 1 1 0 1
2
0
2 - 2
+ 2
g = 2 3 S 1
g -1 g 0 g + 1 g - g 0 g +

This Λ system is embedded in two counterpropagating beams with σ + σ − – light


of equal frequency in one dimension. There is no external field such that the
two Zeeman-ground state levels in the Λ-system are degenerate in energy and
each of the two states |g±1 i couples to the common excited state, |e0 i at the
same frequency.
12.2. OPTICAL PUMPING IN VELOCITY SPACE 85

1) Non-moving atom : A stationary atom sees both laser beams at the same
frequency and a coherent superposition of the states |g+1 i and |g−1 i forms which
is decoupled from |e0 i. This situation evolves when the amplitudes for excitation
from |g+1 i → |e0 i and |g−1 i→ |e0 i interfere destructively. The non-absorbing
state (the dark state) is the superposition state
1
ΨN C = √ (|g+1 i + |g−1 i) (12.2)
2
where the plus sign appears in this definition of the dark state because the re-
spective Clebsch Gordan coefficients in our example happen to have opposite
sign. The dark state is formed in the presence of the appropriate laser field. In
the absence of decoherence effects the dark state persists in this laser field, that
is the atom never reaches the excited state |e0 i. The dark state also persists
when both lasers are turned off concurrently and no external effects decohere
the dark state.

2) Atom moves along the z-axis: In this case the two-photon resonance
condition is no longer fulfilled as the Doppler effect leads to different frequencies
in the moving atomic frame. A simple argument shows that the excitation
amplitudes cannot be fully destructive in this case, only atoms with speed zero
can fulfill condition (12.2) for counter-propagating laser beams.
How can an atom reach the near-dark state condition v ≈ 0 ?
→ By chance the atom may fall into this velocity class in a spontaneous event.

12.2 Optical Pumping in Velocity Space


In each pumping cycle the atom is given the momentum h̄k which is taken away
again by the fluorescence photon. As a consequence there is an arbitrary mo-
mentum change in each absorption-fluorescence cycle. Therefore it may happen
that an atom with vR > v > dv ends after a fluorescence cycle in the velocity
class v < dv. This stochastic form of momentum diffusion may be viewed as
optical pumping process in velocity space. This process may move the atom
from an absorbing state into a nearly non-absorbing state but also vice versa.
Once atoms have landed in the vicinity of v = 0 they stay there for a long time
as this state is nearly dark. In this way a peak of atoms builds up at ever lower
speeds as time goes on.
R
N (v )

0 v 0 v

This cooling mechanism is radically different from the processes discussed so


far. One cannot easily identify a force of friction, rather it is the stochastic
combination of momentum diffusion and CPT which lets the atom get colder
the longer one waits. Limitations appear from processes which destroy the co-
herent dark state upon which the absorption cycle needs to begin again.
86 CHAPTER 12. VSCPT

12.3 Momentum-Families
We have previously always sharply defined the position of the atom, z = vt.
When the momentum uncertainty ∆p is smaller than h̄k, then the spatial co-
herence wavelength ζA = h̄/∆p exceeds the laser wavelength. In this case the
atomic wavefunction must be constructed from the product of a wavefunction
describing the internal state of the atom (discrete Hilbert space) and a second,
continuous one, which describes the translation of the atomic wavepacket.
For this we introduce basis states with quantum numbers for the internal
degrees of freedom and the translational state of the atom. In the absence of
spontaneous emission these basis states form closed families whose members are
coupled to each other by stimulated emission and absorption.
The state |e0 , pi describes an atom in the excited state with the linear mo-
mentum p along the z-direction. Due to momentum conservation this state
couples with a σ + -wave to the ground state |g−1 , p − h̄ki, as σ + light propa-
gates along the positive z-direction. This wave imparts in the atom in stimulated
emission a recoil by −h̄k, hence the change in the external state to |p − h̄ki.
On the other hand the σ − -wave couples the excited state in stimulated emission
only to the ground state |g+1 , p + h̄ki. The family of the three states
n o
F(p) = |e0 , pi, |g−1 , p − h̄ki, |g+1 , p + h̄ki (12.3)

is closed from families with momentum other than p as long as no spontaneous


emission occurs. In this basis the off-diagonal matrix elements from laser-atom
coupling are, see Eq. (6.11),
1 h̄Ω1
VAL |g−1 , p − h̄ki = − √ |e0 , pi (12.4)
2 2
1 h̄Ω1
VAL |g+1 , p + h̄ki = + √ |e0 , pi (12.5)
2 2
where Ω1 is the Rabi frequency (we assume that both lasers are of equal strength
and that p is so small that we may neglect the Doppler detuning of the two laser
beams). Note that the prefactor and the sign of the terms reflects the different
Clebsch-Gordan coefficients (see figure on page 84). If we now introduce the
so-called dark state (N C stands for non-coupled),
1 h i
|ΨN C (p)i = √ |g−1 , p − h̄ki + |g+1 , p + h̄ki (12.6)
2
we find that it is not coupled to the laser
h̄Ω1 h̄Ω1
VAL |ΨN C (p)i = − |e0 , pi + |e0 , pi = 0 . (12.7)
4 4
The absorption amplitudes from the two ground states end in the same final
state |e0 , pi, and they interfere destructively. This destructive interference oc-
curs only for two ground states which differ in momentum by 2h̄k. In the absence
of the two laser fields, an atomic wavepacket in the state |ΨN C (p)i expands in
space at a speed 2h̄k/M . In the presence of the laser field this expansion is
interrupted by the balanced stimulated processes.
12.4. MOTION INDUCED ATOM-LASER COUPLING 87

Equivalently we may define a coupled state (bright state, Hellzustand )


1
|ΨC (p)i = √ [ −|g−1 , p − h̄ki + |g+1 , p + h̄ki ] (12.8)
2
This combination is orthogonal to (12.6). For the coupled state we have con-
structive interference of the absorption amplitudes. An atomic wavepacket in
the bright state state coupled to the laser field,
h̄Ω1 h̄Ω1
VAL |ΨC (p)i = + |e0 , pi + |e0 , pi =
6 0. (12.9)
4 4
We conclude :
• If the atom is in state |ΨN C (p ≈ 0)i it never visits the excited state and
will not participate in flourescence.
• If the atom is in the bright state |ΨC (p ≈ 0)i it is frequently pumped into
the excited state, giving rise to spontaneous emission.

12.4 Motion Induced Atom-Laser Coupling


We assume that the internal energy of the two Zeeman ground states is identical
(no external magnetic field). We define these states to be at the potential
energy Epot = 0. Then the atomic Hamiltonian contains a contribution from
ext int
translational motion HA and an internal part HA :

ext int p2
HA = HA + HA = + h̄ω0 |e0 ihe0 | . (12.10)
2M
In the absence of laser-atom coupling the eigenstates of the family F(p) are
described by the Schrödinger equations
(p ± h̄k)2
HA |g±1 , p ± h̄ki = |g±1 , p ± h̄ki (12.11)
 2M
p2

HA |e0 , pi = + h̄ω0 |e0 , pi . (12.12)
2M
It follows for the dark state that
(p−h̄k)2 (p+h̄k)2
 
1
HA |ΨN C (p)i = √ |g−1 , p−h̄ki + |g+1 , p+h̄ki (12.13)
2 2M 2M
After introducing Eq. (12.6) and (12.8) we obtain with p = M v,
 2 
p
HA |ΨN C (p)i = + ER |ΨN C (p)i + h̄kv |ΨC (p)i (12.14)
2M
 2 
p
HA | Ψ C (p) i = + ER |ΨC (p)i + h̄kv |ΨN C (p)i (12.15)
2M
HA shifts the energy of dark and bright state by p2 /2M + ER and causes a
motional coupling between bright and dark state
hΨC (p)|HA |ΨN C (p)i = h̄kv . (12.16)
of magnitude equal to the Doppler frequency shift kv times h̄.
88 CHAPTER 12. VSCPT

12.5 Decay Due to Spontaneous Emission


We assume that there are no collisions with other atoms which lead to a change
in atomic momentum. Under these conditions and in the presence of the atom-
laser coupling VAL the family of the three states

F(p) = { |e0 , pi, |g−1 , p − h̄ki, |g+1 , p + h̄ki } (12.17)

is closed as long as no spontaneous event occurs. We may equally represent this


family of three states by the orthogonal states ΨN C , ΨC und e0 :
n o
F(p) = |e0 , pi, |ΨC (p)i, |ΨN C (p)i (12.18)

If we introduce the detuning δ = ωL − ω0 the energy of the excited state is in


the dressed state picture h̄δ̄ = h̄ωL − h̄ω0 + ER . The energy gap between |e0 , pi
and |ΨC (p)i and |ΨN C (p)i is equal to −(h̄δ + ER ) = −h̄δ̄.
e 0 ,p

9 1
2 @

k p /M
/ C' / N' C
O C (p ) O 'N C (p )

In this representation two couplings appear between members of the fam-


ily: States |e0 , pi and |ΨC (p)i couple with VAL = h̄Ω1 /2. States |ΨC (p)i and
ext
|ΨN C (p)i couple with HA = h̄kp/M = h̄kv. In the absence of these couplings
ext
the sole unstable state is |e0 , pi with the natural width Γ. Due to VAL and HA
the states |ΨC (p)i and |ΨN C (p)i obtain a component of |e0 , pi and hence finite
energy widths. These we name Γ0C and Γ0N C .
The temporal evolution of this family is given by the effective (non-hermitian)
Hamiltonian [46]:

−δ̄ − iΓ/2 Ω1 /2 0
 
 
Heff = h̄ 
 Ω1 /2 0 kv  (12.19)
 
0 kv 0

which gives us three complex eigenvalues with differing imaginary components.

Physical interpretation of the three decay modes : We assume that the


off-diagonal elements (the couplings) in Eq. (12.19) are small.

1. special case, p = 0
1  
|ΨN C (p = 0)i = √ |g−1 , −h̄ki + |g+1 , h̄ki (12.20)
2
12.5. DECAY DUE TO SPONTANEOUS EMISSION 89

This state is decoupled from the two other family members of (12.18). An atom
in this state stays in this state indefinitely, a perfect trap for atoms as long the
two lasers stay turned on and there are no velocity changing collisions. We find
the decay rates of the two other states by diagonalizing
!
−δ̄ − i Γ/2 Ω1 /2
Heff (p = 0) = h̄ . (12.21)
Ω1 /2 0

The excited state emits at the rate Γ. When Ω1  Γ and Ω1  |δ̄| the complex
eigenvalue of the bright state is
2
0 (Ω1 /2)
δ̄C − i Γ0C /2 = . (12.22)
δ̄ + i Γ/2
This value tends to zero for Ω1 → 0. The imaginary part of this eigenvalue
Ω21
Γ0C = Γ , (12.23)
4δ̄ 2+ Γ2
represents the energetic width of the bright state. The bright state obtains a
finite width due to optical pumping of the magnitude Γ0C ∝ Ω21 . This width rep-
resents a contamination of |ΨC (0)i by the excited state. The real part represents
a small energy shift of magnitude

0 Ω21
δ̄C = δ̄ . (12.24)
4δ̄ 2 + Γ2

2. p is very small but not zero :


Now we account for a motional coupling h̄kv between |ΨC (0)i and |ΨN C (0)i
We assume that it is much smaller than the light shift or the widths from
Eq. (12.23,12.24), which describe the case p = 0. In this case the Eigenvalues
of two of the three states are still very close to those of Eq. (12.21), namely
− δ̄ − i (Γ/2) (12.25)
0
δ̄C − i (Γ0C /2) (12.26)
nd
We derive the energy of the third state by 2 -order perturbation theory under
consideration of the coupling h̄kv between |ΨN C (0)i and the perturbed state
|ΨC (0)i with the eigenvalue (12.25). We obtain

0 0 k2 v2
δN C − i ΓN C /2 = 0 − i Γ0 /2
. (12.27)
−δ̄C C

As energy width of the uncoupled state we obtain in the presence of a small


motional coupling the imaginary part
4 k2 v2 4 k2 v2
Γ0N C (p) = Γ0C 2 =Γ , (12.28)
4 δ¯0 + Γ0 2 Ω21
C C

and a minor energy shift from the real part

0 0 4 k2 v2 4 k2 v2
δ̄N C (p) = δ̄C 0 2 0
= δ̄ . (12.29)
4 δ̄C + ΓC 2 Ω21
90 CHAPTER 12. VSCPT

We see that the state |ΨN C (p)i is contaminated by |ΨC (p)i via the off-diagonal
coupling kv in Eq. (12.19). As a consequence even the dark state obtains a
small contamination by the excited state, the contamination diminishing as the
atomic velocity v gets smaller. By waiting long enough we automatically select
atoms which are near p = 0. These atoms appear then in a linear superposition
of the two ground states |g−1 i und |g+1 i in different motional states, p + h̄k and
p − h̄k. The residual momentum width ∆p decreases the longer we wait, the
waiting time Θ being characterized by
Θ > 1/Γ0N C (p) . (12.30)
This time describes the time the atom was allowed to develop into the dark state
in the two laser fields. Inserting (12.28) we obtain for the residual momentum
M Ω1
∆p < √ √ (12.31)
2k Γ Θ
a value not bound by the recoil momentum.

Spontaneous transfer between families If an atom leaves a family by


spontaneous emission it generally ends up in a family of different momentum.
If we consider the 3-D case but do not interest us for the magnitude of the mo-
menta along the x and y-coordinates, we see that the momentum change along z
must lie between −h̄k and +h̄k. The following diagram shows two possible final
states of the z-component of the momentum after a spontaneous emission of a
photon (wavy line), followed by stimulated absorption / emission (dashed lines).
The momentum component along z lost in spontaneous emission is labelled u.
From F(p) a new family grows with either p − u − h̄k (left) or p − u + h̄k (right).
e 0 ,p -u -h k e 0 ,p e 0 ,p e 0 ,p -u + h k

. (p -u -h k ) . (p ) . (p ) . (p -u + h k )

g -1 ,p -u -2 h k g -1 ,p -h k g -1 ,p -u g -1 ,p + h k g -1 ,p -h k g -1 ,p -u g 1 ,p + h k g -1 ,p -u + 2 h k

If the system lands in a family with smaller momentum, the atom will remain
for a longer time in this family before undergoing renewed absorption. The dis-
tribution of waiting times for an atom to absorb appears as a Lévy distribution.
-1 -1 -1 -1
P (p z) 3  =  5 0 / 3  =  1 5 0  / 3  =  4 0 0  / 3  =  1 0 0  0  /

-h k 0 + h k p z
-h k 0 + h k p z

This model was used in stochastic simulations of the temporal development of


the momentum distribution, see also § 8.2 in the book of Breuer & Petruccione
[37]. Stochastic quantum jumps finally force the atom into the uncoupled dark
state near p = 0, see figure above (time steps in units of the spontaneous lifetime).
12.5. DECAY DUE TO SPONTANEOUS EMISSION 91

The state |ΨN C (p)i is not an eigenstate of the momentum operator. A


measurement of the momentum of atoms in the dark state should reveal two
expectation values, p + h̄k and p − h̄k.

This was observed in VSCPT experiments in


one, two and three dimensions [47, 48, 49].
Shown here is the 2D result [48]. After turn-
ing off both light fields and a gravitational fall
over 5 cm atoms are observed with momentum
positions ±h̄k along the laser beams. The dis-
tance between impact positions is about 2 cm.

How does the excited state know that after spontaneous emission it
should end up in either the dark or the bright state ?
It is perfectly reasonable to ask how the atom, in the above case with p near
zero, chooses to enter either the dark or the bright state. Would the foremost
decision of the atom in the excited state |e0 i not be to either radiate into state
|g−1 i or else |g+1 i? The quantum mechanical answer is into both until either the
photon emitted is detected or the ground state formed is interrogated further.
A very useful quantum-mechanical interpretation was introduced by Aspect
and Kaiser1 as briefly outlined in the following:
Spontaneous emission from state |e0 i is a measurement process with two
possible outcomes, the final state being |g−1 i or |g+1 i. This is equivalent to the
case of a spin 12 particle in a Stern-Gerlach magnet. When oriented along z the
magnet selects, with equal probability spin up | ↑iz or | ↓iz . If we carry out
an additional measurement with a second Stern-Gerlach magnet, now oriented
horizontal along x, the outcome of such a measurement will be | →ix or | ←ix ,
again with equal probability. In fact
 √  √
| ↑ iz = | →ix + | ←ix / 2 | ↓iz = | →ix − | ←ix / 2,
 √  √
| →ix = | ↑ iz + | ↓ iz / 2 | ←ix = | ↑ iz − | ↓ iz / 2 .

In this manner falling into the dark or bright state is the consequence of a two
stage measurement process, since equivalently
 √  √
|g+1 i = |ΨN C i + |ΨC i / 2 |g−1 i = |ΨN C i − |ΨC i / 2,
 √  √
|ΨN C i = |g+1 i + |g−1 i / 2 |ΨC i = |g+1 i − |g−1 i / 2 ,
the second measurement being done by the laser fields. A quantitative experi-
ment was recently made showing that the atom decides with 50 % probability
to end up in the dark state and with 50% probability in the bright state.[51]

1 Linear Momentum Conservation in Coherent Population Trapping: A Case Study for a

Quantum Filtering Process, A. Aspect and R. Kaiser, Foundations in Physics 20 1413 (1990).
[50]
92 CHAPTER 12. VSCPT
Chapter 13

Evaporative Cooling

13.1 Bose-Einstein Condensation


When we consider atoms as quantum mechanical wavepackets of the size
1/2
πh2

h h
hλdB i = = = (13.1)
hpi M hvi M 8kB T

it is easy to accept that at low temperatures and at sufficiently high density ad-
jacent atoms overlap in their spatial wavefunctions and their indistinguishability
comes into play. A quantum mechanical phase-transition occurs when the mean
distance between atoms is of the order of the de-Broglie wavelength. In this sense
identical particles may occupy a coherent macroscopic quantum state. For non-
interacting particles this occurs when the product of the de-Broglie wavelength
and density exceeds the critical value

N · λ3dB = 2.612 . (13.2)

Provided all atoms occupy the same ground state we may describe the macro-
scopic wavefunction as product of single-particle wave functions. The surprising
aspect of Bose-Einstein condensation is that this phase transition occurs in the
absence of an interaction between the particles as we know it from phase transi-
tions in real gases. Interaction as it usually occurs between particles appears as
a perturbing factor. Bose-Einstein like condensation 1 can occur when a small
interaction is present between the atoms and when the mean collision rate is
small enough to avoid molecule formation in three-particle collisions. Indeed it
was believed for a long time that at sufficient high density (as required by the
critical phase density) condensation into a solid phase would occur prior to BEC.

