Professional Documents
Culture Documents
Quantum Mechanics Script
Quantum Mechanics Script
Quantum Mechanics
Ph.W. Courteille
Universidade de São Paulo
Instituto de Fı́sica de São Carlos
11/02/2022
2
.
3
.
4
Preface
This script represents a synthesis of several postgraduate courses given at the
Institute of Physics of São Carlos (IFSC) of the University of São Paulo (USP). The
courses are Quantum Mechanics (SFI5774), Atomic and Molecular Physics (SFI5814),
Quantum Mechanics B (SFI5707), Interaction of Light and Matter (SFI5905), and
Atomic Optics (SFI5887). The topics of the courses are, of course, closely intertwined.
The purpose of this composite script is to emphasize the interconnection of topics
and facilitate the understanding of how they are related. In part I we introduce the
quantum mechanics, which represents the fundamental theory for the rest of the book.
In the second part we focus on the structure of the atom. In the third and fourth
part we study the properties of light, its interaction with individual atoms and atomic
ensembles and how the interaction is influenced by cavities and surfaces. Finally, in
part V we introduce the optics of matter wave.
The course is intended for masters and PhD students in physics. The script
is a preliminary version continually being subject to corrections and modifications.
Error notifications and suggestions for improvement are always welcome. The script
incorporates exercises the solutions of which can be obtained from the author.
Information and announcements regarding the course will be published on the
website:
http://www.ifsc.usp.br/ strontium/ − > Teaching − > Semester
The student’s assessment will be based on written tests and a seminar on a special
topic chosen by the student. In the seminar the student will present the chosen topic
in 15 minutes. He will also deliver a 4-page scientific paper in digital form. Possible
topics are:
- Observation of super- and subradiant spontaneous emission of two ions (Exc. 21.3.3.2),
- Squeezed states (Sec. 3.8),
- The Jaynes-Cummings model (Sec. 15.2),
- Quantum projection noise (Sec. 16.3.2),
- Quantum gates (Sec. ??),
- The method of quantum Monte-Carlo wavefunction simulation (Sec. 16.1.2),
- The quantum Zeno effect (Sec. 16.3.1),
- Bloch equations: derivation and interpretation (Sec. 14.2),
- The quantum jumps, its history and observation (Sec. 16.1.2),
- Schrödinger’s cat (Sec. 16.1.1),
- The Einstein-Podolski-Rosen hypothesis and its experimental falsification (Sec. ??),
- Elitzur and Vaidman bomb testing problem (Sec. ??),
- Topological phases and the Aharonov-Bohm effect (Sec. 16.4),
- Quantum non-demolition measurements (Sec. 16.3.3),
- Calculation of photoelectric effect from Fermi’s golden rule (Exc. 6.4.5.7),
- Quantum correlations and the experiments of Young and Hanbury-Brown-Twiss
(Sec. 15.3.1),
- The Hartree-Fock method (Sec. 10.3.3),
- Temporal evolution of a free particle described by a Gaussian wave packet,
- The WKB approximation (Sec. 6.3),
- Rydberg atoms (Sec. 8.4.4),
- The helium atom (Sec. 10.2),
5
H.A. Bethe, R. Jackiw, Intermediate Quantum Mechanics, 3rd ed. Taylor & Francis
(1997)
J.I. Steinfeld, Molecules and Radiation, The MIT Press, Cambridge (2005)
I Quantum Mechanics 1
1 Antecedents of quantum mechanics 3
1.1 The discovery of the atom . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Democrit’s model . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.2 Thomson’s model and Rutherford’s experiment . . . . . . . . . 6
1.1.3 Emission of radiation in the planetary model . . . . . . . . . . 10
1.1.4 Zeeman effect in the planetary model . . . . . . . . . . . . . . 12
1.1.5 Bohr’s theory and its limitations . . . . . . . . . . . . . . . . . 13
1.1.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2 The discovery of the photon . . . . . . . . . . . . . . . . . . . . . . . . 16
1.2.1 Radiation in a conductive cavity . . . . . . . . . . . . . . . . . 16
1.2.2 Black body radiation . . . . . . . . . . . . . . . . . . . . . . . . 18
1.2.3 Planck’s distribution of modes . . . . . . . . . . . . . . . . . . 20
1.2.4 The corpuscular nature of the photon . . . . . . . . . . . . . . 21
1.2.5 Einstein’s transitions rates . . . . . . . . . . . . . . . . . . . . . 23
1.2.6 Absorption spectrum for a single atom . . . . . . . . . . . . . . 24
1.2.7 Absorption in a gas . . . . . . . . . . . . . . . . . . . . . . . . 27
1.2.8 Saturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.2.9 Specific heat of solids . . . . . . . . . . . . . . . . . . . . . . . 30
1.2.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.3 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7
8 CONTENT
2.2.8 Stationary Schrödinger equation . . . . . . . . . . . . . . . . . 46
2.2.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.3 Abstract formalism of quantum mechanics . . . . . . . . . . . . . . . . 48
2.3.1 Lie algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.3.2 Complete bases . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.3.3 Degeneracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.3.4 Bases as unitary operators . . . . . . . . . . . . . . . . . . . . . 51
2.3.5 Complete set of commuting operators . . . . . . . . . . . . . . 52
2.3.6 Uncertainty relation . . . . . . . . . . . . . . . . . . . . . . . . 53
2.3.7 Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.3.8 Spanning a Hilbert space with several degrees of freedom . . . 55
2.3.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.4 Time evolutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.4.1 Unitary transformations . . . . . . . . . . . . . . . . . . . . . . 65
2.4.2 Schrödinger picture . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.4.3 Heisenberg picture . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.4.4 Interaction picture . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.4.5 Hamiltonian under arbitrary unitary transformation . . . . . . 69
2.4.6 Ehrenfest’s theorem . . . . . . . . . . . . . . . . . . . . . . . . 70
2.4.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.5 Symmetries in quantum mechanics . . . . . . . . . . . . . . . . . . . . 72
2.5.1 Translation and rotation . . . . . . . . . . . . . . . . . . . . . . 72
2.5.2 Kick operator, Galilei and Lorentz boosts . . . . . . . . . . . . 75
2.5.3 Gauge transformations . . . . . . . . . . . . . . . . . . . . . . . 76
2.5.4 Noether’s theorem and conservation laws . . . . . . . . . . . . 77
2.5.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2.6 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.3.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.4 Numerical approaches for arbitrary potentials . . . . . . . . . . . . . . 99
3.4.1 Calculation of free and bound states wavefunctions . . . . . . . 99
3.4.2 The Fourier grid method for bound states . . . . . . . . . . . . 99
3.4.3 Steepest descent ot the ground state . . . . . . . . . . . . . . . 102
3.4.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
3.5 Harmonic oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.5.1 Factorization of the Hamiltonian and Fock states . . . . . . . . 104
3.5.2 Harmonic oscillator in spatial representation . . . . . . . . . . 106
3.5.3 Properties of the harmonic oscillator . . . . . . . . . . . . . . . 108
3.5.4 Time evolution of the unperturbed harmonic oscillator . . . . . 109
3.5.5 Multidimensional harmonic oscillator . . . . . . . . . . . . . . . 110
3.5.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.6 Superposition states of a harmonic oscillator . . . . . . . . . . . . . . . 111
3.6.1 Coherent states . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.6.2 Kicking a harmonic oscillator . . . . . . . . . . . . . . . . . . . 115
3.6.3 Shaking a harmonic oscillator . . . . . . . . . . . . . . . . . . . 119
3.6.4 Forcing a harmonic oscillator . . . . . . . . . . . . . . . . . . . 121
3.6.5 Quantization of the electromagnetic field . . . . . . . . . . . . 125
3.6.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
3.7 (Quasi-)probability distribution functions . . . . . . . . . . . . . . . . 128
3.7.1 The density operator . . . . . . . . . . . . . . . . . . . . . . . . 129
3.7.2 The Wigner function . . . . . . . . . . . . . . . . . . . . . . . . 129
3.7.3 P , Q, and Wigner functions for the harmonic oscillator . . . . 130
3.7.4 Representation of particular states in the Fock and Glauber basis132
3.7.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.8 Squeezed states of the harmonic oscillator . . . . . . . . . . . . . . . . 137
3.8.1 The squeezing operator . . . . . . . . . . . . . . . . . . . . . . 137
3.8.2 Squeezed state in the Fock basis . . . . . . . . . . . . . . . . . 139
3.8.3 Squeezed state in the Glauber basis . . . . . . . . . . . . . . . 140
3.8.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
3.9 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
3.9.1 on the Fourier grid method . . . . . . . . . . . . . . . . . . . . 141
3.9.2 on the harmonic oscillator . . . . . . . . . . . . . . . . . . . . . 142
12 Collisions 347
12.1 Motion of interacting neutral atoms . . . . . . . . . . . . . . . . . . . 347
12.1.1 The collisional phase shift . . . . . . . . . . . . . . . . . . . . . 348
12.1.2 Hard-sphere potentials . . . . . . . . . . . . . . . . . . . . . . . 352
12.1.3 Spherical wells with a flat bottom . . . . . . . . . . . . . . . . 353
12.1.4 Other types of potentials . . . . . . . . . . . . . . . . . . . . . 363
12.1.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
12.2 Scattering theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
12.2.1 Lippmann-Schwinger equation . . . . . . . . . . . . . . . . . . 371
12.2.2 Wave packets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
12.2.3 Born approximation . . . . . . . . . . . . . . . . . . . . . . . . 375
12.2.4 Spherical potentials . . . . . . . . . . . . . . . . . . . . . . . . 376
12.2.5 Scattering phase and length . . . . . . . . . . . . . . . . . . . . 377
12.2.6 Optical theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 378
12.2.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
12.3 Cold atomic collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
12.3.1 Collision cross section, unitarity regime . . . . . . . . . . . . . 383
12.3.2 Collisions between identical particles . . . . . . . . . . . . . . . 384
12.3.3 Collisions between hot atoms . . . . . . . . . . . . . . . . . . . 387
12.3.4 Photoassociation during ultracold collisions . . . . . . . . . . . 387
12.3.5 Ground state collisions . . . . . . . . . . . . . . . . . . . . . . . 389
12.3.6 Hyperfine structure . . . . . . . . . . . . . . . . . . . . . . . . . 389
12.3.7 Scattering length in specific channels . . . . . . . . . . . . . . . 390
12.3.8 Hyperfine coupling in magnetic fields . . . . . . . . . . . . . . . 390
12.3.9 Inelastic collisions . . . . . . . . . . . . . . . . . . . . . . . . . 392
12.3.10 Excited states collisions . . . . . . . . . . . . . . . . . . . . . . 393
12.3.11 Heteronuclear collisions . . . . . . . . . . . . . . . . . . . . . . 395
12.3.12 Heteronuclear electric dipole moment . . . . . . . . . . . . . . 398
12.3.13 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
12.4 Resonances in cold collisions . . . . . . . . . . . . . . . . . . . . . . . . 400
14 CONTENT
12.4.1 Shape resonances . . . . . . . . . . . . . . . . . . . . . . . . . . 400
12.4.2 Feshbach resonances . . . . . . . . . . . . . . . . . . . . . . . . 400
12.5 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
12.5.1 on cold collisions . . . . . . . . . . . . . . . . . . . . . . . . . . 402
12.5.2 on Feshbach resonances . . . . . . . . . . . . . . . . . . . . . . 402
Quantum Mechanics
1
Chapter 1
Antecedents of quantum
mechanics
Developed at the beginning of the 20th century, quantum mechanics is today the
most fundamental and abrageant theory in physics. Despite its long history and its
incomparable success, it is far from being exhausted. Quantum information science
has matured to a point to be able to trigger important technological progress, such
that we are today at the verge of a new quantum revolution 2.0, from which we expect
profound scientific, economic, and societal changes.
This chapter traces briefly the historical conditions and the discoveries that led to
the invention of quantum mechanics.
3
4 CHAPTER 1. ANTECEDENTS OF QUANTUM MECHANICS
second important step was initialized by Niels Bohr, who quantized the hydrogen
atom in 1913 into discrete excitation levels.
Table 1.2: Historical sketch of the quantization of light
uum, was the greatest critic of Democritus’ theory. Aristotle said that the reasoning
that guided Democritus to affirm the existence of atoms was as follows. For a body to
change its shape, it is necessary that its parts can move. This presupposes an empti-
ness (or vacuum) in which the matter moves. But if matter were divided infinitely
into ever smaller parts, it would loose its consistency. Nothing could be formed be-
cause nothing could arise from the ever more infinitely deep dilution of matter into
emptiness. Hence, he concluded that the division of matter can not be infinite, that
is, there is an indivisible limit: the atom. ’There is only atoms and emptiness’, he
said.
Observing dust particles in a whirling motion within a ray of sunlight, Democritus
was led to the idea that atoms would behave in the same way, randomly colliding,
some crowding, others dispersing, others never yet joining with another atom.
The consistency of clusters of atoms, which makes something look solid, liquid,
gaseous, or animated (which is the state of the soul) would then be determined by the
shape of the atoms involved and their spatial arrangement. In this sense, water atoms
are smooth and slippery; the atoms of steel have shapes with sharp edges that hold
them solidly together; the atoms of salt, as their taste shows, are harsh and pointed;
the atoms of air are small and little connected, penetrating all other materials; and
the atoms of soul and fire are spherical and very delicate.
Figure 1.3: Atoms of steel and air, atoms of the soul, and Bohr’s atom model.
We know nowadays that Democritus’ first theory of the structure of matter was
very close to the truth: There really are indivisible particles called atoms composed
of a nucleus and an electronic shell, and the space between the atomic nuclei is, in
fact, quite empty.
6 CHAPTER 1. ANTECEDENTS OF QUANTUM MECHANICS
The atomic hypothesis came to be reborn in the modern age with the scientists
Boyle, Clausius, Maxwell, and Boltzmann due to their successful explanations of the
properties of gases based on the so-called kinetic theory, where they assumed a gas
being constituted of identical molecules that collided elastically with each other and
with the walls of the recipient containing them. The discovery of the atom through
the laws of proportions in chemistry and the establishment of Avogadro’s number
considerably strengthened the atomic hypothesis. The hypothesis was definitely con-
secrated with the various experiments that established the charge of the electron and
the mass ratio between electrons and protons.
By the beginning of the 19th century the atomic nature of matter had definitely
been established, and the basic composition of the atoms was already relatively well
known. It was known, through experiments, that electrons could be removed from
neutral atoms thus creating positively charged ions and that only a certain number
of electrons could be removed from each atom. This number proved to be dependent
on the atomic species and was called the atomic number Z. This information was
fundamental for establishing the basic composition of atoms. The question that arose
at this point concerned the dimensions and configurations of the atomic system. How
would loads and masses be distributed in this entity?
Figure 1.4: Comparison of Rutherford scattering by free electrons and electrons strongly
bound to small nuclei. (a) Thomson’s ’raisin-in-a-pudding’ type atom; (b) Rutherford’s
’planetary’ atom. (c) Rutherford scattering by a raisin pudding atom and (d) by a planetary
atom.
For a fixed scattering angle, the amount of particles scattered into a solid angle
element dΩ is proportional to the thickness of the metal foil.
For a given fixed angle and a given metal sheet the amount of scattered particles
2
in dΩ varies inversely with Ekin , where Ekin is the kinetic energy of the α-
particles.
For a given energy and a given metal sheet, the number of particles scattered
into dΩ is proportional to (sin θ2 )−4 .
For a given energy and sheet thickness, the number of particles scattered into
2
dΩ in a given direction is proportional to Ztg , where Ztg is the atomic number
of the element that constitutes the sheet.
The extremely rare deflection of α-particles and their angular distribution can be
understood by the assumption that the positive charge is concentrated in a very small
volume (∼ 1 fm, that is 10000 times less than the size of the atom itself). This volume
is called the atomic nucleus, hence the denomination of nuclear model. Since most of
the particles pass through the gold sheet without hindrance, there must be a large
gap between the nuclei. The electrons, which move within a large (in comparison with
the diameter of the nucleus) empty space (the vacuum) around the nucleus, shield
the positive nuclear charge, so that the atom appears outwardly neutral.
We now derive Rutherford’s scattering formula from the hypothesis of a point-like
nucleus. Due to the repulsive action of the Coulomb force,
Zα Ztg e2
F = , (1.2)
4πε0 r2
8 CHAPTER 1. ANTECEDENTS OF QUANTUM MECHANICS
Figure 1.5: (a) Trajectory of an α-particle. (b) Illustration of the scattering cross section.
we have for the trajectory of the α-particle (Zα = 2) a hyperbola [see Fig. 1.5(a)].
The large half-axis of the hyperbola can be determined from the following ansatz,
Zα Ztg e2 1
Ekin = , (1.3)
4πε0 2a
where 2a is the minimum distance of the particle α, when it collides with the nucleus
in a central collision 1 . The distance a depends on the kinetic energy and can also
be used for non-central collisions. The collision parameter b is the minimum distance
of the α-particle to the nucleus, if it continued to fly in a straight line. In fact the
α-particle will be deflected by an angle θ. From the geometry of the hyperbola, as
2φ + θ = 180◦ , we obtain the following equation:
tan φ = b
a = tan 90◦ − θ
2 = cot θ2 , (1.4)
and therefore
b 8πε0 Ekin
cot θ2 = = b, (1.5)
a Zα Ztg e2
replacing a with the formula (1.3). Taking the derivative of this latter formula, we
obtain a relation between the width db of the hollow cone and the pertinent width dθ
of the deflection angle θ,
1 8πε0 Ekin
− 2 θ dθ = Z Z e2 db . (1.6)
2 sin 2 α tg
N
Let ntg = Vtg be the density of the particles in the target (Ntg atoms per volume
V ) and x the film thickness. Then σ = NAtg = VN/xtg
= ntg1
x is the average cross-section
per atom sensed by the α-particle on its way through the film. The probability P (θ)dθ
for the α-particle of being within a ring at distance b from the nucleus (whose area is
2πbdb) and being scattered into the angle θ is then given by,
2πbdb
P (θ)dθ = = ntg x2πbdb . (1.7)
σ
1 In a central collision, when the α-particle reaches the minimum distance 2a, its initial kinetic
These particles, i.e., dN of the N particles, are deflected into the hollow cone with
the probability,
dN Zα Ztg e2 θ Zα Ztg e2 1
= P (θ)dθ = ntg x2π cot · · dθ (1.8)
N 8πε0 Ekin 2 8πε0 Ekin 2 sin2 θ2
Zα2 Ztg
2 4
e cos θ2
= ntg x 2 2 · dθ ,
64πε0 Ekin sin3 θ2
where we replaced the parameters b and db with the expressions (1.5) and (1.6). The
solid angle of the cone can be expressed by,
dN 2 4
Zα2 Ztg e 1
= ntg x 2 2 · dΩ . (1.10)
N 2
256π ε0 Ekin sin4 θ2
That is Rutherford’s scattering formula. Often, the formula is expressed with the
dσ
differential cross section dΩ . We get,
dN dσ
= = ntg xdσ , (1.11)
N σ
and therefore
2
dσ dN Zα Ztg e2 1
= = . (1.12)
dΩ N ntg xdΩ 4πε0 · 4Ekin sin4 θ
2
The angle θ = 0 is not defined, since there exists a minimum deflection angle
θmin . This angle is reached, when the α-particle moves at the distance b = bmax
from the atom, that is, at the edge of the circular area of the cross section.
For a greater collision parameter b, the α-particle traverses the field of the next
neighboring atom, and the deflection angle increases again. We have:
A Zα Ztg e2
σ= = πb2max and θmin
2 ' tan θmin
2 = , (1.13)
Ntg 8πε0 Ekin · bmax
simply by inverting the formula (1.5). For very large impact parameters, that
is, when the α-particle passes the atom outside its electronic layer, the electrons
of the atom shield the charge of the nucleus, an effect called screening.
For very high energies, the distribution of the nuclear charge over a finite vol-
ume influences the scattering, calling for corrections in the Rutherford formula.
Moreover, at short internuclear distances, nuclear forces appear additionally to
the electromagnetic interaction.
10 CHAPTER 1. ANTECEDENTS OF QUANTUM MECHANICS
10−24
10−26
N
10−28
10−30
0 50 100 150
θ
Figure 1.6: (code) Angular dependence of the cross-section corresponding to Thomson’s
(green) and Rutherford’s (red) models.
p0 = −er . (1.16)
2 This type of model had already been proposed by Jean Perrin in 1901 and by Hantaro Nagaoka
in 1903, around the same time when Thompson developed his model. The planetary model was later
on rescued by John William Nicholson in 1911.
1.1. THE DISCOVERY OF THE ATOM 11
We calculate in the Exc. 1.1.6.4 the power emitted by the acceleration a = ω 2 r of the
electron on its circular trajectory,
µ0 ω 4 p20
P = . (1.17)
12πc
The initial energy of the electron spinning around the nucleus (for a hydrogen atom
Z = 1),
p2 e2 me ω 2 r 2 e2
E= − = − , (1.18)
2me 4πε0 r 2 4πε0 r
is dissipated by radiation of the power (1.17), i.e.,
dE dr e2 dr dr
−P = = me ω 2 r + = 2me ω 2 r . (1.19)
dt dt 4πε0 r2 dt dt
The latter equation supposes an equilibrium between the centrifugal force and the
Coulomb force,
e2
me ω 2 r = , (1.20)
4πε0 r2
allowing to link the revolution frequency ω to the instantaneous radius of the orbit
r(t). Resolving the Eq. (1.19) by ṙ and replacing the power by the relation (1.17) and
the frequency ω by the relation (1.20), we obtain,
dr P µ0 ω 2 e2 e4 1
=− =− r=− . (1.21)
dt 2
2me ω r 24πme c 96π ε20 m2e c3 r2
2
We now imagine this atom in the presence of a magnetic field B ~ oriented in the
direction that we will call z. This results in a precession of the magnetic moment
around the field (similar to the precession of a spinning top in the presence of a
gravitational field) governed by the equation,
dL ~ = e L×B
~ = −ΩL × L ,
~` × B
=µ
dt 2me
e ~
with ΩL = 2m e
B representing the precession frequency and being called Larmor
frequency. It is evident that the presence of the magnetic field considerably alters
the state of the atom, even producing profound modifications in the frequency of the
orbit of the electron ω0 and therefore in the energetic state of the atom. This change
is called Zeeman effect.
The Zeeman effect can be calculated by imagining that the field has an arbitrary
direction with respect to L. In this case, the equation describing the electronic motion
as resulting from an equilibrium between the centrifugal force and the Coulomb force
needs to be complemented by a Lorentz force,
~.
me r̈ + me ω02 r = FL = −ev × B (1.26)
where mr̈ is the centrifugal force due to the circular motion of the electron and me ω02 r
the centripetal force due to the Coulomb attraction exerted by the nucleus. Assuming
~ = Bêz with B = 2me ΩL /e, the equations
the direction of the magnetic field given by B
of motion can be decomposed into,
The z-direction is not influenced. With the ansatz x = aeıωt and y = beıωt we obtain
the system of equations,
which has a non-trivial solution for a and b only when the determinant of the coeffi-
cients of a and b vanishes:
2
ω0 − ω 2 2ıΩL ω
0= = ω 4 − (2ω02 + 4Ω2L )ω 2 + ω04 . (1.29)
−2ıΩL ω ω02 − ω 2
We get,
r q
1 Ω2L
ω = ω1,2 = ω02 + 2Ω2L ± 2ΩL ω02 + Ω2L = ω0 ± ΩL + + ... , (1.30)
2 ω0
or, as ΩL ω, we get ω1,2 = ω0 ∓ ΩL . The result is a splitting of the energy levels
proportional to the magnetic field,
~e
∆E = 2~ΩL = B = 2µB B , (1.31)
me
where the abbreviation µB = e~/2me ' 9.27 · 10−24 JT−1 is called the magneton
Bohr.
Although the classical derivation shows quantitative deviations from experimental
observations, it is quite interesting, as it illustrates several aspects which have a
quantum mechanical equivalence.
F = −~ ~.
µ · ∇B
where m and n are integers. RH = (1/4π)2 (me e4 /4π~3 c) is the Rydberg constant
and µ = me mat /(me + mat ) the reduced mass.
The discrete nature of spectral lines and the problem of the radiation collapse led
Niels Bohr to formulate the following postulates:
1. There are specific stationary orbits, where electrons do not emit energy.
3. The laws of mechanics can describe the dynamic equilibrium of electrons in sta-
tionary states, but fails to describe the transition of electrons between stationary
orbits.
Thus, Bohr’s model predicts the quantization of energy levels, known as first quanti-
zation of quantum mechanics. The radii of the possible orbits can be calculated from
the postulate that the orbital angular momentum be quantized in units of ~, that is,
the electrons form stationary de Broglie waves along the orbits 3 . We discuss Bohr’s
model in Excs. 1.1.6.6 and 1.1.6.7.
In the picture proposed by Bohr, the radiative decay happens as an abrupt tran-
sition of an electron between an outer (more energetic) orbit and an inner (less ener-
getic) orbit. Since the energies of stationary orbits are very well defined, the emitted
radiation is mono-energetic, i.e., the spectrum consists of discrete characteristic lines.
We note here that the picture of an abrupt transition of electrons between discrete
states, called the quantum jump, did not receive Schrödinger’s blessing. He rather
imagined for electrons, within his theory of quantum wave mechanics, wave-shaped
orbitals instead of planetary trajectories, thus avoiding the problem of radiation due
to charge deceleration and the quantum jump concept. According to him, during a
transition between electronic orbits, the energy is transformed into radiation gradu-
ally 4 .
1.1.6 Exercises
1.1.6.1 Ex: Analysis of Rutherford scattering
a. What conclusions can be drawn from the observation that Rutherford’s formula
describes well the scattering of charged particles traveling through matter over a wide
range of parameters?
b. Why do we see a deviation from Rutherford’s formula for large energies?
c. The scattering of protons with energy E crossing a thin film of thorium is well
described up to energies of E = 4.3 MeV by Rutherford’s formula. Estimate for this
case the range of nuclear forces.
d. For small scattering angles θ we observe large deviations from Rutherford’s formula.
3 A generalization of Bohr’s theory was provided by Arnold Johannes Wilhelm Sommerfeld. As-
suming elliptical orbits for the electrons he managed to explain some features of the fine structure,
provided the motion of the electron was treated relativistically. The basic premises were 1. stable
orbits Rwhen the Coulomb attraction is balanced by the centrifugal force, 2. quantization of phase
space rq dq = nq ~, and 3. quantization of angular momentum Ldθ = nθ ~.
R
4 We note here, that quantum jumps were observed much later!
1.1. THE DISCOVERY OF THE ATOM 15
Explain why?
e. Assume the thorium atoms of item (c) to form a periodic crystal with the lattice
constant d = 10aB . At which minimum angle θ Rutherford’s formula loses its validity.
B0 = 1c E0 . (1.35)
The energy of the electromagnetic field of a standing wave, averaged over one oscilla-
tion of the frequency ω is,
Z
1 ε0 ~ 2 1 ~2
Ū = |E| + |B| dV . (1.37)
2 2 2µ0
Now, the energy density of the oscillating electromagnetic field is given by,
dŪ 1 ~ 2 + 1 |B|
~2
ū = = ε0 |E| . (1.38)
dV 4 µ0
From the equation (1.35) we can see that the contributions of the electric and magnetic
fields are equal. Therefore,
Z
~ 2 dV
Ū = 12 ε0 |E| and ~2.
ū = 12 ε0 |E| (1.39)
Again using the equation (1.35), we find the value averaged over a time period,
I¯ = 12 ε0 c|E|
~2 . (1.41)
18 CHAPTER 1. ANTECEDENTS OF QUANTUM MECHANICS
This quantity, called intensity, describes the fact that the flux is a density of energy
multiplied with the velocity of propagation in vacuum,
~ 2 = I¯ .
ūc = 21 ε0 c|E| (1.42)
1 ∂ 2 E~
∇2 E~ = and ∇ · E~ = 0 . (1.45)
c2 ∂t2
The stationary waves solutions separate into terms oscillating in time and in space.
Figure 1.8: (a) Cavity in position space showing the thermal motion of the electrons inside
the walls. (b and c) Density-of-states in a cavity in momentum space.
However, the periodic boundary conditions for |k| limit the components kx , ky , kz
to positive values (n ≥ 0), that is, the volume under consideration is limited to an
octant. On the other hand, we must multiply the number of states by two because of
the degeneracy of polarizations,
Z n 3 Z k
L 4L3 k 3 4L3 ω 3
4N = 4πn2 dn = 4πk 2 dk = 2 = . (1.50)
0 π 0 π 3 3π 2 c3
With this, we obtain the mode density,
N ω3
= . (1.51)
L3 3π 2 c3
The spectral density of modes % can be given in several units,
Z Z Z
N
%(n)dn = %(k)dk = %(ω)dω = 3 , (1.52)
L
such that,
πn2 k2 ω2
%(n) = or %(k) = or %(ω) = . (1.53)
L3 π2 π 2 c3
The density of oscillating modes within the cavity grows like the square of the
frequency. Now, the mean energy per mode in a sample of oscillators in thermal
equilibrium is, following the equipartition law, equal to,
Ē = kB T , (1.54)
7 See script on Electrodynamics (2020).
20 CHAPTER 1. ANTECEDENTS OF QUANTUM MECHANICS
where kB is the Boltzmann constant. We conclude that the spectral energy density
uRJ (ω) in the cavity is,
ω2
uRJ (ω)dω = kB T %(ω)dω = kB T dω . (1.55)
π 2 c3
T = 3000 K
10000 T = 5000 K
T = 6000 K
u (eV)
5000
0
0 500 1000
ν (THz)
Figure 1.9: (code) Spectral energy density following Rayleigh-Jeans’ and Planck’s laws.
e−En /kB T
pn = . (1.56)
q
Now Planck hypothesized that the energy be quantized, that is, it must be assigned
in discrete portions, proportional to the frequency, such that,
En = n~ω , (1.57)
ω2 ~ω
u(ω)dω = Ē%(ω)dω = dω . (1.61)
π 2 c3 e~ω/kB T − 1
This result, drawn in Fig. 1.9, is much more satisfactory, because now the energy
density has an upper bound, and coincides with the results of experiments. Note, that
for high temperatures or low excitation energies, ~ω kB T , Planck’s distribution
converges to that of Rayleigh-Jeans’, u(ω) → uRJ (ω).
Note that the form of the expression for the energy depends on the parametrization
and must be derived respecting u(ω)dω = u(λ)dλ, etc.. Often the blackbody radiation
is expressed in terms of the spectral radiance,
c
L(ω) ≡ u(ω) , (1.62)
4π
which can be understood as the (isotropic) energy flux into all directions of space.
Solve the Excs. 1.2.10.1 to 1.2.10.7.
E = hν . (1.63)
Assuming a fixed exit work A for the extraction of an electron, we can measure the
constant ~:
2 eV
hν = A + mv2 = A + eV → h= . (1.64)
ν − νg
The energy of the fastest electrons is measured through the decelerating voltage by
varying ν and I. We will discuss the photoelectric effect quantitatively later in the
Exc. 6.4.5.7.
the atomic mass, such that the spectral displacement becomes much smaller. This
limit, which involves bound electrons, is that of Thomson and Rayleigh scattering.
The relevance of this effect lies in the fact that it shows that light exhibits prop-
erties commonly attributed to corpuscles, since Thomson’s scattering model, based
on the classical theory of charged particles accelerated by electromagnetic fields, can
not explain any spectral shift.
dN1 dN2
=− = −R1→2 + R2→1 + S2→1 (1.65)
dt dt
= −N1 B12 u(ω) + N2 B21 u(ω) + N2 A21 .
R1→2 is the absorption rate, R2→1 the stimulated emission rate and S2→1 the sponta-
neous emission rate. The assumption of a third type of transition, called spontaneous
emission, is necessary, if B12 = B21 but N1 > N2 in thermal equilibrium. In thermal
equilibrium we have the condition of stationarity, dN dN2
dt = − dt = 0 for a given energy
1
24 CHAPTER 1. ANTECEDENTS OF QUANTUM MECHANICS
A21
uth (ω) = . (1.66)
N1
N2 B12 − B21
The Boltzmann distribution law controlling the distribution of the number of atoms
in the lower and upper states is given by,
N1 g1
= e−(E1 −E2 )/kB T , (1.67)
N2 g2
where g1,2 are the degeneracies of the lower and upper states and E2 − E1 = ~ω0 . We
find,
A21
uth (ω) = g1 ~ω /k T . (1.68)
g2 e 0 B B
12 − B21
But this result must be consistent with the Planck’s distribution (1.61). Therefore,
by comparing this equation with the equation (1.68), it must be,
g1 B12
=1. (1.69)
g2 B21
and also,
A21 ~ω 3
= 2 03 . (1.70)
B21 π c
This equation shows that, once we know one of the three transition rates, we can
always calculate the others.
It is useful to compare the rate A21 with B21 from the equation (1.68) inserting
the equation (1.69),
A21
= e~ω0 /kB T − 1 . (1.71)
B21 uth (ω)
This expression shows that, when ~(ω2 − ω1 ) kB T , that is, for optical, UV, or X-
ray frequencies, spontaneous emission dominates. But in low-frequency regimes, that
is, IR, microwave, or radio waves, stimulated emission is more important. Note that
even when stimulated emission dominates, spontaneous emission is always present and
plays an important role, for example, in processes ultimately limiting the emission
bandwidth of lasers.
1
g2 2 Γ 4Γ g2 Γ
σ(ω) = λ = λ2 LΓ (ω − ω0 ) . (1.73)
g1 2π (ω − ω0 )2 + 14 Γ2 g1 4
Example 2 (Limiting cases): Let us analyze the two limiting cases when
either one of the spectral distributions I(ω) or σ(ω) is much narrower than the
other.
For a narrow laser, we may assume a δ-peaked spectral intensity distribution,
¯
I(ω) = Iδ(ω − ωL ) . (1.75)
¯
P = Iσ(ωL) . (1.76)
8 At low saturation.
26 CHAPTER 1. ANTECEDENTS OF QUANTUM MECHANICS
g2 λ2 ΓI¯
P (∆) = (LΓ ? Lβ )(∆) . (1.81)
g1 4
Figure 1.11: Absorption spectrum (blue) and spectral energy distribution of the source (red).
This result reproduces the two limiting cases discussed in the above example, since
for narrow transitions, Γ → 0, that is LΓ → δ, we recover the results (1.79), and for
narrow lasers, β → 0, that is, Lβ → δ, we recover (1.76). Obviously, this formula holds
for other line profiles e.g. when the resonance is broadened by some perturbations 9 .
ω2 ~ω
I(ω) = 2 2
(1.82)
π c e ~ω/kBT − 1
Z 4 Z ∞ 3
~ kB T x dx π2
with I¯ = I(ω)dω = 2 2 = (kB T )4 .
π c ~ 0
x
e −1 15c2 ~3
9 Let us here remind the following identities holding for Lorentzian and Gaussian line profiles:
Since the width of the transition is negligibly small in comparison with the blackbody
spectrum, Γ → 0, we may evaluate the scattered power as,
Z
g2 λ2 Γ ω2 ~ω
P = LΓ (ω − ω0 ) 2 2 ~ω/k T dω (1.83)
g1 4 π c e B −1
2
Γ→0 g2 λ Γ ω0
2
~ω0
−→ 2 2 /k
.
g1 4 π c e ~ω0 BT − 1
Using the result (1.69), we can write the amount of power removed from the system
by spontaneous emission as,
g1
P = N2 A21 ~ω0 = u(ω0 )B12 (N1 − g2 N2 )~ω0 . (1.85)
On the other hand, the power absorbed from the radiation field u(ω0 ), by atoms
whose transition is described by the cross section (1.73), is given by,
Z
P = (N1 − gg21 N2 )σ(ω)I(ω)dω . (1.86)
Remembering I(ω) = cu(ω) and assuming a large radiation spectrum, I(ω) ' I(ω0 ),
a comparison of equations (1.85) and (1.86) yields,
Z
c
B12 = σ(ω)dω . (1.87)
~ω0
Here, z is the total distance, over which absorption takes place. The last equation
is the Lambert-Beer law for light absorption. It is very useful for measuring atomic
densities in gas cells or of atomic beams [482, 456, 736]. Solve the Excs. 1.2.10.9 to
1.2.10.11.
1.2.8 Saturation
Strong driving of a transition leads to its saturation and causes line broadening. To
see this we go back to Einstein’s rate equations in steady-state (1.65) additionally
simplifying by assuming g1 = 1 = g2 , such that B12 = B21 . Resolving these equations
by N1 and N2 and using N1 + N2 = N and N1 − N2 ≡ ∆N , we get,
B21 u(ω) + A21 B12 u(ω)
N1 = N and N2 = N . (1.92)
2B21 u(ω) + A21 2B21 u(ω) + A21
u→0 u→0
For vanishing pump rate we expect, N1 −→ N and N2 −→ 0. In contrast, when
u→∞
the pump rate becomes much larger than the relaxation rates, N1 , N2 −→ 12 . This
means that the absorption coefficient α = σ(N1 − N2 ) goes to zero, and the medium
becomes completely transparent,
A21 N
∆N = N = . (1.93)
2B21 u(ω) + A21 1 + s(ω)
The saturation parameter
2B12 u(ω)
s(ω) ≡ (1.94)
A21
represents the ratio of pumping rate to the relaxation rate. The pump rate due to a
monochromatic wave with intensity I¯ is obtained by comparing (1.85) with (1.89),
σ(ω)I¯
B12 u(ω) = . (1.95)
~ω
We obtain for the saturation parameter,
2σ(ω)I¯
s(ω) = . (1.96)
~ωA21
According to (1.85) and (1.93) the power absorbed per unit volume on the transition
by atoms with the populations N1,2 in a radiation field with a broad spectral profile
and spectral energy density u(ω) is,
N
P = ~ωB12 u(ω)∆N = ~ωB12 u(ω) . (1.97)
1 + s(ω)
1.2. THE DISCOVERY OF THE PHOTON 29
(Γ/2)2
s(ω) = s(ω0 ) . (1.99)
∆2 + (Γ/2)2
Substituting this into (1.98) yields the frequency dependence of the absorbed radiation
power per unit frequency interval dω,
A21 N s(ω0 )(Γ/2)2 ¯ (Γ/2)2
P = ~ω = N Iσ(ω0 ) , (1.100)
2 (ω − ω0 )2 + (Γ/2)2 [1 + s(ω0 )] ∆2 + (γs /2)2
where we introduced the increased halfwidth of the Lorentzian profile,
p
γs ≡ Γ 1 + s(ω0 ) . (1.101)
Apparently, the halfwidth of the saturation-broadened line increases with the resonant
saturation parameter s(ω0 ). If according to (1.94) the induced transition rate at
resonance equals the total relaxation rate A21 /2, the resonant √
saturation parameter
becomes s(ω0 ) = 1, which increases the linewidth by a factor 2, compared to the
unsaturated linewidth Γ for weak radiation fields. Starting from (1.100) we can define
a saturated absorption cross section,
g1 2π 2 c~
Isat = Γ . (1.105)
g2 3λ30
10 Some authors define the saturation for s = 2, as happens when Ω = Γ.
30 CHAPTER 1. ANTECEDENTS OF QUANTUM MECHANICS
1
s=0
s=1
s = 10
σs (ω)
0.5
0
-2 -1 0 1 2
Δ/Γ
I¯
Ω2 = σ(ω0 ) Γ . (1.107)
~ω0
gives the specific heat. The disappearance of the specific heat at low temperatures,
3N (~ω)2 −~ω/(kB T )
CV ' e , (1.111)
kB T 2
which is related to the finite localization energy of harmonic oscillators, does not
describe experimental observations very well, and the model had to be refined by
Debye, later on.
where v is the velocity of sound propagation, the formula is totally equivalent to the
density-of-states for photons in a cavity. Assuming thatR there is an upper bound νm
ν
for the vibrational frequencies, we normalize as 3N0 = 0 m ρ(ν)dν. The energy now
is 11 ,
Z νm Z θ/T
~ω 4πV T4 x3 dx
E= νdν = 9N kB . (1.113)
0 e~ω/kB T − 1 v3 θ3 0 ex − 1
The Debye temperature θ = hνm /kB is characteristic for the metal. The derivative is
then, " Z #
3 θ/T
T x3 dx θ 1
CV = 9N kB 4 − . (1.114)
θ 0 ex − 1 T eθ/T − 1
12π 4 3
CV = N kB (T /θ) . (1.115)
5
1.2.10 Exercises
1.2.10.1 Ex: Resistance of vacuum
p
Show that µ0 /ε0 has the dimension of a resistance and the value of 376.7 Ω.
a. The light power emitted by a laser (λ = 633 nm) be P = 1 nW. How many photons
does the laser emit per second? How many laser photons are in a mode volume of
L = 10 cm length?
b. How many photons on average are inside an optical cavity having the same mode
volume at ambient temperature, when there is no incident light?
a. How many modes do fit into a cubical box of 10 cm size for a frequency interval of
1000 Hz centered at a wavelength of 500 nm?
b. How many photons are in the box supposing it has a temperature of T = 300 K,
respectively, T = 6000 K?
a. Calculate the total number of photons per area per unit time emitted by a black-
body at temperature T .
b. The linewidth of a helium-neon laser is ∆ν = 1000 Hz. The operating wavelength
is λ = 632.8 nm, the power is P = 1 mW, and the beam size w0 = 1 mm. How many
photons are emitted per second?
c. What would be the temperature of a blackbody radiator emitting the same number
of photons from an equal area and over the same frequency interval as the laser?
1.3. FURTHER READING 33
Foundations and
mathematical tools
In this chapter, we will introduce step by step the formalism of quantum mechanics
by gradually increasing the degree of abstraction. Applications of the formalism will
be shown in consecutive chapters (watch talk).
∂2
A − c2 ∇2 A = 0 , (2.2)
∂t2
reproduces the dispersion relation.
On the other hand, we have slow massive particles possessing kinetic energy,
p2
E= . (2.3)
2m
With the hypothesis of de Broglie that even a massive particle has wave quality, we
can try an ansatz 1 of a wave equation satisfying the dispersion relation (2.3). From
1 Trial, working hypothesis.
35
36 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
Planck’s formula, E = ~ω, and the formula R of Louis de Broglie, p = ~k, describing
the particle by a wave packet ψ(r, t) = eı(k·r−ωt) ϕ(k)d3 k not subject to external
forces, it is easy to verify that the equation,
∂ ~2 2
ı~ ψ = − ∇ ψ, (2.4)
∂t 2m
reproduces the dispersion relation. If the particle is subject to a potential, its total
energy is E = p2 /2m + V (r, t). This dispersion relation corresponds to the famous
Schrödinger equation,
∂ ~2 2
ı~ ψ = − ∇ + V (r, t) ψ . (2.5)
∂t 2m
∂
E −→ ı~ and p −→ −ı~∇ . (2.11)
∂t
We will discuss this later in the context of Ehrenfest’s theorem in Secs. 2.1.6, 2.1.7,
and 2.4.6.
ψ(r, t)
ψ̃(r, t) ≡ qR , (2.14)
R3
|ψ(r, t)|2 d3 r
Starting from the Schrödinger equation we can easily derive the continuity equation
(see Exc. 2.1.8.1),
ρ̇(r, t) + ∇ · j(r, t) = 0 , (2.16)
or in the integral form,
Z Z I
d
− ρd3 r = ∇ · jd3 r = j · dS , (2.17)
dt V V ∂V
R
using Gauß’ law. With I ≡ S j · dS, the probability
R current which flows through the
surface S delimiting the probability charge Q ≡ V ρ(r, t)d3 r, we obtain,
−Q̇ = I . (2.18)
The continuity equation is obviously similar to that of electromagnetism.
with h ≡ 2π~. At time t = 0, this expansion is nothing more than a Fourier transform,
Z
1
ψ(r, 0) = h3/2 d3 pϕ(p)eır·k , (2.20)
In the absence of forces the momentum distribution becomes stationary. We can now
use the momentum distribution ϕ(p) as coefficients of the expansion of the tempo-
ral wavefunction ψ(r, t), as shown above. Thus, the expansion represents a general
solution of the time-dependent Schrödinger equation. The magnitude |ϕ(p)|2 is the
probability density in momentum space.
Example 3 (Normalization of the wave function in momentum space): It
is easy to show that the probability density in momentum space is also normal-
ized:
Z Z Z Z
0
|ϕ(p)|2 d3 p = h13 d3 p d3 rψ ∗ (r)eır·k d3 r0 ψ(r0 )e−ır ·k
Z Z Z
0
= d3 r d3 r0 ψ ∗ (r)ψ(r0 ) (2π)
1
3 d3 keık·(r−r )
Z Z Z
= d3 r d3 r0 ψ ∗ (r)ψ(r0 )δ 3 (r − r0 ) = |ψ(r)|2 d3 r = 1 ,
2.1. BASIC NOTIONS 39
knowing that the Fourier transform of a plane wave is nothing more than the
Dirac distribution.
Since the probability distributions |ψ(r)|2 and |ϕ(p)|2 are interconnected by Fourier
transform, we already know that we can not localize 2 both simultaneously. If one is
well localized, the other is necessarily delocalized. Do the Exc. 2.1.8.2.
2.1.6 Eigenvalues
We have already seen that the position and momentum distributions of a particle are
spread. We calculate the mean values of these distributions, denoted by hri and hpi,
as first moments of the respective distributions:
Z Z
hri = d3 r|ψ(r, t)|2 r and hpi = d3 p|ϕ(p, t)|2 p . (2.22)
This calculation shows that the expectation value, called eigenvalue, of the momentum
can be expressed through an operator p̂ ≡ (~/ı)∇ acting on the wavefunction 3,4 .
More generally, we can compute the eigenvalue of a function in r and p via,
Z
hf (r, p)i = d3 rψ ∗ (r)f (r, p̂)ψ(r) . (2.23)
However, it is important to note that the operators r̂ and p̂ do not necessarily com-
mute.
~ d ~ ~ d ~ d
p̂x xψ = xψ = ψ + x ψ 6= x ψ = xp̂x ψ .
ı dx ı ı dx ı dx
(2.28)
In summary, we found a law,
d2
hF̂i = hmr̂i , (2.29)
dt2
much like the Newton law, but instead of applying to localized particles, the law
applies to the eigenvalues of probability distributions. Similar laws can be derived for
angular momentum and energy conservation.
The observation made by Paul Ehrenfest, that in quantum mechanics the mean
values follow the same laws of classical mechanics, is called Ehrenfest’s theorem.
2.1.8 Exercises
2.1.8.1 Ex: Conservation of probability
Demonstrate the conservation of local probability through the definitions of proba-
bility densities, ρ(r, t), and probability current j(r, t).
2.2. POSTULATES OF QUANTUM MECHANICS 41
But the notation has other advantages. For example, let us suppose that we know
the three possible states of a system, |1i, |2i, and |3i, which are linearly independent.
Then we can define the states as vectors:
1 0 0
|1i = 0 , |2i = 1 ,
|3i = 0 . (2.35)
0 0 1
These three states can be interpreted as the basis of a vector space representing the
system. Now, each wavefunction can be expanded on this basis and expressed by a
vector. An arbitrary ket state of this system will then be,
c1
|ψi = c2 . (2.36)
c3
Now we can easily calculate the probability for a system to be in a state |ψi,
c1
||ψi|2 = hψ|ψi = c∗1 c∗2 ∗
c3 · c2 = |c1 |2 + |c2 |2 + |c3 |2 . (2.38)
c3
the Hilbert space of wavefunctions, |ψi 7→ p̂|ψi, where p̂ would be the operator of the
linear momentum. To better distinguish the observables, we decorate their symbols
with a hat. We will see more ahead (see Sec. 2.3.5) that every quantum system is
completely described by a complete set of observables.
To find the current values aψ of any observable  in a specific situation given by
a wave function ψ, we need to solve an equation of eigenvalues,
We can rewrite the equation as aψ = hψ|Â|ψi. The values an are real numbers, if the
observable is a Hermitian operator, that is,
The eigenvalue of a projector, hP̂k i = hψ|P̂k |ψi = |hk|ψi|2 , is nothing more than the
probability of finding a system, whose general state is |ψi, in the particular state,
since expanding as done in (2.33), we have,
X
hP̂k i = c∗m cn hm|kihk|ni = |ck |2 . (2.44)
m,n
Using the matrix formalism we can define other interesting operators and verify
their properties,
1 0 0 0
|1ih1| = , |2ih2| = ,
0 0 0 1
. (2.45)
0 1 0 0
|1ih2| = = σ̂ − , |2ih1| = = σ̂ +
0 0 1 0
are also called Pauli spin matrices, because they were introduced by Wolfgang Pauli.
The vector,
+
σ̂x σ̂ + σ̂ −
ˆ ≡ σ̂y = ı(σ̂ + − σ̂ − )
~σ (2.47)
σ̂z [σ̂ − , σ̂ + ]
is called Bloch vector 6,7 . The eigenvalue of the Bloch vector has a fixed length (see
Exc. 2.2.9.2). An important property of the Pauli matrices is their behavior under
commutation and anti-commutation,
relation for massive particles. Heisenberg invented a mechanics (detailed in later sections), which he
called mechanics of matrices. Later, he showed the formal equivalence of both theories.
2.2. POSTULATES OF QUANTUM MECHANICS 45
∂
ı~ |ψi = Ĥ|ψi . (2.51)
∂t
A closed system, disconnected from the rest of the world (we will now call the
rest of the world reservoir) is not subject to dissipation, i.e., it does not lose energy
to the reservoir. Such a system is always described by a hermitian Hamiltonian.
Unfortunately, this system also does not allow information leakage, that is, we can
not measure the system. This is reflected in the fact that the Schrödinger equation
does not allow to describe the process of a quantum measurement. This is because
before the measurement, the system can be in several states or even in a superposition
of states, while after the measurement we know exactly the state. This amounts to a
reduction of entropy, which is not allowed in a closed system.
The famous postulate of state reduction or projection of the wavefunction for-
mulated by John von Neumann describes the quantum measurement process as a
sequence of two distinct steps 9 . In a first step, the measuring apparatus projects
the measured operator  on an eigenvector basis. That is, if the measurement is
compatible with the operator 10 , we obtain a distribution of probability amplitudes
of the results,
X X X
 y hÂi = hψ|Â|ψi = hψ|Â| ck |ki = ak ck hψ|ki = ak |ck |2 , (2.52)
k k k
P 2 2
with hψ|ψi = k |ak | = 1. Therefore, we can understand |hk|ψi| as the probability
of the system to be in the eigenstate |ki. In other words, this first step of the measure-
ment Pprocess removes in an irreversible manner all coherences from the observable,
 y k |kihk|Â|kihk|.
8 Considering the Weyl order.
9 For simplicity, we only consider pure state, here.
10 To understand the meaning of compatible, we must establish a more complete theory of mea-
In a second step, the observing scientist will read the measuring device and note
the result, which will necessarily be one of the possible ak ,
hÂi y ak . (2.53)
If the state is stationary, it will never change any more. That is, each subsequent
measurement will yield the same result. The Exc. 2.2.9.4 illustrates the process of
quantum measurement at the example of a measurement of the excitation energy of
a two-level atom.
is nothing more than an eigenvalue equation. This means that the Schrödinger wave
mechanics is equivalent to the mechanics of the Heisenberg matrices. The Excs. 2.2.9.5
and 2.2.9.6 are first simple calculations of the eigenvalues and eigenvectors of a two-
level system.
2.2.9 Exercises
2.2.9.1 Ex: Reality of eigenvalues
Show that the eigenvalues of an observable are real.
Using the stationary Schrödinger equation, calculate the eigenvalues and eigenvectors.
a. If the system is initially in state |1i, will it remain in that state at a later time?
How about if the initial state is |2i?
b. Obtain the eigenvalues EI and EII and the respective eigenvectors |Ii and |IIi of
48 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
d. Based on the above result, we can predict at least one possible electromagnetic
radiation emission frequency for an ammonia sample. What is this frequency?