BEC was first achieved with alkali atoms in 1995 [52, 53, 54], then with
atomic hydrogen in 1998 [55], in 2001 with He-atoms [56] then Cromium atoms
[57] and recently also with molecules such as Li2 [58]. 2

1 In the sense of existing for a particular time span.


2A list of atoms for which BEC was achieved to date can be found in [59].

93
94 CHAPTER 13. EVAPORATIVE COOLING

All BEC experiments begin with a pre-cooling stage in a magneto-optic trap.


In the MOT the temperature is limited by spontaneous emission and the max-
imal density by radiation trapping (reabsorption of spontaneous photons), as
well as by accelerating collisions of excited atoms with ground state atoms. In
a standard MOT the product N · λ3dB is about six orders of magnitude below
the critical value of Eq. (13.2) A key feature in finally raising the phase space
density was so far always achieved by evaporative cooling in the complete ab-
sence of resonant light sources. Evaporation is usually performed in magnetic
traps, but also in optical dipole traps.
In a gas at high temperature the
mean distance between atoms
H ig h T :
(∝ N 1/3 ) is typically much larger
than the size of the atom.
" B illia r d b a lls "
v
d

At low temperature the atomic size


l d B L o w T : may be defined in terms of the scat-
"W a v e p a c k e ts " tering length for s-wave scattering.

T = T c : The macroscopic containment is


B E C l dB » d
formed by an external trap. For a
" M a tte r w a v e o v e r la p "
harmonic potential the spatial extent
of the ground state wavefunction (the
T = 0 :
defining size of an ideal condensate)
P u re B o s e c o n d e n s a te is
" G ia n t m a tte r w a v e "
p
aHO = h̄/M ωHO . (13.3)

Evaporative cooling does not have the temperature and density limits which
we know from laser cooling. In evaporation the high energy component of the
atoms energy distribution is selectively removed from the sample and by ther-
malization of the residual atoms a lower temperature is achieved. This method
is always connected with a substantial (> 99%) loss of atoms. The magnitude
of the elastic cross section is crucial in how fast the re-thermalization occurs.
This method is applied repeatedly and in the case of magnetic traps a runaway
condition can be achieved in the sense that the increase in atom density associ-
ated with lowering the temperature, accelerates the re-thermalization. Typical
time spans are (1 − 30 s). Longer time spans are impractical as the residual
background gas density is usually leading to heating and further loss of atoms.
The largest condensates shown today are made of about 2 × 107 atoms in the
case of sodium and by about 1 × 109 atoms in the case of atomic hydrogen.
13.2. ATOM-ATOM INTERACTIONS 95

Bosonic atoms
Atoms with integer total spin are bosons. This happens when the sum of pro-
tons, neutrons and electrons is even. With the exception of beryllium all neu-
tral elements have a bosonic member in their isotopes. Under which conditions
can we consider particles which are built from fermions as pointsize bosons ?
When the internal excitation energy is much larger than kB T then the internal
degrees of freedom are practically frozen and irrelevant for a thermodynamic
description of the particle. Bosons of one isotope may appear with different
spin structure, specifically hyperfine states with small energy differences. An
example are H(1s)-Atoms which appear in F = 1 und F = 0 total angular
momentum quantum numbers. This provides the possibility of forming BECs
of different total angular momentum components.

13.2 Atom-Atom Interactions


Dilute gases differ in their Bose-Einstein condensation from the BEC examples
in condensed matter (suprafluidity of 4 He, BEC of Cooper-pairs) in that the
strong and complex multi-particle interaction in condensed matter plays an only
minor role in a dilute gas. For example when spin-polarized hydrogen atoms
approach each other they form the molecular state H2 3 Σ+ u . This state is only
very weakly bound due to van-der-Waals interaction. The potential shows a
minimum of about −5 cm−1 at around 8 a.u. and supports no bound vibrational
states. Hence spin-polarized thermal H-atoms can approach to about 4 a.u.
without forming a bound diatomic molecule. By comparison the zero point
energy in the electronic ground state of hydrogen, H2 1 Σ+ g is of the order of
−36500 cm−1 , corresponding to -4.52 eV. Only by spin-changing collisions may
spin-polarized hydrogen atoms transfer into the singlet combination to form the
tightly bound singlet ground state 1 Σ+g . For this reason one originally assumed
that spin-polarized H-atoms were ideal objects to realize BEC in a dilute gas.

40 000 4000

3 +
20 000 Su 2000
energy Hcm-1 L

energy Hcm-1 L

3 +
HH1sL + HH1sL Su RbH5sL + RbH5sL
0 0
M 30
-20 000 1 + -2000
Sg 1 +
Sg M 31
-40 000 -4000
0 2 4 6 8 10 5 10 15 20 25
R HbohrL R HbohrL

Alkalis have potential energy curves that are similar to those of H2 but their
triplet form is more tightly bound and due to their higher mass their zero point
energies are much smaller. In alkali condensates the three-body recombination

Rb + Rb + Rb → Rb2 + Rb + ∆E (13.4)

limits the maximal density of the condensate and its lifetime, but not the fact
that condensation can be reached. In this sense the three-body reaction also
limits the lifetime of the condensate. The rate coefficient for reaction (13.4) is
96 CHAPTER 13. EVAPORATIVE COOLING

near 10−28 cm6 s−1 . Typical lifetimes for condensates are in the range from sec-
onds to minutes. This is basically the time that it takes at typical temperatures
(0.5-2 µK) and densities of 1014 -1015 cm−1 until the energy gain in molecule
and cluster formation leads to the evaporation of the condensate. In process
13.4 the binding energy of the molecule is released into translational energy. For
Rb2 1 Σ+
g the energy release is about 0.6 eV.

13.3 Collisions of Cold Atoms


Scattering of two atoms which interact via a central potential U (R) with each
other corresponds to the scattering of one particle of the reduced mass µ in the
relative coordinates of the two particles. In classical scattering one introduces
a continuous impact parameter b. If the relative speed of the collision partners
is v0 and the impact parameter is b the angular momentum of the scattered
particle is defined as

Lclass = µ b v0 . (13.5)

In quantum mechanics one describes scattering as the superposition of an in-


coming plane wave and a spherical wave emerging from the scattering center

eikR
Ψ ∝ eikz + f (k, θ) , (13.6)
R

where f (k, θ) is the scattering amplitude. As shown in Appendix A-4 the scat-
tering amplitude is defined p in terms of the phase shifts δ` of the partial waves
of angular momentum L = `(` + 1) h̄


1 X h i
f (k, θ) = (2` + 1) e2iδ` (k) − 1 P` (cos θ) (13.7)
2ik
`=0

The phase shifts result from a solution of the radial equation in the central
potential

d2
 
2 `(`+1) 2µ
uk,` (R) + k − + U (R) uk,` (R) = 0 (13.8)
dR2 R2 h̄2

for the amplitudes of the partial waves with the boundary condition uk,` (0) = 0.
The second term in Eq. (13.8) gives the centrifugal potential. The following
picture gives the effective potential experienced by the collision partner

`(` + 1) 2
Ueff (R) = U (R) + h̄ (13.9)
2µR2

for three values of the quantum mechanical angular momentum.


13.3. COLLISIONS OF COLD ATOMS 97

Effective potential for the triplet


8
6 {=2 ground state of Rb2 . As long-range po-
Ueff HMHzL
tential we use the van-der-Waals form
4
2 {=1
U (R) = −C6 /R6 (C6 = 4700 a.u.)
0
-2 {=0 The respective centrifugal barriers are
-4 shown dashed. An ` = 2 partial wave
0 100 200 300 400 at 4 MHz kinetic energy never gets to
R HbohrL know the bound part of the potential.

The total elastic scattering cross section is given by the integral over the
squared scattering amplitude. In terms of the partial waves as the sum over the
contributions of all scattering phases δ` :

Z ∞ ∞
8π X
σ(k) = 2π |f (k, θ)|2 sin θ dθ = (2` + 1) sin2 δ` (k) (13.10)
0 k2
`(gerade)

The sum in (13.10) is written for identical bosons. In this case only even partial
waves contribute (a consequence of the indistinguishability of scattering angles θ
and π +θ), see Appendix A-4. The value δ` gives the phase shift √ of the scattered
wave relative to the incoming wave. The wavenumber k = 2µE/h̄ gives the
number of oscillations of the continuum function per unit length. The effective
potential is specific for each partial wave and so is each phase shift. We may
accept from the above picture that at low temperature3 the higher partial waves
do not feel the molecular potential, they are reflected at the long range repulsive
wall of the effective potential which is barely modified by molecular forces. At
low temperature the contribution from s-wave scattering dominates (` = 0). In
this case the total cross section is


σ0 (k) = sin2 δ0 (k) . (13.11)
k2

To illustrate the meaning of the phase shift we show in the following numer-
ical solutions of the radial Schrödinger equation for a potential energy curve
of 87 Rb2 . At the left solutions for bound states are shown (E < 0). These
states correspond to vibrational states of the Rb2 -molecule.4 The right picture
shows examples for continuum wavefunctions (E > 0) for the triplet ground state
87
Rb(5s)+87 Rb(5s). At large distances the continuum wavefunctions appear as
sin(kR + δ), a wave shifted by δ.

3 The relative energy of Rb-atoms at T = 100 µK corresponds to ≈ 3 MHz.


4 The potential used in the left picture corresponds to the interaction of Rb(5p)+Rb(5s) for
which the long range attractive potential U (R) = −C3 /R3 . The solutions for Rb(5s)+Rb(5s)
are similar but are restricted to much shorter range. The highest bound state in the potential
shown at the right (which is the triplet ground state) is at -1.3 cm−1 ≈ −39000 MHz!
98 CHAPTER 13. EVAPORATIVE COOLING

20

0 16
15
3
10
GHz 6 MHz 8
5
9 4
2
1
12 0

100 200 300 400 500 100 200 300 400 500
R bohr R bohr

Without potential the solution of the radial equation (13.8) is



2µE
uk (R) ∝ sin kR mit k= (13.12)

for E > 0. With inclusion of the potential we have at large separation for s-wave
scattering for E > 0
δ0
uk,0 (R) ∝ sin k(R + ) (13.13)
k
The scattering length a0 (for the partial wave ` = 0) is defined as the limit
δ0
− lim ( ) = a0 (13.14)
k→0 k

12
20
∆0 k
10 15

10
MHz 8 MHz
5

6
0

10 20 30 40 50 60 70 80 100 200 300 400 500 600 700


R bohr R bohr

We see at the right picture that the scattering length is about a0 = +90 bohr.
The potential pushes the wave from the origin. For positive scattering lengths
the potential appears repulsive.

There is also the case of a negative scattering length. The exact magnitude
and sign of the scattering length depends on the location of the highest bound
state. If the highest bound state is just below the dissociation limit (E = 0), the
scattering length is positive and the potential acts repulsive for low energy par-
ticles. If however an additional bound state would just fit into the bound region,
if the potential minimum was slightly more attractive, the scattering length is
negative and the potential appears attractive. If the last bound state5 is exactly
5 The location of the last bound state in the potential depends on the area suspended by
13.3. COLLISIONS OF COLD ATOMS 99

at the limit, then the scattering phase is δ = π/2 and the scattering length is
a0 → −∞. When a bound resonance passes through E = 0 the phase shift jumps
by π and the scattering length jumps from −∞ to +∞.

With the definition (13.14) the total cross section for s-wave scattering
σ0 = 8πa20 . (13.15)
87
The cross section for Rb-atom is about
σ0 = 8π 902 = 2 × 105 bohr2 = 5 × 10−12 cm2 (13.16)
This cross section determines the rate of thermalization of cold atoms
Rtherm = N v̄ σ0 (13.17)
At a density of N = 1012 cm3 the rate is
Rtherm ≈ 5 × 10−12 3 × 101 × 1012 = 15 s−1 (13.18)
To calculate the rate at room temperature at least all partial waves for which
the maximum in the effective potential is below the collision energy E have to
be included. For U (R) = −C6 /R6 the maximum in the effective potential is

dUeff (R) 6C6 `(` + 1)h̄2


= 7 − =0 (13.19)
dR R µ R3
From this we obtain for the position of the potential barrier in the effective
potential
 6µC 1/4
6
RB = (13.20)
`(` + 1)h̄2
and for the height of the barrier
3/2
`(` + 1)h̄2

EB = −C6 (13.21)
6µC6
` values which need to be included at room temperature (EB ≤ E ≈ 0.001 a.u.)
6
are at least
 2/3
EB 6µC6 6 · 1836 · 87 · 4700
`(` + 1) ≥ ≈ ≈ 4002 (13.22)
C6 h̄2 (4700/0.001)
2/3

At room temperature partial waves up to (` > 400) have to be included. The


elastic cross section at room temperature determines the rate of loss of cold
atoms from a trapped cold sample (Rloss = Nrg v̄ σ). Nrg is the residual (un-
cooled) gas density. At 10−8 mb and typical MOT-Parameters the loss rates are
of the order of 0.1−1 s−1 .
the potential energy curve below the energy E < 0. A semiclassical quantization condition is
  Z Rmax r
1 µ
h̄ v + = dR [E − U (R)]
2 Rmin 2
where v is the vibrational quantum number and Rmin , Rmax are the classical turning points
in the potential.
6 300 K corresponds to 30 meV, hence ≈ 1/1000 of an atomic energy units.
100 CHAPTER 13. EVAPORATIVE COOLING

13.4 Mean-Field Approximation


BEC was formulated to occur in an ideal gas. In real experiments one nearly
always deals with a real gases, the approximation ideal gas being only valid at
extreme dilution. At extreme dilution the phase space density N λ3dB can only
be reached for extremely cold atoms. s-wave scattering dominates for cold atoms
and a criterion
√ for dilution is N a30  1, saying that the mean distance between
3
atoms (∝ 1/ N ) is very much larger than the s-wave scattering length a0 . In
this case one may approximate the interaction between particles in a condensate
as the sum of all pair interactions (never are three particles close together). The
pair interaction is accounted for via the scattering length for s-wave scattering
through a contact potential term
V (~r) = g δ(~r) (13.23)
where g is a measure for the strength of the pair interaction energy density,
4π h̄2 a0
g= . (13.24)
M
To describe the ground state of a BEC which is trapped in an external potential
Ve (~r) under inclusion of the pair interaction one may use the Gross-Pitaevskii-
equation [60, 61]
h̄2
 
∂ 2
ih̄ Φ(~r, t) = − ∆ + Ve (~r) + g|Φ(~r, t)| Φ(~r, t) . (13.25)
∂t 2M
This equation describes the motion of the condensate wavefunction in the ex-
ternal field and in the molecular field g|Φ(~r, t)|2 of all condensed atoms. The
density distribution in the condensate is given by N (~r, t) = |Φ(~r, t)|2 . The
strength and the sign of the interaction term gN (~r) in comparison with the
external potential determines the stability of the condensate. For negative scat-
tering lengths we have g < 0 and a stable condensate is only observed at low
density for which Ve (~r) + gN (~r) > 0. A nice intuitive picture of the influence of
the size of the scattering length on the effective size of the trap can be found in
Burnett et al. [62].

13.5 Magnetic Trapping


Most observations of Bose-Einstein Condensation in dilute gases were made
following rf-induced evaporative cooling in magnetic traps. To understand this
concept we first discuss magnetic trapping.
The first report on magnetic trapping of neutral atoms appeared in 1985
[63]. Atoms with unpaired electrons (such as alkalis) have magnetic moments
µm of the order of 1 µB . In an external magnetic field the energy of a magnetic
dipole is
EmF = −~ ~ = −|µm ||B|cosθ
µm · B
= gF mF µB |B| (13.26)
where the classical view of a fixed value of the angle θ (the angle between the
magnetic moment and the external field) corresponds to the quantum mechan-
ical concept that an atom at a stationary value of B is found in a discrete state
13.5. MAGNETIC TRAPPING 101

mF where −F ≤ mF ≤ F .
Requirement for trapping in static magnetic fields is the presence of a local mini-
mum of the potential energy (13.26). The so called weak-field seeking states (the
states for which gF mF > 0) are accelerated towards the minimum of potential
energy. On the other hand the strong-field seeking states (gF mF < 0) are accel-
erated towards the maximal potential energy. In the case of static fields there
is no maximum of B in free space. Hence strong-field seeking states cannot be
trapped in static magnetic fields.
An atom moving in an inhomogeneous ' $ % ( #

field sees the field changing in mag- ! " "


' $ % ( )
nitude and direction. This variation # " "

may induce transitions among mF - $ % # ' $ % "

components. The atom is stable in a &# " "

fixed mF state only if the rate at which ' % &)

- . / 01 2 345 6 7 8
$
&! " "
the direction of the field vector changes
' % &#
is small compared to the classical rate
$

&* ! " "


of precession of the magnetic moment. ' % ( )
&* * " " $

dθ gF µB |B|
< ωLarmor =
$ % ) ' % "
$

dt h̄ &+ " " "


' $ % &)
This adiabatic condition is not given ) " " # " " , " " ! " "
5 9 1 . / :;< 3=;/ >? 3' 9 1 . ;:@ ? / 34 A 9 @ B B 8
for atoms which visit the center of a
quadrupole trap where B → 0.
In this region atoms may undergo a spin flip (Majorana transition) and change
from a low field seeking state to a high field seeking state and be lost. In order
to avoid this loss additional static magnetic fields are introduced.

Such a magnetic field configuration can


be realized in a Ioffe-Pritchard trap [64].
This DC trap is cylindrically symmetric,
The bias field B0 is oriented along the
symmetry axis [65] and is generated by
the two circular coils. The four straight
line segments carry DC current in oppos-
ing direction and establish the radial gra-
dient.
The different strength of confinement in
the radial and axial directions is im-
printed in the shape of the condensate.
102 CHAPTER 13. EVAPORATIVE COOLING

Shown are the Zeeman energies ! " # $


of F = 2 as a function of trap ( & &

position for nonzero value of $ & &


! " # %

the magnetic field at the cen-

) * + ,- . /01 2 3 4
ter, (B0 6= 0). Now atoms in the & ! " # &

mF = 2 state are trapped but


'$ & &
rf-transitions between Zeeman ! " # '%

states may be induced by appro- '( & &


! # '$
priate rf-fields. In this fashion
"

one may control the forced evap- 5 6 7 8986 * /8* /9, : ; ,

oration of hot atoms.


13.6. EVAPORATIVE COOLING 103

13.6 Evaporative Cooling


Hess [66] proposed this concept in 1986. By selective removal of the hottest
atoms, residual atoms will re-thermalize at a lower temperature provided colli-
sions between atoms are active in thermalization.

Ti Tf Ti
Ecut
fE fE fE

E E E
rf-evaporation : A Zeeman transition can be induced by an external rf-field
of suitable frequency. In this way a transition from a trapped state into an un-
trapped state may occur. The spatial dependence of resonance energy for such
transitions is governed by the Zeemann-effect. Therefore the choice of frequency
governs the height of the potential barrier above which atoms are transferred
from bound to unbound states. The rf-knife selectively removes the hottest
atoms, see black arrows in the picture on page 102. The radial coordinate for
the resonance condition is given by

µB |gF | B = h̄ωrf (13.27)

The potential energy of an atom in the Zeeman state mF is

V (mF , ~r) = mF µB gF [B(~r) − B0 ] . (13.28)

In this fashion atoms with total energy above

Etot > h̄mF (ωrf − ω0 ) (13.29)

can be selectively removed. The energy h̄ω0 corresponds to the energy sepa-
ration of Zeeman-components at the trap center, the position of the minimum
magnetic field.7 By reducing the rf-frequency ωrf with time one may lower the
height of the barrier without changing the form of the magnetic trap potential.
The deviation from the linear Zee-
man Effect (dashed lines in figure 9 ) : ;
( ) *+ , -./ !
on page 102) plays a significant role.
9 ) : <
Due to the interaction of states with ' "
#
equal quantum number mF which $
% , / 6> ?

9 ) : 0

originate from F = 2 and F = 1, %


&

states with equal mF are pushed 9 ) : =<

apart stronger than predicted by the


linear Zeeman-effect (transition to 9 ) : =;

Paschen-Back effect). This causes 0


1 2 3 -4-2 , *-, *45 / *46 7 8 !

resonance transitions to lie at differ-


ent spatial positions.
For example after preparation of atoms in state mF = 2 they are trapped in
the external potential V (mF = 2, ~r) = 2 µB gF [B(~r) − B0 ] In the figure shown
the rf-knife is resonant at the positions A, C, D, and E. Hot atoms which
reach A can be pumped to B. The translational motion of such atoms in space
7 The energy separation for ∆mF = 1 in 87 Rb (F = 2) is equal to 0.7 MHz/G.
104 CHAPTER 13. EVAPORATIVE COOLING

is indicated in the figure [67]. Here the atom is pumped again (by the same
frequency field at position C and again at D until the atom leaves the trapping
areas along the energy curve of mF = −1.
Evaporation is effective in cooling if thermalization of the remaining atoms
leads to lower temperature. From Monte-Carlo simulations one has learned how
to optimize this process [68]. An rf-induced cooling scheme was also attempted
in an optical dipole trap but with only poor results [69]. More successful is
evaporation by successive lowering of the trap barrier of the dipole potential
[70, 71, 72].

13.7 Diagnostics on BE Condensates


The macroscopic size of a BEC can be observed in absorptive or dispersive
imaging. In situ observation is frequently limited by the camera resolution and
blurring due to imperfect focussing. In this case observation after a specific time
of free fall is made when the cloud of atoms has expanded slightly.
The following phase-contrast images show the phase transition from a ther-
mal cloud to a BE-condensate. Here the radio-frequency for evaporative cooling
was sequentially lowered from 1.45 MHz to 1.10 MHz. In the figure on page 103
this corresponds to a change in position of the rf-knife from initially point A
(1.45 MHz) to a value of half this distance r from the trap center. The uncon-
densed portion gradually diminishes and the condensate has a length of about
300 µm. The cigar shape is caused by the different magnetic field curvature in
the axial and radial directions. The radial eigenfrequency is about 230 Hz, the
axial frequency about 17 Hz [73].

Indications of macroscopic coherence appear when two or more condensates in-


terfere [74, 75, 76]. A condensate can be cut into two with a blue detuned laser.
After turning off the laser the two condensates will flow into each other in the
trap and interfere. The interference image (right) has dimensions of 1 × 0.5 mm.

Atom-Laser : A mirror-like device which can couple out a portion of a con-


densate in a similar way as an output coupler releases light from a laser resonator
was demonstrated in 1997 [77]. Here radio frequency was used to pump a small
portion of the condensate into unbound Zeeman states. For atoms with F = 1,
which are bound in state mF = −1 an rf-pulse generates from the condensate
N
|BECi = |mF = −1i the superposition
rf N
|BECi −→ α|mF = −1i + β|mF = 0i + γ|mF = +1i (13.30)
This process can be treated in terms of a Landau-Zener crossing, see Ap-
pendix A-7. The unbound part of the condensate (β|mF = 0i + γ|mF = +1i)
falls due to gravity in one piece as a coherent packet of atoms.
Chapter 14

Sideband Cooling

This concept originated in experiments with trapped ions [78] and was further
developed by the groups of C. Salomon in Paris [79], and S. Chu in Stanford [80].
If atoms (ions) are trapped in an external potential, approximated by a harmonic
potential with eigenfrequency ωHO and the width of the excited state is smaller
than the energy gap h̄ωHO , then sidebands can be resolved in the excitation
spectrum, corresponding to electronic transitions with concurrent change in
vibrational quantum number v.
E E E

v1 v1 v1


v v v
e v1 e v1 e v1

v1 v1 v1


v v v
g v1 g v1 g v1

x x x

Efficient sideband cooling is possible if the spatial motion of the particles is


restricted to a range smaller than the wavelength of the cooling laser,
p
λ  h̄/(M ωHO ) (14.1)
The measure at the right is the spatial extent of the ground state vibrational
wave function in the harmonic trapping potential hx2 i = h̄/(M ωHO ). It can be
expressed by the so-called Lamb-Dicke parameter
h̄k 2 h̄/(M ωHO )
η2 = = (14.2)
2M ωHO λ2
This parameter gives the ratio between the recoil energy h̄2 k 2 /(2M ) and the
vibrational energy h̄ωHO . It is also equal to the squared ratio of the spatial

105
106 CHAPTER 14. SIDEBAND COOLING

extent of the vibrational wavefunction in relation to the wavelength of absorp-


tion/emission. When η  1 one speaks of the so called Lamb-Dicke Limit.
When η is small one may selectively excite sideband transitions as they are
shown in the figures on page 105 (center and right). Following excitation from
a state with vibrational quantum number v to v ± 1 spontaneous emission will
preferentially end up in v±1, whereby we have changed the vibrational quantum
number by ±1.
If ω0 characterizes the electronic transition energy in the free atom, a cooling
transition in the red sideband occurs at ω0 − ωHO

|g, vi + h̄(ω0 − ωHO ) → |e, v − 1i (14.3)

At small values of η the transition back into the electronic ground state is most
probably diagonal (with no change in vibrational quantum number)

|e, v − 1i → |g, v − 1i + h̄ω0 (14.4)

In this fashion one may reduce the vibrational energy by one. A cooling limit
appears from non-resonant transitions
|g, vi + h̄(ω0 − ωHO ) → |e, vi (14.5)
|g, vi + h̄(ω0 − ωHO ) → |e, v + 1i (14.6)
which may lead to heating. At high laser intensity such transitions are difficult
to suppress unless Γ  ωv . The concept of EIT cooling, introduced by Morigi
in 2000 [82] circumvents this problem and was successfully demonstrated in the
laboratory [83] for ions, but not yet for neutral atoms.

14.1 Lamb-Dicke Parameter


In spontaneous emission a photon with wave-vector ~k and polarisation ~ appears.
In this process the photon number in the mode {~k,~} increases by one and the
atom falls from the internal state |ei into the deeper lying state |gi. The initial
state of the combined system is described by the atom |ei with momentum h̄K ~
and a field in the vacuum state
~ 0i
|φi i = |e, K; (14.7)

~ 0 and a free photon


In the final state the atom is in |gi with the momentum h̄K
in the field
~ 0 ; ~k,~ i
|φf i = |g, K (14.8)

Due to energy conservation we request

h̄2 K 2 h̄2 K 0 2
Ee + = Eg + + h̄ω (14.9)
2M 2M
and due to momentum conservation

h̄K ~ 0 + h̄~k
~ = h̄K (14.10)
14.1. LAMB-DICKE PARAMETER 107

~ 0 = −~k.
~ = 0 and hence K
In the center of mass frame of the atom we have K
Then the photon’s energy is

h̄ω = Ee − Eg − Erec (14.11)

where
h̄2 k 2 h̄2 ω 2
Erec = = . (14.12)
2M 2M c2
Since h̄ω  M c2 the recoil term leads to a change of wavelength from the reso-
nance wavelength by a tiny amount which is usually neglected. In this case we
have h̄ω ≈ h̄ω0 = Ee −Eg , consistent with the assumption M = ∞.

~
In a reference frame which moves with the speed of the atom ~v = h̄K/M we
have from (14.9) and (14.10)

h̄ω = h̄ω0 − Erec + h̄ ~k · ~v (14.13)

The Doppler-effect leads to a specific frequency of emission which depends on


the angle between the atomic velocity ~v and ~k.

In these considerations we have assumed that the center of mass of the atom
is free to move in space. If the atom is trapped in an external potential then
relation (14.10) is no longer valid and we have instead of (14.9)
(i) (f )
Ee + Ecm = Eg + Ecm + h̄ω (14.14)
()
where Ecm is the translational energy of the center-of-mass motion before and
(i) (f )
after the spontaneous event. Under certain circumstances we have Ecm = Ecm .
In this case the photon appears at the frequency ω0 and no recoil and no Doppler
effect appears.
Such a situation occurs when an atom is firmly trapped in a crystalline
matrix and the phonon energy in the crystal is larger than the recoil energy
(Mößbauer effect). It also appears when the mean free path of the atom is
smaller than the wavelength of the emitted light, for example when an atom is
trapped in an external potential of suitable dimensions or when a molecule/atom
is embedded in a high pressure gas. In the latter case this effect was first ob-
served by Dicke [84]. Dicke observed a narrowing of vibrational lines of molecules
for which the vibrational frequency is much smaller than the frequency of colli-
sions with gas atoms, an effect which is known as the Dicke-effect.

In the following we consider a particle trapped in a harmonic oscillator and


we describe the state of the particle by the product wavefunction

|ii ⊗ |χn i = |i, χn i . (14.15)

Here |ii describes the internal state and |χn i the external motion of the center-
of-mass. The vibrational states, |χn i are at the energy En . Initially our atom
is in the electronic ground state, |gi in the vibrational level |χn i

|φi i = |g, χn i . (14.16)


108 CHAPTER 14. SIDEBAND COOLING

After absorbing one photon we find the atom in the excited state |ei in the
vibrational state |χm i

|φf i = |e, χm i . (14.17)

Now we explore the probability that the vibrational state of the atom changes
in such a transition. We characterize the electronic interaction by the Rabi
frequency Ω1 and explicitly account for the spatial variation of the light field
h̄ ~
H1 = − Ω1 eik·~x (|eihg| + |gihe|) (14.18)
2
Applied to our product state we have
h̄ ~
hφf |H1 |φi i = − Ω1 hχm | eik·~x |χn i (14.19)
2
a product of an electronic transition matrix element and an element which
describes the change in vibrational motion. The expression
~
hχm | eik·~x |χn i = F (n, m, η) (14.20)

may be regarded as a Frank-Condon factor. The magnitude and variation of


this factor is controlled by the Lamb-Dicke parameter. In order to see this we
write for k from (14.2)
√ p
k = η 2 M ωHO /h̄ , (14.21)

and recall that the HO may be described in ladder operators. The reduced
spatial variable of the HO is
r
M ωHO 1
x = √ a† + a ,

X̂ = (14.22)
h̄ 2
Combining (14.21) and (14.22) we write for the Frank-Condon factor
~ †
hχm | eik·~x |χn i = hχm | eiη(a +a)
|χn i (14.23)

where for simplicity we have assumed ~k || ~x). For small values of η we may
expand the exponential function

eiη(a +a)
≈ 1 + iη a† + a + ...

(14.24)

and see that in the limit η → 0 due to the orthogonality of the vibrational
wavefunctions χ only diagonal transitions (m = n) will occur. In this case the
atom is firmly trapped and cannot absorb any recoil energy. In fact the entire
trap takes up the recoil energy, analogous to the Mößbauer effect.
At low values of η sidebands appear, m = n ± 1 weighted with a probability
of about η 2 .
14.2. DEGENERATE RAMAN SIDEBAND COOLING 109

14.2 Degenerate Raman Sideband Cooling


The group of S. Chu [80] succeeded in cooling atomic Cs to temperatures compa-
rable to polarization gradient cooling but at much higher densities of 1013 cm−3 .
Laser cooling is done in only one dimension but strong collisional coupling leads
to effective 3D-cooling. The initial state is a spin polarized ground state atom
in |F = 3, mF = 3; vi where v is the vibrational quantum number (high).
An external magnetic field is ap-
plied to shift this level to be res-
onant with |F = 3, mF = 2; v−1i.
In a harmonic trap this degen-
eracy between |3, 2; v − 1i and
|3, 3; vi holds for all vibrational
levels v. A cooling cycle consists
of a degenerate Raman transition
from |3, 3; vi to |3, 2; v − 1i, fol-
lowed by optical pumping to an
F = 2 excited state which may
radiate back to |3, 3i.
Since the atom is tightly bound, the recoil momentum from the scattered
photon is unlikely to change its vibrational state in the spontaneous event, and
the atom preferentially returns to |3, 3; v − 1i with one quantum of vibration
energy removed. The Raman transitions use two photons of the same energy,
they can be driven by the trapping light itself. The Raman coupling strength is
given by the matrix element of the two-photon operator h3, 2, v − 1|U (r)|3, 3; vi.
Note that the two photons in the Raman transition connect m states which
differ by only ∆m = 1. For a general polarization of the light field this operator
is nonzero only when the external magnetic field B ~ is oriented at some angle
β relative to the wave vector ~k. Different center-of-mass wave functions are
coupled by the two-photon recoil associated with the spatial dependence of U ,
whose parity with respect to the potential wells is determined by the polarization
configuration of the standing wave. To drive transitions with ∆v = −1, the
authors used two counter-propagating running waves with linear polarizations
under an angle α. The coupling strength for this configuration is [81]

h3, 2, v − 1|U (r)|3, 3; vi ≈ v (δ) U0 η sin α sin β (14.25)

where η is the Lamb-Dicke parameter, (δ) characterizes the relative strength of


the Raman transition, dependent on the detuning δ, and U0 is the trap depth.
The experiment was carried out in a 1D optical lattice generated by a 20
Watt Nd:YAG with a vertical standing wave with waist of 260 µm. This leads
to U0 = 3 MHz or 160 µK at the position of the atomic cloud and the eigenfre-
quencies fax = 130 kHz and frad = 120 Hz. The scattering rate of the trapping
light is 2 s−1 . The degenerate Raman sideband cooling is applied only along
the steep axial direction. Temperatures of 2 µK were achieved in 1 second.1

1 In this 1D lattice the authors trapped a total of 107 atoms in a cigar-shaped cloud with a

vertical FWHM of 2.5 mm. This length corresponds to 4700 individual pancake shaped traps
each with an aspect ratio of 1000, spaced by 532 nm. The vertical density distribution is
approximately Gaussian, yielding a population of ≈2000 atoms per trap in the central region.
110 CHAPTER 14. SIDEBAND COOLING

14.3 EIT Cooling


This method is based on electromagnetically induced transparency of a three-
level system. In this case the excitation (14.20) at ∆v = m − n = 0 can be
suppressed and the transition ∆v = m − n = +1 be made much weaker than
∆v = m − n = −1. The concept of EIT cooling was introduced by Morigi in
2000 [82] and considered in some detail by Roghani et al. [85, 86].

We consider a system containing a ground state |gi, an excited state |ei and
a third state |ri. The excited state has the natural width Γ and transitions
from it to |gi and |ri are dipole allowed. We now pump |ri → |ei with a strong
and blue detuned coupling laser at ωr = ωre + δr and observe the absorption
spectrum on the transition |gi → |ei at the frequency ωg = ωge + δg .

In the field of the coupling laser we have


Absorption
pairs of dressed states {|r, N i, |e, N − 1i},
Detuning δg

at an energy separation of δr . Here N


|e> is the number of photons in the coupling
laser field. Under interaction with this
δg,Ω g,k g laser at the Rabi frequency Ωr the dressed
δr,Ωr,kr states shift by ±∆ where
|g> |r> p
2∆ = δr2 + Ω2r − |δr |

The new perturbed states

|+i = cos θ|r, N i + sin θ|e, N − 1i (14.26)


|−i = sin θ|r, N i − cos θ|e, N − 1i (14.27)

are defined via the mixing angle 2θ = arctan Ωr /δr . When Ωr is chosen small
compared to δr , the state |+i contains only a weak admixture of |ei and a width
Γ0  Γ. The weak cooling laser observes these light shifted states and sees the
absorption spectrum indicated in the figure above. Two states at the energies
h̄ωge − ∆ and h̄ωge + δr + ∆ appear with substantially different widths.