EUF 2014, 1 Semestre, Mecânica Quântica - Q1 1
/ )
)
)
)
)
/
)
|1 > >
|2
which can be 6= 0. The sum of two Hermitian operators is Hermitian, but the product
is not, since,
On the other hand, the following relations of Hermitian operators are always Hermi-
tian,
ÂB̂ + B̂ Â and ı(ÂB̂ − B̂ Â) . (2.64)
We define the scalar product as,
hψ|ϕi . (2.65)
Two states are called orthogonal, if hψ|ϕi = 0. The norm is written as,
the functions are called linearly independent. A set of linearly independent functions
may form a basis. The space opened by a set of linearly independent functions is
called Hilbert space.
An operator  is completely characterized by its eigenvalues and eigenfunctions.
If a set of eigenfunctions |ni is complete, every allowed state of the system can be
expanded in these eigenfunctions,
X
|ψi = cn |ni and Â|ni = an |ni . (2.69)
n
that is, we calculate the eigenvalues and eigenvectors. When all eigenvalues are differ-
ent, an 6= am , we know that the corresponding eigenvectors are orthogonal, hn|mi = 0,
Â|ni = an |ni , Â|mi = an |mi , ∀{n, m} an 6= am (2.70)
=⇒ ∀{n, m} hn|mi = δm,n .
Exc. 2.3.9.1 asks for demonstrating this.
Frequently, for example, in the case of a particle confined to a potential, there
exist discrete eigenvalues (for E < 0) simultaneously with continuous eigenvalues (for
E > 0). Assuming hm|m0 i = δm,m0 , hm|ki = 0 and hk|k0 i = δ (3) (k − k0 ), with a
complete base, Z
X
|mihm| + d3 k|kihk| = 1̂ , (2.71)
m
an arbitrary vector can be expanded on an orthogonal basis,
X Z
|ψi = |mihm|ψi + d3 k |kihk|ψi . (2.72)
m
2.3.3 Degeneracy
The eigenvectors form a natural basis for the Hilbert space. However, a problem arises
in the case of degeneracy, that is, when some eigenvalues are equal, an = am . In this
case, the eigenvectors that correspond to degenerate eigenvalues are not completely
defined, and we have to construct a basis verifying that all constructed eigenvectors are
orthogonal. For this, there exists the method of orthogonalization by Schmidt, which
works like this: We assume that we have already solved the eigenvalue equation, that
we found a degenerate eigenvalue, Â|ak i = a|ak i for every k = 1, .., gk , where gk is
the degree of degeneracy, and that we also found a complete basis of eigenvalues |am i,
but which is not orthogonal, that is, ∃{m, n} with han |am i 6= 0. The task is to build
another basis |bm i satisfying hbn |bm i = δn,m .
The first vector of the orthogonal base can be chosen freely, e.g.,
|b1 i ≡ |a1 i . (2.75)
Since the basis {|ak i} is assumed to be complete, the second vector is necessarily a
linear combination of vectors |ak i, that is, |b2 i = |a2 i + λ|b1 i. With the condition
hb1 |b2 i = 0 = hb1 |a2 i + λhb1 |b1 i we can determine the parameter λ, and obtain for the
second vector,
hb1 |a2 i
|b2 i ≡ |a2 i − |b1 i . (2.76)
hb1 |b1 i
2.3. ABSTRACT FORMALISM OF QUANTUM MECHANICS 51
In the same way, we can derive for a third vector, |b3 i = |a3 i + µ|b1 i + ν|b2 i, the
conditions, hb1 |b3 i = 0 = hb1 |a3 i + µhb1 |b1 i and hb2 |b3 i = 0 = hb2 |a3 i + νhb2 |b2 i, and
obtain,
hb1 |a3 i hb2 |a3 i
|b3 i ≡ |a3 i − |b1 i − |b2 i . (2.77)
hb1 |b1 i hb2 |b2 i
An overall way of writing this down is,
|b1 ihb1 | |b2 ihb2 | |bk−1 ihbk−1 |
|bk i ≡ 1 − − − ... − |ak i . (2.78)
hb1 |b1 i hb2 |b2 i hbk−1 |bk−1 i
In the Exc. 2.3.9.2 we practice the orthogonalization of a set of three linearly inde-
pendent but non-orthogonal vectors, and in the Exc. 2.3.9.3 we find an orthogonal
basis for a partially degenerate three-level system.
Therefore,
U † U = 1̂ =⇒ U † U U −1 = 1̂U −1 =⇒ U † = U −1 (2.82)
U † U = 1̂ =⇒ U U † U U −1 = U 1̂U −1 =⇒ U U † = 1̂ .
Also,
Â|ni = Ê|ni and ÂU = U Ê . (2.83)
That is, by knowing the unitary matrix (or transformation matrix) U , we can solve
the eigenvalue problem simply by Ê = U −1 ÂU .
Note, that this does not apply to a non-orthonormal basis. In this case, we need
to do a Schmidt orthogonalization and use the condition det U = 1. We apply the
technique detailed in this section to solve Excs. 2.3.9.4, 2.3.9.5, and 2.3.9.6.
52 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
The fact that commuting operators have a common system of eigenvectors autho-
rizing sharp eigenvalues can be used to construct and characterize a state.
Example 5 (Measuring momenta in orthogonal directions): For example,
the obvious solutions of the eigenvalue equations,
~ d ~ d
p̂x |ψpx i = |ψpx i = px |ψpx i and p̂y |ψpy i = |ψpy i = py |ψpy i
ı dx ı dy
are the plane waves eıpx x/~ and eıpy y/~ . Therefore, the total state of the particle
can be described by,
However, these eigenfunctions are infinitely degenerate, since the linear momen-
tum in z-direction is not specified. A third operator p̂z |ψi = pz |ψi commutes
with the others,
[p̂k , p̂m ] = 0 .
Hence,
|ψpx ,py ,pz i = e(ı/~)(px x+py y+pz z) ,
2.3. ABSTRACT FORMALISM OF QUANTUM MECHANICS 53
Also, the number needed for a complete set of commuting operators (CSCO) de-
pends on the number of degrees of freedom and the symmetry of the system. In the
case of the free particle in one dimension it is enough to consider one observable only,
for example, x̂ or p̂. In three dimensions, we already need at least three commuting
observables. In Exc. 2.3.9.8 we will try to find a CSCO for a matrix with partially
degenerate eigenvalues.
This is Heisenberg’s famous uncertainty principle. For example, [p̂, x̂] = −ı~,
and hence, ∆p∆x ≥ ~/2. We will see later (see Sec. 4.3.1), that [ˆlx , ˆly ] = ı~ˆlz such
that ∆lx ∆ly ≥ ~|hlz i|/2. More difficult to show, since time has no simple quantum
operator, is ∆E∆t ≥ ~/2. In the Exc. 2.3.9.9 we will show the Schwartz inequality,
and in the Exc. 2.3.9.10 we ask for a formal derivation of Heisenberg’s uncertainty
principle.
2.3.7 Representations
2.3.7.1 Spatial representation
A Hilbert space can be discrete or, as in the case of the momentum of a free particle,
continuous. In this latter case, the eigenvalues are continuously distributed, since the
equation,
−ı~∇r ψ(r) = pψ(r) , (2.89)
has solutions for each value of E. The eigenfunctions are ψ(r) = aeıp·r/~ . Eq. (2.89)
clearly has the form of an eigenvalue equation, for which we have already introduced
the Heisenberg matrix formalism. The question now is how these descriptions com-
bine.
54 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
The quantities hr|ψ(t)i = ψ(t, r) Schrödinger wave functions. We can also say that the
wavefunctions are the coordinates of the state in the particular base |ri. Consequently,
hr|r̂|r0 i = rδ 3 (r − r0 ) (2.93)
0 3 0
hr|f (r̂)|r i = f (r)δ (r − r ) .
where the quantity A(r, r0 ) ≡ hr|Â|r0 i is called kernel of the operator. The transition
from Heisenberg’s abstract mechanics to Schrödinger’s wave mechanics is done by the
substitutions |ψ(t)i → ψ(t, r) and  → A(r, r0 ).
where hp|ψ(t)i = ϕ(t, p). The formulas are analogous to the ones in the spatial
representation. In particular, in the momentum representation the position operator
is r = ı~∇p .
The representations follow from one another by Fourier transformation. Since
−ı~∇r hr|pi = phr|pi, we know,
1
hr|pi = ~3/2
exp( ~ı r · p) , (2.98)
where the prefactor ~−3/2 is introduced to take account of the unit of the states 11 .
ψ and ϕ are different representations of the same quantum state related by,
Z Z
3
hr|ψ(t)i = 1
hr|pihp|ψ(t)id p = h3/2 eır·p/~ ϕ(p, t)d3 p = ψ(r, t) (2.99)
R3 R3
Z Z
hp|ψ(t)i = hp|rihr|ψ(t)id3 r = h3/2
1
e−ır·p/~ ψ(r, t)d3 r = ϕ(p, t) .
R3 R3
defining the function ϕ̃(k) ≡ ~3/2 ϕ(p). Applying the Fourier transform to functions
of operator we can calculate,
Z Z
hr|G(p̂)|r0 i = d3 phr|G(p̂)|pihp|r0 i = d3 pG(p)hr|pihp|r0 i (2.102)
Z
0
1
= ~3/2 d3 pG(p)eık·(r−r ) = ~13 (FG)(r − r0 ) .
In Exc. 2.3.9.11 we will show hr|p̂|ψi = (~/ı)∇r hr|ψi, thus justifying that we
can understand an operator as a rule to determine what happens to a function. For
example, the rule p̂x asks for a derivation of the wavefunction by x.
Introducing the parenthesis [...] to extract the unit of a physical quantity, we find, [|ri] = [ψ(r)] =
[r−3/2 ] and [|pi] = [ϕ(p)] = [p−3/2 ]. We do not assign a unit to the abstract state |ψi, that is,
[|ψi] = 1.
56 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
spaces, which allowed us to treat the dynamics of the degrees of freedom separately.
In this chapter, we will establish the theoretical foundations allowing us to analyze
systems, where degrees of freedom can not be separated because they are entangled
or interact. In particular, we will consider the system of two spins and the coupling
of angular momenta in general.
The projector,
1 0 0
P̂ = 0 1 0
0 0 0
reduces the Hamiltonian to a two-level transition,
ω1 Ω12 0
ĤR = Ω12 ω2 0 .
0 0 0
Obviously, the concatenation (2.107) only serves to increase the Hilbert space of
a given degree of freedom described by a given observable, e.g., when we add one
more level of energy to the spectrum of an atom described by a Hamiltonian. If, in
contrast, we want to add another degree of freedom, we need the external sum or
external product discussed below.
The symbol ⊗ denotes the outer tensorial product of two vectors (states) 12 . Now, in
order to represent the multidimensional space HA ⊗ HB by a matrix, we use the fact
12 The tensorial product of two states should not be confused with other definitions of products,
Figure 2.3: Illustration of the isomorphism between HA ⊗ HB and HI ⊗ HA⊗B for matrices
of rank 2.
For observables we proceed in the same way: The external product of two com-
mutators spans a Hilbert product space with the dimension corresponding to product
of the dimensions of the sub-spaces. Assuming that,
X X
 ≡ |iiAij hj| and B̂ ≡ |kiBkl hl| . (2.111)
i,j k,l
then X
 ⊗ B̂ ≡ |ikiAij Bkl hjl| , (2.112)
(ik)(jl)
such that
dim  ⊗ B̂ = dim  dim B̂ .
the outer product,
α1 β1 α1 β2
|αihβ| ≡
α2 β1 α2 β2
†
A11 B11 A11 B11 A11 B12 A11 B21 A11 B22
A12 B12 A12 B11 A12 B12 A12 B21 A12 B22
A21 B21
=
,
A21 B11 A21 B12 A21 B21 A21 B22
A22 B22 A22 B11 A22 B12 A22 B21 A22 B22
the exterior (cross) product,
α2 β3 − α3 β2
A11 B11 A11 B12 A12 B11 A12 B12
A11 B21 A11 B22 A12 B21 A12 B22
 ⊗ B̂ =
A21 B11
, (2.113)
A21 B12 A22 B11 A22 B12
A21 B21 A21 B22 A22 B21 A22 B22
A11 A12 B11 B12
A11 A12 B21 B22
 ⊗ B̂ = ( ⊗ I)(I ⊗ B̂) =
.
A21 A22 B11 B12
A21 A22 B21 B22
(2.114)
The concept (2.112) can be generalized to more degrees of freedom like,
X
 ⊗ B̂ ⊗ Ĉ ≡ |ikmiAij Bkl Cmn hjln| , (2.115)
(ikm)(jln)
Obviously, the external product is associative (Â ⊗ B̂) ⊗ C = Â ⊗ (B̂ ⊗ C), but
does not commute, even though the operators acting on different spaces do commute,
[Â, B̂] = 0. Nevertheless, we can reverse the order of the product of two operators
using,
1
0 1
 ⊗ B̂ = S(B̂ ⊗ Â)S with S≡
.
(2.116)
1 0
1
We note, that it is important to distinguish from what space the vector came
from. In our notation, the vector before the symbol of the tensorial product (⊗) is
belongs to the space HA , and the one after the ⊗ belongs to the space HB . With the
definition (2.111) we can verify that the operators only act on their respective states:
defined by,
(Â ⊗ B̂)(α ⊗ β) = Â(α) ⊗ B̂(β) .
In this way, the tensor product becomes a bifunctor from the category of vector
spaces to itself, covariant in both arguments. If  and B̂ are both injective,
2.3. ABSTRACT FORMALISM OF QUANTUM MECHANICS 61
0
surjective or (in the case that Hα , Hα , Hβ , and Hβ0 are normed vector spaces
or topological vector spaces) continuous, then  ⊗B̂ is injective, surjective or
continuous, respectively.
By choosing bases of all vector spaces involved, the linear maps  and B̂ can be
represented by matrices. Then, depending on how the tensor α⊗β is vectorized,
the matrix describing the tensor product  ⊗B̂ is the Kronecker product of
0
the two matrices. For example, if Hα , Hα , Hβ , and Hβ0 above are all two-
dimensional and bases have been fixed for all of them, and  and B̂ are given
by the matrices,
! !
A11 A12 B11 B12
 = , B̂ =
A21 A22 B21 B22
The resultant rank is at most 4, and thus the resultant dimension is 4. Note that
rank here denotes the tensor rank i.e. the number of requisite indices (while the
matrix rank counts the number of degrees of freedom in the resulting array).
Note Tr  ⊗ B̂ = Tr Â × Tr B̂.
A dyadic product is the special case of the tensor product between two vectors
of the same dimension.
that is,
A11 + B11 A11 + B12 A12 + B11 A12 + B12
A11 + B21 A11 + B22 A12 + B21 A12 + B22
 ⊕ B̂ =
A21 + B11
. (2.120)
A21 + B12 A22 + B11 A22 + B12
A21 + B21 A21 + B22 A22 + B21 A22 + B22
Again, using the definition (2.116) of the unitary operator S, we can reverse the order
of the operator by,
 ⊕ B̂ = S(B̂ ⊕ Â)S . (2.122)
62 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
..
~ω
2
~ω
2
+ ~Ω 0 0 .
..
! ~ω + ~Ω ~ω
+ ~∆ 0 0 .
2 2
0 ~Ω ..
Ĥ = ~ω(n+ 21 )⊕ = .
~Ω ~∆ 0 0 3~ω
2
3~ω
2
+ ~Ω .
..
0 0 3~ω
+ ~Ω 3~ω
+ ~∆ .
2 2
..
··· ··· ··· ··· .
It acts on the product state |ni|ii, where the first ket denotes the vibrational
level and the second ket the electronic excitation of the atom.
2.3.8.4 Trace
The trace of an operator over a subspace reduces its domain to the remaining dimen-
sions (the · -symbol is a place holder for the dimension over which we do NOT want
to trace):
X X
TrB Â ⊗ B̂ = h· m|ikiAij Bkl hjl| · mi = |iiAij Bkl hj|δkm δlm
(ik)(jl)(·m) (ik)(jl)(·m)
(2.123)
X X
= |iiAij Bmm hj| = Â Bmm = Â TrB B̂ .
(i)(j)(m) m
For example, Trρ Â ⊗ ρ̂ = Â. See the Excs. 2.3.9.15 and 2.3.9.16.
It can be shown,
Tr ÂB̂ = Tr B̂ Â . (2.124)
2.3.9 Exercises
2.3.9.1 Ex: Orthogonality
Show that two eigenvectors of a Hermitian operator associated with two different
eigenvalues are orthogonal.
That is, exchanging |ψi by U |ψi and at the same time  by U ÂU † , we obtain quan-
tities describing the same physical reality, since the eigenvalues are unchanged. This
allows us to choose the best mathematical representation for a specific problem. As
an example, we will apply the temporal unitary transformation to solve the dynamics
of a coupled two-level system in Exc. 2.4.7.1.
d ∂ ÂS 0 ∂ Â 0 ∂ Â
S S
ÂS = + p̂˙S + r̂˙S . (2.127)
dt ∂t ∂pS ∂rS
This is,
d ∂
ÂS (t) = ÂS (t) . (2.128)
dt ∂t
In this case, the formal solution of the Schrödinger equation,
d
ı~ |ψS (t)i = ĤS |ψS (t)i , (2.129)
dt
can be written,
|ψS (t)i = e−(ı/~)ĤS t |ψS (0)i ≡ U (t)|ψS (0)i . (2.130)
Apparently, the temporal dynamics is completely within the wave functions.
66 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
|ψS (t)i −→ U (t)† |ψS (t)i ≡ |ψH i ÂS (t) −→ U (t)† ÂS (t)U (t) ≡ ÂH (t)
and
(2.132)
with the transformation defined by equation (2.130), is equivalent. The subscript H
means the Heisenberg picture. In particular, we obviously have,
Thus, the matrix element of the operator ÂS in Schrödinger’s picture with the time-
dependent base {|ψS i} is equal to the matrix element of the operator ÂH = U † ÂS U
in Heisenberg’s picture with the time-independent base {|ψH i}. In this picture the
wavefunctions are independent of time,
d d
|ψH i = |ψS (0)i = 0 , (2.134)
dt dt
d d dU † dU ∂ ÂS (t)
ÂH (t) = U (t)† ÂS (t)U (t) = ÂS (t)U (t) + U (t)† ÂS (t) + U (t)† U (t)
dt dt dt dt ∂t
ı −ı ∂ ÂS (t)
= Ĥ † U (t)† ÂS U (t) + U (t)† ÂS ĤU (t) + U † (t) U (t) . (2.135)
~ ~ ∂t
That is,
d ı ∂ ÂH (t)
ÂH (t) = [Ĥ, ÂH (t)] + . (2.136)
dt ~ ∂t
This so-called Heisenberg equation, which describes the temporal evolution of an op-
erator acting on time-independent states in the Heisenberg picture, is equivalent to
the Schrödinger equation, which expresses the temporal evolution of a quantum state
in Schrödinger’s picture.
According to equation (2.136), the rate of temporal variation of an operator in
the Heisenberg representation is given by the commutator of that operator with the
2.4. TIME EVOLUTIONS 67
In the Exc. 2.4.7.3 we will use the Heisenberg picture to derive the equations of
motion for a particle confined to a potential.
Analogously to Eq. (2.130), we define a time evolution operator in terms of the time-
independent part of the total Hamiltonian,
|ψI (t)i = eıĤ0 t/~ |ψS (t)i and ÂI (t) = eıĤ0 t/~ ÂS e−ıĤ0 t/~ . (2.139)
Now we are interested in the temporal dependence of quantum states and operators
in the interaction picture. Replacing the inverse function |ψS (t)i = e−ıĤ0 t/~ |ψI (t)i
in the Schrödinger equation (2.129) we immediately see,
∂
V̂ (t)|ψI (t)i = ı~ |ψI (t)i . (2.140)
∂t
for all m. Indeed, this situation can always be arranged by absorbing the diag-
onal elements of V̂ into Ĥ0 , thus modifying its eigenvalues to,
0
Em = Em + hm|V̂ |mi . (2.143)
Ĥ 0 = Ĥ0 + V̂ + [ıS, Ĥ0 ] + [ıS, V̂ ] + 12 [ıS, [ıS, Ĥ0 ]] + 21 [ıS, [ıS, V̂ ]] + . . . . (2.146)
which confirms the validity of the Schrödinger equation (2.140) in the interaction
picture, provided the Hamiltonian is taken to be the perturbation part V̂ (t),
only. In the Heisenberg picture V̂ (t) = 0, such that,
Ĥ 0 = 0 . (2.156)
70 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
δF (r̂)
[p̂, F (r̂)] = −ı~ . (2.158)
δr̂
δ Ĥ δ Ĥ
[r̂, Ĥ] = ı~ and [p̂, Ĥ] = −ı~ , (2.159)
δ p̂ δr̂
δ Ĥ δ Ĥ
r̂˙ = and p̂˙ = − . (2.160)
δ p̂ δr̂
We will demonstrate this in Exc. 2.4.7.7 for the case of a harmonic potential.
In the Schrödinger picture the equation of motion for the eigenvalues of the ob-
servables takes the form,
d ∂ ı
hÂS i = h∂t ψ|ÂS |ψi + hψ|∂t ÂS |ψi + hψ|ÂS |∂t ψi = hÂS i + h[Ĥ, ÂS ]i . (2.161)
dt ∂t ~
The eigenvalues behave as Heisenberg observables in Eq. (2.135), that is, they follow
the laws of Hamilton’s and Newton’s mechanics.
The important result now is that the equations that govern the eigenvalues of the
observables are identical in the both pictures, since from the Heisenberg picture we
obtain with Eq. (2.135),
d ∂ ı
hÂH i = hÂH i + h[Ĥ, ÂH ]i .
dt ∂t ~
2.4.7 Exercises
2.4. TIME EVOLUTIONS 71
where ∆ = ω−ω0 is the detuning between the frequency of the light and the frequency
of the transition and Ω the Rabi frequency. Help: Determine the matrix of the
eigenvalues Ê and the unitary transformation U given by U † ĤU = Ê and use the
†
formal solution of the Schrödinger equation: |ψ(t)i = e−ıĤt/~ |ψ0 i = e−ıU ÊU t/~ |ψ0 i =
U † e−ıÊt/~ U |ψ0 i 13 .
echo in NMR) of Bragg diffraction pulses leading to a splitting and recombination of an atomic
wavefunction in momentum space. Every pulse transfers one unit of photonic recoil to the atomic
center-of-mass.
72 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
The temporal evolution has already been discussed extensively in Sec. 2.4.
Ttr r ≡ a + r . (2.164)
Before this, we need to derive the following calculation rule for commutators, which
will be done in Exc. 2.5.5.1:
Applying this formula to the two operators p̂ and r̂ related by the commutation
rule (2.50), we obtain,
0
e(ı/~)a·p̂ r̂e(−ı/~)a·p̂ = r̂ + [(ı/~)a · p̂, r̂] + 1
2! [(ı/~)a · p̂, a] + ... = r̂ + a . (2.166)
and forms a group since Utr (a)Utr (b) = Utr (a + b). Summarizing the impact of the
translation on the operators of space,
† †
Utr (a)r̂Utr (a) = r̂ + a , Utr (a)p̂Utr (a) = p̂ , (2.169)
Hence,
Utr (a)|ri = e(−ı/~)a·p̂ |ri = |r + ai . (2.171)
This relation is easier to remember as ψ 0 (r + a) = ψ(r), which can be read as: The
value of the new wavefunction at the new point equals the value of the old wavefunction
at the old point.
we obtain
~
∇|ri .
p̂|ri = (2.175)
ı
Finally, we note that the momentum operator can be defined via the translation
operator,
p̂ = ı~ ∇a Utr (a)|a=0 . (2.176)
74 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
α) = e(−ı/~)L·~α .
Urt (~ (2.183)
The explicit form of L follows from its action on a state |ψi projected into position
space. In analogy with the derivation of the result (2.175), comparing the expansion
of the operator (2.183),
ı
α)|ψi = he(ı/~)~α·L r|ψi = 1 + α
hr|Urt (~ ~ · L + .. hr|ψi , (2.184)
~
with the Taylor expansion of the wavefunction,
heα~ × r|ψi = ψ(r + α
~ × r + ..) = [1 + (~
α × r) · ∇r + ..] ψ(r) , (2.185)
we find,
ı ı
~ · L = (~
α α × r) · ∇r = α
~ · (r × ∇r = α~ · (r × p̂) , (2.186)
~ ~
that is,
L̂ = r̂ × p̂ . (2.187)
Therefore, the observable L̂ is the orbital angular momentum of the particle producing
the rotations.
2.5. SYMMETRIES IN QUANTUM MECHANICS 75
Tv r = r + vt and Tv p = p + mv . (2.188)
This is the prescription for the so-called kick operator, which will play a prominent
role in the discussion of photonic recoil (see 3.6.2).
By defining
Ĝ = p̂t − r̂m = ı~∇v UG (v)|v=0 , (2.191)
the unitary transformation
UG (v) = e(−ı/~)v·Ĝ (2.192)
[G · a, G · b] = 0 , (2.195)
for any vectors a and b, and with that, using the Baker-Haussdorff formula (3.108),
we verify,
2
UG (v1 )UG (v2 ) = e(−ı/~)(v1 +v2 )·G−[v1 ·G,v2 ·G]/2~ = UG (v1 + v2 ) . (2.196)
Obviously, for very high velocities, the Galilei-boost should be replaced by the
Lorentz transform (or Lorentz boost) [374]. Here, we only note, that the additivity
of velocities expressed by equation (2.195) does not hold for non-collinear relativistic
velocities 16 .
15 We need to keep in mind that, following Glauber’s rule, e eB̂ = eÂ+B̂+[Â,B̂]/2 with  ≡ v · p̂t
and B̂ ≡ v · r̂m. However, the commutator [v · p̂t, v · r̂m] = −ı~mtv2 is independent of p̂ and r̂,
such that it only contributes an unimportant phase.
16 See script on Electrodynamics (2020).
76 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
~ t) = −∇Φ − ∂t A
E(r, and ~ t) = ∇ × A .