0.003 Wr =1 0.003 Wr =1 0.003 Wr =1


The cooling laser also
Im@ Ρ3,2 D

Im@ Ρ3,2 D

Im@ Ρ3,2 D

0.002 Wg =0.02
∆r =1
0.002 Wg =0.02
∆r =10
0.002 Wg =0.02
∆r =25
sees a Fano-like pro-
0.001 0.001 0.001
0.000 0.000 0.000
file. A broad absorp-
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
detuning ∆g HMHzL detuning ∆g HMHzL detuning ∆g HMHzL
tion peak appears near
resonance δg ≈ 0 with a
0.03 Wr =2.8 0.03 Wr =6.5 0.03 Wr =10
width ≈ Γ and a very
Im@ Ρ3,2 D

Im@ Ρ3,2 D

Im@ Ρ3,2 D

0.02 Wg =0.2 0.02 Wg =0.2 0.02 Wg =0.2

M 34 0.01
∆r =1
0.01
∆r =10
0.01
∆r =25
narrow resonance, near
0.00 0.00 0.00
-30 -20 -10 0 10 20 30
detuning ∆g HMHzL
-30 -20 -10 0 10 20 30
detuning ∆g HMHzL
-30 -20 -10 0 10 20 30
detuning ∆g HMHzL
δg ≈ δr with a width
Γ0  Γ. The narrow
0.03 0.03 0.03
peak mirrors the small
Im@ Ρ3,2 D

Im@ Ρ3,2 D

Im@ Ρ3,2 D

0.02 Wr =2.8 0.02 Wr =6.5 0.02 Wr =10


Wg =0.2 Wg =0.2 Wg =0.2
0.01
∆r =1
0.01
∆r =10
0.01
∆r =25 admixture of |ei to the
0.00 0.00 0.00
-2 -1 0 1 2 3
detuning ∆g HMHzL
4 7 8 9 10 11 12
detuning ∆g HMHzL
13 22 23 24 25 26 27
detuning ∆g HMHzL
28 dressed excited state.

The top row shows the development of the Autler-Townes profile with in-
creasing blue detuning at constant intensity of the coupling laser Ωr = 1 MHz.
The center row shows spectra where the intensity is chosen such that the dis-
tance between the absorption zero and the blue detuned narrow peak is exactly
14.3. EIT COOLING 111

1 MHz. The required value of Ωr increases from 2.8 to 10 MHz. The bottom
row gives a close-up near the absorption zero.
We now introduce the translational motion of our atom in an external po-
tential by a vibrational quantum number v. We assume that the vibrational
spacings are identical for the three electronic states and are given by the har-
monic oscillator energy h̄ωHO . This is a good assumption for trapped ions where
the trap potential derives from the electric charge. This is not a good assump-
tion for neutral atoms in an optical dipole trap where the trap potential derives
from the polarizability of the electronic state.
For trapped ions typical eigenfrequencies are of the order of ωHO = 2π MHz.
This may be compared to Γ = 2π × 5 MHz.
When including the translational motion the EIT spectra shown above are
appropriate for diagonal vibrational transitions v → v 0 . For off-diagonal transi-
tions v → v 0 ±∆v the effective detuning of the cooling laser is δgeff → δg ∓∆v ωHO .
This is shown in the figure for ∆v = +1, 0, −1 where the detuning of the cou-
pling laser was chosen δr = +25 MHz.

v  v 1
Χg
The probability for transitions with ∆v 6= 0
is controlled pby the Lamb-Dicke parameter
22 23 24 25 26 27 28
η = |~kg − ~kr | hx2 iwhere hx2 i = h̄/(M ωHO ) is
∆eff
g MHz
the size of the HO ground state wavefunction and
~kg (~kr ) are the wave vectors for the cooling and
coupling lasers. For small values of η the only
v  v significant off-diagonal transitions are ∆v = ±1.
Χg

The cooling laser is now tuned to dark resonance


22 23 24 25 26 27 28 δg = δr and the intensity of the coupling laser
∆eff
g MHz is chosen such that the light shift ∆ amounts to
ωHO , here 1 MHz.

v  v 1 In this way diagonal vibrational transitions are


Χg
suppressed while transitions with v → v 0 −1 are
enhanced compared to those with v → v 0 +1.
22 23 24 25 26 27 28
∆eff
g MHz
112 CHAPTER 14. SIDEBAND COOLING
Chapter 15

Appendices

A-1 Ammonia Molecule and Maser


The ammonia molecule (NH3 ) has the shape of a pyramid. We call ~x the axis
pointing along the distance between the center of the 3 hydrogen atoms and the
nitrogen atom. There is a potential minimum for a specific value of |x| = a.
There are two symmetric realizations when we move the nitrogen atom along ~x.
The energy of the system shows a double-well potential for this coordinate.1

UHxL
M 35
E0

x
-a 0 +a

We restrict ourselves to the vibrational mode of nitrogen atom in the double


well and consider the case that the temperature of the molecule is so low that
only the lowest vibrational level needs to be considered.
Classically we may think of the nitrogen atom sitting either to the left or to the
right of the plane of the hydrogen atoms. Approximating the double-minimum
potential as two harmonic oscillator (HO) states we would associate the wave-
function ΨL with the Gaussian ground state vibrational wavefunction of the HO
located in the left minimum and likewise the wavefunction for ΨR would be the
Gaussian ground state vibrational wavefunction of the HO located in the right
minimum. These two states should be degenerate in energy and indeed they
are as long as the barrier between the two minima is very high. For a barrier
of finite height the nitrogen atom may tunnel from one side to the other, the
1 The actual energy spectrum of ammonia is very complicated. In addition to the many

vibrational states in which NH3 can be realized (as a molecule with n=4 atoms there are
(3n-6)=6 vibrational normal modes), the rotational levels ought to be taken into account... .

113
114 AMMONIA MOLECULE and MASER

rate of tunneling being given by the rate of vibration in each well (that is the
eigenfrequency of the HO, ω) multiplied with the transmission probability2 , T ,
h 2√2m Z x2 p i
Rt = ωT with T ∝ exp − U (x) − E0 dx (A-1.1)
h̄ x1

where m is the mass of the nitrogen atom, E0 is the energy of the vibrational
ground-state level and U (x) is the potential energy curve describing the double
minimum potential. If we identify ΨL and ΨR with the solutions of H0 and the
coupling energy h̄Rt with the perturbation W, with W = h̄Rt , we find for the
perturbed wavefunctions (mixing angle β = π/4)
1  
|ψA i = √ |ΨL i − |ΨR i
2
1  
|ψS i = √ |ΨL i + |ΨR i
2
which are named symmetric and antisymmetric for obvious reason.

The perturbed energies of the two states


are (pink and green)
UHxL
EA
EA = E0 + W E0
ES
ES = E0 − W . (A-1.2) x
M 36
ÈΨA \
The energy splitting EA − ES amounts
to 24 GHz, equivalent to λ = 1.25 cm.
The equilibrium distance is a = 0.4 Å.
ÈΨS \

-a 0 +a

In the perturbed basis, |ψS i, |ψA i , we now have a diagonal (unperturbed)


Hamiltonian
 
E0 + W 0
H= . (A-1.3)
0 E0 − W

In the following we explore the action of an external electric field on this system
and attempt to explain the principle of the ammonia maser.
In the classical configuration our ammonia molecule has a dipole moment d~0 .
In the presence of an electric field E~ the classical potential energy is −d~0 · E~ .
For simplicity we assume the external electric field to point along the ~x-axis.
With the interaction η = |d~0 · E|
~ our Hamiltonian is
 
E0 + W η
H= . (A-1.4)
η E0 − W
p
Solutions are E± = E0 ± W 2 + η 2 with the wavefunctions
|ψ+ i = sin β |ψS i + cos β |ψA i
|ψ− i = cos β |ψS i − sin β |ψA i , (A-1.5)
2 The so-called Gamov-factor.
APPENDIX 115

where β is the mixing angle, β = η/W . At low fields we have

d2 E 2
 
E± ≈ E0 ± W + 0 . (A-1.6)
2W

In an external electric field ammo- ÈΨ+ \ ÈΨL \

Energy Harb.unitsL
nia molecules are either strong- ÈΨA \
field seeking, |ψ− i, or weak-field EA
seeking, |ψ+ i. Thus we may
ES M 37
separate molecules in the upper ÈΨS \
state from those in the lower state ÈΨ- \ ÈΨR \
with the help of an inhomoge-
neous electric field. 0 2 4 6 8 10
Electric Field Strength HkVcmL

Due to the quadratic Stark shift


we may construct an electric
quadrupole which spatially sepa-
rates molecules in the two states. M 38
It keeps the low field seeking
molecules |ψA i near the axis,
where the field goes to zero.

At the end of the field region these molecules adiabatically return to |ψA i and
are guided into a resonator optimized for 24 GHz. In this way one achieves
inversion in the two-level system. After the resonator is filled with a few spon-
taneous photons formed in the transition |ψA i → |ψS i, the electromagnetic
field inside the resonator is amplified by stimulated emission from subsequent
molecules and the resonator turns into a source of microwave radiation at 24
GHz. This radiation can be extracted with a suitable antenna.

We have just reconstructed


the ammonia maser, the in-
strument with which Gordon,
Zeiger and Townes for the
first time realized amplifica-
tion by stimulated emission in
1954 [87].
116 BLOCH VECTOR

A-2 Bloch Vector


In Equations (7.41) only three real parameters appear. These form the three
components of the Bloch vector
~ = {U, V, W }
R (A-2.1)

U and V are the real and the imaginary parts of the coherence, W is the
population inversion

U = ( ρ̃eg + ρ̃ge ) /2 = <(ρ̃eg )


V = i ( ρ̃eg − ρ̃ge ) /2 = =(ρ̃eg ) (A-2.2)
W = ( ρ̃ee − ρ̃gg ) /2
With this definition and in the absence of dissipation the Bloch equations (7.41)
are written as
U̇ = −δ V
V̇ = +δ U − Ω1 W (A-2.3)
Ẇ = +Ω1 V

W
We now define the additional vector
~ = (−Ω1 , 0, −δ)
Ω (A-2.4)

which characterizes the driving field and its -V


detuning. This vector allows to write the -W1
equations (A-2.3) in the simple form -U
~
dR ~ ×R
=Ω ~. (A-2.5)
dt -∆
W
~
The length of the vector Ω is equal to the generalized Rabi frequency
q
Ω = Ω21 + δ 2 . (A-2.6)

In this Feynman-Vernon-Hellwarth picture [88] the vector R ~ precesses around


~
the vector Ω. In the absence of dissipation the length of the Bloch vector is con-
~ = 1/2). The vector Ω
stant in time (|R| ~ acts just like a torque, controlling the
axis and rate of precession.
The power of this representation lies in the fact that all possible quantum states
of the two-level system can be projected onto a point on the surface, or inside
of the Bloch sphere. It is customary to define the W -axis as the z−axis: At
the south pole the atom is in its ground state, at the north pole in its excited
state. All points in between represent coherent superpositions between ground
and excited state. Any position along the equator represents a balanced super-
position of ground and excited state with varying phase difference between the
two states.
In case that δ  Ω1 the axis of precession Ω ~ lies near the poles, hence the
APPENDIX 117

atom stays either in its ground or in its excited state. A so called π-pulse at
resonance (δ = 0) transfers the entire population from the ground state to the
excited state (left figure), or else back from the excited to the ground state. In
the non-resonant case (right figure) the excited state is only partially populated.
The path of the Bloch vector and the Ω-vector are shown in blue.

W W

-V -V
M 39

W
-U -U

For δ = 0 the equations (A-2.3) have exact solutions for arbitrary pulse shapes
U (t, δ = 0) = 0 (A-2.7)
2 V (t, δ = 0) = + sin θ(t) (A-2.8)
2 W (t, δ = 0) = − cos θ(t) (A-2.9)
with the definition of the angle
Z t
θ(t) = dt0 Ω1 (t0 ) . (A-2.10)
−∞

The angle θ(t) is referred to as the pulse area. In case the pulse area is equal to
π the population is inverted (δ = 0 is assumed). In case the pulse area is equal
to 2π the population is brought back entirely to its initial state.

Including dissipation the Bloch equations take the form


U̇ = − δ V − Γ U/2 (A-2.11)
V̇ = + δ U − Ω1 W − Γ V /2 (A-2.12)
Ẇ = + Ω1 V − ΓW − Γ/2 (A-2.13)
In this situation the length of the Bloch vector decreases with time and in the
absence of an external driving field the Bloch vector, now lying inside the Bloch
sphere, eventually spirals back to the south pole.

In the following we consider a pulse of


W1 = 1
area θ = 0.9π which pumps population Θ = 9 Π 10
from the ground state at a detuning of
EHtL

δ = 0.3 and a Rabi frequency Ω1 = 1.


Following this excitation pulse, the sys-
tem decays due to spontaneous emis- 0 Π 2Π
sion with the rate Γ = 0.05. time
118 BLOCH VECTOR

W W

-V -V

W -U W -U

This figure shows the action of the After excitation the system decays
pump pulse. The path of the Bloch back into the ground state along the
vector is shown blue. The pulse trans- red path. The radius of the spiral di-
fers population from the ground state minishes with time, a sign of loss of
to a value slightly below the maximal coherence. The inversion finally drops
inversion possible at this detuning. to zero, W = −1/2.
Now the excitation field is turned on at t=0
and stays on indefinitely. Parameters are W
Ω1 = 1, δ = 0.2, and Γ = 0.2. The length
of the torque vector is shown reduced by a
factor of 2. The population of the excited
state initially oscillates, but eventually settles
to a value slightly below 1/2, W (t → ∞) < 0. -V
Note that a small coherence remains at long
times, corresponding to the competition be- W -U
tween the coherent driving and the stochas-
tic spontaneous emission (compare with the
Monte-Carlo result for many atoms, page 32).

Note that the stationary point to which the system converges for t → ∞ should
agree with the stationary solution predicted by Equations (7.42) and (7.43):

Ω21 1
ρee = = = 0.4545 (A-2.14)
4δ 2 + Γ2 + 2Ω21 0.16 + 0.04 + 2

and
2Ω1 (δ − i Γ/2) 0.4 − 0.2 i
ρ̃eg = − =− = −0.18 + 0.09 i (A-2.15)
4δ 2 + Γ2 + 2Ω21 2.2

The stationary result for the Bloch vector is {U = −0.18, V = 0.09, W = −0.045}.
This agrees with the numerical solution of Eq. (A-2.11)-(A-2.13), visible as the
endpoint of the blue spiral in the figure above.
APPENDIX 119

A-3 Clebsch-Gordan Coefficients


The angular momentum of the ground state j1 is coupled with that of the photon
j2 = 1 to yield the angular momentum of the excited state J. We search the
eigenvalues of the operators
2
J2 = (j1 + j2 )
Jz = j1z + j2z (A-3.1)
and their respective eigenvectors |JM i, under the assumption that the eigenval-
ues of the operators j21 , j22 , j1z , j2z and their eigenfunctions |j1 m1 i and |j2 m2 i
are known. The total angular momentum numbers are restricted to the range
J = j1 + j2 , j1 + j2 − 1, j1 + j2 − 2, . . . , |j1 − j2 | (A-3.2)
M = m1 + m2 (A-3.3)
The product state |j1 j2 ; J M i contains all values of m1 and m2 for which the
relationship M = m1 + m2 holds. It is therefore appropriate to describe the
final state (in our example the excited state) as a sum of uncoupled states,
j1
X j2
X
|j1 j2 ; J M i = C(j1 j2 J; m1 m2 M ) |j1 m1 ; j2 m2 i
m1 =−j1 m2 =−j2
j1
X j2
X
= hj1 m1 , j2 m2 |J M i |j1 m1 ; j2 m2 i (A-3.4)
m1 =−j1 m2 =−j2

The expressions C(j1 j2 J; m1 m2 M ) are called vector-coupling coefficients or


Clebsch-Gordan coefficients. Formally the coefficients give the projection of the
individual |j1 m1 ; j2 m2 i wavefunctions on the state |j1 j2 ; J M i. The coefficients
give the amplitude with which each product state |j1 m1 ; j2 m2 i contributes to
the coupled state |j1 j2 ; J M i. The coefficients form a unitary matrix of size
n × n where n = (2j1 + 1)(2j2 + 1). Clebsch-Gordon coefficients are defined for
integer and half-integer quantum numbers.

In modified form these coefficients are represented by Wigner 3j-symbols.


The relation between the 3j-symbols and the Clebsch-Gordan coefficients is
 
J j2 j1 1
= (−1)2j1 +J+M √ C(j1 j2 J; m1 m2 M )
−M m2 m1 2J + 1
(A-3.5)
The triangle rule has to be fulfilled for the angular momentum quantum num-
bers, that is we require that the vector with lengths corresponding to J, j1 and
j2 can form a triangle. In addition we require the m1 + m2 − M = 0. The sums
in Eq. A-3.4 define a state of specific value of M . only those coefficients are non
zero for which M = m1 + m2 . The analytic expression for the Clebsch-Gordan
coefficients is rather complex.3
Clebsch-Gordan coefficients are easily evaluated using Mathematica. M 40
3 See for example Eq. 7.51 in ref. [89]. Tables for Clebsch-Gordan and 3j-Symbols can be

found in many books, for example [90] pages 57-63.


120 CLEBSCH-GORDAN COEFFICIENTS

Relative Intensity of Zeemann Components We consider the case that


a magnetic field is oriented along z and we observe spontaneous emission from
atoms along the z-axis or along the x or y-axes.
The probability for spontaneous emission into the solid angle dΩ is [91]

ω03 ~ 0 j1 m1 i |2 dΩ
dWρ (γJM ; γ 0 j1 m1 ) = | ˆρ hγJM |D|γ (A-3.6)
hc3

where D ~ is the electric dipole moment, ˆρ the polarization vector and γ char-
acterizes the electronic state. Integration over all space and summation over
the possible polarizations and possible m-values gives for W the Einstein A-
coefficient.