B(r, (2.197)
Also, we know that the fields are invariant under the substitution,
Φ → Φ0 ≡ Φ − ∂t χ and A → A0 ≡ A + ∇χ , (2.198)
obviously must keep the Schrödinger equation invariant. Transforming operators and
wave functions as,
−1
Ĥ → Ugg ĤUgg ≡ ĤU and |ψi → Ugg |ψi ≡ |ψU i , (2.200)
in accordance with the transformation rule (2.153) for time-dependent unitary trans-
formations. For the momentum, we get analogously,
−1
p̂U |ψU i = Ugg (−ı~∇)Ugg −1
|ψU i = Ugg (−ı~)Ugg (∇|ψU i) + Ugg (−ı~) −ıq −1
~ Ugg ∇χ |ψU i
= (−ı~) ∇ − ıq~ (∇χ) |ψU i , (2.202)
18
This corresponds to the substitutions ,
d −1 d dχ −1
Ugg ı~ U = ı~ + q and Ugg p̂Ugg = p̂ − q∇χ . (2.203)
dt gg dt dt
This shows that the gauge transformation applies to the minimal coupling rule,
Ugg Ugg
Ĥ = Ĥkin +qΦ y Ĥkin +qΦ+q∂t χ and mv = p−qA y p−qA−q∇χ , (2.204)
confirming the rules (2.198). That is, the Hamiltonian of a particle carrying the
charge q and interacting with an electric potential Φ and a magnetic vector potential
A is,
1
Ĥ = 2m (p − qA − q∇χ)2 + qΦ + q∂t χ . (2.205)
17 See script on Electrodynamics (2020).
18 In quadrivetorial notation ı~∂µ −→ ı~∂µ + q∂µ χ.
2.5. SYMMETRIES IN QUANTUM MECHANICS 77
Therefore, to find a conservation law, i.e., an invariable observable (also called con-
stant of motion), we must verify that the observable and the transformed wavefunc-
tions simultaneously satisfy the same fundamental equations (that is, Schrödinger’s
or Heisenberg’s equation) as the original observable and wavefunctions. For example,
if the wavefunction |ψi satisfies the Schrödinger equation, the wave function U |ψi
must do this too,
! d dU d dU
ĤU |ψi = ı~ U |ψi = ı~ |ψi + ı~U |ψi = ı~ |ψi + U Ĥ|ψi . (2.207)
dt dt dt dt
Consequently, we obtain the relation,
As shown in (2.161) and (2.162), an operator that commutes with the Hamiltonian
does not explicitly depend on time, that is, it is conserved.
d (ı/~)Êτ
Since dt e = 0, this means [e(ı/~)Êτ , Ĥ] = 0, which implies conservation of
energy [Ê, Ĥ] = 0. This will be verified in the Exc. 2.5.5.3.
19
This is equivalent to momentum conservation [p̂, Ĥ] = 0 .
third law, body A attracts body B, more than the body B attracts the body A. In that case after a
while the two bodies have different momenta. With the unitary transformation Utr (a) = e−ıp̂·a/~ '
1 − ıp · a/~ + ... we have,
defined by the mirroring of the wavefunction through a point in space, for example
r = 0,
P̂ |ψ(r)i ≡ |ψ(−r)i . (2.212)
with
P̂ 2 = P̂ . (2.213)
We talk about even parity when P̂ |ψ(r)i = |ψ(r)i and odd parity when P̂ |ψ(r)i =
−|ψ(r)i. See Exc. 2.5.5.4.
L(xµ ) = 1
4µ0 F
µν
Fµν − Aµ j µ = ε20 E~2 − 2µ1 0 B
~2 − Aµ j µ (2.215)
ε0 2 1 2
= 2 [∇Φ + ∂t A] − 2µ0 [∇ × A] − Φρ +A·j ,
With this result, we can calculate the difference between the actions under gauge
transformation and recall, that they can not be different:
Z Z
! 0
0 = S − S = [(∂t χ)ρ + ∇χ · j]dV dt = − χ[∂t ρ + ∇ · j]dV dt , (2.218)
80 CHAPTER 2. FOUNDATIONS AND MATHEMATICAL TOOLS
using partial integration 20 and choosing volumes so large, that every charge is inside.
This is the continuity equation derived from the gauge invariance of the action. The
calculation really is nothing more than an application of Noether’s theorem from
which we could have derived directly the continuity equation, ∂µ j µ = 0.
In summary, the conservation of charge means invariance with respect to gauge
transformations,
Ucl (χ) ≡ e−ıqχ(r,t)/~ , (2.219)
where χ is the gauge field. We note that q and χ are conjugated observables. There-
fore, if [χ, Ĥ] = 0, then the charge q is a conserved quantity.
Transformations can be combined. For example, we believe that nowadays all laws
are invariant with respect to CPT transformation, that is, a combination of charge
conjugation, parity inversion, and θ-transform.
2.5.5 Exercises
2.5.5.1 Ex: Calculus with commutator
Derive the rule (2.165) via a Taylor expansion of the operator Ĝ(τ ) ≡ eτ Â B̂e−τ Â .
Ph.W. Courteille (2020), Script on Quantum mechanics applied to atomic and molec-
ular physics [http]
J.J. Sakurai, J.J. Napolitano, 2nd ed. Springer (2011), Modern Quantum Mechanics
[ISBN]
H.A. Bethe, R. Jackiw, 3rd ed. Taylor & Francis (1997), Intermediate Quantum
Mechanics [ISBN]
~2 d2
Ĥ = − . (3.1)
2m dx2
Therefore, the general solution of the Schrödinger stationary equation,
is, q
ψ(x) = Aeıkx + Be−ıkx with k= 2mE
~2 . (3.3)
R∞
Note that the eıkx functions are not quadratically integrable, since −∞ |eıkx |2 dx =
R∞
−∞
dx → ∞. On the other side, they do not represent actual physical systems. In
practice, we need to consider wave packets or specify a finite volume for the particle.
Note also that the spectrum of eigenvalues is continuous. Do the Exc. 3.1.4.1.
This means that the wavefunction can not be infinite inside a finite volume. But it
can be infinite within an infinitely small volume. Also, since the Schrödinger equation
contains the second derivative by position, the wavefunction must be continuous and
have a continuous derivative.
83
84 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
This is the case, for example, for a rectangular well with Vx (x) = Vy (y) = Vz (z) =
V0 /3 inside the well and V (x, y, z) = 0 outside. It also holds for a harmonic potential,
m 2 2
V (r) = ωx x + ωy2 y 2 + ωz2 z 2 . (3.6)
2
In these cases, the following ansatz for the wavefunction is generally useful,
~2 ψx00 (x)
− + Vx (x) = const. ≡ Ex , (3.9)
2m ψx (x)
and the same for y and z. Since, E = Ex + Ey + Ez may have the same value
for different combinations of Ex , Ey and Ez , multidimensional systems are often
degenerate.
3.1.4 Exercises
3.1.4.1 Ex: Trapped particle
Consider the problem of a particle of mass m forced to move in a single direction and
completely confined to a box, with walls placed at the positions x = 0 and x = a.
a. The particle be in the ground state, what is its energy and its wavefunction?
b. Suppose the particle has the following wavefunction:
π π
1 2π
ψI (x) = √ 2 cos (6x − a) − 3ı sin x + cos (2x − a) ,
7a 2a a 2a
2 2
what is the probability that a measurement of the energy yields the result E = 2π ma2 ?
~
c. Considering again the ground state of item (a), what is the probability distribution
for the momentum of the particle in this state?
d. Still starting from the ground state, suppose we remove (instantaneously) the walls,
3.2. RECTANGULAR POTENTIALS 85
leaving the particle free (Ĥ = p̂2 /2m). What is the energy of this free particle?
Formulae:
Z L nπx nπL[1 − (−1)n eıBL ]
eıBx sin dx = for n = 1, 2, 3, ...
0 L n2 π 2 − B 2 L2
Z ∞
x2 π2
2 2
cos2 πx
2 dx =
−∞ (1 − x ) 4
As the potential barriers are high, the walls are hard, that is, the particle, even being
a quantum particle, can not penetrate. The wavefunction and the possible energy
values are,
r
2 nπx n 2 ~2 π 2
ψ(x) = sin and En = . (3.11)
L L 2mL2
The Exc. 3.2.5.1 asks to demonstrate the result (3.11) illustrated in Fig. 3.1.
Obviously the spectrum of eigenvalues is now discrete. They can be enumerated by
an integer n called quantum number. Note that the energy levels are not equidistant.
2 2
~ π
Example 17 (Localization energy ): There is a minimal energy E1 = 2mL 2
which is called zero point energy or localization energy. This energy can be
understood as a consequence of Heisenberg’s uncertainty principle. We can make
the following gross estimation of the zero point energy. Obviously, the particle
is localized with an uncertainty lower than ∆x < L. Hence, ∆p > ~/∆x > ~/L.
The average kinetic energy is,
The fact that the numerical value is different from the value calculated by the
formula (3.11) comes from the particular geometry of the box potential.
86 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
40 (a) 40 (b)
0
35 35
-5
30 30
(E − V0 )/E1 , ψ(x)
(E − V0 )/E1 , ψ(x)
-10
E/E1 , ψ(x)
25 25
-15
20 20
-20
15 15
-25
10 10
-30
5 5
-35
0 0
0 0.5 1 0 0.5 1 0 0.5 1
x/L x/L x/L
Figure 3.1: (code) (a) Wavefunctions and energies in the box potential. (b) The rectangular
potential well with the reference energy set to the height of the well.
For every transition between two sections a = 1 and a = 2 we require the boundary
conditions,
ψ1 (x) = ψ2 (x) and ψ10 (x) = ψ20 (x) . (3.14)
R∞ 2
Together with the normalization, 1 = −∞ |ψ| dx, these conditions are sufficient to
determine the wavefunction unambiguously.
Figure 3.3: (a) Scheme of a two-sided and (b) one-sided potential well.
For confined particles, E < 0, the problem is totally symmetric. In addition, the
wavefunction must disappear for x → ±∞. Therefore, we can simplify,
A1 = 0 = B 3 and A3 = B 1 . (3.17)
88 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
Since the amplitudes are real, the imaginary part of the quotient (3.19) should dis-
appear, which is the case when,
-0.2
(V0 )
-0.4
E, En
-0.6
-0.8
-1
-10 0 10
2κ1 k2
tan k2 L ,
−κ21 + κ22
Figure 3.4: (code) Graphical solution for a finite bilateral potential well. The red dotted
curves represent the tangents (left side of the equation (3.21)), the solid green curves the
hyperbolas (right side of the equation), the circles in cyan are the eigenenergies. When
0 < E − V0 E, they converge to the eigenenergies of the infinitely deep well (black crosses
and vertical black line).
then,
~2 ~2 π 2 2
E − V0 = 2 2m = n . (3.22)
k2 2mL2
Apply the notions obtained in this section to solve Excs. 3.2.5.3 and 3.2.5.4.
3.2.5 Exercises
3.2.5.1 Ex: Particle in a box
Obtain the wavefunctions and associated energy levels of a particle confined in a box,
where V (x) = 0 for 0 ≤ x ≤ l and V (x) = ∞ outside.
~ d
hψ|p̂|ψi = hψ| |ψi = ~k . (3.23)
ı dx
Therefore, this particle propagates towards +∞. On the contrary, the particle Be−ıkx
propagates towards −∞. Thus, the two solutions (3.13) of the Schrödinger equation
(3.12) correspond to propagating particle waves. From here on we will use the letter
A (B) to denote the amplitudes of waves propagating in direction ∞ (−∞).
In locations where the potential changes abruptly, the particle can be partially
reflected.
90 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
with the scattering matrix T for a particle with energy E (see Fig. 3.2),
1 + kk12 eı(k1 −k2 )L 1 − kk21 eı(−k1 −k2 )L
T = 12 (3.26)
1 − kk21 eı(k1 +k2 )L 1 + kk12 eı(−k1 +k2 )L
−ık L !
1 e
2
0 1 + kk12 1 − kk12 eık1 L 0
=2 .
0 eık2 L 1 − kk21 1 + kk12 0 e−ık1 L
If there are more zones with different depths, we may concatenate the scattering
matrices. Denoting by Tm→n the scattering matrix describing a transition at position
Lm,n of a potential of the depth Vm to another potential Vn , we write,
Multiplying the first line with T22 and the second with −T12 and adding them,
This equation resolved by A2 along with the second equation (3.29) resolved by B1
give,
A2 B2 T12 /T22 T11 − T12 T21 /T22 B2
=S = . (3.31)
B1 A1 1/T22 −T21 /T22 A1
3.3. POTENTIAL BARRIER 91
The matrix S describes the causality of scattering process more adequately: The
amplitude A2 in region (2) results from the superposition of a wave B2 being reflected
by the barrier and a wave A1 being transmitted by the barrier. The amplitude B1 in
region (1) results from the superposition of a wave A1 being reflected by the barrier
and a wave B2 being transmitted by the barrier. Therefore, the matrix S is more
appropriate for the description of the quantum reflection, as we will discuss in the
next section. However, it has the disadvantage that it can not be concatenated in the
same way as the T matrices.
Unlike the T matrix the S matrix is unitary, since
T11
det S = S11 S22 − S12 S21 = − = −e2ık1 L . (3.32)
T22
Also, it is possible to show,
∗ ∗
S11 S21 S11 S12 1 0
S †S = ∗ ∗ = . (3.33)
S12 S22 S21 S22 0 1
we find that one part of the wave is reflected into the region (1), another is transmitted
into the region (2),
!
(1+k1 /k2 )2 −(1−k1 /k2 )2
A2 0 T11 − T12 T21 /T22 2(1+k1 /k2 )
=S = = (3.35)
B1 1 −T21 /T22 − 1−k 1 /k2
1+k1 /k2
1 2k1
= .
k1 + k2 k1 − k2
We use B2 = 0, since no wave comes from the side of region (2), and A1 = 1,
because it simplifies the formulas and does not affect the generality of the results.
The interesting results are:
Even when E2 < V2p , the particle enters the classically prohibited region: ψ2 (x) ∝
e−κ2 x with κ2 = ~1 2m(V2 − E2 ), i.e. the transmission is non-zero, |A2 | > 0.
Even with E2 > V2 , the particle has a probability of being reflected at the step,
|B1 | > 0.
In Exc. 3.3.7.1 we calculate the behavior of a Broglie wave passing through a po-
tential step and entering a classically forbidden region. In Exc. 3.3.7.2 we investigate
a model describing the collision between attracting or repelling particles via a partial
reflection at a potential step.
1 = |B1 |2 + k2
k1 |A2 |
2
=R+T , (3.39)
problem in the same way as in the previous section. For example, we can calculate
the transmission and reflection probabilities. The formula is derived in Exc. 3.3.7.3,
−1
16E/V0 (1 − E/V0 )
R=1−T = 1+ , (3.41)
(eκL − e−κL )2
3 (b) 3
2.5 2.5
2 2
E/V0
E/V0
1.5 1.5
1 1 β = 10
0.5 0.5
β=3
0 0
0 0.5 1 0 0.5 1
R T
Figure 3.5: (code) (a) Tunnel effect and quantum transmission and reflection at a
potential barrier. (b) Coefficients of transmission and reflection (horizontal) through
the shown potential barrier as a function of the energy normalized to the height
√ of the
barrier E/V0 . The dashed red curve corresponds to a low barrier, β ≡ ~1 L 2mV0 = 3,
the blue solid curve corresponds to a deep barrier β = 10.
~2 d2 ψ(x)
− + αδ(x)ψ(x) = Eψ(x) , (3.42)
2m dx2
94 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
d2 ψ 2mE
=− 2 ψ . (3.43)
dx2 ~
where, in the case of positive energies (real k), eıkx represents a wave traveling to
the right, and e−ıkx one traveling to the left. One obtains a relation between the
coefficients by imposing that the wavefunction be continuous at the origin,
~2 0
− [ψ (0) − ψ10 (0)] + αψ(0) = 0 . (3.47)
2m 2
~2
− ık(A1 − B1 − A2 + B2 ) + α(A1 + B1 ) = 0 . (3.48)
2m
The boundary conditions thus give the following restrictions on the coefficients,
A1 + B1 − A2 − B2 = 0
2mα . (3.49)
A1 − B1 − A2 + B2 = (A1 + B1 )
ık~2
3.3. POTENTIAL BARRIER 95
Figure 3.6: (a) The δ-potential (green) and the bound state wavefunction (blue). (b) Double
δ-potential (green).
The bound state wavefunction solution to the δ-function potential is continuous ev-
erywhere, but its derivative is not defined at x = 0.
In any one-dimensional attractive potential there will be a bound state.
p To find its
energy, note that for negative energies, E < 0, the wavenumber k = ı 2m|E|/~ = ıκ
is imaginary and the wavefunctions are exponentially increasing or decreasing func-
tions of x (see above). Requiring that the wavefunctions do not diverge at infinity
eliminates half of the terms: A2 = B1 = 0. The wavefunction is then an evanescent
wave,
(
ψ1 (x) = A1 eκx for x < 0
ψ(x) = −κx
. (3.50)
ψ2 (x) = B2 e for x > 0
~2 κ2 mα2
Eb = − =− 2 . (3.52)
2m 2~
The δ-potential well and its wavefunction are exhibited in Fig. 3.6(a).
3.3.6.2 Scattering
For positive energies, the wavefunctions are oscillating functions of x. That is, the
particle is free to move in either half-space: x < 0 or x > 0, but it may be scattered at
the δ-potential. The quantum case can be studied in the following situation: a particle
incident on the barrier from the left side (A1 ) may be reflected (B1 ) or transmitted
(A2 ). To find the amplitudes for reflection and transmission for incidence from the
left, we set in the equations (3.49) A1 = 1 (incoming particle), B1 = r (reflection),
B2 = 0 (no incoming particle from the right), and A2 = t (transmission), and solve
96 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
for r and t,
1 1
t= , r= 2 . (3.53)
mα ı~ k
1− 2 −1
ı~ k mα
Due to the mirror symmetry of the model, the amplitudes for incidence from the right
are the same as those from the left. The result is that there is a non-zero probability,
1 1
R = |r|2 = = . (3.54)
~4 k 2 2~2 E
1+ 2 2 1+
m α mα2
for the particle to be reflected. This does not depend on the sign of α, that is, a
barrier has the same probability of reflecting the particle as a well. This is a significant
difference from classical mechanics, where the reflection probability would be 1 for
the barrier (the particle simply bounces back), and 0 for the well (the particle passes
through the well undisturbed). The probability for transmission is,
1 1
T = |t|2 = 1 − R = 2 2
= . (3.55)
m α mα2
1+ 4 2 1+ 2
~ k 2~ E
10
E/Eb
0
0 0.5 1
R, T
Figure 3.7: (code) Transmission (red) and reflection (blue) probability of a δ-potential well.
The energy E > 0 is in units of Eb = mα2 /2~2 .
~2 d2 ψ
− (x) + V (x)ψ(x) = Eψ(x) ,
2m dx2
where the potential is now:
R R
V (x) = −qλ δ x + 2
+δ x− 2
where 0 < R < ∞ is the ’internuclear’ distance with δ-function (negative) peaks
located at x = ±R/2 (shown in brown in the diagram). Keeping in mind the
relationship of this model with its three-dimensional molecular counterpart, we
use atomic units and set ~ = m = 1. Here 0 < λ < 1 is a formally adjustable
parameter. From the single well case, we can infer the ’ansatz’ for the solution
to be:
ψ(x) = Ae−d|x+ 2 | + Be−d|x− 2 | .
R R
which has two solutions d = d± . For the case of equal charges (symmetric
homonuclear case), λ = 1 and the pseudo-quadratic reduces to:
d± = q[1 ± e−d± R ] .
The ’+’ case corresponds to a wave function symmetric about the midpoint
(shown in red in the diagram) where A = B and is called gerade. Corre-
spondingly, the ’-’ case is the wavefunction that is anti-symmetric about the
midpoint where A = −B is called ungerade (shown in green in the diagram).
They represent an approximation of the two lowest discrete energy states of the
three-dimensional H+ 2 and are useful in its analysis. Analytical solutions for the
energy eigenvalues for the case of symmetric charges are given by:
d± = q + W (±qRe−qR )/R ,
where W is the standard Lambert function. Note that the lowest energy corre-
sponds to the symmetric solution d+ . In the case of unequal charges, and for
that matter the three-dimensional molecular problem, the solutions are given
by a generalization of the Lambert function (see section on generalization of
Lambert function and references herein).
One of the most interesting cases is when qR ≤ 1, which results in d− = 0.
Thus, one has a non-trivial bound state solution with E = 0. For these specific
parameters, there are many interesting properties that occur, one of which is
the unusual effect that the transmission coefficient is unity at zero energy.
98 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
3.3.7 Exercises
3.3.7.1 Ex: Tunneling
A rubidium-87 atom moves in free space (region 0) with velocity v = 1 cm/s (see
diagram). Suddenly it encounters a gap with depth V1 = −kB · 1µK.
a. What is the particle’s Broglie wavelength in region 1?
b. Now the atom encounters a barrier of height V2 = −V1 . What is the probability
that the particle will enter region 2?
c. What is the probability of finding the particle inside region 2 up to a depth of
x2 = 10 nm?
~2 00
− ψ (x) + αδ(x)ψ(x) = Eψ(x) .
2m
The energy spectrum may be a discrete spectrum of bound states and a continuum
of free states.
a. Calculate the transmission coefficient for the case of a particle with energy E thrown
against the potential energy barrier V (x) = αδ(x). Does the result change for the
case when V (x) = −αδ(x), with α > 0?
b. For this last potential, find the energy of the bound state and its corresponding
wavefunction.
Consider a particle with energy E thrown (in the direction êx ) against a potential
energy barrier of finite height and width, such that V (x) = 0 for x < 0 or x > L and
V (x) = V0 for 0 ≤ x ≤ L.
a. Obtain the reflection and transmission coefficients R and T for the case E > V0 .
Discuss the result.
b. Do the same for the case E < V0 .
3.4. NUMERICAL APPROACHES FOR ARBITRARY POTENTIALS 99
Figure 3.9: (code) (a) Free particle (magenta) and bound state wavefunction (red) calculated
using a MATLAB ODE solver. (b) Bound state wavefunction of the harmonic oscillator cal-
culated using a MATLAB ODE solver. For the red curve the eigenvalue has been slightly
overestimated and for the blue curve underestimated. (c) Calculation of the harmonic oscil-
lator energies and wave functions using the Fourier grid method.
The highest harmonics of the Fourier expansion N is set by the distance between two
grid points. The expansion coefficients are,
Z N N
1 2π
1 X 1 X
ck = f (x)e−ıkx dx −→ f (xj )e−ıkxj ∆x = f (xj )e−ıkxj .
2π 0 2π 2N + 1
j=−N j=−N
(3.61)
3.4. NUMERICAL APPROACHES FOR ARBITRARY POTENTIALS 101
Now, substituting x by the discrete variable xi , we take the derivative f 0 (xi ) of the
function (3.60), insert the expansion coefficients (3.61), and compare the resulting
expression to the linear combination (3.59) of the functional values f (xj ),
N
X N
X N
X
ı
f 0 (xi ) = ıkck eıkxi = k f (xj )eık(xi −xj ) (3.62)
2N + 1
k=−N k=−N j=−N
N
X N
X N
X
2
= k f (xj ) sin k(xj − xi ) ≡ Dij f (xj ) .
2N + 1
k=1 j=−N j=−N
From this we see, that the matrix elements Dj 0 j are given by,
X N X N
2 2 2πk(i − j)
Dij = k sin k(xj − xi ) = k sin . (3.63)
2N + 1 2N + 1 2N + 1
k=1 k=1
(−1)i−j (−1)i−j
Dii = 0 and Di6=j = xi −xj = . (3.64)
2 sin 2 2 sin π(i−j)
2N +1
yielding,
N N
−2 X 2 −2 X 2 2πk(i − j)
(D2 )ij = k cos k(xj − xi ) = k cos . (3.66)
2N + 1 2N + 1 2N + 1
k=1 k=1
Again, the sum can be evaluated analytically, as shown in Exc. 3.4.4.3 [677],
N2 + 2 (−1)i−j
(D2 )ii = − and (D2 )i6=j = − π(i−j)
. (3.67)
12 2 sin2 2N +1
h2
Ĥψ = Eψ with Ĥij = − (D2 )ij + V (xj )δij . (3.68)
2mL2
This eigenvalue problem can easily be solved on a computer, see Fig 3.9(b).
102 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
continuously reinserting the solution into the equation. This Newton method does not
converge quickly (dt should be chosen small enough, when Ĥ(t) varies rapidly), but
there are other more sophisticated methods like the Runge-Kutta method.
A variation of this method is called steepest descent method. This method is
similar to the Newton Eq. (3.70), but replaces the time dt with an imaginary time.
Thus, the coherent temporal evolution of the Schrödinger equation is replaced by a
dissipative evolution. The loss of energy automatically takes the system to the ground
state. However, in order to preserve the normalization of the wavefunction, it must
be renormalized at each iteration step,
|ψ(t + dt)i
|ψ(t + dt)i → p . (3.71)
hψ(t + dt)|ψ(t + dt)i
The method also applies to more complicated equations than the Schrödinger equa-
tion, for example, the Gross-Pitaevskii equation. We will deepen this technique in
Sec. 26.3.3.
5
4
3/2
3
Ψ/aho
2
1
0
-1 0 1
x/aho
Figure 3.10: (code) Condensate wavefunction (red) in a double-well potential (blue) numer-
ically calculated using the steepest descent method. Also shown is the chemical potential
(green).
Another numerical method often used in quantum mechanics is the method called
the quantum Monte Carlo simulation of the wavefunction [512]. This method sim-
ulates trajectories of quantum systems treating intrinsic quantum noise as random
processes disrupting the uniformity of the trajectory. The advantage of this method
is that it also applies to dissipative systems.
3.4. NUMERICAL APPROACHES FOR ARBITRARY POTENTIALS 103
3.4.4 Exercises
3.4.4.1 Ex: Numerical resolution of the Hermite differential equation
Solve the Hermite differential equation (3.90) numerically for n = 8, e.g. using the
’ode45’ ordinary differential equation solver of Matlab, or similar. Plot the wavefunc-
tion of the 8-th vibrational level of a harmonic oscillator.
Derive (a) the formula (3.64) and (b) the formula (3.67) for N 1 using the formula
(1.352) from [290].
3.4.4.4 Ex: Derivation of the Fourier grid method from the FFT
N
X −1
Hn = e−2πınk/N hk
k=0
N −1 N/2−1
X X
= e−2πınk/(N/2) h2k + e−2πık/N e−2πınk/(N/2) h2k+1 = even + odd .
k=0 k=0
1 XN −1 2πınk/N
hk = e Hn .
N k=0
2 XN −1
Sn = sk sin πnk/N .
N k=1
commutator is,
"r r r r #
mω 1 mω 1 ı
[â, ↠] = x̂ + ı p̂, x̂ − ı p̂ = [x̂ + p̂, x̂ − p̂]
2~ 2m~ω 2~ 2m~ω 2~
ı
= ~ [p̂, x̂] = 1 . (3.74)
Knowing Ĥ|ψi = E|ψi is it clear that ↠â is an observable with the eigenvalue n ≡
E 1
~ω − 2 ,
↠â|ψi = ~ω
E
− 12 |ψi ≡ n|ψi =⇒ |ψi = |ni . (3.75)
Now, we show that the states â|ψi are eigenstates of the operator defined as n̂ ≡ ↠â,
since,
↠ââ|ψi = (â↠− [â, ↠])â|ψi = (â↠â − â)|ψi = â(↠â − 1)|ψi = (n − 1)â|ψi
=⇒ â|ψi ∝ |n − 1i ≡ C|n − 1i
=⇒ n = hn|↠â|ni = C 2 hn − 1|n − 1i
√
=⇒ C = n . (3.76)
↠â↠|ψi = ↠([â, ↠] + ↠â)|ψi = ↠(1 + ↠â)|ψi = (n + 1)↠|ψi
=⇒ ↠|ψi ∝ |n + 1i ≡ C|N + 1i
=⇒ n + 1 = hn|↠â + [â, ↠]|ni = C 2 hn + 1|n + 1i
√
=⇒ C = n + 1 . (3.77)
Therefore, this new state is also an eigenvector |n + 1i, with a quantum number
increased by one unit. ↠and â are creation and annihilation operators of an energy
packet,
√ √
↠|ni = n + 1|n + 1i and â|ni = n|n − 1i . (3.78)
The matrix representation of the field operators is,
X√ X√
↠= n + 1|n + 1ihn| and â = n|n − 1ihn| . (3.79)
n n
Now we start looking for asymptotic solutions. For x̃ → ±∞, that is, when the
particle enters the classically forbidden region, we can neglect the total energy of the
particle,
d2 2
− 2 + x̃ ψ̃∞ (x̃) ' 0 . (3.87)
dx̃
2
The solution of this equation is ψ̃∞ (x̃) = Ce−x̃ /2 , since
d2 2 d 2 2
− 2 + x̃ e−x̃ /2 = − (−x̃)e−x̃ /2 + x̃2 e−x̃ /2
2
(3.88)
dx̃ dx̃
2 2 2 2
= −x̃2 e−x̃ /2
+ e−x̃ /2
+ x̃2 e−x̃ /2
= e−x̃ /2
'0.
2
This motivates the ansatz ψ̃(x̃) ≡ e−x̃ /2
H(x̃) for the complete differential equation
(3.86),
2 2
2
d2 de−x̃ /2 dH(x̃) d2 e−x̃ /2
2 −x̃2 /2 −x̃2 /2 d H(x̃) 2
− 2
+ x̃ e H(x̃) = −e 2
− 2 − 2
H(x̃) + x̃2 e−x̃ /2 H(x̃)
dx̃ dx̃ dx̃ dx̃ dx̃
2
2 d H(x̃) 2 dH(x̃) h 2 2
i 2
= −e−x̃ /2 − 2(−x)e−x̃ /2 + −x̃2 e−x̃ /2 + e−x̃ /2 H(x̃) + x̃2 e−x̃ /2 H(x̃)
dx̃2 dx̃
2
≡ (2n + 1)e−x̃ /2
H(x̃) . (3.89)
Thus, the functions H(x̃) must satisfy the differential equation,
H 00 (x̃) = 2x̃H 0 (x̃) − 2nH(x̃) . (3.90)
We can verify that the Hermite polynomials defined by,
dn −x̃2 2
Hn (x̃) = (−1)n ex̃
e , (3.91)
dx̃n
transform the differential equation into a recursion formula,
Hn+1 (x̃) = 2x̃Hn (x̃) − 2nHn−1 (x̃) , (3.92)
which allows us to easily calculate the polynomials,
H0 (x̃) = 1 , H1 (x̃) = 2x , H2 (x̃) = 4x2 − 2 , ... (3.93)
In summary, the eigenfunction of a harmonic oscillator in the state of excitation
n is,
2
/2a2ho
hx|ni = ψn (x) = Cn e−x Hn (x/aho ) , (3.94)
where the constant Cn is determined by the normalization condition (7.22), hψm |ψn i =
δm,n ,
1
Cn = p √ . (3.95)
aho π2n n!
The Hermite functions, Hn , are found in mathematical tables, see Sec. 7.3. The
spatial and momentum wavefunctions for the vibrational ground state are,
r
1 −x2 /2a2ho 1 aho −a2ho p2 /2~2
hx|0i = 1/4 √ e and hp|0i = 1/4 e . (3.96)
π aho π ~
Here we will only show the graphical representation of |ψ|2 in Fig. 3.12. The Exc. 3.5.6.1
asks to evaluate HO in a classically forbidden region and in Exc. 3.5.6.2 we will cal-
culate the spectrum of a semi-harmonic HO.
108 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
E/h̄ωho , ψ(x)
4
0
-5 0 5
x/aho
We see that for high energy values the wavefunction approaches the classical expec-
tation.
We already mentioned that there exist solutions only for certain energies En =
~ω(2n + 1). Consequently, the energy levels are equidistant, En+1 − En = ~ω, as if
there were a box into which we add, one after the other, particles with the energy ~ω
until we have accumulated n portions of energy. These particles are called phonons
in the case of vibrations of massive particles, and photons in the case of a radiation
field.
The fact that the energy distribution is the same as the one proposed by Planck
for the black-body radiation suggests the use of the harmonic oscillator to describe
the second quantization.
3.5. HARMONIC OSCILLATOR 109
we can write,
X
e−iĤt/~ = |nie−iωt(n+1/2) hn| . (3.101)
n
P
If the initial state is |ψ(0)i = m cm |mi, the final state and the eigenvalue of any
observable will be,
X X
|ψ(t)i = |nie−ıωt(n+1/2) hn|ψ(0)i = e−ıωt(n+1/2) cn |ni (3.102)
n n
X X X
hψ(t)|Â|ψ(t)i = hm|eıωt(m+1/2) c∗m |Â| e−ıωt(n+1/2) cn |ni = c∗m cn eiωt(m−n) hm|Â|ni .
m n m,n
that is, the state remains stationary. Motion needs non-diagonal elements of Â.
Another observation is that the populations do not change, even in the case of an
initial superposition, since,
We conclude that
That is, the total energy of the oscillator is the sum of the energies of the
states weighted with the populations of those states. In the case of the position
operator,
X
hψ(t)|x̂|ψ(t)i = aho
√
2
c∗m cn eiωt(m−n) hm|â + ↠|ni
m,n
X √ √
= aho
√
2
c∗n−1 cn e−iωt n + c∗n+1 cn eiωt n + 1
n
m,n→∞ √ X√
−→ aho 2 n|cn |2 cos ωt .
n
That is, the particle can only oscillate, if there are populations in consecutive
states. If this is not the case, hψ(t)|x̂|ψ(t)i = 0. The oscillation frequency
is always ω, independent of the energy of the particle. We will study this
in Excs. 3.5.6.3, leaving the discussion of the temporal evolution of perturbed
oscillators to later sections.
where k = x, y, z.
3.5.6 Exercises
3.5.6.1 Ex: Ground state of a harmonic oscillator
Equating the ground state energy of quantum HO to that of its classical analog, obtain
the maximum elongation xm . Now, knowing that the ground state wavefunction is
2 2
proportional to the Gaussian ψ0 ∝ e−x /2xm , obtain the expression for the probability
of finding the HO outside the classical limits and estimate its value.
3.6. SUPERPOSITION STATES OF A HARMONIC OSCILLATOR 111
Ĝ0 (τ ) = (Â + B̂)eτ (Â+B̂) e−τ B̂ e−τ Â − eτ (Â+B̂) B̂e−τ B̂ e−τ Â − eτ (Â+B̂) e−τ B̂ Âe−τ Â
h i h i
= eτ (Â+B̂) Âe−τ B̂ − e−τ B̂ Â e−τ Â = eτ (Â+B̂) Â − e−τ B̂ Âeτ B̂ e−τ B̂ e−τ Â
h i
= eτ (Â+B̂) Â − Â + [−τ B̂, Â] + 2! 1
[−τ B̂, [−τ B̂, Â]] + .. e−τ B̂ e−τ Â ,
using the formula (2.165). If now [Â, [Â, B̂]] = 0 = [B̂, [Â, B̂]], then,
Ĝ0 (τ ) = eτ (Â+B̂) τ [B̂, Â]e−τ B̂ e−τ Â = −τ [Â, B̂]eτ (Â+B̂) e−τ B̂ e−τ Â = −τ [Â, B̂]Ĝ(τ ) .
The state resulting from the action of the operator D̂(α) onto the ground state of
the HO is,
X∞
2
/2 α↠−α∗ â (α↠)n 2
|αi ≡ D̂(α)|0i = e−|α| e e |0i = e−|α|
|0i /2
(3.112)
n=0
n!
−|α|2 /2 † (α↠)2 −|α|2 /2 α√ α2 √
=e 1 + αâ + + .. |0i = e |0i + 1|1i + 2!|2i + .. ,
2! 1! 2!
3 The operator acts on the phase space spanned by the operators â and ↠, that is, x̂ ∝ Re â and
p̂ ∝ Im â.
3.6. SUPERPOSITION STATES OF A HARMONIC OSCILLATOR 113
that is, the state |αi is a superposition distributed according to the Poisson distribu-
tion,
X∞
2 αn
|αi = e−|α| /2 √ |ni . (3.113)
n=0 n!
that is,
â|αi = α|αi . (3.115)
We can also write,
† †
hα|↠= (â|αi) = (α|αi) = hα|α∗ .
The state |αi is called coherent state or Glauber state 4 . We note that, in spite of
its appearance, the equation (3.115) is not an eigenvalue equation, since â is not
observable.
Figure 3.14: Illustration of the action of displacement operators D(α) and D(β) on Glauber
states in the phase space spanned by α = Re α+ıIm α = |α|eıφ . The uncertainty of Glauber
states ∆x and ∆p is represented by a finite distribution of the amplitude α in the phase space.
Furthermore, the product of two displacement operators is, apart from a phase factor,
another displacement operator satisfying,
∗
−α∗ β)/2
D̂(α)D̂(β) = e(αβ D̂(α + β) , (3.117)
4 We can also define a Bargmann state as the eigenstate corresponding to the step-up operator
as will be verified in Exc. 3.6.6.1. When acting on an eigenket, the phase factor
∗ ∗
e(αβ −α β)/2 appears in each term of the resulting state, which makes it physically
irrelevant.
√
2
aho hα|x̂|αi = hα|â + ↠|αi = α + α∗ (3.118)
√
ıaho 2
~ hα|p̂|αi = hα|â − ↠|αi = α − α∗ .
a2ho
∆x2 = hα|x̂2 |αi − hα|x̂|αi2 = 2 (3.120)
2 2 2 ~2
∆p = hα|p̂ |αi − hα|p̂|αi = 2a2ho
.
∆p∆x = ~
2 . (3.121)
Comparing with the uncertainty relation (3.85) derived for Fock states, we con-
clude that the uncertainty is always smallest for Glauber states. In this sense, the
Glauber states are the ones which are closest to classical states characterized by the
absence of uncertainty.
We leave the demonstration for Exc. 3.6.6.2, but we note here already that for |α −
β| 0 the states are approximately orthogonal. The reason for this is, that the
respective population distributions through the Fock states, |hn|αi|2 and |hn|βi|2 , do
not overlap and hence do not interfere. The state |αi + | − αi is sometimes called
Schrödinger cat state. In Exc. 3.6.6.4 we will show why such states are very difficult
to detect.
3.6. SUPERPOSITION STATES OF A HARMONIC OSCILLATOR 115
we find,
ıkaho
√ (â+↠) †
−α∗ â
eıkx̂ = e 2 = eαâ = D(α) , (3.127)
which is nothing else than the displacement operator (3.112) generating coherent
states, |αi = D(α)|0i. That is, the momentum kick attempts to put the oscillator
116 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
1 2 2
into a coherent state with the mean occupation number n = |α|2 = 2 k aho . Note
that the same effect is obtained setting,
√
α ≡ −b/(aho 2) (3.128)
and applying the translation operator
−b
√ (â−↠) †
−α∗ â
e−ıbp̂/~ = e aho 2
= eαâ = D(α) . (3.129)
Now,
X∞
2 αm
hn|eıkx̂ |0i = hn|αi = hn|e−|α| /2
√ |mi (3.130)
m=0 m!
2 αn 2 (ıka )n
= e−|α| /2
√ = e−|kaho | /4 √ ho .
n! 2n n!
This formula tells us that for small α, transitions may only occur to the vibrational
states |0i and |1i, since,
ka →0 αn
hn|eıkx̂ |0i −→
ho
√ ' δn,0 + αδn,1 . (3.131)
n!
On the other hand for very large α, the exponential in (3.130) pulls the transition
rate to a particular vibrational state |ni to zero, but this is simply due to the fact
that the population is redistributed over many states, since,
X X∞
2 |α|2n 2 2
|hn|eıkx̂ |0i|2 = e−(kaho ) /2
= e−(kaho ) /2 e|α| = 1 . (3.132)
n n=0
n!
1
n=0
1
|n|eikx̂ |0|2
2
3
0.5 4
0
0 0.5 1
√1.5 2 2.5
kaho / 2
Figure 3.15: (code) Transition matrix element |hn|eıkx̂ |0i|2 as a function of kaho . The curves
are Poisson distributions of photon numbers in coherent states.
For small momentum kicks we may approximate the transition matrix elements
by,
hm|eıkx̂ |ni ' hm|1 + ık a√ho2 (â + ↠)|ni (3.133)
√ √
= δm,n + ıka
√ ho ( nδm,n−1 +
2
n + 1δm,n+1 ) ,
3.6. SUPERPOSITION STATES OF A HARMONIC OSCILLATOR 117
such that, X
|hm|eıkx̂ |ni|2 ' k 2 a2ho (n + 12 ) . (3.134)
m6=n
Formula (3.134) tell us that the probability for a transition between vibrational
states depends on the parameter kaho , which we will discuss in the next section.
For small kaho it gets increasingly more difficult for the system to leave the original
vibrational state |ni and to form a coherent state. After a momentum kick, the
population is coherently distributed over several vibrational states in a way to fulfill
momentum and energy conservation. Let us consider, for simplicity, an initial state
|0i. If the kick is weak or the trap strong (i.e. if |α| < 1), the atom will stay in |0i with
a high probability amplitude and go to |1i only with a small probability amplitude.
We will derive in Exc. 3.6.6.5 the general expression for the transition matrix element
for arbitrary values of α,
s
n
X
ıkx̂ |α|2 /2 n m
hm|e |ni = e hn − k|αihm − k|αi . (3.135)
k k
k=0
Kicking a harmonic oscillator initially in state |0i we obtain for the expectation
value of position and momentum,
where we understand the recoil-energy ~ωrec as the kinetic energy gained through
the recoil acceleration. In this form the Lamb-Dicke parameter tells us that, in the
Lamb-Dicke regime, the energy of the momentum kick is not sufficient to efficiently
excite vibrational states,
(2n + 1)~2 k/m ωho . (3.139)
That is, cold particles in low vibrational states n can not accommodate the recoil
shift within the vibrational spectrum of the trap. Consequently, the recoil cannot be
transferred to the particle itself, but must be absorbed by the entire trap. This is the
case of the strong binding regime in ion traps, which is analogous to the Mößbauer
effect discussed in Sec. 18.3.3.
Finally, the Lamb-Dicke parameter can be rewritten in terms of the inverse Doppler
modulation index,
kvmax m 2
η= with v = ~ωho . (3.140)
2ωho 2 max
Accelerated by the momentum kick, the atom will execute harmonic oscillations with
frequency ωho and with the maximum velocity-excursions vmax . Hence, any quantity
depending on the atomic velocity, e.g. the Doppler-shift, will be modulated. We will
discuss this in Sec. 18.3.3, as well.
~2 d2 ~2 d2 µ 2
Ĥ = − − + ωho (x1 − x2 )2 ,
2m1 dx21 2m2 dx22 2
3.6. SUPERPOSITION STATES OF A HARMONIC OSCILLATOR 119
~2 d2 ~2 d2
µ 2
− 2
− 2
+ ωho (x̂1 − x̂2 )2 ψ(x1 , x2 ) = Etot ψ(x1 , x2 ) ,
2m1 dx1 2m2 dx2 2
we calculate,
d dR d dr d m1,2 d d
= + = ±
dx1,2 dx1,2 dR dx1,2 dr M dR dr
d2 m21,2 d2 m1,2 d d d2
2
= 2 2
±2 + 2
dx1,2 M dR M dR dr dr
~2 d2 ~2 d2 ~2 d2 ~2 d2
− 2
− 2
=− 2
− .
2m1 dx1 2m2 dx2 2M dR 2µ dr2
Hence, we get,
~2 d2 ~2 d2
µ 2 2
− Θ(R) = (Etot − E)Θ(R) , − + ωho r̂ φ(r) = Eφ(r) .
2M dR2 2µ dr2 2
m2
Now, we analyze the kick operator eikx̂1 with x̂1 = R̂ − M
r̂. We find that,
because of [R̂, r̂] = 0,
m2
eikx̂1 = eikR̂ e−ik M
r̂
.
This means that the kick operator simultaneously acts on both: the center-of-
mass receives a recoil accelerating it by an amount pcm = ~k, and the vibrational
relative motion receives a kick of an amount prl = ~km2 /M following the dy-
namics described in Sec. 3.6.2.
Obviously, the total system conserves momentum, which is imparted to the
center-of-mass motion. Whether the kick also excites the relative motion de-
pends on mass ratio. In the limit m1 m2 we find prl = ~k, while in the limit
m1 m2 we get prl → 0. In terms of the Lamb-Dicke parameter, we find,
s s
~ ~
prl aho km2 km2 1 1
η= √ = √ = + .
~ 2 M 2 µωho m1 + m2 2ωho m1 m2
finite amount of time, e.g. the duration of a radiative π-pulse required to excite an
atomic transition.
As a first example let us consider a harmonic oscillator subject to periodic forcing,
~2 d2 m 2 2 ~Ω
Ĥ(t) = Ĥ (0) + Ĥ (1) = − + ωho x̂ + (âeıνt + ↠e−ıνt ) , (3.141)
2m dx2 2 2
and study its time evolution,
X an =δn,0
|ψ(t)i = e−(ı/~)Ĥt an |ni −→ e−(ı/~)Ĥt |0i , (3.142)
n
when the oscillator is initially in the state |0i. We rewrite the time evolution propa-
gator as,
ıνt † −ıνt
e−(ı/~)Ĥt = e−ıωho t(n̂+1/2)−(ıΩt/2)(âe +â e )
. (3.143)
To simplify the propagator, we first have a closer look at the terms âeıωho t and
↠e−ıωho t , which we evaluate through their action on the complete system of unper-
turbed eigenfunctions,
(0) (0) (0) (0)
e−(ı/~)Ĥ t
âe(ı/~)Ĥ t
|ni = e−(ı/~)Ĥ t
âe(ı/~)En t |ni (3.144)
(0) (0)
−(ı/~)En−1 t (ı/~)En t
=e e â|ni = eıωho t â|ni .
Hence, defining the detuning ∆ ≡ ν − ωho ,
The dynamics is illustrated in Fig. 3.18. In Exc. 3.6.6.6 we will study how to
generalize the problem to non-resonant excitation, ∆ 6= 0.
5 Note, that this formula can not be simplified using the Baker-Hausdorff formula, because
0.2
1
0.15
ρkk (tf )
ρkk (t)
0.1 0.5
0.05
0 0
0 5 10 15 20 0 0.5 1 1.5
n Ωt
Figure 3.18: (code) (a) Populations of the vibrational states after a given interaction time.
(b) Time evolution of the lowest populations.
where g denotes the acceleration. That is, we expect that a harmonic oscillator,
~2 d2 m 2 2
Ĥ(t) = − + ωho x̂ − mgx̂ Θ[0,∆t] (t) , (3.151)
2m dx2 2
where the last constant term plays no role in the dynamics. Knowing the eigenvalues
and -states of the unperturbed Hamiltonian, the obvious solution of the perturbed
eigenvalue problem is,
and using the representation (3.83) of the momentum operator by the field operators
and introducing the abbreviation β ≡ ω2 ag √2 we can rewrite,
ho ho
†
g
hx − 2 |ni
ωho
= e−β(â−â ) hx|ni . (3.155)
Note, that the transition from (3.151) to (3.153) can also be obtained by a redefinition
of the field operators according to,
mga√
b̂ ≡ â − ho
~ωho 2
, (3.156)
since,
mga mg 2
Ĥ = ~ωho (↠â + 12 ) − √ ho (â
2
+ ↠) = ~ωho (b̂† b̂ + 21 ) − 2
2ωho
(3.157)
To simplify the first three exponential functions, we use the relationships (3.147) and
(3.144) 6 ,
â−e−ıωho t â†
= e−β (e ),
† −(ı/~)Ĥt
(â−↠)e(ı/~)Ĥt ıωho t
e−(ı/~)Ĥt e−β(â−â ) e(ı/~)Ĥt = e−βe
(3.159)
and write the temporal solution,
ıωho t
−↠e−ıωho t )
ψn(1) (x, t) = e−β(âe e−(ı/~)Ĥt ψn (x) (3.160)
−β(â−↠) cos ωho t−ıβ(â+↠) sin ωho t −(ı/~)En
(0)
t
=e e ψn (x) .
Finally,
g
|ψn(1) (x, t)|2 = |ψn (x − x̄(t))|2 where x̄(t) ≡ 2
ωho
cos ωho t . (3.162)
6 We can use (3.144), because the shifted harmonic oscillator has, except from a constant offset
(1)
This means that the spatial distribution of ψn around x̄(t) is the same as of ψn
around x̄ = 0. The entire distribution oscillates without deformation. The momentum
distribution follows from the Fourier transform,
Z
1
(1)
φn (p, t) = √ dxe−(ı/~)px ψn(1) (x, t) (3.163)
2π~
Z g
1 −(ı/~)umωho 2 sin ωho t ıγ(p,t)
=√ du e−(ı/~)pu e ωho e ψn
2π~
= eiγ(p,t) φn (p + mωho ωg2 sin ωho t) ,
ho
∗
where the abbreviation γ = γ contains all unitary transformations of (3.161), that
do not depend on x. We obtain,
|φ(1) 2
n (p, t)| = |φn (p − p̄(t))|
2
where p̄(t) ≡ −m ωgho sin ωho t . (3.164)
or with (3.167),
1
hα(t)|x̂|α(t)i = hα(0)|x̂|α(0)i cos ωho t + mωho hα(0)|p̂|α(0)i sin ωho t
. (3.169)
hα(t)|p̂|α(t)i = mωho hα(0)|x̂|α(0)i sin ωho t + hα(0)|p̂|α(0)i cos ωho t
We note, that the position and momentum wavefunctions hx|ni and hp|ni of the
excited vibrational states are complicated Hermite polynomials, but we don’t have
to write them down explicitly to get to the last result. We will derive some further
properties in 3.6.6.8.
when thePoscillator is initially in the state |0i. Note that, without perturbation,
|ψ(t)i = n e−ıωho (n̂+1/2)t an |ni.
Various quantities now are interesting to plot as a function of time. Firstly, we
calculate the populations |hn|ψ(t)i|2 of the various states and the total energy from,
X
hn|ψ(t)i = hn|e−(ı/~)Ĥt am |mi
m
X . (3.172)
Etot (t) = ~ωho (n + 12 )|hn|ψ(t)i|2
n
weighed with instantaneous population (3.172) of this state. We thus obtain for the
total wavefunction,
Furthermore, we may calculate the expectation values of the position and the
momentum of the harmonic oscillator from the total wavefunction hx|ψ(t)i,
Z Z
hψ(t)|x̂|ψ(t)i = hψ(t)|x̂|xihx|ψ(t)idx = x|ψ(x, t)|2 dx (3.176)
Z Z
hψ(t)|p̂|ψ(t)i = hψ(t)|p̂|pihp|ψ(t)idp = p|φ(p, t)|2 dx ,
The temporal evolution of the forced harmonic oscillator defined by (3.171) can
be numerically propagated. In cases where the force varies in time, the propagation
can be done iteratively by subdividing time in intervals tk − tk−1 for k ∈ N sufficiently
small that the force can be considered constant within them,
0.4 6
(d)
|n|ψ(t)|2
(a)
0.2
5
0
1 2 3 4 5 6 7 8
E/h̄ωho , |x|nn|ψ(x)|2
n
4
1 (b)
|n|ψ(t)|2
n=0
1
2
0.5 3 3
4
0
0 1 2 3
ωho t/π 2
x̂ , p̂ , h̄k(t)
2 ψ(t)|p̂|ψ(t) (c)
0 1
-2 ψ(t)|x̂|ψ(t)
0 1 2 3 0
ωho t/π -2 0 2
x/aho
| n hn|ψ(p)i|2 . The horizontal yellow line is the total energy of the system. You may also
P
run a movie of the simulation clicking on (watch movie)! Furthermore, a presentation about
the topic of recoil on trapped atoms is available at (watch talk)!
3.6.6 Exercises
3.6.6.1 Ex: Sum of displacements operators
Prove formula (3.117).
c. What is the population of the state |ni of a harmonic oscillator in a Glauber state?
Show that the annihilation operator â does not reduce the photon number of a Glauber
state in contrast to a Fock state.
Calculate the probability of finding n photons in Schrödinger’s cat state |ψi = 2−1/2 (|αi±
| − αi).
(↠)n 2 αn
|ni = √ |0i and hn|αi = e−|α| /2
√
n! n!
X∞ n
2 t
e2xt−t = Hn (x) .
n=0
n!
different approaches. We have seen in Sec. 3.6.1 how to expand the state of a harmonic
oscillator on a Fock state basis. On the other hand, we have seen that Glauber states
are more ’natural’ states for a harmonic oscillator, so that we would like to visualize
expansions of arbitrary states into a coherent state basis.