For the representation of the vector operator D ~ it is most convenient to


transform from the cartesian basis ˆx , ˆy , ˆz , to the spherical unit vectors [92]

ˆ 0 = ˆz
1
ˆ±1 = ∓ √ (ˆx ± iˆ
y )
2
~ is represented by4
In the spherical basis the operator D

D0 = Dz
1
D±1 = ∓ √ (Dx ± iDy ) (A-3.7)
2
The advantage lies in the simplicity with which a general matrix elements of
the vector operator can be developed. With the abbreviation q = −1, 0, 1 we
express the product
X
~ = D cos θ ~ =
ˆρ · D ∗q Dq (A-3.8)
ˆD
q

and hence
X
~ 0 j1 m1 i =
ˆρ hγJM |D|γ ∗q hγJM |Dq |γ 0 j1 m1 i (A-3.9)
q

where q and Dq are the spherical components of the vectors r and D. This
matrix element can now be separated into two parts. One expression stands
for the geometry, symmetry and selection rules of the system, a second for the
electron dynamics. This separation is the essence of the Wigner-Eckart theorem
according to which
 
J 1 j1
hγJM |Dq |γ 0 j1 m1 i = (−1)J−M (γJ||D||γ 0 j1 )
−M q m1 | {z }
 
J 1 j1
= (−1)J−M D (A-3.10)
−M q m1
4 With ~ = e~
D r expression (A-3.7) is proportional to spherical harmonics Y`m :
p p
D 0 = 4π/3 |~r| Y10 und D±1 = 4π/3 |~ r| Y1±1
APPENDIX 121

The reduced dipole matrix element D contains only electronic properties of the
system. All angular dependent terms enter in the pre-factor and the 3j-symbol.
Also the selection rules for electric dipole radiation follow from the properties
of the 3j-symbols. We have ∆M = −1, 0, +1, corresponding to q = −1, 0, 1 as
well as ∆J = 0, ±1.

Observation along z-axis : If the k-vector of the emitted light is along


z the polarization vectors are in the x − y plane and we obtain from (A-3.6)
dW ∝ |hγJM |Dx |γ 0 j1 m1 i |2 + |hγJM |Dy |γ 0 j1 m1 i |2 dΩ

(A-3.11)
or
X
dW ∝ |hγJM |Dq |γ 0 j1 m1 i |2 dΩ (A-3.12)
q=±1

In this case right- (∆M = +1) and left- (∆M = −1) circularly polarized light
is observed. These components are abbreviated as σ + and σ − . Their intensity
is proportional to the square of the 3j-symbols.
 2
J 1 j1
∆M = +1 , dW ∝ dΩ (A-3.13)
−M 1 M − 1
 2
J 1 j1
∆M = −1 , dW ∝ dΩ (A-3.14)
−M −1 M + 1

Observation perpendicular to z-axis : for example along the x-axis.


In this case the directions y and z may be considered as independent polarization
axes:
dW ∝ |hγJM |Dz |γ 0 j1 m1 i |2 + |hγJM |Dy |γ 0 j1 m1 i |2 dΩ

(A-3.15)
or
 1 X 
dW ∝ |hγJM |D0 |γ 0 j1 m1 i |2 + |hγJM |Dq |γ 0 j1 m1 i |2 dΩ
2 q=±1
(A-3.16)
Here we observe σ-components and a π-component (∆M = 0), which is polar-
ized parallel to the magnetic field axis (along z) with the intensity
 2
J 1 j1
∆M = 0 , dW ∝ dΩ . (A-3.17)
−M 0 M

M-dependent dipole matrix elements : We consider a ground state with


quantum numbers j1 m1 and an excited state with J M . The relative amplitudes
for absorption and emission from the different Zeeman levels are from Eq. (A-
3.10)
 
− 0 J−M J 1 j1
σ : hγJM |D−1 |γ j1 m1 i = (−1) D
−M −1 M + 1
 
J 1 j1
σ+ : hγJM |D+1 |γ 0 j1 m1 i = (−1)J−M D
−M +1 M − 1
 
J 1 j1
π: hγJM |D 0 |γ 0 j1 m1 i = (−1)J−M D
−M 0 M
122 CLEBSCH-GORDAN COEFFICIENTS

where D is the reduced matrix element. For σ − polarization M = m1 + 1, for


σ + polarization M = m1 − 1, and for π polarisation M = m1 .

Clebsch-Gordan coefficients for j = 1/2 → j = 3/2 transition : As an


example we consider the transitions in a 1/2 → 3/2 system. At a given laser
intensity the squared Rabi frequency for a σ + -transition from m1 = +1/2 is
three times stronger than from m1 = −1/2. The squared Rabi frequency for a
π-transition from m1 = +1/2 is twice as strong as for a σ − -transition.

! "# $% ! "& $% ! ' & $% ! ' # $%

% & & %
& # # &
# #

( "& $% ( ' & $%


APPENDIX 123

A-4 Partial Wave Expansion


The scattering of two particles which interact via a central potential U (R) cor-
responds to the scattering of a particle with the reduced mass µ in the relative
co-ordinates of the two particles. In a classical scattering event we may define
the impact parameter b in terms of the angular momentum,

Lklass = µ b v0 = b p . (A-4.1)

Here v0 is the relative speed and p the relative momentum of the particles.
Classically the impact parameter b can change continuously, as can Lklass . In
quantum mechanics we demand that the angular momentum is an integer mul-
tiple of h̄, L = h̄`. With the de-Broglie wavelength λ and the wavenumber
(momentum) for the relative motion k = 2π/λ we may define a transition from
the classical to the quantum picture

h̄` = b p = b h̄k → ` = b k = 0, 1, 2, 3, ..... (A-4.2)

At a fixed momentum the impact parameter can only change in discrete units,
a consequence of the uncertainty relation.5
In quantum mechanics we describe the incoming particle as a plane wave,
along the z-direction. The scattering center is at the origin and R gives the
distance from the origin, z = R cos θ. A plane wave can be expanded in terms
of an infinite number of spherical waves
~
X
eik ·~r = eikR cos θ = i` (2` + 1) j` (kR) P` (cos θ) (A-4.3)
`

where the radial functions j` (kR) are spherical Bessel functions and the angular
term are Legendre polynomials. Below we show the real part of Eq. (A-4.3)
for various values of the maximum ` included in the sum against two spatial
coordinates. The wave vector is taken as k = 2π.
.

M 41

5 We consider two Rb atoms which collide with a relative speed of 1 cm/s (the temperature

equivalent is 263 nK). The corresponding de-Broglie wavelength is 72 nm and the wavenumber
is 8.7×107 m−1 . For ` = 1 we obtain b = 11 nm. In words: for classical impact parameters
b < 11 nm is the angular momentum zero!
124 PARTIAL WAVE EXPANSION

After scattering a portion of the amplitude is emitted as spherical wave6

~ eikR
Ψ ∝ eik ·~r + f (θ, φ) (A-4.4)
R
where f (θ, φ) is the scattering amplitude. The sum of squares of the scattering
amplitude defines the total elastic scattering cross section
Z
σ= dΩ |f (θ, φ)|2 (A-4.5)

For elastic scattering the energy of the wave does not change and the scattering
amplitude is determined by the Schrödinger equation

h̄2
 
− ∆ + U (R) ~ Ψk (R) ~ = Ek Ψk (R)
~ (A-4.6)

Here the kinetic energy is Ek = h̄2 k 2 /(2µ) = 12 µv02 . For a central potential the
angular momentum is conserved and an expansion of Ψk (R) ~ in terms of partial
waves is possible, such that each scattered angular momentum component can
be compared with the incoming term from the plane wave expansion.

X u` (kR)
Ψk (R, θ) = P` (cos θ) (A-4.7)
R
`=0

The scattering process of each partial wave is solved in a radial Schrödinger


equation,

d2
 
2 `(`+1) 2µ
u` (kR) + k − + 2 U (R) u` (kR) = 0 . (A-4.8)
dR2 R2 h̄

Solutions for U (R) = 0 are spherical Bessel functions. At large values of R


these functions have the asymptotic form
−ikR
eikR
 
(0) 1 `+1 e
j` (kR) ≈ (2` + 1) (−1) + (A-4.9)
2ik R R

This corresponds to a pairwise superposition of an incoming and an outgoing


wave, weighted with a phase factor 0 or π, depending on the parity of `. For
sufficiently large values of R we have U (R) → 0, such that the solution A-4.9
can be considered as approximate solution for Eq. (A-4.8). In an elastic collision
the particle number and the angular momentum are conserved. We demand in
addition that for each partial wave the amplitudes of incoming and outgoing
wave must be the same. As a consequence the potential scattering may only
appear in a phase shift of the outgoing wave e2iδ` (k) . This leads us to the
solution at large values of R is

e−ikR eikR
 
1 X
Ψk, (R, θ) ≈ (2` + 1) P` (cos θ) (−1)`+1 + e2iδ` (k) (A-4.10)
2ik R R
`=0

6 See for example C. Cohen-Tannoudji, Quantenmechanik, Band 2, page 120-132.

Michael Fowler, http://galileo.phys.virginia.edu/classes/752.mf1i.spring03/Scattering II.htm


APPENDIX 125

M 42

Examples of Abs[Ψk, (R, θ)] are shown for δ` =0 with k = 2π. Contours
are given for values Abs[Ψk, (R, θ)] equal to 0.1 and 1. The effect of the
phase shift δ on Abs[Ψk, (R, θ = 0)] for ` = 0 is shown below.

1 1 1
Π Π
ÈYk HR,Θ=0LÈ

ÈYk HR,Θ=0LÈ

ÈYk HR,Θ=0LÈ
∆=0 ∆= ∆=
4 2

0 0 0
0 1 2 0 1 2 0 1 2
R R R

By comparing the coefficients of Eq. (A-4.10) with Eq. (A-4.4) we obtain for
the scattering amplitude

1 X h i
f (θ, φ) = f (θ) = (2` + 1) e2iδ` (k) − 1 P` (cos θ) . (A-4.11)
2ik
`=0

We see that contributions with zero phase shift (or of multiples of π) do not
contribute to the scattering amplitude. To determine the cross section we need
to evaluate the scattering phase shifts from the radial equations with the bound-
ary conditions from Eq. (A-4.10) and u` (R = 0) = 0. Introducing (A-4.11) in
(A-4.5) one obtains for the elastic cross section
4π X
σ= (2` + 1) sin2 δ` (k) . (A-4.12)
k2
`

˜ ˜ ˜ ˜ ˜ ˜ ˜ ˜ ˜ ˜
For identical particles scattering by
angles of θ and π + θ cannot be dis-
!" # !$ #
tinguished. A consequence is that
!" # !" #
for identical bosons
Z !$ # !$ #
σ= dΩ |f (θ) + f (π + θ)|2 !$ #
% #
!" #
& #

and that partial waves of odd parity do not contribute to scattering of identical
bosons. The total cross section for identical bosons appears as
8π X
σ= (2` + 1) sin2 δ` (k) . (A-4.13)
k2
`(even)
126 ROTATION OF BASIS

A-5 Rotation of Basis


In Chapter 11, polarization gradient cooling with σ + − σ − light, we require a
transformation of the basis to a rotated coordinate system. Here we discuss
general aspects of rotation and follow the book by R. N. Zare, chapter 3 [90].

The general rotation operation Rn (α) is distinguished by three parameters:


two fix the position of the rotation axis n̂ and one the rotation angle. We first
assume the rotation axis to be the z-axis. The transformation Rz (φ) transforms
the system from φ0 to φ0 = φ0 + φ

Rz (φ) |φ0 i = |φ0 i (A-5.1)

For an infinitesimal rotation of the state |φ0 i we express in a Taylor expansion


with the property (Rz (φ→0) → 1)

∂ 1 ∂2
|φ0 i = |φ0 i + (−φ) |φ0 i + (−φ)2 2 |φ0 i + . . . (A-5.2)
∂φ 2! ∂φ
The minus sign expresses the fact that for a positive rotation of the coordinates
we examine the portion of the function |φ0 i at negative angles. This series
expansion corresponds to the exponential form
   
∂ i
|φ0 i = exp −φ |φ0 i = exp − φJz |φ0 i (A-5.3)
∂φ h̄
The z-component of the angular momentum operator is the generating function
of an infinitesimal rotation around the z-axis. For an arbitrary rotation around
an axis n̂ we have
 
i ~
Rn (α) = exp − αJ · n̂ , (A-5.4)

where J~ · n̂ = −ih̄ ∂α

. In the following we suppress h̄.

Euler Angles : The most common description of an arbitrary rotation is


through the definition of the three Euler angles φ, θ und χ as shown in the
figure below. In order to transform from the old coordinate system X, Y, Z to
the new one (x, y, z), we execute in sequence the following three rotations : All
positive rotations are clockwise if we view the positive direction of the respective
axis from the origin.
RZ (φ): rotation by φ around Z-axis transforms the Y -axis into the new axis N
RN (θ): rotation about the new y-axis (=N ) by the angle θ
Rz (χ): rotation about the new z-axis by an angle χ
This rotation, first by φ, then by θ and then by χ is equal to the product

R(φ, θ, χ) = exp (−iχJz ) exp (−iθJN ) exp (−iφJZ ) (A-5.5)


APPENDIX 127

3
z

M 43
O
X
. y
?
.
Y N
x

This expression is difficult to handle as we need to define new axes. Euler angles
are distinguished in that the rotation (A-5.5) is equivalent to

R(φ, θ, χ) = exp (−iφJZ ) exp (−iθJY ) exp (−iχJZ ) , (A-5.6)

first a rotation by χ around the original Z-axis, followed by a rotation by θ


around the old Y -axis, and then again a rotation around the same Z-axis by φ.

In the old co-oordinate system the new z-axis has the polar angles (θ, φ).

Rotation matrix : We now examine how a rotation of the co-ordinate system


effects the representation of the eigenfunctions |JM iZ of J~2 und JZ . If the
rotation occurs about the common origin of both systems the magnitude of J~2
stays constant. For the (2J + 1) M quantum numbers in the Z-basis we have
an equivalent number in the new basis. In such a rotation a state |JM iZ is
transformed into a linear combination of states of the new z-axis.
X
R(φ, θ, χ)|JM iZ = |JM 0 iZ DM
J
0 M (φ, θ, χ) (A-5.7)
M0

The rotated state has in general no well defined projection of M . The amplitudes
J
of the rotation matrix DM 0 M (φ, θ, χ) weigh the contributions of unrotated kets
0
|JM iZ in a state of the new basis |JM iz . We obtain the expansion coefficients
(the full rotation matrix, in German: Kreiselfunktionen) with the help of Eq. (A-
5.6)
J 0
DM 0 M (φ, θ, χ) = Z hJM |R(φ, θ, χ)|JM iZ
0
= Z hJM |e−iφJZ e−iθJY e−iχJZ |JM iZ . (A-5.8)

The factorization into individual rotations by the Euler angles leads to simplified
expressions. Since M (and M 0 ) are eigenvalues of JZ , the first and last term in
Eq. (A-5.8) can be expressed in simple exponential functions
X 1
e−iαJZ |JM i = (−iα)ν (JZ )ν |JM i
ν
ν!
X 1
= (−iαM )ν |JM i
ν
ν!
= e−iαM |JM i
128 ROTATION OF BASIS

As a result we obtain
J
DM 0 M (φ, θ, χ) = hJM 0 |e−iφJZ e−iθJY |JM i e−iχM
0
= e−iφM hJM 0 |e−iθJY |JM i e−iχM
0
= e−iφM dJM 0 ,M (θ) e−iχM (A-5.9)
More problematic is the evaluation of the reduced rotation matrix
dJM 0 ,M (θ) = hJM 0 |e−iθJY |JM i , (A-5.10)
Wigner succeeded first in calculating this real matrix (pages 84-87 in [41]).
For the example J = 1 one obtains (page 89 of [90])

d11,+1 = d1−1,−1 = + cos2 θ/2


d11,−1 = d1−1,+1 = +psin2 θ/2
d10,+1 = d1−1, 0 = +p1/2 sin θ
d10,−1 = d1+1, 0 = − 1/2 sin θ
d10, 0 = + cos θ
Basis rotation in orientational cooling : We defined for J = 1 eigenstates
in two orthogonal axes, |gM 0 iz of Jz and |gM iy of Jy (see page 76)
   
|g−1 iz , |g0 iz , |g+1 iz , |g−1 iy , |g0 iy , |g+1 iy (A-5.11)

# #
" ) " ))
" ! ) # )) # )))

! ! )) ! ))) " )))


! $ %& ' ( " $ %& ' ( ! $ %* %& ' (

When we rotate the quantization axis (z-axis) by an angle of 90o around the
x-axis into the y-axis, then this is equivalent to the rotation by the Euler angles

φ = 90o , θ = 90o , χ = −90o (A-5.12)

as apparent from the drawing above. We have according to (A-5.7)


X 1
|gM iy = DM 0 M (φ, θ, χ) |gM 0 iz

M0
X 0
= e−iχM dJM 0 ,M (θ) e−iφM |gM 0 iz (A-5.13)
M0
X π 0
= e+iM π/2 1
dM 0M e−iM π/2
|gM 0 iz . (A-5.14)
2
M0

Explicitly we obtain for example


π
n π π π
π π o
|g−1 iy = e−i 2 ei 2 d−1,−1
1
|g−1 iz + e−i 2 d1,−1
1 1
|g+1 iz + d0,−1 |g0 iz
2 2 2
1 n √ o
= +|g−1 iz − |g+1 iz + i 2|g0 iz ,
2
the expression which appears as Eq. (11.12) in [39].
APPENDIX 129

A-6 Slowly-Varying Amplitude Approximation


We consider the propagation of an electric field along the z-direction. The
one-dimensional wave equation reads
 2
1 ∂2 1 ∂2


− E = P (A-6.1)
∂z 2 c2 ∂t2 0 c2 ∂t2

where P is the polarization density. For the field and polarization we write
1h i
E = E0 e−i(ωt−kz) + E0? ei(ωt−kz)
2
1h i
P = P0 e−i(ωt−kz) + P0? ei(ωt−kz) (A-6.2)
2
where ω and k = ω/c are the frequency and wave number of the field, respec-
tively. In terms of the new variables

ζ=z and τ = t − z/c (A-6.3)

these fields are written as


1
E0 e−iωτ + E0? eiωτ

E =
2
1
P0 e−iωτ + P0? eiωτ

P = (A-6.4)
2
The partial derivatives written in the new variables are
∂ ∂ 1 ∂
= −
∂z ∂ζ c ∂τ
∂2 ∂2 2 ∂2 1 ∂2
= − + 2 2
∂z 2 ∂ζ 2 c ∂ζ∂τ c ∂τ
∂ ∂ ∂2 ∂2
= and = (A-6.5)
∂t ∂τ ∂t2 ∂τ 2
Substituting these equations into (A-6.1) we obtain
 2
2 ∂2
  2 
−iωτ ∂ 2iω ∂ −iωτ ∂ ∂ 2 P0
e 2
− + E0 +c.c. = e 2
− 2iω +ω +c.c.
∂ζ c ∂ζ∂τ c ∂ζ ∂τ ∂τ 0 c 2

This equation is satisfied when


 2
2 ∂2
  2 
∂ 2iω ∂ ∂ ∂ 2 P0
2
− + E0 = 2
− 2iω +ω . (A-6.6)
∂ζ c ∂ζ∂τ c ∂ζ ∂τ ∂τ 0 c2

Assuming that the amplitudes E0 and P0 vary slowly in time and space
we neglect all but the lowest non-vanishing order on each side of (A-6.6) and
obtain
∂ iω
E0 = P0 . (A-6.7)
∂z 20 c
130 Slowly varying amplitude approximation

In the linear-optics approximation we introduce the linear but complex suscep-


tibility χ = χ0 + iχ00 which relates the induced polarization to the driving field

P0 = 0 χ E0 (A-6.8)

and obtain from Equation (A-6.7)

∂ ω
E0 = ( i χ0 − χ00 ) E0 . (A-6.9)
∂z 2c
The real part χ0 determines the shift of the phase front of the wave and the
imaginary part χ00 describes absorption. If we introduce the refractive index
1
nr = 1 + χ0 (A-6.10)
2
and the absorption coefficient
ω 00
α= χ (A-6.11)
c
we obtain the wave equation in the slowly-varying amplitude approximation
∂ ω h αi
E0 = i k(nr − 1) − E0 . (A-6.12)
∂z 2c 2
At low light level the dependence of nr and α on the field intensity can be
neglected and the solution for the field envelope is

E0 = Aeik(nr −1)z e−αz/2 (A-6.13)

with the complex amplitude A = |A|eiφ . With the definition of the field as
1h i
E= E0 e−i(ωt−kz) + E0? ei(ωt−kz) (A-6.14)
2
we obtain for the field in the presence of a medium, characterized by nr and α

E = |A| e−αz/2 cos (ωt − knr z − φ) . (A-6.15)


APPENDIX 131

A-7 Landau-Zener Crossing


In 1932 Landau [93], Zener [94], and Stückelberg [95] discussed the dynamic case
of a forbidden crossing between two quantum states, a concept which is termed
Landau-Zener crossing. They consider a two level system in which the unper-
turbed (diabatic) states vary along some coordinate (for example the spatial
coordinate x) according to
H11 = +f1 (x − x0 ) and H22 = +f2 (x − x0 )
The two states are degenerate at x = x0 . The interaction between the states is

H12 = H21 = W = const


This case is shown in the following picture for f1 = 0, f2 = 1 and W = 1.
The right hand figure gives the energy separation between the diabatic
curves (dotted, blue) and between the adiabatic states (red).
8 7
6 È2> 6
4 5
adiabatic
2 4 adiabatic
E 0 È1> È1> DE
3
-2 M 44
diabatic 2
-4 È2> diabatic
1 W
-6
0
-6 -4 -2 0 2 4 6 8 -6 -4 -2 0 2 4 6 8
x x

The difference in slopes of the diabatic energies


p is ∆f = f1 −f2 , The distance
between the adiabatic states is ∆E(x) = ∆f 2 (x−x0 )2 + 4W 2 . For a constant
velocity v at which this crossing is passed the transition probability is
 W2 
P12 = 1 − Exp − 2π and P11 = 1 − P12 (A-7.1)
h̄ ∆f v
Meaning of the exponential form: At high speed the time for passing the crossing
region, W/(∆f v) is very much smaller than the characteristic interaction time
h̄/W , the system stays diabatic, P12 = 0. Branching into both channels is
significant if the energy uncertainty (inverse residence time in the crossing area)
is of the order of the energy separation between the adiabatic states (2W ).

1.0 1.0
0.8 W = 0.4 0.8 v=1
0.6 0.6 2
P12 P12
0.4 0.2 0.4 4
0.2 0.1 0.2
0.0 0.0
0.0 0.5 1.0 1.5 0.0 0.2 0.4 0.6 0.8 1.0
speed of crossing v interaction energy W
At very low speeds the energy uncertainty is small and the system chooses the
adiabatic path, P12 → 1. At high speeds the systems tends to diabatic passage.
Parameters: h̄ = 1, ∆f = 1.
132 LANDAU-ZENER

Output coupler for the atom laser : In the atom laser [96] a mirror
equivalent for coupling out a portion of the condensate into the continuum
is used which is based on a dynamic transition across a LZ crossing. The
condensate which we call state |1i is in a magnetic bottle. The continuum
is state |2i and RF pumping connects the two states. At a specific distance
from the trap center the energy gap between the trapped Zeeman state and
the continuum Zeeman state is E0 . If we irradiate the system with the radio
frequency ωRF = E0 /h̄ with a Rabi frequency Ω1 = |d12 |E0RF /h̄ over a time
period τ a superposition state forms between the bound and the continuum
state according to Eq. (5.31) and (5.32)

cos (Ω1 τ /2) |1i + sin (Ω1 τ /2) |2i . (A-7.2)

Here Ω1 τ is the pulse area from Eq. (A-2.10). To control the strength of cou-
pling the authors use an RF signal which changes in frequency with time. In
the frame of dressed states we have a curve crossing at a fixed position in space
between the bound state |1i an the continuum state dressed with one photon
|2i + h̄ωRF (t). The curves cross at a time when E0 = h̄ωRF (t).
In the atom laser experiment we have the following equivalences with respect
to the LZ picture introduced above x ≡ t, ∆f v ≡ dωRF /dt and W ≡ Ω1 .

Variation of dωRF /dt is the experimental control of the magnitude of P12 .

Images of coherent atom beams (pulsed and CW) are shown below [97].
APPENDIX 133

A-8 Hyperfine Zeeman Structure of Rb


The 5s electron in Rb has orbital angular momentum, ` = 0. The projection of
its spin S may take two values mS = ±1/2. 87 Rb has nuclear spin I = 3/2. The
projection of the nuclear spin I has four values, mI = ±1/2 und mI = ±3/2.
As a consequence the ground state level is eightfold degenerate in the absence of
interaction between S and I and in the absence of an external field. A possible
basis for describing the ground state is

{n = 5, ` = 0, mL = 0, mS = ±1/2, mI = ±1/2, ±3/2} . (A-8.1)

Without external field the sole interaction term for hyperfine structure is the
contact term, which accounts for the fact that the magnetic field inside the
nucleus is different from that outside:

Whf = A I · S (A-8.2)

where A ∝ ge gn µB µn |R5s (r = 0)|2 derives from the probability of the 5s-


electron residing at the origin (inside the nucleus). To calculate the operator
I · S we introduce the total angular momentum

F=I+S (A-8.3)

and introduce the basis functions |I = 3/2, S = 1/2; F, MF i The possible values
for the total angular momentum in this case are F = 1 und F = 2. In this case
we may rewrite Eq. (A-8.2) as
1
A F2 − I2 − S2

AI · S = (A-8.4)
2
The eigenvalues of the I · S operator in the |F, MF i basis are

Ah̄2
A I · S |F, MF i = [F (F + 1) − I(I + 1) − S(S + 1)] |F, MF i . (A-8.5)
2
where we use the quantum numbers F, I, and S. From this equation we obtain
the following energy corrections due to the contact term Eq. (A-8.2)

Ah̄2 3
+ for F = 2 , MF = 0, ±1, ±2
2 2
Ah̄2 5
− for F = 1 , MF = 0, ±1 . (A-8.6)
2 2
The degeneracy of the 5s-level is thus removed: we obtain a threefold generate
level F = 1 and a five-fold degenerate level F = 2. The experimental hyperfine
splitting of the 5s-level in 87 Rb is [98]

2Ah̄2 = 6.834 682 610 904 30 (10) GHz (A-8.7)

Zeeman-Effect und Hyperfine structure We next explore the effect of


a magnetic field along the z-axis. Without magnetic field the total angular
momentum vector is constant in time F = L + S + I. In the presence of an
external magnetic field the vector of the magnetic moment associated with the
134 ZEEMAN STRUCTURE

total angular momentum MF = M` + MS + MI precesses around the direction


of B. The Zeeman-Hamilton-operator for the interaction of the three magnetic
moments with the external field is
WZ = −B0 (M` + MS + MI )
= ω0 (Lz + 2Sz ) + ωn Iz (A-8.8)
where the Larmor frequencies are greatly different:
q µB
ω0 = − B0 = − B0 (A-8.9)
2me h̄
q gn µn
ωn = gp B0 = B0 . (A-8.10)
2mp h̄
where Bohr’s magneton is µB = 1.399 624 624(56) MHz/Gauss, the nuclear g-
factor for 87 Rb is gn = −0.0009951414(10) [98] and the nuclear magneton is
µn = 762.268 Hz/G.

Zeeman-effect in the ground state in the |mI mS i basis Since ` = 0,


Lz = 0 and we are left with the contact term interaction and the effect of
precession of the spins:

W = A I · S + 2ω0 Sz + ωn Iz (A-8.11)

We may separate this interaction into the terms

W1 = A Iz Sz + 2ω0 Sz + ωn Iz (A-8.12)

and
1
W2 = A (I+ S− + I− S+ ) (A-8.13)
2
where we have as usual

I± = Ix ± i Iy and S± = Sx ± i Sy (A-8.14)

As coupled basis we use the representation |I S; F MF i, which we abbreviate as


|F MF i. In the uncoupled basis we have the eigenfunctions |I mI ; S mS i, which
we abbreviate as |mI mS i.
The contact term A I · S is diagonal in the |F MF i-basis, with the eigenvalues
of Eq.(A-8.6). However in the uncoupled basis we have the nondiagonal con-
tributions from Eq.(A-8.13). Applying the spin operators we obtain the matrix
elements
1
h±, mI |ASz Iz |±, mI i = ± mI h̄2 A
2
h±, mI |2Sz |±, mI i = ±h̄
h∓, mI |S± |±, mI i = h̄
p
hms , mI |I+ |ms , mI + 1i = h̄ 15/4 − mI (mI + 1)
p
hms , mI |I− |ms , mI − 1i = h̄ 15/4 − mI (mI − 1) .
from which we obtain the energy matrix given in Table 15.1. We have set h̄ = 1
and we have neglected third term in Eq.(A-8.8) since |ωn |  |ω0 |.
APPENDIX 135

Table 15.1: Energy matrix for coupled spin 1/2 + spin 3/2 states in a magnetic field.
1 3
ms mI 2
+ 2
− 21 + 3
2
1
2
+ 1
2
− 12 + 1
2
− 12 − 1
2
1
2
− 1
2
1
2
− 3
2
− 12 − 3
2
1 3 3A
2
+ 2 4
+ ω0 0 0 0 0 0 0 0

− 12 + 3
2
0 − 3A
4
− ω0 3 A/2 0 0 0 0 0
1 1
√ A
2
+ 2
0 3 A/2 4
+ ω0 0 0 0 0 0
− 12 + 1
2
0 0 0 −A
4
− ω0 0 A 0 0
A

− 12 − 1
2
0 0 0 0 4
− ω0 0 3 A/2 0
1
2
− 1
2
0 0 0 A 0 −A
4
+ ω0 0 0
1 3

2
− 2
0 0 0 0 3 A/2 0 − 3A
4
+ ω0 0
− 12 − 3
2
0 0 0 0 0 0 0 3A
4
− ω0

After diagonalizing we obtain the eigenenergies (the labels |F MF i are good


quantum numbers only at low magnetic field) for the quintet system
3
E(F = 2, MF = +2) + Ah̄2 + h̄ω0
=
4
1
q
2
E(F = 2, MF = +1) = − Ah̄2 + Ah̄2 + Aω0 h̄3 + (h̄ω0 )2
4
1 2
q
2
E(F = 2, MF +
= 0 ) = − Ah̄ + Ah̄2 + (h̄ω0 )2
4
1 2
q
2
E(F = 2, MF = −1) = − Ah̄ + Ah̄2 − Aω0 h̄3 + (h̄ω0 )2
4
3
E(F = 2, MF = −2) = + Ah̄2 − h̄ω0 (A-8.15)
4
and for the triplet system
1
q
2
E(F = 1, MF = +1) = − Ah̄2 − Ah̄2 − Aω0 h̄3 + (h̄ω0 )2
4
1 2
q

2
E(F = 1, MF = 0 ) = − Ah̄ − Ah̄2 + (h̄ω0 )2
4
1 2
q
2
E(F = 1, MF = −1) = − Ah̄ − Ah̄2 + Aω0 h̄3 + (h̄ω0 )2 (A-8.16)
4
At very weak magnetic fields we may neglect terms with ω02 under the square
root. Hence the states with MF = 0 stay independent of the magnetic field
while the F = 1 terms show a linear dependence on the magnetic field:
5 1
E(F = 1, MF = +1) = − Ah̄2 + h̄ω0
4 2
− 5 2
E(F = 1, MF = 0 ) = − Ah̄
4
5 1
E(F = 1, MF = −1) = − Ah̄2 − h̄ω0 (A-8.17)
4 2
The Figures below show the Zeeman-effect at various values of the magnetic
field. At high field the uncoupling of I and S, the Paschen-Back effect, is clearly
apparent.
136 ZEEMAN STRUCTURE

8 8

6 6

4 4

2 2
The linear Zeeman effect
MHz

MHz
0 0
appears for the low field
!2 !2 states of F = 1 and F = 2.
!4 !4

!6 !6

B !Gauss" B !Gauss"
0 2 4 6 8 10 0 2 4 6 8 10

6000

4000

2000

The F = 1 triplet mani-


0
fold and the F = 2 quintet
MHz

!2000
manifold is spaced by 6.8
GHz at low field.
!4000

!6000

B !Gauss"
0 500 1000 1500 2000 2500 3000

30000

20000

10000

Paschen-Back effect at
MHz

high field.
!10000

!20000

!30000

B !Gauss"
0 5000 10000 15000 20000
APPENDIX 137

A-9 Atom Interferometry


Peters and Chu reported in 2001 [4] on a
Mach-Zehnder atom-interferometer capable of !
measuring the local gravitational acceleration
!
with a precision of 10−9 g.
To realize the beam splitters and mirrors of " ! " "
the interferometer the instrument employed
stimulated Raman transitions between hyper- "
fine states of the Cs atom. " # $%
#
Three different concepts are at work :
• The two-photon transition |1i ↔ |2i leads to a transfer of momentum to
the atom of h̄(k1 +k2 ) = h̄keff if two counter-propagating beams are used.7
The internal atomic energy changes by h̄(ω1 − ω2 ) = h̄ωhfs .
Rt
• The transition probability depends on the pulse area 0 Ω1 (t0 )dt0 . It can
be chosen such that atoms which traverse a region where both beams are
present experience a beam splitter (pulse area of π/2) or a mirror (pulse
area of π), see Eq. (5.36).

• The quantum mechanical phase of the superposition state at the exit of


the interferometer depends on the local phases at the Raman transitions
at the beam splitters, H1 ∝ Ω1 exp[i(kz − ωL t)].

A scheme of the interferometer is shown in the figure on page 138. An atomic


beam in state |1i is sent into the interferometer, all atoms travel along the hor-
izontal direction with momentum p. The z-axis is vertical and points against
the gravitational
√ force. The atoms are transferred into the superposition state
(|1i + |2i)/ 2 at the first beam splitter. The component |2i suffered a momen-
tum transfer of h̄keff perpendicular to its original direction and travels along
path A. The momentum transfer along z by h̄keff permits component |2i to
separate spatially from path B where component |1i propagates.
After a time T both components experience a Raman pulse with pulse area π.
Now the component on path A is transferred into state |1i and looses the mo-
mentum h̄keff , while the component travelling along B is transferred into state
|2i and gains the vertical momentum h̄keff .
After a further period T the paths of both components meet in a beam
splitter with the pulse area π/2. Here atoms are transferred into either state |1i
or |2i, pending on the phase of the Raman pulse. A modulation of interference
occurs when the two components |1i and |2i, show a phase difference in their
respective paths of this trapezoidal arrangement.
The two atomic states can be selectively detected by either photoionization
or selective fluorescence excitation and the rate of transfer from the initial state
|1i into state |2i be determined. In the experiment the two states are the
7 The plus sign appears here because photons are absorbed from one beam and stimulated

in the opposing beam.


138 ATOM INTERFEROMETRY

hyperfine-ground states of Cs with F = 3, m = 0 and F = 4, m = 0.