The following sections are devoted to introducing various such distribution func-
tions [634] and to calculating them for a selection of particular states, such as the Fock
and the Glauber state, the thermal state, and the Schrödinger cat state. We will re-
strict the discussion to pure states, postponing a discussion of the representation of
statistical mixtures to Sec. 15.2, where we will also extend the discussion to quantum
correlations in light fields resulting from a Jaynes-Cummings type interaction of an
atom with a light mode.
ρ̂ ≡ |ψihψ| . (3.182)
A detailed discussion of the density operator for non pure states, that is, statistical
mixtures will be given in Sec. 14.1. The definition (3.182) is independent from a
chosen basis, but a proper representation can be found by simply expanding the state
|ψi on a proper basis.
hr|ψi = √1
V
eık·r ,
For a harmonic oscillator, the Wigner function can be embedded into a formalism
that will be detailed in the following sections.
130 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
these three coherent distribution functions are expressed by inverse Fourier transforms
of the following characteristic functions 8 ,
† ∗
χN (λ) = hψ|eλâ e−λ â |ψi , P = F −1 χN
†
−λ∗ â
χS (λ) = hψ|eλâ |ψi , W = F −1 χS .
∗ †
χA (λ) = hψ|e−λ â eλâ |ψi , Q = F −1 χA
Here, the index N denotes normal order, S symmetric order, and A antisymmetric
† ∗
order. D(α) = eλâ −λ â is the displacement operator defined in (3.109). We see that
the distribution functions χN,S,A are related to each other. Using the Baker-Hausdorff
formula (3.108) we derive in Exc. 3.7.5.1 the so-called disentangling theorem [277],
2 2
e−|λ| /2
χN (λ) = χS (λ) = e|λ| /2
χA (λ) . (3.185)
we find,
2 −2|λ|2 2 −2|λ|2 1 2
W =P? e and Q=W? e = P ? e−|λ| . (3.187)
π π π
The distribution functions can be interpreted as follows: P (α) is the probability of
finding the coherent state |αi within the statistical mixture given by,
Z
ρ̂ = P (α)|αihα|d2 α , (3.188)
8 For statistical mixtures, such as for example thermal states, we need to generalize the definition
1
Q(α) = π hα|ρ̂|αi . (3.189)
The various states that a light field can adopt can now be expressed either by
photon number distribution in a Fock state basis, or by two-dimensional weighting
functions P, Q, W in a coherent state basis. Here, are some examples for these repre-
sentations.
the normally ordered characteristic function (3.185) will be composed of terms like
† ∗
hm|eλâ e−λ â |ni. To evaluate these terms, we begin calculating,
s
n!
âk |ni = |n − ki for k ≤ n and âk |ni = 0 for k ≥ n , (3.191)
(n − k)!
and, s
∞
X Xn
−λ∗ â (−λ∗ â)k (−λ∗ )k n
e |ni = |ni = √ |n − ki . (3.192)
k! k! k
k=0 k=0
Hence, assuming m ≥ n,
s
Xn
λ↠−λ∗ â (−λ∗ )k λm−n+k n m
hm|e e |ni = p (3.193)
k!(m − n + k)! k m−n+k
k=0
r n r
n! m−n X m (−|λ|2 )k n! m−n m−n
= λ = λ Ln (|λ|2 ) ,
m! k+m−n k! m!
k=0
where Lm−n
n are Laguerre polynomials. Now, exploiting the fact that
† ∗ † ∗
hm|eλâ e−λ â |ni = (hn|e−λâ eλ â |mi)† = hn|e−λ↠eλ∗ â |mi (3.194)
r
m!
= (−λ∗ )n−m Ln−m
n (|λ|2 ) ,
n!
and with the expansion (3.190) we obtain for the normally-ordered characteristic
function,
X † ∗
χN (λ) = c∗m cn hm|eλâ e−λ â |ni (3.195)
m,n
X † ∗
= (1 − 12 δm,n ) c∗m cn hm|eλâ e−λ â |ni + c∗m cn hm|e−λ↠eλ∗ â |ni .
m≥n
132 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
finally yielding,
r
X ∗ n! m−n
χN (λ) = (1 − 1
2 δm,n ) cm cn λm−n + cm c∗n (−λ∗ )m−n L (|λ|2 ) .
m! n
m≥n
(3.196)
2
Remembering that the symmetrically ordered function is given by χS (λ) = e−|λ| χN (λ)
we may obtain the Wigner function by a numerical two-dimensional FFT,
Z
1 2 ∗
α−λα∗ 2
W (α) = 2 e−|λ| χN (λ)eλ d λ. (3.197)
π
We will use this result in Sec. 15.2.2 to characterize correlations in an optical mode
emanating from a Jaynes-Cummings type coupling to an atom. In Exc. 3.7.5.2 we
will try to find an analytic solution for this integral.
Figure 3.20: (code) Starting from a cat-state photon distribution, as shown in (a), we
calculate the symmetrically ordered characteristic function (b) from the expression (3.195),
and the Wigner function (c) by an FFT.
We have seen earlier that coherent states |βi can be expanded on a Fock state basis
|ni. For the state function, the density operator, and the photon number distribution
we have,
X βn
|βi = e−|β|/2 √ |ni
n n!
2 X β n β ∗m
ρ̂ = e−|β| √ |nihm| . (3.198)
m,n m!n!
2 |β|2n
Pn = |hn|βi|2 = e−|β|
n!
3.7. (QUASI-)PROBABILITY DISTRIBUTION FUNCTIONS 133
Expanding a coherent state on a basis of Glauber states, we will derive in Exc. 3.7.5.3
the P -function, the density matrix, the Q-function, and the Wigner function,
|ψi = |βi
2 −2|α−β|2
W|βi (α) = πe
Figure 3.21: (code) (a-c) Q functions and (d-f) Wigner functions of a Glauber state (a,d), a
Fock state (b,e), and a Schrödinger cat state (c,f).
134 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
X e−|β0 |2 /2 β n ± e−|β1 |2 /2 β n
0 1
|ψ± i = √ |ni
n 2n!
. (3.203)
2n n 2
2 |β0 | |1 ± (−1) |
Pn = |hn|ψ± i|2 −→ e−|β0 | for β0 = −β1
n! 2
Expanding a Schrödinger cat state on a basis of Glauber states, we will derive in
Exc. 3.7.5.4 the P -function, the density matrix, the Q-function, and the Wigner
function,
√
|ψ± i = C1 (|β0 i ± |β1 i) with C = 2 ± 2e−|β0 −β1 |2
ρ̂ = |ψ± ihψ± |
. (3.204)
1 −|α−β0 |2 2
Q|β0 i|β1 i (α) = πe + π1 e−|α−β1 |
2 2
2
W|β0 i|β1 i (α) = C2π e−2|α−β0 | + e−2|α−β1 |
∗ ∗
2
/2−|β1 |2 /2 )+β0∗ β1
± 2e−|β0 | Re e−2(β1 −α)(β0 −α
We know from microscopic systems, for instance a two-level atom, that they can
live in superposition states |1i + |2i. In contrast, the Schrödinger cat states discussed
above occur in continuous Schrödinger fields involving many particles. Macroscopic
quantum interferences (i.e. interferences that are detectable with macroscopic appa-
ratuses, for example in heterodyne schemes) are named fuzzy Schrödinger cats, if the
interfering states are conspicuously separated in phase space.
At this point, we have to emphasize the fundamental difference between Schrödinger
cats and superpositions of modes,
|ψi = |αi + |βi 6 |αi|βi
=
. (3.205)
ρ̂ = |αihα| + |βihβ| + |αihβ| + |βihα| =
6 |αβihαβ|
Schrödinger cats exhibit interferences in phase space, whereas for mode superposi-
tions, interferences only appear when a parameter is varied (e.g., the length of an
interferometer arm).
Schrödinger cat states are very sensitive to dissipation and easily converted into
statistical mixtures. For example, |αi ± | − αi contains only odd (even) photon num-
bers in the distribution function Pn . After some time ∼ τcav /N , the distribution is
converted into a Poisson distribution. The higher the particle number N , the faster
the decoherence will be. Consequently, truely macroscopic cat states have never been
observed.
3.7. (QUASI-)PROBABILITY DISTRIBUTION FUNCTIONS 135
ρ̂ = |nihn| . (3.206)
Pk = δnk
This state can be expanded into Glauber states |αi by following procedure. For large
n we first calculate the P -distribution function,
√
P|ni (α) = δ (1) (|α| − n) , (3.207)
because it allows us to derive the density matrix via the formula (3.188),
Z
√
ρ̂|ni = δ (1) (|α| − n)|αihα|d2 α (3.208)
Z ∞ Z 2π Z 2π
√ √
= δ (1) (|α| − n)|αihα| |α|d|α|dϕα = n |αihα|dϕα .
0 0 0
where Lm (4|α|2 ) are Laguerre polynomials. In Exc. 3.7.5.5 we will learn how to
derive the above distribution functions directly from the characteristic functions. In
summary, we have,
|ψi = |ni
√
P|ni (α) = δ (1) (|α| − n)
Z 2π
√
ρ̂ = n |αihα|dϕα
0 . (3.211)
|α|2n −|α|2
Q|ni (α) = e
πn!
2 −2|α|2
W|ni (α) = e (−1)n Ln (4|α|2 )
π
136 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
X n̄n
ρ̂therm = |nihn| . (3.212)
n
(1 + n̄)1+n
1 −|α|2 /n̄
Ptherm (α) = π n̄ e
1
R 2
ρ̂ = π n̄ e−|α| /n̄
|αihα|d2 α
. (3.213)
1 −|α|2 /(n̄+1)
Qtherm (α) = π(n̄+1) e
1 −|α|2 /(n̄+1/2)
Wtherm (α) = π(n̄+1/2) e
3.7.5 Exercises
3.7.5.1 Ex: Relationship between the characteristic functions
Derive the relationship (3.185) between the characteristic functions of the Glauber
state basis.
3.7.5.2 Ex: Wigner function for arbitrary HO states in the Fock basis
Search an analytic solution for the integral (3.196).
3.7.5.3 Ex: P -, Q-, and Wigner distribution functions for Glauber states
Starting from the characteristic functions χA,S,N (λ) derive for a coherent state |βi
(a) the Husimi representation, (b) the Glauber-Sudarshan representation, and (c) the
Wigner representation.
3.7.5.4 Ex: P -, Q-, and Wigner distribution functions for cat states
Starting from the characteristic functions χA,S,N (λ) derive for a normalized cat state
C −1 (|β0 i + |β1 i) (a) the Husimi representation, (b) the Glauber-Sudarshan represen-
tation, and (c) the Wigner representation.
3.7.5.5 Ex: P -, Q-, and Wigner distribution functions for Fock states
Starting from the characteristic functions χA,S,N (λ) derive for a number state |ni
(a) the Husimi representation, (b) the Glauber-Sudarshan representation, and (c) the
Wigner representation.
3.8. SQUEEZED STATES OF THE HARMONIC OSCILLATOR 137
∂ ∂
↠− α∗ = and ↠− α = ∗
.
∂→
α
− ∂←
α
−
Note, that with the bosonic operators we can construct the observables,
√
x̂ + ıp̂ = 2â , ∂x − ı∂p = 2∂α , ∂x + ı∂p = 2∂α∗ .
−
→ ←−
In two dimensions [247], d2 α = d(Re α)d(Im α) = dxdp.
where the index S denotes symmetric ordering. Symmetric or Weyl ordering means
that all products r̂p̂ must be replaced by the symmetric expressions, such as (r̂p̂ +
p̂r̂)/2.
The non-linear, i.e. quadratic, appearance of the field operators suggests that the
interaction should include correlation pair production, as is the case for parametric
processes or four-wave mixing. We will see later that cavities are good for producing
squeezing. However, the unused ports of a cavity let vacuum fluctuations come in,
which partially overrule squeezing.
For now, we study, in analogy with the displacement operator (3.109), the propa-
gator e−ıĤsqz t/~ , i.e the operator given by,
†2
/2−ξ ∗ â2 /2
Ŝ(ξ) ≡ eξâ , (3.216)
which we will call the squeezing operator because, applied to the vacuum state, |ξi =
S(ξ)|0i, this operator will compress the uncertainty of one quadrature component, as
we will see shortly.
In analogy with the calculation (3.110), using the commutation rules, it is possible
to verify the unitarity of this operator [457] (see Exc. 3.8.4.1). In particular, using
∗
the relationship (2.165) and the abbreviation  ≡ 2ξ â†2 − ξ2 â2 , we can show [686]
(see Exc. 3.8.4.2),
h i h h ii
Ŝ † (ξ)âŜ(ξ) = e âe− = â + [Â, â] + 2!
1 1
Â, [Â, â] + 3! Â, Â, [Â, â] ... (3.217)
|ξ| †
= â + ξ↠+ 1 ∗
2! ξξ â + 1 ∗
3! ξξ ξâ
†
+ ... = â cosh |ξ| + â sinh |ξ| ,
ξ∗
and similarly for ↠, such that with ξ ≡ reıϕ ,
The formulas (3.218) describe a Bogolubov transform, as they can be cast into the
form,
b̂ ≡ uâ + v↠, b̂† ≡ u∗ ↠+ v ∗ â , (3.219)
for complex numbers u and v. By postulating the same commutation relation for new
operators, [b̂, b̂† ] = 1, we immediately get the condition,
|u|2 − |v|2 = 1 . (3.220)
Comparing with the hyperbolic identity cosh2 r − sinh2 r = 1, we can parametrize the
constants as,
u = cosh r and v = eıϕ sinh r . (3.221)
This is interpreted as a linear simplectic transformation in phase space between
pairs of annihilation and creation operators satisfying the same commutation rela-
tion [â, ↠] = 1.
However, let us take a look at the uncertainties separately. They relate to the
Glauber mode via,
x̂b = 21 (b̂ + b̂† ) = 21 (â cosh r + ↠eıϕ sinh r) + 21 (↠cosh r + âe−ıϕ sinh r) (3.226)
ϕ=0
= x̂a cosh r + 1
2 sinh r(↠eıϕ + âe−ıϕ ) −→ x̂a er
ŷb = 1
2ı (b̂ − b̂† ) = 1
2ı (â cosh r + ↠eıϕ sinh r) − 1
2i (â
†
cosh r + âe−ıϕ sinh r)
ϕ=0
= ŷa cosh r + 1
2ı sinh r(↠eıϕ − âe−ıϕ ) −→ ŷa e−r .
With hâi = 0,
1 †
∆x̂2b = e2r 1 2 1
2 Re hâ i + 4 + 2 hâ âi (3.228)
∆ŷb2 = e−2r − 12 Re hâ2 i + 41 + 21 h↠âi .
with the abbreviation ζ ≡ eıθ tanh r. For squeezed vacuum α = 0, noting that
Hn (0) = (−2)n/2 (n − 1)!! for even n and Hn (0) = 0 for odd n, we find,
(−ζ)n/2 (n − 1)!!
hn|0, ξi = √ (3.231)
n! cosh r
for even photon number. Odd photon numbers are excluded. Hence,
X
|α, ξi = |nihn|α, ξi
n
. (3.232)
ρ̂ = |α, ξihα, ξ|
Pn = |hn|α, ξi|2
In the photon representation we can easily see that the squeezed vacuum is (unlike
the coherent and the Fock vacuum) not empty,
α→0
hα, ξ|n̂|α, ξi = |α|2 + sinh2 |ξ| −→ sinh2 |ξ| (3.233)
α→0
∆α,ξ n̂ = |α| + 2 cosh2 |ξ| sinh2 |ξ| −→ 2 cosh2 |ξ| sinh2 |ξ| .
Figure 3.22: (code) Starting from a squeezed state photon distribution with r = 0.5, as
shown in (a), we calculate the symmetrically ordered characteristic function (b) from the
expression (3.196), and the Wigner function (c) by an FFT.
For the squeezed vacuum state the photon number distribution displays odd-even-
oscillations. This can be explained by the mathematical form of the squeezing opera-
tor, that resembles the operator for two-photon generation and annihilation processes.
Photons in a squeezed vacuum state are more likely to appear in pairs.
|ψi =
P|βi (α) =
ρ̂ = .
2
+|β|2 )+(α∗ β+αβ ∗ ) sech r− 21 [eıθ (α∗2 −β ∗2 )+e−ıθ (α2 −β 2 )] tanh r
Q|βi (α) = sech r
π e−(|α|
∗ 2
(α−α∗ )2
W|βi (α) = 1
2π exp(− (α+α )
2e−2r + 2e2r )
(3.235)
3.8.4 Exercises
3.8.4.1 Ex: Unitarity of the squeezing operator
Calculate hÊi and ∆Ê.
O. Dulieu, Coupled channel bound states calculations for alkali dimers using the
Fourier grid method [DOI]
142 CHAPTER 3. LINEAR MOTION / SEPARABLE POTENTIALS
J. Stare et al., Fourier Grid Hamiltonian Method for Solving the Vibrational Schroedinger
Equation in Internal Coordinates Theory and Test Applications [DOI]
C. Clay Marston et al., The Fourier grid Hamiltonian method for bound state eigen-
values and eigenfunctions [DOI]
−~2 2 −~2 2
Ĥ = ∇r1 + ∇ + V (r1 − r2 ) , (4.1)
2m1 2m2 r2
where r1,2 are the positions of the proton and the electron. With the ansatz Ξ(t, r1 , r2 ) =
Ξ(r1 , r2 )e−ıEtot t/~ , the time-dependent Schrödinger equation
d
ĤΞ(t, r1 , r2 ) = ı~ Ξ(t, r1 , r2 ) , (4.2)
dt
becomes stationary,
2
−~ 2 −~2 2
∇ + ∇ + V (r1 − r2 ) Ξ(r1 , r2 ) = Etot Ξ(r1 , r2 ) . (4.3)
2m1 r1 2m2 r2
Now we transform into the center-of-mass system making for the total wavefunction
the ansatz Ξ(r1 , r2 ) = e−ıP·R/~ Ψ(r) with R ≡ m m2
M r1 + M r2 and r ≡ r1 − r2 and
1
143
144 CHAPTER 4. ROTATIONS / CENTRAL POTENTIALS
wave, describing the linear motion of the center of the masses, and a radial wave
function, which describes the relative motion of the atom. The kinetic energy of one
mass is:
−~2 2 −ıP·R/~
∇ e Ψ(r) (4.4)
2m1 r1
−~2 h −ıP·R/~ 2 −ıP·R/~ 2 −ıP·R/~
i
= e ∇r1 Ψ(r) + 2(− ım 1P
)e ∇ r 1
Ψ(r) + Ψ(r)(− ım1
2
~ M P) e
2m1 ~M
2 2
−~ 2 ı~P m1 P
= e−ıP·R/~ ∇r1 Ψ(r) + ∇r1 Ψ(r) − Ψ(r) .
2m1 M 2M 2
Hence, for two atoms,
Etot Ξ(r1 , r2 ) − V (r)Ξ(r1 , r2 ) (4.5)
2 2 2
−~ 2 −~ 2 ı~P P
= e−ıP·R/~ ∇ Ψ(r) + ∇ Ψ(r) + (∇r1 + ∇r2 )Ψ(r) + Ψ(r) .
2m1 r1 2m2 r2 M 2M
Using ∇r1 = −∇r2 = ∇r , we see that the third term cancels, such that,
P2 −~2 2 −~2 2
Ψ(r) + ∇r Ψ(r) + ∇ Ψ(r) + V (r)Ψ(r) = Etot Ψ(r) . (4.6)
2M 2m1 2m2 r
P2
Subtracting the energy of the center-of-mass motion with E = Etot − 2M and intro-
ducing the abbreviation m−1 = m−1 −1
1 + m2 , we finally get,
−~2 2
∇r + V (r) Ψ(r) = EΨ(r) . (4.7)
2m
and obtain,
2
1 ~2 ∂ 1 ∂ ~2 1 ∂ 2 ~2 1 ∂2
− + + V (ρ) R(ρ)− ζ(z)− ξ(ϕ) = E .
R(ρ) 2m ∂ρ2 ρ ∂ρ 2m ζ(z) ∂z 2 2mρ2 ξ(ϕ) ∂ϕ2
(4.12)
First, we separate the axial motion,
ζ 00 2mEz
− = const ≡ ≡ kz2 , (4.13)
ζ ~2
the solution of this equation being a superposition of two plane waves counterprop-
agating along the axis z, ζ(z) = Aeıkz z + Be−ıkz z . Now, we separate the azimuthal
motion,
ρ2 ∂R2 (ρ) ρ ∂R(ρ) 2mρ2 ξ 00
+ + 2 [E − V (ρ)] − ρ2 kz2 = − = const ≡ m2ϕ . (4.14)
R(ρ) ∂ρ 2 R(ρ) ∂ρ ~ ξ
The solution of the right-hand part of the equation is ξ(ϕ) = Ceımϕ ϕ + De−ımϕ ϕ .
Finally, we have the radial equation,
1 ∂R(ρ)2 1 ∂R(ρ) 2m 2
m2ϕ
+ − [E − V (ρ)] − k z − =0, (4.15)
R(ρ) ∂ρ2 ρR(ρ) ∂ρ ~2 ρ2
~2 m 2
with the effective potential Vef f = V (ρ) + 2mρϕ2 . For a homogeneous potential,
V (ρ) = V0 , the solution will be a superposition of Bessel functions.
Example 26 (Rigid rotor in cylindrical coordinates): To give an example,
we disregard the potential, V (ρ) = 0, and we consider for the particle an orbit
with constant radius, ρ = const such that R(ρ) = δ(ρ − ρ0 ). In this case, we
only need to treat the orbital motion described by the right part of Eq. (4.14).
For the solution of this equation, ξ(ϕ) = Aeımϕ ϕ , to be well-defined, we need
ξ(ϕ) = ξ(ϕ + 2π). This implies,
mϕ = 0, ±1, ±2, ..
and
~2 m2ϕ
Eϕ = .
2mρ2
The allowed energies Emϕ = Eϕ can be obtained by letting the Hamiltonian
~2 ∂ 2
Ĥ = − ,
2I ∂ϕ2
with the moment of inertia I = mρ2 actuate on the azimuthal wavefunction
ξ(ϕ). We now define the operator,
ˆ ~ ∂
lz = .
ı ∂ϕ
This operator acts on the wavefunction ξ as follows,
ˆ
lz ξ(ϕ) = ~mϕ ξ(ϕ) .
as 1 ,
1 ∂ ∂ 1 L̂2 L̂2 1 ∂ 1 ∂2 ∂
∇2r = 2 2
r + 2 2 where ≡ ,
sin ϑ +
r ∂r ∂r r ~ ~ 2 sin ϑ ∂ϑ sin2 ϑ ∂ϕ2 ∂ϑ
(4.17)
is an abbreviation called Legendre operator. For an isotropic potential, V (r) = V (r),
we can try the ansatz,
Ψ(r) = R(r)Y (ϑ, ϕ) (4.18)
to solve the Schrödinger equation (2.57),
r2 ~2 1 ∂ 2 ∂ −1 L̂2 Y (ϑ, ϕ) ~2
− r + V (r) − E R(r) = = const ≡ − `(`+1) ,
R(r) 2m r2 ∂r ∂r 2m Y (ϑ, ϕ) 2m
(4.19)
where we choose a separation constant, `(` + 1), the significance of which we shall
soon learn. Considering only the angular part,
we obtain,
2 1 1 ∂ ∂ 1 ∂2
sin ϑ sin ϑ Θ(ϑ) + `(` + 1) = − Φ(ϕ) = const ≡ m2 ,
Θ(ϑ) sin ϑ ∂ϑ ∂ϑ Φ(ϕ) ∂ϕ2
(4.22)
where we choose a separation constant, m2 . Introducing another abbreviation,
~ ∂
L̂z ≡ , (4.23)
ı ∂ϕ
the azimuthal equation takes the form
As in the case of the cylindrical potential, the solution of the azimuthal equation is,
using the normalization,
Φ(ϕ) = √12π eımϕ , (4.25)
with the magnetic quantum number m = 0, ±1, ±2, ...
1 We L2
may also write: p2 = (êr · p)2 + (êr × p)2 = p2r + r2
, where p2r is the radial part of the
2
L
Laplace operator and r2
the angular part.
4.1. PARTICLE IN A CENTRAL POTENTIAL 147
1 1 ∂ ∂ m2
sin ϑ Θ(ϑ) + `(` + 1) = , (4.26)
Θ(ϑ) sin ϑ ∂ϑ ∂ϑ sin2 ϑ
is called Legendre’s differential equation and can be solved by a power series in cosk ϑ.
For m = 0, the solutions are the Legendre polynomials, P` (cos ϑ) with
1 d`
P` (z) = [(z 2 − 1)` ] . (4.27)
2` `! dz `
The first polynomials are,
dm (−1)m 2 m/2 d
`+m
P`m (z) = (−1)m (1 − z 2 )m/2 P ` (z) = (1 − z ) [(z 2 − 1)` ]
dz m 2` `! dz `+m
(4.29)
(` − m)! m
P`−m (z) = (−1)m P (z) .
(` + m)! `
The functions Y`m (ϑ, ϕ) are the spherical harmonics. They form an orthonormal
system,
Z π Z 2π
Y`∗0 m0 (ϑ, ϕ)Y`m (ϑ, ϕ) sin ϑdϑdϕ = δ`0 ` δm0 m . (4.31)
0 0
Finite solutions only exist when the angular momentum quantum number is ` = 0, 1, ..
and for |m| ≤ `.
=0 =1 =2 =3
90 1 90 1 90 1 90 1
120 60 120 60 120 60 120 60
Figure 4.1: (code) Angular wavefunctions. Shown are the Legendre polynomials P`m (cos ϑ)
for ` = 0, 1, 2, 3 and m = 0, .., `. Red: m = 0, green: |m| = 1, blue: |m| = 2, and magenta:
|m| = 3.
148 CHAPTER 4. ROTATIONS / CENTRAL POTENTIALS
The solutions of the angular part of the Schrödinger equation for the hydrogen
atom are finally,
s
1 2` + 1 (` − m)! ımϕ
Y`m (ϑ, ϕ) = √ P`m (cos ϑ) e . (4.32)
2π 2 (` + m)!