F /2 p u ls e F p u ls e F /2 p u ls e
z ( b e a m s p litte r ) ( m ir r o r ) ( b e a m s p litte r )

A
2
B A B 3
2
?
1
A
th
A P a
B 1 B B
2
?
1
P a th B

t
0 T 2 T

A phase difference in the two paths arises from the different magnitude of action
along the two paths. The classical action corresponds to the time integration
over the particle energy
Z 2T  
1
Sclass = M v 2 − M gz dt . (A-9.1)
0 2
In the experiment this classical action ist much larger than h̄. The phase dif-
ference in the two paths appears as
A B

∆φpath = Sclass − Sclass /h̄ . (A-9.2)
This phase difference in Eq. (A-9.2) vanishes when g is independent of z, an
assumption well justified on the basis of the small size of the interferometer.
A second contribution to the phase difference appears from the local phase
of the Raman transitions. Whenever the atomic state changes due to the inter-
action
H1 = h̄Ω1 ei(k1 z−ω1 t) |3ih1| + h̄Ω2 ei(−k2 z−ω2 t) |3ih2| + c.c. (A-9.3)
a phase information [100]
φi = ±(keff zi − ωeff ti ) (A-9.4)
is imprinted in the atom where zi is the position of the atom at time ti and
where the sign depends on the initial state of the atom. Four critical transfers
of phase occur as they are marked in the Figure on page 138. The phase φA 1
in imprinted in the atom when it leaves the original flight direction and travels
along path A. This atom experiences a second phase shift φA 2 , when it is pumped
back into state |1i to continue along the horizontal direction. An atom which
passes the first beam splitter undeflected experiences a phase shift φB 2 at the
π-pulse and continues in the interferometer as state |2i. If this atom decides at
the second beam splitter to change the direction to horizontal, it experiences a
phase shift φA3 . The phase difference of the superposition state which emerges
at the exit of the interferometer is
∆φL = φA A B A
 
1 − φ2 − φ2 − φ3 (A-9.5)
APPENDIX 139

In the absence of gravitation this phase difference would vanish due to the
symmetry of the interferometer.
The gravitational field destroys this symmetry as atoms in the second time
period fall three times as deep, a consequence of the relationship s = gt2 /2.
In order to calculate (A-9.5) we determine the spatial positions of the atom in
the different interferometer parts. We set z = 0 and vz = 0 at the entrance to
the interferometer. An atom along path A will reach after a time T the height
z2A = T h̄keff /M − gT 2 /2 with the vertical velocity v2A = h̄keff /M − gT . An atom
on path B on the other hand will sink to a position z2B = −gT 2 /2 where its
vertical velocity is v2B = −gT . Immediately after the mirror Raman pulse we
have v2A = −gT and v2B = h̄keff /M − gT . At the end of the second segment of
the interferometer atoms reach the position z3A = h̄keff T /M − 2gT 2 . The sum
of these contributions gives a phase difference

∆φL = keff g T 2 . (A-9.6)

Hence the phase shift is directly related to the local value of g. A precise mea-
surement of g requires T to be chosen as large as possible. This is achieved best
in an atomic fountain in which atoms are thrown up vertically.
For this purpose atoms were first collected
in a MOT and then, after a phase of
polarization gradient cooling the trapping
beams from above were detuned red and
the beams from below were detuned blue
by ∆ω. This causes a moving polariza-
tion gradient in which atoms√move with a
stationary speed v = c ∆ω/( 2 ω0 ) in the
vertical direction. Stationary, because at
this speed the atoms see both beams at
the same frequency. After 200 µs at ∆ω =
3 MHz, all atoms are pumped into state |1i
and all lasers are turned off. In this fash-
ion an atomic fountain emerges with initial
speed of a few meters/second which can be
operated at a repetition rate of 1 Hz.
In this fountain the interferometer dis-
cussed above was realized by sending the
Raman beams parallel to the beam in the
fountain.
The wavepackets separate vertically in the time between the first and second
π/2-pulse (path A and path B). At the turning point in the gravitational field
the spatial separation of the two paths amounts to about 1.1 mm, at T = 160 ms.
The following pictures show the vertical trajectory and the spatial separation of
the wavepackets as function of time. The starting speed is vz = 1.5 m/s. The
height at the turning point is 12.5 cm at a flight time of 160 ms.
In the arrangement on page 138 the pulse area is determined by the time of flight
of atoms through the laser beam. In the fountain the pulse area is controlled by
the pulse duration. The three Raman pulses were realized using square pulses
of 40, 80 and 40 µs duration, the pulses being tuned on resonance as dictated
by the momentary speed. Due to the gravitational acceleration the atoms see
140 ATOM INTERFEROMETRY

the second π/2-pulse at a gravitation-induced Doppler shift ∆ω = keff · g T . For


Cs we have keff · g ≈ 2π × 23 M Hz/s.
12
1.0
10 trajectory
0.8
8

cm 0.6
6 mm

0.4
4

2 0.2 wave packet separation

0 0.0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30
seconds seconds

The phaseshift ∆φRF = ∆ω T is tuned


experimentally to record the interference
pattern. The pattern shown here is taken
from figure 2 of the reference [99]. It shows
40 measurements, recorded in over a one
minute interval. The phase shift π cor-
responds to a detuning ∆ω/(2π) ≈ 3 Hz.
Peters and Chu observed this pattern to
change with time due to temporal changes
of g owing to the tide of the Pacific Ocean,
30 miles from Stanford. Temporal changes
on the scale of minutes in g can be recorded
with a precision of 10−9 .
APPENDIX 141

A-10 Multiphoton Bragg Scattering


Light-pulse atom interferometers use atom-photon interactions to coherently
split, guide, and recombine freely falling matter-waves. Substantial improve-
ments in atom interferometry were recently achieved in employing a Ramsey-
Bordé type interferometer and multiphoton Bragg scattering.8 We consider two
laser beams counter-propagating along the vertical (gravitational) axis with fre-
quencies ω1 (from below) and ω2 (from above). We have ω1 ≈ ω1 and hence ~k1 ≈
−~k2 . If an atom scatters n photons from each of the two laser beams and ends
up in the initial state, the process transfers a momentum nh̄(~k1 − ~k2 ) ≈ 2nh̄k to
the atom. Herer M is the mass of the cesium atom. The kinetic energy of the
atom changes by ∆Ekin = 4n2 h̄ωR where ωR = h̄(~k1 − ~k2 )2 /2M . This energy
change must equal the change of energy in the two laser fields nh̄(ω1 − ω2 ). The
condition for such a multiphoton resonance is therefore

ω1 − ω2 = 4nωR . (A-10.1)

Now we consider a recent experiment9 where atoms in a 1.5 meter tall foun-
tain travel against the gravitational field with a speed of ≈ 450 m/s and a
transfer with n = 10 is used. The two lasers are well detuned from resonance by
δ = 4 GHz. Due to the free fall the resonance condition for ω1 − ω2 changes by
23 MHz/s which is accounted for by an EOM throughout the experiment. At
time t1 a π/2 pulse (duration of 100 µs) transfers the momentum of 2n photons
to the atom with 50% probability, thus the atom wavepacket separates with
a relative velocity of 2nvR (vR is the one-photon recoil speed). Depending on
whether momentum was transferred or not, the atom follows trajectory 1 or 6.
The path separations are greatly exaggerated in the figure below.
Π Π Π Π
2 2 2 2
After a time T a second pulse stops
2 this relative motion and after a further
time interval T 0 a third pulse directs
1 3 the wavepacket portions against each
other. Finally at time t4 = t1 + 2T + T 0
4
a fourth pulse overlaps the atoms again.
height ™

a
6 5 b
In this way two conjugate interferom-
eters appear with outputs labeled (a),
(b) and (c), (d). Because of the mo-
tion of the atoms, which gives rise to a
T T' T c Doppler frequency shift, addressing the
d upper and lower interferometer requires
t1 t2 t3 t4
two separate laser frequencies for ω2 .
time ™
The probability that an atom arrives at output (c) is proportional to the phase
by cos2 φ where φ now contains two contributions, φ = φF + φI . A classical
term appears from the free evolution of the atom φF = Sclass /h̄, which accounts
8 H. Müller, S. Chiow, Q. Long, S. Herrmann, and S. Chu, Atom Interferometry with up

to 24-Photon-Momentum-Transfer Beam Splitters, Phys. Rev. Lett. 100 180405 (2008).


9 S. Lan, P. Kuan, B. Estey, P. Haslinger, and H. Müller, Influence of the Coriolis Force

in Atom Interferometry, Phys. Rev. Lett. 108 090402 (2012).


142 ATOM INTERFEROMETRY

for the energy change associated with the recoil. A second term from the inter-
action with the local light field φI , because whenever a photon is absorbed, the
laser phase is added to the matter wave phase and subtracted for emission of a
photon.
The fluorescence f1 , f2 of the two outputs of each interferometer is detected
as the atoms pass a photomultiplier tube located near the initial launch position.
Herer 1 and 2 stand for (a) and (b) and for (c) and (d) respectively. To take
out fluctuations of the initial atom number, the normalized fluorescence f =
(f1 − f2 )/(f1 + f2 ) is used to monitor the fringe pattern when scanning the
phase of the last beam splitter, the fringe period being 2π/n. The arrival times
are separated on a scale of greater than about 1 ms and hence the individual
ports can be monitored separately.
The setup allows for improved precision as the momentum-space splitting of
the wavepacket (usually 2h̄k) increases by a factor of n2 . This was exploited in
a recent implementation of this interferometer10 the non-vertical deflection of
the atoms by the Coriolis acceleration ( Ω × ~v ) caused by the Earth’s rotation.

In this experiment the Coriolis force


was actively compensated for by tilt-
ing a mirror with piezoelectric actua-
tors. This allowed to give the momen-
tum transfer ~k1 −~k2 a constant direction
as seen from an inertial frame, one that
does not rotate with the Earth, with a
substantial improvement of fringe con-
trast.

The increased momentum transfer in multiphoton Bragg scattering leads to


larger areas enclosed between the interferometer arms and hence greater sensi-
tivity. It can be combined with common-mode noise rejection between simulta-
neous interferometers. This permits to balance out mechanical vibrations and
laser noise to even further increase sensitivity.
These light-pulse atom interferometers are important (and will become even
more so) in measurements of the local gravity, gravity gradients, the gravi-
tational constant G, gravitational waves, the Sagnac effect, the fine structure
constant11 , and tests of fundamental laws of physics, such as the most precise
measurement of the gravitational redshift to date12 .

10 S. Lan, P. Kuan, B. Estey, P. Haslinger, and H. Müller, Influence of the Coriolis Force

in Atom Interferometry, Phys. Rev. Lett. 108 090402 (2012).


11 A. Wicht. J. M. Hensley, E. Sarajlic and S. Chu, A Preliminary Measurement of the Fine

Structure Constant Based on Atom Interferometry, Physica Scripta T102 82-88 (2002).
12 H. Müller, A. Peters, and S. Chu, A precision measurement of the gravitational redshift

by the interference of matter waves, Nature (London) 463, 926-929 (2010).


APPENDIX 143

A-11 Feshbach Resonances


In Section A-4 we considered the collision of two atoms with masses m1 and m2
under the interaction of the potential U (R), where R is the distance between
the two atoms. The separated atoms are prepared in a plane wave with relative
kinetic energy E = h̄2 k 2 /(2µ) and relative momentum h̄k, with the reduced
mass is µ = m1 m2 /(m1 + m2 ). The plane wave is expanded in spherical partial
waves with angular momentum ` = 0, 1, 2.. with designation s-, p-, d-,.. waves.
For a central potential U (R) (depending only on the magnitude of R) there is
no coupling among partial waves, each of which is described by a solution to
the Schrödinger equation (A-4.8).

h̄2 d2 h̄2 `(`+1)


 
− u` (kR) + + U (R) u` (kR) = E(k) u` (kR) . (A-11.1)
2µ dR2 2µR2

For a given value of ` this equation has a disrecte spectrum of N bound state
solutions at energies Ev,` for E < 0 and a continuous spectrum of scattering
states with E > 0. Bound states for a given ` are labeled by the vibrational
quantum number v = 0, ..., N − 1 when counting up from the bottom of the
potential. To label threshold bound states we use the quantum number n =
−1, −2, ... counting down from the top of the potential for the last, next to
last, etc. The bound state solutions En,` are normalized to unity, hn `|n `i2 =
1, as they tend to zero for R → ∞. The scattering solutions represent the
incident plane plus a scattered wave and approach (A-4.10) where δ` (E) is
the scattering phase shift, the key parameter that incorporates the effect of
the potential interaction in the collision. The continuum wave function |E `i
is normalized per unit energy, hE `|E 0 `i = δ(E − E 0 ). For van der Waals
potentials, U (R) ∝ R−6 , the tangens of the phase shift varies as k for s waves
near threshold, 13

tan δ0 (E) = −ka0 , (A-11.2)

where a0 is the s-wave scattering length.


When the scattering length is positive and sufficiently large, that is, large
compared to the characteristic length scale of the molecular potential, the last
s-wave bound state of the potential, labeled by index n = −1 and ` = 0, is just
below threshold with a binding energy Eb = −E−1,0 given by 14

E−1,0 = h̄2 /2µa20 (A-11.3)

from which we see that the scattering length tends to ±∞ as E−1,0 → 0.

Alkali- metal atoms which are commonly used in Feshbach resonance ex-
periments with Bose-Einstein condensates have a 2 S1/2 electronic ground state.
The singlet and triplet states which emerge from their interaction, see page 95,
characterize the chemical bonding at short-range. Due to the hyperfine struc-
ture of the atomic states the singlet and triplet terms evolve from the hyperfine
split dissociation limits in a way shown below for 85 Rb.
13 H.Friedrich, Theoretische Atomphysik, Springer (1994) Chapter 1.4
14 C. Chin, R. Grimm, P. Julienne and E. Tiesinga, Feshbach resonances in ultracold gases,
Rev. Mod. Phys. 82 1225-1258 (2010).
144 ATOM INTERFEROMETRY

The atomic hyperfinestructure F = 1 and F = 2 splits the two potential


energy curves ”singlet” and ”triplet” into 5+4+3=12 molecular states, distin-
guished by the projection of the total angular momentum on the internuclear
axis. The twelve states converge at large separation into the separated atom
limits (F = 2 + F = 2), (F = 1 + F = 2) and (F = 1 + F = 1), as shown in the
figure on the right. The second of these limits is degenerate as either atom 1 or
atom 2 might be in the state (F = 1, 2).
2
0.4 triplet
1 3 +
Su È2\a +È2\b

Hcm-1 L
0.2
U H 1000 cm-1 L

0 a b
Rb H5sL+Rb H5sL
0.0 È1\a +È2\b , È2\a +È1\b

U-ECoulomb
-1
-0.2
-2 1 +
Sg È1\a +È1\b
-3 -0.4
singlet
-4
6 8 10 12 14 16 18 20 20 22 24 26
R HbohrL R HbohrL

Adding a magnetic field, see A-8, the atomic hyperfine states (F = 1, 2) split
into 3 and 5 Zeeman states respectively and the twelve molecular states shown
above split into 64 molecular states, the dissociation limits now being character-
ized by the magnetic quantum numbers of the two free atoms, |F1 m1 i+ |F2 m2 i.
Note that the Ground state Zeeman shift in Rb is typically 0.70 MHz/Gauss, by
comparison the hyperfine splitting (F = 1, 2) is 6.8 GHz. Hence at low magnetic
field the 64 molecular states practically coincide with the 12 molecular states
shown above.
In the context of multiple bound states emerging from different dissociation
limits it is clear that the following schematic situation (shown below) may ap-
pear: A bound state in the potential UB (R) converging to the dissociation limit
|F10 m01 i + |F20 m02 i is embedded in the continuum channel of the potential VA (R)
arising from |F1 m1 i + |F2 m2 i.
2
Collision channels may be ÈF'1 m'1 \+ÈF'2 m'2 \
grouped into two classes, 1 v
”closed” (bound) and ”open” 0
EHkL
ÈF1 m1 \+ÈF2 m2 \
(continuum) channels. A third
energy

-1000
distinct class, ”quasi-bound”,
-2000 B A
may also appear but does not in
the case shown here. -3000
In the absence of interaction -4000
between the potentials A and B, 8 12 16 20
R HbohrL
the scattering in the continuum
of channel of A is separate and independent of the presence of possible bound
states in B which might occur at the scattering energy E. An example of such
a bound state is the red vibrational level labelled v.
Very generally interactions appear which were not part of the Hamiltonian
used to generate the potential functions VA (R) and VB (R). Such interactions
allow the continuum channel to feel the presence of bound state channels equiv-
alent to the coupling in the Landau-Zener case of actual curve crossing (A-7).
In this case the situation may be handled by defining two coupled Schrödinger
APPENDIX 145

equations,13 written here for when ` = 0,


h̄2 d2
 
− + UA (R) uA B
0 (kR) + VAB (R) u0 (kR) = E(k) uA
0 (kR)
2µ dR2
h̄2 d2
 
− + U B (R) uB A
0 (kR) + VAB (R) u0 (kR) = E(k) uB
0 (kR) .
2µ dR2
VAB is the interaction between bound and continuum states and the solutions
are at the same energy, E(k), referenced to the lower dissociation limit.
In the absence of an interaction the phase shift in the open channel A varies
smoothly with the collision energy, E(k). In the event of interaction the con-
tinuum phase shift in the open channel A changes by π whenever the scattering
energy surpasses a bound state in the closed channel B.
This is termed Breit-Wigner resonance.
The width over which the phase
Π
changes by π is related to the life-
∆ time of the bound resonance v in
pre-dissociation into the lower dis-
phase shift ∆

Π
sociation limit. With φv we desig-
d∆
2
dE
nate the bound state wavefunction
in potential B in the absence of in-
G
teraction. We have for the predisso-
ciation rate13
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
energy Γ = 2π|hφv |VAB |uA 2
0 i| .

When we apply a magnetic field the Zeeman-splitting among the various disso-
ciation limits |F1 m1 i + |F2 m2 i changes. Hence we can shift the absolute energy
of the level v with respect to the dissociation limit of potential A with an ex-
ternal magnetic field. We may even make the bound level cross from E > 0 to
E < 0. When the field is tuned such that the bound state in B is at the en-
ergy E = 0 in the open channel, the s-wave scattering length diverges (A-11.2).
Hence one may magnetically tune the scattering length, a topic of many recent
studies with Bose-Einstein Condensates.15

15 M. Theis, G. Thalhammer, K. Winkler, M. Hellwig, G. Ruff, R. Grimm, and J. Hecker

Denschlag, Tuning the Scattering Length with an Optically Induced Feshbach Resonance,
Phys. Rev. Lett. 93 123001 (2004).
146 Atomic fountain clock

A-12 Atomic Fountain Clock


Zacharias in the 1950s [101]

The first rf spectroscopy experiment in an atomic fountain using laser cooled


atoms was reported in 1989 [102]

improved by Gibble and Chu [103].

PTB construction details [105]

operational video: Nist F1 fountain clock.mpeg

The ultimate limitation to the accuracy of an atomic fountain clock are colli-
sions between Cs atoms in the beam. These lead to a measurable frequency
shift.
Because of the extremely low relative velocities of the atoms, the cross sections
are large. By varying the density of Cs atoms in the fountain, Gibble found fre-
quency shifts of the order of a few mHz for an atomic density of 109 cm−3 [104].
Bibliography

[1] C. Cohen-Tannoudji and W.D. Phillips. New Mechanisms for Laser Cooling.
Phys. Today 43, October, 33-40 (1990). 1
[2] W. Neuhauser, M. Hohenstatt, P. E. Toschek, and H. Dehmelt, Localized visible
Ba+ mono-ion oscillator. Phys. Rev. A 22 1137 (1980). 2
[3] C. Amole et al., Resonant quantum transitions in trapped antihydrogen atoms.
Nature 483 439-443 (2012). 3
[4] A. Peters, K. Chung, and S. Chu, High-precision gravity measurements using
atom interferometry. Meterologica 38 25-61 (2001). 3, 137
[5] V. S. Letokhov and V. G. Minogin, Cooling, trapping, and storage of atoms by
resonant laser fields. J. Opt. Soc. Am. 69 413-419 (1979). 3
[6] C. Salomon, J. Dalibard, A. Aspect, H. Metcalf, and C. Cohen-Tannoudji. Chan-
neling Atoms in a Standing Wave. Phys. Rev. Lett. 59, 1659 (1987). 5, 50
[7] J. C. Maxwell, Lehrbuch der Elektricität und des Magnetismus, 792 Deutsch
von B. Weinstein, Berlin 1883. 4
[8] P. Lebedev, Untersuchungen über die Druckkräfte des Lichtes. Annalen d.
Physik, 11 433-458 (1901). 4
[9] A. Ashkin, History of Optical Trapping and Manipulation of Small-Neutral Par-
ticle, Atoms and Molecules. IEEE Journal on Selected Topics in Quantum Elec-
tronics 6 841-856 (2000). 5
[10] V. S. Letokhov, Narrowing of the Doppler Width in a Standing Light Wave.
JETP Letters 7 272 (1968). 5, 8
[11] A. Ashkin, Trapping of Atoms by Resonance Radiation Pressure. Phys. Rev.
Lett. 40 729 (1978). 5, 11
[12] S. Stenholm, The semiclassical theory of laser cooling. Rev. Mod. Phys. 58
699-739 (1986). 5
[13] D. J. Wineland and H. Dehmelt, Proposed 1014∆ν < ν Laser Fluorescence Spec-
troscopy on Tl+ Mono-Ion Oscillator III (side band cooling). Bull. Am. Phys.
Soc. 20, 637 (1975). 8
[14] T. W. Hänsch and A. L. Schawlow, Cooling of gases with laser radiation. Opt.
Commun. 13 68 (1975). 8
[15] J. Prodan, W. D. Phillips, and H. Metcalf, Laser Production of a Very Slow
Monoenergetic Atomic Beam. Phys. Rev. Lett. 49 1149 (1982). 10
[16] K. Sengstock und W. Ertmer, Laser manipulation of atoms. Adv. At. Molec.
Opt. Physics 35 1 (1995). 10
[17] J. Dalibard und C. Cohen-Tannoudji, Dressed-atom approach to atomic motion
in laser light: the dipole force revisited. JOSA B 2 1707-1720 (1985). 11, 51, 64

147
148 BIBLIOGRAPHY

[18] R. P. Feynman, Lectures on Physics Vol I, 34-10 12


[19] Steven Chu, J. E. Bjorkholm, A. Ashkin, and A. Cable, Experimental Observa-
tion of Optically Trapped Atoms. Phys. Rev. Lett. 57 314 (1986). 13
[20] W. D. Phillips, Nobel Lecture: Laser cooling and trapping of neutral atoms.
Rev. Mod. Phys 70 721 (1998). 19
[21] P. Lett, R. Watts, C. Westbrook, W. Phillips, P. Gould, and H. Metcalf. Ob-
servation of Atoms Laser Cooled below the Doppler Limit. Phys. Rev. Lett. 61
169 (1988). 20, 66
[22] Y. Shevy, D. S. Weiss, P. J. Ungar, and Steven Chu, Bimodal speed distributions
in laser-cooled atoms. Phys. Rev. Lett. 62 1118 (1989). 20, 66
[23] E. L. Raab, M. Prentiss, A. Cable, S. Chu, D. E. Pritchard, Trapping of Neutral
Sodium Atoms with Radiation Pressure. Phys. Rev. Lett. 59 2631 (1987). 21
[24] T. Walker, D. Sesko, and C. Wieman, Collective behavior of optically trapped
neutral atoms. Phys. Rev. Lett. 64 408 (1990), 23, 24
[25] W. Ketterle, K. B. Davis, M. A. Joffe, A. Martin, and D. E. Pritchard, High
densities of cold atoms in a dark spontaneous-force optical trap. Phys. Rev. Lett.
70 2253 (1993). 23
[26] C. Monroe, W. Swann, H. Robinson, and C. Wieman, Very cold trapped atoms
in a vapor cell. Phys. Rev. Lett. 65 1571 (1990). 24, 25
[27] F. Shimizu, K. Shimizu, and H. Takuma, A High-Intensity Metastable Neon
Trap. Chem. Phys. 145 327 (1990). 24
[28] E. Riis, D. S. Weiss, K. A. Moler, and S. Chu, Atom funnel for the production
of a slow, high-density atomic beam. Phys. Rev. Lett. 64 1658 (1990). 24
[29] J. Nellessen, J. Werner, and W. Ertmer, Magneto-optical compression of a mo-
noenergetic sodium atomic beam. Opt. Commun. 78 300-308 (1990). 24
[30] Z. T. Lu, K. L. Corwin, M. J. Renn, M. H. Anderson, E. A. Cornell, and C. E.
Wieman, Low-Velocity Intense Source of Atoms from a Magneto-optical Trap.
Phys. Rev. Lett. 77 3331 (1996). 24
[31] M. O. Scully und M. S. Zubairy, Quantum Optics Cambridge Univ. Press (1997)
29, 46
[32] C. Cohen-Tannoudji and S. Reynaud, Dressed-atom description of resonance
fluorescence and absorption spectra of a multi-level atom in an intense laser
beam. J. Phys. B 10 345 (1977). 33
[33] B. R.Mollow, Power Spectrum of Light Scattered by Two-Level Systems. Phys.
Rev. 188 1969 (1969). 38, 54
[34] C. Cohen-Tannoudji, J. Dupont-Roc und G. Grynberg, Atom-Photon Interac-
tions, John Wiley (1992). 148, 149
[35] Reference [34], pages 621-627. 35
[36] Reference [34], Excercise 17, page 597. 36
[37] H. P. Breuer and F. Petruccione, The Theory of Open Quantum Systems, Oxford
University Press, (2002), page 202. 46, 90
[38] A. Aspect, J. Dalibard, A. Heidmann, C. Salomon, and C. Cohen-Tannoudji,
Cooling Atoms with Stimulated Emission. Phys. Rev. Lett. 57 1688 (1986). 64
[39] J. Dalibard und C. Cohen-Tannoudji, Laser cooling below the Doppler limit by
polarization gradients: simple theoretical models. J. Opt. Soc. Am. B 6 2023-
2045 (1989). 65, 66, 73, 74, 75, 81, 82, 128
BIBLIOGRAPHY 149

[40] P. J. Ungar, D. S. Weiss, E. Riis, and S. Chu, Optical molasses and multilevel
atoms: theory. J. Opt. Soc. Am B, 6 2058-2071 (1989). 66
[41] A. Lindner, Drehimpulse in der Quantenmechanik Teubner (1984) 78, 128, 151
[42] Manuel Gessner, private communication. 79, 82
[43] A. Aspect et al., Laser Cooling below the one-photon recoil energy. J. Opt. Soc.
Am. B 6 2112 (1989). 83
[44] C. Cohen-Tannoudji, in Laser Manipulation of Atoms and Ions, Proceedings of
the Varenna Summer School 1991, NorthHolland, p. 99-169. 83
[45] C. Cohen-Tannoudji, Laserkühlung an der Grenze des Machbaren. Physikalische
Blätter 51 91-95 (1995). 83
[46] Reference [34], Chapter III, page 201. 88
[47] A. Aspect, E. Arimondo, R. Kaiser, N. Vansteenkiste, and C. Cohen-Tannoudji,
Laser Cooling below the One-Photon Recoil Energy by Velocity-Selective Co-
herent Population Trapping. Phys. Rev. Lett. 61 826 (1988). 91
[48] J. Lawall et al.,Two-Dimensional Subrecoil Laser Cooling. Phys. Rev. Lett. 73
1915 (1994). 91
[49] J. Lawall et al., Three-Dimensional Laser Cooling of Helium Beyond the Single-
Photon Recoil Limit. Phys. Rev. Lett. 75 4194 (1995). 91
[50] A. Aspect and R. Kaiser, Linear Momentum Conservation in Coherent Popula-
tion Trapping: A Case Study for a Quantum Filtering Process, Foundations in
Physics 20 1413 (1990). 91
[51] F. Meinert, C. Basler, A. Lambrecht, S. Welte, and H. Helm, Quantitative anal-
ysis of the transient response of the refractive index to conditions of electromag-
netically induced transparency. Phys. Rev. A 85 013820 (2012). 91
[52] M.H. Anderson, J.R. Ensher, M.R. Matthews, C.E. Wieman, and E.A. Cornell,
Observation of Bose-Einstein Condensation in a Dilute Atomic Vapor. Science
269 198 (1995). 93
[53] C. C. Bradley, C. A. Sackett, J. J. Tollett, and R. G. Hulet, Evidence of Bose-
Einstein Condensation in an Atomic Gas with Attractive Interactions. Phys.
Rev. Lett. 75 1687 (1995). 93
[54] K. B. Davis, M. -O. Mewes, M. R. Andrews, N. J. van Druten, D. S. Durfee,
D. M. Kurn, and W. Ketterle, Bose-Einstein Condensation in a Gas of Sodium
Atoms. Phys. Rev. Lett. 75 3969 (1995). 93
[55] D. G. Fried, T. C. Killian, L. Willmann, D. Landhuis, Bose-Einstein Conden-
sation of Atomic Hydrogen. S. C. Moss, D. Kleppner, and Thomas J. Greytak,
Phys. Rev. Lett. 81 3811 (1998). 93
[56] F. P. Dos Santos, J. Lonard, J. Wang, C. J. Barrelet, F. Perales, E. Rasel, C. S.
Unnikrishnan, M. Leduc, and C. Cohen-Tannoudji, Bose-Einstein Condensation
of Metastable Helium. Phys. Rev. Lett. 86 3459 (2001). 93
[57] A. Griesmaier, J. Stuhler and T. Pfau, Production of a chromium Bose-Einstein
condensate. Appl. Phys. B 82 211 (2006). 93
[58] S. Jochim, M. Bartenstein, A. Altmeyer, G. Hendl, S. Riedl, C. Chin, J. Hecker
Denschlag, and R. Grimm, Bose-Einstein condensation of molecules. Science 302
2101 (2003). 93
[59] http://www.uibk.ac.at/exphys/ultracold/atomtraps.html 93
[60] E. P. Gross, Structure of a quantized vortex in boson systems. Nuovo Cimento
20, 454 (1961). 100
150 BIBLIOGRAPHY

[61] L. P. Pitaevskii, Vortex lines in an imperfect Bose gas. Sov. Phys. JETP-USSR
13, 451 (1961). 100
[62] K. Burnett, P. S. Julienne, P. D. Lett, E. Tiesinga, and C. J. Williams, Quantum
encounters of the cold kind. Nature 416 225-232 (2002). 100
[63] A. Midgall, J. V. Prodan, W. D. Phillips, Th. H. Bergeman and H. J. Metcalf,
First Observation of Magnetically Trapped Neutral Atoms. Phys. Rev. Lett. 54
2596 (1985). 100
[64] D. E. Pritchard, Cooling neutral atoms in a magnetic trap for precision spec-
troscopy. Phys. Rev. Lett. 51 1336 (1983). 101
[65] T. Bergeman, G. Erez, and H. J. Metcalf, Magnetostatic trapping fields for
neutral atoms. Phys. Rev. A 35 15351546 (1987). 101
[66] H. F. Hess, Evaporative cooling of magnetically trapped and compressed spin-
polarized hydrogen. Phys. Rev. B 34 3476 (1986). 103
[67] W. Ketterle et. al., Proceed. Int. School of Physics, Enrico Fermi 1998 Course
CXL, Bose-Einstein Condensation of Atomic Gases, editors: M. Iguscio, S.
Stringari, and C. E. Wieman, (IOS Press Ohmsha, 1999) p. 67-176. 104, 150
[68] E. Arimondo et. al. in Reference [67] p. 573-590. 104
[69] K. W. Miller. S. Dürr, and C. Wieman, RF-induced Sisyphus Cooling in an
Optical Dipole Trap. Phys. Rev. A 66 023406 (2002) 104
[70] M. D. Barrett, J. A. Sauer, and M. S. Chapman, All-Optical Formation of an
Atomic Bose-Einstein Condensate. Phys. Rev. Lett. 87 010404 (2001). 104
[71] T. Weber, J. Herbig, M. Mark, H.-C. Nägerl, R. Grimm, Bose-Einstein Conden-
sation of Cesium. Science 299 232 (2003). 104
[72] G. Cennini, G. Ritt, C. Geckeler, and M. Weitz, All-Optical Realization of an
Atom Laser. Phys. Rev Lett. 91 240408 (2003). 104
[73] D. S. Durfee and W. Ketterle, Experimental studies of Bose-Einstein condensa-
tion. Optics Express 2 299 (1998). 104
[74] B. P. Anderson and M. A. Kasevich, Macroscopic quantum interference from
atomic tunnel arrays. Science 282 1686 (1998). 104
[75] M. R. Andrews, C. G. Townsend, H.-J. Miesner, D. S. Durfee, D. M. Kurn, W.
Ketterle, Observation of Interference Between Two Bose Condensates. Science
275 637 (1997). 104
[76] D. S. Hall, M. R. Matthews, C. E. Wieman, and E. A. Cornell, Measurements of
Relative Phase in Two-Component Bose-Einstein Condensates. Phys. Rev. Lett.
81 1543 (1997). 104
[77] W. Ketterle, Experimental studies of Bose-Einstein condensation. Physics Today
52 30-35 1999. 104
[78] D. J. Wineland, W. M. Itano, J. C. Bergquist, and R. G. Hulet, Laser-cooling
limits and single-ion spectroscopy. Phys. Rev. A 36 2220 (1987). 105
[79] H. Perrin, A. Kuhn, I. Bouchoule and C. Salomon, Sideband cooling of neutral
atoms in a far-detuned optical lattice. Europhys. Lett. 42 395-400 (1998). 105
[80] V. Vuletic, C. Chin, A. J. Kerman, and S. Chu, Degenerate Raman Sideband
Cooling of Trapped Cesium Atoms at Very High Atomic Densities. Phys. Rev.
Lett. 81 5768-5771 (1998). 105, 109
[81] I. H. Deutsch and P. S. Jessen, Quantum-state control in optical lattices. Phys.
Rev. A 57 1972-1986 (1997). 109
[82] G. Morigi, J. Eschner, and Christoph H. Keitel, Laser Cooling Using Electro-
magnetically Induced Transparency. Phys. Rev. Lett. 85 4458 (2000). 106, 110
BIBLIOGRAPHY 151

[83] F. Schmidt-Kaler, J. Eschner, G. Morigi, C. F. Roos, D. Leibfried, A. Mundt,


and R. Blatt, Laser cooling with EIT: application to trapped samples of ions or
neutral atoms. Appl. Phys. B 73 807-814 (2001). 106
[84] R. H. Dicke, The Effect of Collisions upon the Doppler Width of Spectral Lines.
Phys. Rev. 89, 472 (1953). 107
[85] M. Roghani and H. Helm, Trapped-atom cooling beyond the Lamb-Dicke limit
using EIT , Phys. Rev. A77 043418 (2008). 110
[86] M. Roghani, H.-P. Breuer and H. Helm, Dissipative light scattering by a trapped
atom approaching EIT conditions, Phys. Rev. A81 033418 (2010). 110
[87] J. P. Gordon, H. J. Zeiger, and C. H. Townes, The MaserNew Type of Microwave
Amplifier, Frequency Standard, and Spectrometer. Phys. Rev. 99, 12641274
(1955). 115
[88] R. P. Feynman, F. L. Vernon, and R. W. Hellwarth, Geometrical representation
of the Schrödinger equation for solving maser problems. J. Appl. Phys. 28 49-52
(1957). 116
[89] W. J. Thomson, Angular Momentum, John Wiley, New York, page 268. 119
[90] R. N. Zare, Angular Momentum, John Wiley, New York, 1988. 119, 126, 128
[91] I. I. Sobelmann, Atomic Spectra and Radiative Transitions, chapters 4.1.4, 4.2,
9.2, Springer, Berlin (1979). 120
[92] Reference [41], page 15. 120
[93] L. D. Landau, On the theory of transfer of energy at collisions. Phys. Z. Sowje-
tunion 2 46 (1932). 131
[94] C. Zener, Nonadiabatic crossing of energy levels. Proc. Roy. Soc. A137 696-702
(1932). 131
[95] E. C. G. Stueckelberg, Theorie der unelastischen Stosse zwischen Atomen. Helv.
Phys. Acta, 5 369-422 (1932). 131
[96] M. -O. Mewes, M. R. Andrews, D. M. Kurn, D. S. Durfee, C. G. Townsend, and
W. Ketterle, Output Coupler for Bose-Einstein Condensed Atoms. Phys. Rev.
Lett. 78 582-585 (1997). 132
[97] Images from a no longer active webpage of W. Ketterle : this webpage. 132
[98] D. A. Steck, Rubidium 87 D Line Data [http://steck.us/ alkali-
data/rubidium87numbers.1.6.pdf] 133, 134
[99] A. Peters, K. Y. Chung, and S. Chu, Measurement of gravitational acceleration
by dropping atoms. Nature 400 849-852 (1999). 140
[100] M. Kasevich and S. Chu, Measurement of gravitational acceleration of an atom
with a light-pulse atom interferometer. Appl. Phys. B 54 321-332 (1992). 138
[101] R. A. Nauman and H. Henry Stroke, Apparatus Up-ended: A Short History of
the Fountain A-Clock. Physics Today 49 89 (1996). 146
[102] M.A. Kasevich, E. Riis, S. Chu, and R.G. Devoe, RF Spectroscopy in an Atomic
Fountain. Phys. Rev. Lett. 63 612-616 (1989). 146
[103] K. Gibble and S. Chu, Future Slow-Atom Frequency Standards. Metrologia 29,
201-212 (1992) and Phys. Rev. Lett. 70, 1771 (1993). 146
[104] K. Gibble and B. Verhaar, Eliminating Cold-Collision Frequency Shifts. Phys.
Rev. A 52, 3370 (1995). 146
[105] R. Schröder, U. Hübner, and D. Griebsch, Design and Realization of the Mi-
crowave Cavity in the PTB Caesium Atomic Fountain Clock CSF1. IEEE Trans.
on Ultrasonics, Ferroelectric, and Frequency Control, 49 383-392 (2002). 146

You might also like