Now, we make the substitution R(r) = u(r)/r and the radial equation becomes,
" #
~2 ∂ 2 L̂2
− + + V (r) u(r) = Eu(r) . (4.34)
2m ∂r2 2mr2
L̂2
V` (r) ≡ . (4.35)
2mr2
For example, for the potential of an electron orbiting a proton, we have,
~2 ∂ 2 Ze2 ~2 `(` + 1)
− − + − E uE` (r) = 0 . (4.36)
2m ∂r2 4π0 r 2mr2
We will discuss this equation intensely in the context of the hydrogen atom.
In Exc. 4.1.5.2 we derive the radial Gross-Pitaevskii equation for a Bose-Einstein
condensate trapped in a spherical potential. In the Exc. 4.1.5.3 we will study particles
inside a central potential of zero depth, in the Excs. 4.1.5.4 and 4.1.5.5 we consider
3D spherical box potentials and in Exc. 4.1.5.6 a spherical harmonic potential.
V (r) 0 5 10
r/aB
Figure 4.2: (code) Sum of a Coulomb potential and centrifugal potential for ` = 0 (lower
curve), ` = 1 (center curve), and ` = 2 (upper curve).
to the radial motion and the potential, both being constant. In this case the
radial Schrödinger equation is,
~2 `(` + 1)
uE` = E` uE` .
2mr2
~2 `(` + 1)
E` = ,
2I
4.1.5 Exercises
4.1.5.1 Ex: Parity of the spherical harmonic functions
P
We consider the parity transformation P with (x, y, z) −→ (−x, −y, −z). Use spheri-
P
cal coordinates to show that Y`m −→ (−1)` Y`m , and therefore that a spherical surface
function has even parity when ` is even, and odd parity, when ` is odd.
where the factor g depends on the force of the interatomic interaction and Vtrp is the
potential trapping the atoms. Derive the stationary Gross-Pitaevskii equation via the
transform ψ(r, t) = ψ(r)e−ıµt/~ , where the constant µ is called the chemical potential.
For V (r) = V (r) the wavefunction will have radial symmetry, ψ(r) = φ(r) r . Rewrite
the Gross-Pitaevskii equation for the function φ.
150 CHAPTER 4. ROTATIONS / CENTRAL POTENTIALS
V = 12 mω 2 (x2 + y 2 + z 2 ) .
a. Obtain the energy levels of this particle. That is, determine the eigenvalues of
~2 2
− ∇ ψ + V ψ = Eψ .
2m
b. Consider the fundamental level and the first two excited levels. Set up a table
showing for each of these three levels the energy value, the degeneracy, and the re-
spective states in terms of the quantum numbers.
c. Using
2 1 ∂ 2 ∂ L2
∇ ψ= 2 r − 2 2 ψ
r ∂r ∂r ~ r
and remembering L̂2 Y`m (θ, φ) = ~2 `(` + 1)Y`m , write down the differential equation
of item (a) for the radial part of the wavefunction (it is not necessary to solve it).
Identify in this equation the effective potential Vef f (r).
d. Solve the differential equation of the previous item for the case where ` = 0 and
determine the corresponding eigenvalue. To do this, allow for a solution of the type
2
e−αr and determine α.
light negatively charged charge electronic cloud. Since the nucleus is very small com-
pared to the electronic cloud, we treat it as an entity with mass M and charge Ze,
where Z is the number of protons and corresponds to the order of the element in the
periodic system.
The canonical procedure for calculating all properties of an atom is to establish
its Hamiltonian, that is, to determine the kinetic energies of all components and all
interaction energies between them, and to solve the Schrödinger equation. For each
component we write the kinetic energy,
XZ
P2 p2i
Tncl = and Tele = . (4.37)
2M i=1
2m
Here, (R, P) are the nuclear coordinates and (ri , pi ) those of the electrons. The ener-
gies that corresponds to the interactions, that is, Coulombian attraction or repulsion,
between the components of the atom are,
Z
X Z
X
Ze2 e2
Vncl−ele = − and Vele−ele = . (4.38)
i=1
4πε0 |R − ri | 4πε0 |ri − rj |
i6=j=1
There are also interactions due to the spin of the particles, which we will deal with
later.
Obviously, the solution to this many-body problem is very complicated. For this
reason, we will in this chapter, based on the Schrödinger equation, calculate the
complete spectrum of the simplest possible atom, hydrogen. This atom consists of a
proton and an electron, only.
Figure 4.3: The hydrogen model applies to other atoms having a single valence electron
occupying a sufficiently large space, that it sees the nucleus together with rest of the electrons
shielding the nucleus as a single positive charge.
Paschen
Brackett
Ballmer
Lyman
abgd
The discussion of the hydrogen atom within quantum mechanics can start from
the radial Schrödinger equation (4.36) with the Coulomb attraction potential,
~2 ∂ 2 Ze2 ~2 `(` + 1)
− − + − E uE` (r) = 0 . (4.41)
2m ∂r2 4πε0 r 2mr2
In order to facilitate comparison with Bohr’s classical model, let us express the energy
in terms of Bohr’s energy, E ≡ En = E1 /n2 , and write the radius in units of aB , that
is, r̃ ≡ Zr/aB . This yields,
`(` + 1) 2 1
u00n` (r̃) + − + − un` (r̃) = 0 . (4.42)
r̃2 r̃ n2
To ensure that for large radii, r → ∞, the solution is finite, we need an asymptotic
behavior like un` (r̃ → ∞) = e−r̃/n . To ensure that for small radii, r → 0, the
solution is finite, we need un` (r̃ → 0) = r̃`+1 . We derive the asymptotic solutions in
Exc. 4.2.3.1. The resulting differential equation only has solutions for an integer and
positive main quantum number n and when ` = 0, 1, .., n − 1. That is, in the relation
E = E1 /n2 the parameter n is integer and positive, such that energy levels remain
degenerate in ` and m. This means that Bohr’s postulate of discrete (i.e. quantized)
energy levels is valid (uff!)
Substituting the ansatz,
it’s easy to show (see Exc. 4.2.3.2), that the differential equation (4.42) reduces to,
r̃L00 (r̃) + 2 (` + 1) − n1 r̃ L0 (r̃) + 2 1 − n1 (` + 1) L(r̃) = 0 . (4.44)
4.2. QUANTUM TREATMENT OF HYDROGEN 153
un` (r) ~2 Z 2
ψn`m (r, θ, φ) = Y`m (θ, φ) and En = − , (4.47)
r 2ma2B n2
(a) (b)
0.4 u10
[rRn (r)]2
0.04
Rn (r)
R10
0.3
0.2 0.02 u20
3/2
0.1
aB
u30
aB
R20 R30
0
0
0 10 20 0 10 20
r/aB r/aB
(c) (d) 0.03
0.1
[rRn (r)]2
Rn (r)
aB
0
0
0 10 20 0 10 20
r/aB r/aB
Figure 4.6: (code) (a,c) Radial wavefunctions Rn` (r) and (b,d) square of the radial wave-
functions un` (r) for various quantum numbers n and `.
times degenerate.
From the normalization condition,
Z
1= |ψn`m (r, θ, φ)|2 d3 r (4.49)
3
ZR∞ Z Z ∞
= |un` (r)|2 dr |Y`m (θ, φ)|2 sin θdθdφ = |un` (r)|2 dr ,
0 R2 0
s 3
2Z (n − ` − 1)!
Dn` = , (4.50)
naB 2n(n + `)!
`
2r̃ (2`+1) (2`+1)
Rn` (r) = Dn` e−r̃/n Ln−`−1 ( 2r̃ ` −ρ/2
n ) = Dn` ρ e Ln−`−1 (ρ) , (4.51)
n
3/2
ψ100 = √1
π
Z
aB e−r̃ (4.52)
3/2
ψ200 = √1
4 2π
Z
aB (2 − r̃)e−r̃/2
3/2
ψ210 = √1
4 2π
Z
aB r̃e−r̃/2 cos θ
3/2
ψ21±1 = 1
√
8 π
Z
aB r̃e−r̃/2 sin θe±ıϕ
3/2
ψ300 = 1
√
81 3π
Z
aB (27 − 18r̃ + 2r̃2 )e−r̃/3
√ 3/2
ψ31±1 = √2
81 3π
Z
aB (6 − r̃)r̃e−r̃/3 sin θe±ıϕ
3/2
ψ320 = 1
√
81 6π
Z
aB r̃2 e−r̃/3 (3 cos2 θ − 1) ,
where we use the abbreviation r̃ ≡ Zr/aB . Using these wavefunctions we can now
calculate important eigenvalues such as, for example,
h1in`m = 1 (4.53)
δ`,0
hδ (3) (r̃)in`m =
πn3
1 `(` + 1)
hr̃in`m = n2 1 + 1−
2 n2
3 `(` + 1) − 13
hr̃2 in`m = n4 1 + 1−
2 n2
35 35 15 3
hr̃3 in`m = n6 − 2 − 2 (` + 2)(` − 1) + 4 (` + 2)(` + 1)`(` − 1)
8 8n 4n 8n
63 35 5 12
hr̃4 in`m = n8 + 2 (2`2 + 2` − 3) + 4 5`(` + 1)(3`2 + 3` − 10) + 8
8 8n 8n n
1 1
= 2
r̃ n`m n
1 1
= 3
r̃2 n`m n (` + 12 )
1 n
= 4
r̃3 n`m n `(` + 12 )(` + 1)
3 2 1
1 2 n − 2 `(` + 1)
= .
r̃4 n`m n5 (` + 32 )(` + 1)(` + 21 )`(` − 21 )
These results will become important later. In Exc. 4.2.3.4 we will calculate the eigen-
value hri for several orbitals |ψn`m i.
156 CHAPTER 4. ROTATIONS / CENTRAL POTENTIALS
In Exc. 4.2.3.5 we calculate the eigenvalues hr−1 i and hp2 i and we verify the virial
theorem. Finally, in Exc. 4.2.3.6 we calculate transition matrix elements between
different orbitals.
4.2.3 Exercises
4.2.3.1 Ex: Asymptotes of Laguerre’s polynomials
Derive the asymptotic solutions of equation (4.42).
a. calculate the normalization constant Dn,l for a hydrogen-like atom with atomic
number Z;
b. calculate the mean value
n2 aB 1 `(` + 1)
hrinlm = 1+ 1− ;
Z 2 n2
ĵ × ĵ = ı~ĵ , (4.63)
or [ĵm , ĵn ] = ı~kmn ĵk using the Levi-Civita symbol, we obtain a consistent algebra.
Since ĵ2 and ĵz commute (we show this from Eq. (4.63) in Exc. 4.3.4.4), they have
common eigenfunctions |j, mi. We can write the eigenvalues as,
where, for now, we only know that m is real and j ≥ 0. But since hj, m|ĵ2 |j, mi ≥
hj, m|ĵz2 |j, mi, it is clear that j(j + 1) ≥ m2 .
[ĵz , ĵ± ] = ±~ĵ± and [ĵ2 , ĵ± ] = 0 and ĵ∓ ĵ± = ĵ2 − ĵz2 ∓ ~ĵz . (4.66)
ĵz ĵ± |j, mi = ([ĵz , ĵ± ] + ĵ± jz )|j, mi = ~(m ± 1)ĵ± |j, mi (4.67)
2 2 2
and ĵ j± |j, mi = ĵ± ĵ |j, mi = ~ j(j + 1)ĵ± |j, mi .
That is, ĵ± |j, mi is a eigenstate of ĵ2 and ĵz with the eigenvalues j and m ± 1,
respectively, if j± |j, mi =
6 0. Hence,
In order not to violate the condition m2 ≤ j(j + 1), we need to fix ĵ± |j, ±ji = 0.
Therefore, for a specified j, the m can have only one of the 2j + 1 possible values
m = −j, −j + 1, .., j. Since 2j + 1 is an integer, j can only have values j = 0, 12 , 1, 32 , ...
Thus, the eigenvalue equation of the observables ĵ2 , ĵ is solved since we could have
chosen instead of ĵz any one of the components of ĵ, knowing that the others do not
commute with the chosen one.
All spin components ĵz and the scalar ĵ2 can only have discrete eigenvalues. The
smallest unit is ~/2. With the normalization hj, m|j 0 , m0 i = δj,j 0 δm,m0 we have,
hj, m|ĵ∓ ĵ± |j, mi = hj, m|(ĵ2 − ĵz2 ∓ ~ĵz )|j, mi = ~2 [j(j + 1) − m(m ± 1)] , (4.69)
160 CHAPTER 4. ROTATIONS / CENTRAL POTENTIALS
and
p
ĵ± |j, mi = ~ j(j + 1) − m(m ± 1)|j, m ± 1i . (4.70)
In Exc. 4.3.4.5 we calculate the uncertainty of the angular momentum components,
in Exc. 4.3.4.6 we write the operator ĵx in a matrix form, and in Excs. 4.3.4.7 and
4.3.4.8 we calculate projections of the spin of the electron in different directions of
the quantization axis.
ŝ × ŝ = ı~ŝ , (4.71)
and
4.3.4 Exercises
4.3.4.1 Ex: Properties of the angular orbital momentum
Show that l̂ × l̂ = ı~l̂ and [ˆlx , ˆly ] = ı~ˆlz .
Initially (t = 0) the particle is in the state ψ(t = 0)i = |+i and is subject to a uniform
magnetic field B = Bêy , so that:
Ĥ = −~
µ · B = −γB Ŝy .
a. What are the possible measurements of the spin projection on the y-axis?
b. Find the eigenvectors of Ŝy .
c. Get |ψ(t)i at t > 0 in terms of the eigenvectors |+i and |−i defined above.
d. Obtain the mean expectation values of the observables Ŝx , Ŝy and Ŝz as a function
of time.
0 1 1 0 0 −1 −1 0 1 0 0 0
1 0 0 1
ı~ 1 0 0 −1 0 0 0 0
Ŝx = ~2
1
, Ŝy = , Ŝz = ~ .
0 0 1 2 1 0 0 −1 0 0 0 0
0 1 1 0 0 1 1 0 0 0 0 −1
The operator for the square of the absolute value of the total angular momentum is
calculated as follows:
2 0 0 0
0 1 1 0
Ŝ2 = Ŝx2 + Ŝy2 + Ŝz2 = ~2
0
. (4.76)
1 1 0
0 0 0 2
4.4. COUPLING OF ANGULAR MOMENTA 163
Now, we look for the eigenvalues of the total angular momentum. The equation for
the eigenvalues of Ŝz ,
Ŝz |γk i = MS |γk i , (4.77)
is already diagonal in the introduced basis {γk } with the eigenvalues,
MS = ~, 0, 0, −~ . (4.78)
For Ŝ2 the situation is more interesting: The states |γ1 i and |γ4 i are eigenstates of
S2 for the eigenvalue 2~2 , but the states |γ2 i and |γ3 i are not eigenstates. On the
other hand, we know that the linear combination of two eigenstates with the same
eigenvalue is also a eigenstate. Therefore, the states
|γ̃1 i ≡ |γ1 i , |γ̃4 i ≡ |γ4 i , |γ̃a i ≡ √1 (|γ2 i−|γ3 i) , |γ̃s i ≡ √1 (|γ2 i+|γ3 i) , (4.79)
2 2
are still eigenstates of Ŝz , but they also are eigenstates of Ŝ2 , since we can easily
verify,
Ŝ2 |γs i = 2~2 |γs i and Ŝ2 |γa i = 0~2 |γa i , (4.80)
using the matrices (4.75). In summary, for the eigenvalue hŜ2 i = 2~2 there exist the
following three states:
|γ1 i Ms = 1
|γ4 i Ms = −1 triplet , S = 1 (4.81)
|γs i Ms = 0
By exchanging the two electrons, the vectors |γ1 i and |γ4 i retain their shape,
while the mixed vectors change their shape: γ2 ←→ γ3 . Under particle exchange |γa i
reverses its sign, that is, it is antisymmetric, while |γ1 i, |γ4 i and |γc i conserve their
signs, that is, they are symmetrical.
In summary, the triplet states have the quantum number of the total angular
momentum (with the expected value for Ŝ2 of ~2 S(S + 1) = 2~2 ), and they are
symmetrical about the exchange of particles. The singlet state has the quantum
number of the total angular momentum S = 0, and it is antisymmetric about the
exchange of particles. The transition from the original basis {|γ1 i, |γ2 i, |γ3 i, |γ4 i}
to the new basis {|γ̃1 i, |γ̃s i, |γ̃a i, |γ̃4 i} described by Eqs. (4.79) is done by a unitary
transformation,
1
√1 √1
2 2
|γ̃i = UCGC |γi with UCGC = 1 1 , (4.83)
√
2
− 2√
1
In this case, the perturbation Hamiltonian does not commute with the total angular
momentum,
[Ŝ2 , Ĥ] 6= 0 . (4.86)
Another type of perturbation affects both spin symmetrically (e.g., Dicke superra-
diance with two atoms in the same radiative mode or two counterpropagating modes
in a ring cavity). The interaction Hamiltonian is now the sum of the individual
perturbations,
Ω∗ Ω∗
Ω Ω∗
Ĥ = Ĥ1 ⊗ I + I ⊗ Ĥ1 =
Ω
. (4.87)
Ω∗
Ω Ω
This Hamiltonian commutes with the total angular momentum,
S now is a good quantum number. Singlet states do not couple with triplets. This
is the idea behind Dicke’s superradiance. The absolute value of the total angular
momentum is conserved. The quantum number S is called Dicke cooperativity [203].
Ĥint = − q
m
A(r1 ) · (px êx σ̂ x ⊗ I + py êy σ̂ y ⊗ I) (4.90)
x y
− q
m
A(r2 ) · (px êx I ⊗ σ̂ + py êy I ⊗ σ̂ ) ,
4.4. COUPLING OF ANGULAR MOMENTA 165
The angular momenta of two particles or two angular momenta of different origins
in a single particle represent independent degrees of freedom, [j1 , j2 ] = 0. Without
interaction between angular momenta the Hilbert spaces are orthogonal:
H1 0
H1 ⊗ H2 = . (4.94)
0 H2
That is, there is a complete set of commuting operators {ĵ21 , ĵ1z , ĵ22 , ĵ2z }. Therefore,
we can specify quantum numbers j1 , j2 , mj1 , and mj2 simultaneously. On the other
hand, the group {ĵ21 , ĵ22 , ĵ2 , ĵz } also represents a complete set of commuting operators,
as shown in Exc. 4.4.5.2. It has the basis
To describe the two angular momenta simultaneously, we must opt between the
decoupled picture |j1 , mj1 ; j2 , mj1 i and the coupled picture |(j1 , j2 )j, mj i. For now,
the choice of the picture makes no difference, but we will see later that there may be an
energy associated with the coupling 3 . In this case, as we will show, the choice of the
coupled base is more natural, because the energy commutes like [Ĥ, ĵ2 ] = 0 = [Ĥ, ĵz ],
but [Ĥ, ĵ21 ] 6= 0 6= [Ĥ, ĵ22 ].
3 That is, the Hamiltonian of the system does not contain terms of type ĵ · ĵ , but may have
1 2
terms proportional to ĵ1 + ĵ2 .
4.4. COUPLING OF ANGULAR MOMENTA 167
Now we need to find the possible values of j and mj in the coupled picture. The
values of mj follow immediately from ĵ1 + ĵ2 = ĵ,
|mj | ≤ j1 + j2 . (4.99)
We often know the two angular momenta j1 and j2 and all their projections in the
decoupled base,
|mj1 | ≤ j1 and |mj2 | ≤ j2 . (4.100)
To find the quantum numbers in the coupled base, we arrange the states ordering
them by their total magnetic quantum number mj . We can, without loosing general-
ity, concentrate on the situation j1 ≥ j2 . The following table reproduces the possible
combinations. The x-symbols represent Clebsch-Gordan coefficients:
The possible values for j are all those allowing for j ≥ |mj | = |mj1 + mj2 |, that
is,
|j1 − j2 | ≤ j ≤ j1 + j2 . (4.101)
Each value of j has the degeneracy 2j +1. Therefore, as will be verified in Exc. 4.4.5.3,
the total degeneracy is,
jX
1 +j2
Figure 4.9: Possible spin states of two electrons occupying the (5p)2 orbital. Spin-orbit
coupling L · S leads to a splitting of the energy levels.
In the Excs. 4.4.5.4 and 4.4.5.5 we write all possible states of two angular mo-
menta in decoupled and coupled bases. In Excs. 4.4.5.6, 4.4.5.7, we derive the matrix
representation of two spins in the decoupled and the coupled base. In 4.4.5.8, and
4.4.5.9 we practice the transformation between decoupled and coupled bases, and
in Exc. 4.4.5.10 we verify a rule guaranteeing the unitarity of the Clebsch-Gordan
transformation. Finally in 4.4.5.11, 4.4.5.12, and 4.4.5.13 we study L · S-coupling.
4.4.4.3 jj-coupling
There is also the case that for each electron its spin couples to its own orbital angular
momentum, jk P = lk + sk , before coupling to the total angular momenta of other
electrons, J = k jk . This is called jj-coupling. In the case of two electrons the
recoupling of the four involved spins
l1 + l2 = L
+ + +
s1 + s2 = S (4.111)
= = =
j1 + j2 = J
170 CHAPTER 4. ROTATIONS / CENTRAL POTENTIALS
l1 l2 L
is described by {9j} = s1 s2 S -symbols,
j1 j2 J
X
|[(l1 , s1 )j1 , (l2 , s2 )j2 ]Ji = {9j}|[(l1 , l2 )L, (s1 , s2 )S]Ji . (4.112)
L,S
4.4.5 Exercises
4.4.5.1 Ex: Addition/subtraction of angular momenta
Show that ĵ1 + ĵ2 is an angular momentum, but not ĵ1 − ĵ2 .
Be ĵ = ĵ1 + ĵ2 . Show that {ĵ21 , ĵ22 , ĵ2 , ĵz } is a CSCO; that is, show that
a. ĵ2 commutes with ĵ21 and ĵ22 ;
b. ĵ2 does not commute with ĵ1z or ĵ2z and that we can not specify mj1 or mj2
together with j.
Find all possible states with the angular momenta j1 = 1 and j2 = 1/2 in decoupled
and coupled pictures.
85
4.4.5.5 Ex: Fine and hyperfine structure of the rubidium atom Rb
1. The rubidium atom 85 Rb has one valence electron. In the first excited state this
electron has the orbital angular momentum, L = 1. What are the possible states?
2. In the fundamental state of this atom the total electronic angular momentum J
couples with the spin of the nucleus, I = 5/2, to form the total angular momentum
F = J + I. Determine the possible values for the angular momentum F and the
magnetic quantum number mF .
Determine for the states S1/2 and P3/2 of an atom with nuclear spin I = 3/2 and
hyperfine coupling Ĵ · Î how the eigenstates of the coupled base expand into the
decoupled base. Do not consider external magnetic fields.
4.4. COUPLING OF ANGULAR MOMENTA 171
3/2 1 + cos θ θ
d 3/2,3/2 = cos
2 2
√ 1 + cos θ ³ 1 + cos θ ´2
3/2 θ d 22,2 =
d 3/2,1/2 = − 3 sin 2
2 2
√ 1 − cos θ θ 1 + cos θ
3/2
d 3/2,−1/2 = 3 cos d 22,1 = − sin θ
2 2 2
√ 1 + cos θ
3/2 1 − cos θ θ 6 d 21,1 = (2 cos θ − 1)
d 3/2,−3/2 = − sin d 22,0 = sin2 θ 2
2 2 4 r
3/2 3 cos θ − 1 θ 1 − cos θ 2 3
d 1/2,1/2 = cos d 22,−1 = − sin θ d 1,0 = − sin θ cos θ
2 2 2 2
3 cos θ + 1 θ ³ 1 − cos θ ´2 1 − cos θ ³3 1´
3/2
d 1/2,−1/2 = − sin d 22,−2 = d 21,−1 = (2 cos θ + 1) d 20,0 = cos2 θ −
2 2 2 2 2 2
Periodic systems
Many physical systems treat quantum particles in periodic potential. Examples are
electrons in crystals or cold atoms in optical lattices. The periodicity gives rise to a
wealth of new phenomena, such as Bragg scattering, the formation of energy bands,
Mott insulators, or Bloch oscillations. Various theoretical approaches have been de-
veloped leading to a deep understanding of the features of crystals, metals, and optical
lattices for atomic gases.
In the following sections we will mainly focus on the Bloch model, which is based
on introducing periodic boundary conditions to the Schrödinger equation. In the
Secs. 5.1 we will give a brief introduction into the Bloch model for electrons in solids,
where the periodicity is imposed by the crystalline structure of the material. We
will, however, no go into details leaving a comprehensive treatment to specialized
textbooks [26, 411]. In the Secs. 5.2 we will concentrate on optical lattices for atomic
gases formed by counterpropagating laser beams. The physical realization of these
lattices, which can be realized with nearly perfect periodicity, have permitted the
experimental observation of phenomena which had been elusive in solid state physics.
Since V and ∇ are invariant under translations Utr (R)ψ(r) ≡ ψ(r + R) by a fixed
distance R, where the translation operator has been defined in Eq. (2.171), we have,
175
176 CHAPTER 5. PERIODIC SYSTEMS
This relation holds for all vectors R of the lattice, such that,
The relationship f (R1 + R2 ) = f (R1 )f (R2 ) is satisfied by the ansatz f (R) ≡ eık·R ,
where k is an arbitrary vector of reciprocal space. We get the famous Bloch theorem,
automatically satisfies Bloch’s theorem. That is, the wave function of the electron
ψ is a plane wave eık·r modulated by a function uk having the same periodicity as
the lattice [418]. Although the vector of the electronic wave is arbitrary, it is possible
(and useful) to restrict its value to the first Brillouin zone defined by k ∈ [−π/a, π/a],
where a is an elementary vector of the lattice. The reason is that we can reduce a
wavevector k in a wavefunction trespassing the first Brillouin zone by an appropriate
vector G of reciprocal lattice,
k0 = k + G , (5.9)
yielding, 0
ψk (r) = uk (r)eık·r = uk (r)e−ıG·r eık ·r . (5.10)
We now define another function uk0 (r) ≡ uk (r)e−ıG·r , which also satisfies the require-
ment (5.8), knowing that G · R = n2π, we see,
Hence, 0
ψk (r) = uk0 (r)eık ·r = ψk0 (r) . (5.12)
That is, we neglect superposition states of the electron at various sites of the lattice.
The atom is subject to a potential Uat (r − Ri ) located near the atom at the position
1 This also holds true for degenerate eigenvalues if we choose suitable basis of eigenvectors.
5.1. THE BLOCH MODEL FOR ELECTRONS 177
Ri , and it is described by the eigenfunction φ(r − Ri ) (only defined for the site i)
with energy E0 ,
~2 2
− ∇ + Uat (r − Ri ) φ(r − Ri ) = E0 φ(r − Ri ) . (5.14)
2m
Even so, the function ψk (r) must satisfy Bloch’s theorem. This is the case when
ci (k) = eik·Ri and therefore,
X
ψk (r) = eık·Ri φ(r − Ri ) . (5.15)
i∈lattice
We now calculate the energy E(k) of an electron with the wavevector k inserting
the function ψk (r) of (5.15) in the Schrödinger equation and obtain,
X X
~2 2
− ∇ + U (r) eık·Ri φ(r − Ri ) = E(k) eık·Ri φ(r − Ri ) . (5.16)
2m i i
U (r) is the potential energy of the electron illustrated in Fig. 5.1 together with the
energy Uat (r − Ri ) of a free electron.
Substituting the kinetic energy term of (5.16) by the kinetic energy of (5.14), we
calculate,
X
eık·Ri [−Uat (r − Ri ) + E0 + U (r) − E(k)]φ(r − Ri ) = 0 . (5.17)
i
P
Now, multiplying this equation with ψk∗ (r) = j eık·Rj φ∗ (r − Rj ) and integrating
over the volume of the crystal, we obtain,
X Z
[E(k) − E0 ] eık·(Ri −Rj ) φ∗ (r − Rj )φ(r − Ri )dV (5.18)
i,j
X Z
= eık·(Ri −Rj ) φ∗ (r − Rj )[U (r) − Uat (r − Ri )]φ(r − Ri )dV .
i,j
The functions φ∗ (r − Rj ) and φ(r − Ri ) overlap only a little, even for adjacent atoms,
such that we can neglect the terms i 6= j on the left side. The sum then corresponds
to the number N of sites in the lattice. On the right side we can not neglect the
terms involving other sites, because even if the wavefunctions of adjacent sites overlap
little, the contribution of the potential difference |U (r) − Uat (r − Ri )| is much lower
for r = Ri than for r = Rj . On the other hand, as the wavefunctions φ(r − Ri )
178 CHAPTER 5. PERIODIC SYSTEMS
disappear quickly when |r − Ri | > |Rm − Ri |, we can focus on adjacent sites (called
Rm ),
Z
N [E(k) − E0 ] =N φ∗ (r − Ri )[U (r) − Uat (r − Ri )]φ(r − Ri )dV (5.19)
X Z
+N eık·(Ri −Rm ) φ∗ (r − Rm )[U (r) − Uat (r − Ri )]φ(r − Ri )dV .
m=adjacent
Figure 5.1: Potential energy U (r) of a crystal electron (red) and potential energy UA (r − Ri )
of the electron of a free atom (blue).
Now we further suppose that the eigenfunction φ exhibits radial symmetry corre-
sponding to an s orbital. We obtain for the eigenvalues from the Schrödinger equation,
X
E(k) = E0 − α − γ eık·(Ri −Rm ) (5.20)
m adjacent of i
Z
∗
with α= φ (r − Ri )[Uat (r − Ri ) − U (r)]φ(r − Ri )dV
Z
and γ= φ∗ (r − Rm )[Uat (r − Ri ) − U (r)]φ(r − Ri )dV .
Consequently, we can make for Bloch functions (5.10) the following periodic ansatz,
X
ψk (r) = uk (r)eık·r with uk (r) = √1V uG (k)eiG·r , (5.22)
c
G
Now multiplying with √1V eı(k+G)·r and integrating over the volume of the crystal
R c
(knowing V1c Vc eıG·r dV = δG,0 ), we obtain,
h 2 i X
2
2m (k + G) − E(k) uG (k) + UG−G0 uG0 (k) = 0 , (5.27)
~
G0
~2
[(k + G)2 − k 2 ]uG (k) − U0 uG (k) + U0 uG (k) + UG u0 (k) = 0 (5.28)
2m
UG u0 (k)
uG (k) = ~2 . (5.29)
2 2
2m [k − (k + G) ]
Since the Fourier coefficients UG have, for G 6= 0, small values, the function uG (k) is
not negligible only for k 2 ' (k + G)2 that is,
∗
Since the potential U (r) is real, U−G = UG . Therefore, introducing the eigenenergies
E0 (k) of free electrons (5.24),
Under the influence of the periodic perturbation potential we find at the surfaces of
a Brillouin zone an energy splitting developing a forbidden gap in the spectrum. We
can understand this observation as follows: In the crystal all electronic waves with
wavevectors ending on a surface of a Brillouin zone are reflected by Bragg reflection.
In the example of a one-dimensional lattice we understand that the superposition
of an incident wave (k = nπ/a) with the reflected one (k = −nπ/a) produces a
standing electronic probability density wave ρ being proportional to ρ1 ∝ cos2 nπ/a
or ρ2 ∝ sin2 nπ/a. The charge density ρ1 is maximal at the location of the atom
in this site, which corresponds to an increased interaction energy; the density ρ2 is
minimal at the location of the atom. This explains the splitting.
The Bloch model can explain many properties of metals, semiconductors and
insulators.
and we insert these expansions into Schrödinger’s stationary equation Ĥψ = εψ,
yielding, " #
−~2 ∂ 2 X X X
ıKz
2
+ UK e cq eıqz = ε cq eıqz . (5.37)
2m ∂z q q
K
In matrix notation,
Ĥc = εc . (5.39)
5.1. THE BLOCH MODEL FOR ELECTRONS 181
Figure 5.2: (a) Approximation of a periodic potential by a potential harmonic. (b) Dispersion
relation in momentum space.
2 2
At the edges of the Brillouin zone, k = 12 K, we get the eigenvalues ε = ~ mK −V0 ± V20 ,
that is, the band gap is ∆ε = V0 . Bloch’s theorem says that Schrödinger’s equation
can be solved for any Bloch states. These are superpositions of plane wave momentum
states [26],
ψk (z) = eıkz uk (z) , (5.44)
182 CHAPTER 5. PERIODIC SYSTEMS
Figure 5.3: (code) Bloch bands with coupling potential (continuous red line) and without
potential, V0 = 0 (dotted black line). The parameters are ωr = (2π) 20 kHz, ωho =
(2π) 12 kHz, λL = 689 nm, and V0 = 0.2~ωr .
dk
~ = −eE , (5.46)
dt
5.2. OPTICAL LATTICES 183
E = A cos ak , (5.49)
where a is the lattice parameter and A a constant. Then, v(k) is given by,
Aa
v(k) = − sin ak , (5.50)
~
and the position of the electron by,
Z
A aeE
x(t) = v(k(t))dt = − cos t . (5.51)
eE ~
This shows, that the electron is oscillating in real space. The oscillation frequency,
called Bloch frequency is given by,
ae|E|
ωblo = . (5.52)
~
where |cn (t)|2 are the population of momentum states separated by 2kl in momentum
space. Inserting this expansion and the potential (5.53), after removing the constant
energy V0 /2, into the stationary Schrödinger equation,
~2 ∂ 2 ψ V 0
Eψ = − + cos(2kl z)ψ , (5.55)
2m ∂z 2 2
we obtain, collecting terms oscillating with the same exponential e2ınkl z ,
2
q V0
En cn = 4~ωrec n + cn + (cn+1 + cn−1 ) , (5.56)
kl 4
~ωrec is the kinetic energy of an atom having received two units of photonic momenta.
5.2. OPTICAL LATTICES 185
~2 ∂ 2 V0
Ĥ = − 2
+ cos 2kẑ + mgẑ . (5.62)
2m ∂z 2
We define the transformation U ≡ e−ımgẑt/~ and find with (2.153) and (2.190) the
Hamiltonian in the accelerated frame,
p̃2 V0 2ıkẑ
H̃ = U † ĤU + ı~U̇ † U = + e + e−2ıkẑ . (5.63)
2m 4
with
p̃ ≡ p̂ − mgt = U † p̂U . (5.64)
Now, we expand the operators using the rules (??) and (??),
Z X
p̃ = p|pihp|dp = (p + 2n~k)|p + 2n~kihp + 2n~k| (5.65)
n
Z X
e2ıkẑ = |p + 2~kihp|dp = |p + 2(n + 1)~kihp + 2n~k| ,
n
based on the assumption that every atom can only exist in a superposition of discrete
momentum states separated by 2~k, but can have an ’offset’ momentum p = mv,
e.g. due to thermal motion. Disregarding thermal motion, we may set in the acceler-
ated frame p = 0. Then,
X X
p̃ = 2n~k|2n~kih2n~k| and e2ıkẑ = |2n~k + 2~kih2n~k| , (5.66)
n n
such that,
X X
H̃ = 4n2 ~ωrec |2n~kih2n~k|+ V40 (|2n~k + 2~kih2n~k| + |2n~kih2n~k + 2~k|) .
n n
(5.67)
3 For Bose-Einstein condensates, the procedure should be generalized taking into account the
P
With the expansion of the wavefunction |ψ̃i = n cn |2n~ki the Schrödinger equation
becomes [619, 620],
d X
ı~ |ψ̃i = ı~ ċn |2n~ki (5.68)
dt n
X
= 4n2 ~ωrec cn + V0
4 (cn−1 + cn+1 ) |2n~ki = H̃|ψ̃i ,
n
that is,
ċn = −4ın2 ωrec cn + V0
4ı~ (cn−1 + cn+1 ) . (5.69)
In the lab frame the wavefunction reads,
X
|ψi = U |ψ̃i = e−ımgẑt/~ cn |2n~ki (5.70)
n
X X
= cn |2n~k − mgti = cn |2~k(n − νblo t)i ,
n n
mg
νblo = . (5.71)
2~k
Finally,
ċn = −4ı(n − νblo t)2 ωrec cn + V0
4ı~ (cn−1 + cn+1 ) . (5.72)
hpilab X
= n|cn (t)|2 + νblo t . (5.73)
~kl n
The Bloch oscillations can be understood in various pictures. The first one, illus-
trated in 5.4(a), is based on Bragg reflection: A resting atom has infinite de Broglie
wavelength. Being constantly accelerated by gravity, the matter wave reduces its de
Broglie wavelength from ∞ to a value, where it becomes commensurate with the peri-
odicity of the standing light wave potential. At this moment Bragg scattering comes
into play, reflecting the atomic motion back into upward direction, and the process
starts over again. The atoms evolve like jumping on a trampoline with a frequency
given by νblo .
To understand the second picture, we need to address a question we have left aside
so far, as this requires a notion of optical forces. It is the question in which way the
matter wave interacts with the standing light wave. For the present discussion, it is
sufficient to know that the atom must have an internal transition capable of scattering
photons from the light beams. As any absorption and emission process transfers a
recoil momentum of ~kl to the atom, we can understand the Bragg scattering process
as a so-called Raman scattering process: a photon of the laser beam generating the
optical lattice coming from the left is absorbed and re-emitted to the left. This is best
illustrated in the momentum domain sketched in Fig. 5.4(b). This Raman scattering
transfers twice the photonic recoil to the atom. The requirement for commensurability
5.2. OPTICAL LATTICES 187
of the Broglie wavelength and wavelength of the standing light wave is equivalent to
saying that the matter wave momentum is equal to the recoil of a single photon. In
other words, the matter wave always Bragg-reflected at the edge of a Brillouin zone.
Finally, in the Bloch state picture, the dispersion relation of a free particle is
distorted due to the periodicity of the potential generated by the standing light wave
such as to open a forbidden band. As a consequence, instead of being accelerated
without limits, the atom enters the second Brillouin zone, which is to say that it is
reflected to the other side of the first Brillouin zone.
Figure 5.4: Illustration of Bloch oscillations (a) in real space, (b) in momentum space, and
(c) in the moving frame. You may also run a movie of Bloch oscillations clicking on (watch
movie)!
The additional term, which contains the frequency of the Bloch oscillation νblo ,
increases linearly over time. As time goes by, a resonance is crossed when t = −nτblo ,
and the crossing is periodically repeated at every n = −1, −2, 0, ... Tracing the
matter wave evolution in the laboratory system, we see that whenever the resonance
is crossed, the momentum undergoes a change of sign corresponding to a reflection
of its motion. We expand the population of the momentum states into plane (Bloch)
waves with |cn (t)|2 .
Of course there are some conditions that need to be met to observe Bloch oscil-
lations. The transfer of momentum is efficient only in the rapid adiabatic passage
(ARP) regime characterized by the conditions 2(νblo /ωrec ) (W0 /4ωrec )2 16.
The first condition requires that the force that drives the atoms to perform the Bloch
oscillations must be weak enough to avoid transitions between Bloch bands, which
guarantees the adiabaticity of the process. The other condition requires that the
optical lattice be weak enough so that the dynamics involves only two adjacent mo-
mentum states at the same time and the transfer between the two is successful. A
talk on this subject can be watched at (watch talk). Do the Exc. 5.2.3.1.
We will come back to the topic of optical lattices holding matter waves in Sec. 27.4.2
in the context of Mott insulating states of Bose-Einstein condensates.
Example 33 (Dicke model of Bloch oscillations in the two-mode ap-
proximation): In the limit where Bloch oscillations can be modeled by a two-
level system [620] we may try a representation within the Dicke model and
188 CHAPTER 5. PERIODIC SYSTEMS
(a) 1 (b) 1
0
lab frame
p/2h̄k
-1
|cn |2 0.5
-2 moving frame
-3
0 -4
0 2 4 0 2 4
νblo t νblo t
Figure 5.5: (code) Dynamics of Bloch’s oscillations calculated from Eq. (5.57). (a) The
colored curves show successively populated momentum states. (b) Center-of-mass motion of
the wave packet in the lab frame (red) and in the moving frame (blue). The used parameters
are ωrec = (2π) 4.8 kHz, νblo = 0.05ωrec , and W0 = −0.7ωrec .
4 Note that, In order to simulate a series of Bloch oscillations, we must change the basis by hand
5.2.3 Exercises
5.2.3.1 Ex: Purification of Bloch oscillations
(a) 1 (b)
0 lab frame
p/2h̄k
|cn |2
0.5
-2 moving frame
0 -4
(c) 1 (d)
p/2h̄k 0 lab frame
|cn |2
0.5
-2 moving frame
0 -4
(e) 1 (f)
0 lab frame
p/2h̄k
|cn |2
0.5
-2 moving frame
0 -4
0 2 4 0 2 4
νblo t νblo t
Figure 5.6: (code) Dynamics of Bloch oscillations (a-b) for W0 = 0.4ωrec , (c-d) for W0 =
0.8ωrec , and (e-f) for W0 = 1.6ωrec . Furthermore, ωrec = (2π) 4.8 kHz and νblo = 0.05ωrec .
Consider a standing light wave producing a dipolar potential of the shape V (x) =
W0 sin2 kx = W20 − W20 cos 2kx with W0 Erec = (~k)2 /2m.
a. Approximate the potential of a single lattice site around x = 0 by a harmonic
potential. Calculate for which depth of the dipolar potential the approximated po-
tential supports at least one bound state. What is the spacing of the levels and the
characteristic size of the harmonic oscillator aho . Write down the normalized ground
state wavefunction of the harmonic oscillator.
b. Now, let us consider a 3D cubic lattice made of three identical orthogonal standing
light waves. Approximate the ground state Wannier function of a lattice site by ap-
propriately normalized ground state wavefunctions of a harmonic oscillator. Derive
190 CHAPTER 5. PERIODIC SYSTEMS
3/4
8 W0
the formula U 3d = π kad Erec starting from,
Z
U 3D = g 3D 4
ω(0,0) (x, y, z)d3 r
4π~ad
with g 3D = m .
Inserting into the Schrödinger equation the plane wave ansatz ψ = AeıKx + Be−ıKx
for the wavefunction in the valley, 0 < x < a, and ψ = CeQx + De−Qx in the
peak, −b < x < 0, we obtain ε = ~2 K 2 /2m and U0 − ε = ~2 Q2 /2m. Choosing the
constants A, B, C, D such that ψ and ψ̇ are continuous in x = 0, a, we derive, using
the periodicity of the Bloch wave ψ(a < x < a + b) = ψ(−b < x < 0)eık(a+b) ,
1 1 −1 −1 A
ıK −ıK −Q Q B
= 0 . (5.75)
eıKa e−ıKa −e−Qb+ık(a+b) −eQb+ık(a+b) C
ıKeıKa −ıKe−ıKa −Qe−Qb+ık(a+b) QeQb+ık(a+b) D
Q2 − K 2
sinh Qb sin Ka + cosh Qb cos Ka = cos k(a + b) . (5.76)
2QK
5.3. THE KRONIG-PENNEY MODEL 191
For δ-shaped peaks, we let b = 0 and U0 = ∞ such that Q2 ba/2 = P , this simplifies
to,
P
sin Ka + cos Ka = cos ka . (5.77)
Ka
In one dimension,
L L
dN = 2 dk = dω . (5.80)
2π πc
R πc/L
The density is normalized 0 D(ω)dω = 1 and the total energy, if all states are
R πc/L
populated, is E ≡ 0 ωD(ω)dω = πc/L. However, this applies only if ω = ck. If
the dispersion relation is more complicated, for example, inside a cavity or a forbidden
photonic band, ω = ω(k), we must generalize,
L dk
D(ω) = . (5.81)
π dω
Assuming that the dispersion relation is given by the Kronig-Penney model, we obtain
gaps in the density-of-states for those values of ω which do not belong to any k.
ω0
ω0
Re θ ρ(ω)
Figure 5.7: Dispersion ratio and state density for a one-dimensional optical lattice.
192 CHAPTER 5. PERIODIC SYSTEMS
5.3.2 Exercises
5.3.2.1 Ex: Mass of photons in a lattice
The effective mass of particles in a lattice is defined as,
1 1 d2 E(k)
= .
m∗ ~2 dk 2
Calculate the effective mass of a rubidium atom in a far-detuned optical standing
wave.
Approximation methods
Virtually every problem going beyond the potential well, the harmonic oscillator, or
the hydrogen atom without spin and external fields is impossible to solve analytically.
In this chapter we will talk about techniques to solve approximately problems in more
realistic situations. There are a number of methods of which we will discuss the only
following: 1. The stationary or time-dependent perturbation method is useful for
evaluating small perturbations of the system, for example, caused by external electric
or magnetic fields; 2. the variational method, which serves to find and improve trial
wavefunctions, the initial shapes of which are generally motivated by the symmetries
of the system; 3. the semi-classical WKB method; 4. and finally the method of self-
consistent fields, which is an iterative method of solving the Schrödinger equation.
The contributions ∝ λn are the corrections of order n. The equation we need to solve
now is,
Ĥ|ψi = E|ψi . (6.5)
193
194 CHAPTER 6. APPROXIMATION METHODS
By inserting all the expansions above and segregating all orders of λk , we find the
following system of equations,
We insert this expansion into the second equation (6.6) and multiply with hn|,
X
hn|(Ĥ (0) − En(0) ) |mihm|ψn(1) i = 0 = hn|En(1) − Ĥ (1) |ni . (6.8)
m
We obtain for the first order correction of the energy of unperturbed states,
As a first example we will calculate in Exc. 6.1.3.1 the first order correction for
the energy of a slightly deformed one-dimensional box potential.
This procedure simulates the distortion of the state by blending it with other states.
The perturbation induces virtual transitions to other states. The perturbation is large
when the blended levels are close.
6.1. STATIONARY PERTURBATIONS 195
See Exc. 6.1.3.2. In Exc. 6.1.3.3 we calculate the first order correction due to the
finite extension of the hydrogen nucleus. In Exc. 6.1.3.4 we treat the coupling of the
energy levels of a two-level system as a first order perturbation, and compare the
result with the exact solution. The Stark effect for an electron confined in a box can
be discussed (see Exc. 6.1.3.5) in first order TIPT.
import it into the third equation (6.6) and multiply with hn|,
X X
hn|(Ĥ (0) −En(0) ) |mihm|ψn(2) i = hn|(En(2) −Ĥ (2) )|ni+hn|(En(1) −Ĥ (1) ) |mihm|ψn(1) i .
m m
(6.14)
Now,
X X
hm|ψn(2) i(En(0) −Em
(0)
)δnm = 0 = En(2) −hn|Ĥ (2) |ni+ hm|ψn(1) i En(1) δnm − hn|Ĥ (1) |mi .
m m
(6.15)
The left-hand side of this equation disappears. Also, on the right-hand side, for
(1)
n 6= m, the term En δnm disappears, and for n = m the whole parenthesis disappears.
(1)
Therefore, we can discard the term En and restrict the sum to terms with n 6= m.
(1)
Inserting the coefficients hm|ψn i calculated in (6.11), we finally obtain,
The first term is similar to the first order correction; the eigenvalue of the second
order perturbation calculated in the base of the unperturbed states. The second term
describes the shift of the energies through possible temporary transitions to other
states.
In Exc. 6.1.3.6 we treat a system of three coupled levels up to the second pertur-
bative order. The Stark effect discussed in Exc. 6.1.3.7 needs the TIPT calculation
up to the second order.
(0)
When the perturbation Ĥ (1) is applied, we assume that the state |ψnµ i is distorted
(0)
towards the similar state |ψnµ i, and the energy changes from En to Enµ . We now
need the index µ to label the energy, since the degeneracy can be removed by the
perturbation. As before, we write now,
Ĥ = Ĥ (0) + λĤ (1) + .. (6.19)
(0) (1)
|ψnµ i = |ψnµ i + λ|ψnµ i + ..
Enµ = En(0) + λEnµ
(1)
+ .. .
The replacement of these expansions in Ĥ|ψnµ i = Enµ |ψnµ i, and a collection of the
terms in λ up to first order gives,
Ĥ (0) |ψnµ
(0)
i = En(0) |ψnµ
(0)
i (6.20)
(En(0) − Ĥ (0) (1)
)|ψnµ i = (1)
(Enµ − Ĥ (1) (0)
)|ψnµ i .
As before, we try to express the first-order corrections for the wavefunctions
(0)
through degenerate unperturbed wavefunctions |ψnµ i and non-degenerate wavefunc-
(0)
tions 2 |ψm i: X X
(1) (0) (0)
|ψnµ i= bµν |ψnν i+ anm |ψm i. (6.21)
ν m
Inserting this into the first-order equation (6.20), we obtain,
X X
bµν (En(0) − En(0) )|ψnν
(0)
i+ (0)
anm (Em − En(0) )|ψm
(0) (1)
i = (Enµ − Ĥ (1) )|ψnµ
(0)
i . (6.22)
ν m
and multiplying the two sides with hn, µ|, we get zero on the left-hand side, since we
can choose the non-degenerate states to be orthogonal hn, ν|mi = δm,n . Hence,
X h i
(1)
cµν Enµ hn, µ|n, νi − hn, µ|Ĥ (1) |n, νi = 0 . (6.24)
ν
This secular equation (one for each µ) represents, in fact, a set of r linear equations
for the coefficients cµν . The condition for having non-trivial solutions is,
det hn, ν|Ĥ (1) |n, µi − Enµ
(1)
δµ,ν =0. (6.25)
µ,ν
(1)
The solution of this secular determinant yields the solicited energies Eµ . Now, the so-
lution of the secular equation (6.24) for each energy value produces those coefficients,
which represent the best linear combinations adapted to the perturbation. Unlike in
previous calculations with degenerate states, here we consider linear combinations of
vectors of the degenerate subspace prior to switching on the perturbation.
In practice, we apply perturbation theory only to the lowest relevant order. That
is, we only calculate the second order correction if first order corrections vanish. One
famous example is the quadratic Stark effect discussed in Sec. 9.3. In the case of
eigenvalues, which are degenerate in the absence of perturbation, the first order will
always produce a remarkable correction, as in the example of the linear Stark effect,
also discussed in Sec. 9.3. For this reason, we need not discuss higher perturbation
orders in the case of degenerate eigenvalues.
That is, the correction of the energy vanishes in first order, while the correction
of the wavefunction diverges. Obviously, the |νi obtained by the diagonalization
of the matrix Ĥ (0) is not adapted to the calculation of the matrix elements Ĥ (1) .
Now, applying degenerate perturbation theory, we obtain by the secular deter-
minant,
!
(1)
h
(1)
i −Eµ Ω
0 = det hν|Ĥ |µi − Eµ(1) δµ,ν = det (1) = (Eµ(1) )2 − Ω2 ,
Ω −Eµ
(1) (1)
eigenvalues are E1 = Ω and E2 = −Ω allowing the establishment of the
secular equation,
h i
(1)
c11 E1 − h1|Ĥ (1) |1i − c12 h1|Ĥ (1) |2i = c11 [Ω − 0] − c12 Ω = 0
h i
(1)
−c21 h2|Ĥ (1) |1i + c22 E2 − h2|Ĥ (1) |2i = −c21 Ω + c22 [−Ω − 0] = 0 .
coincides with the exact calculation made at the beginning. The eigenfunctions
(0)
|ψ1 i should be already corrected in zeroth order, which we verify by calculating,
(0) (0)
(1) (0) hψ1 |Ĥ (1) |ψ2 i (1)
|ψ1 i = |ψ1 i = 0 = |ψ2 i .
E1 − E2
In Exc. 6.1.3.8 we study a partially degenerate three-level system and the break-
down of the degeneracy due to a perturbation. And in Exc. 6.1.3.9 we will treat a
perturbation in a box potential with degenerate energy levels.
6.1.3 Exercises
6.1.3.1 Ex: One-dimensional well with a deformation in the centre
Consider a one-dimensional potential well between −L/2 and L/2 with infinitely high
walls. In the center of the well is a small deformation,
ε for −a
2 ≤x≤ 2
a
H (1) =
0 outside that region .
Calculate the correction for the eigenenergies in first order and discuss the limits
a L and a → L.
6.1. STATIONARY PERTURBATIONS 199
The expression V (r) = −e2 /4π0 r for the potential energy of an electron in the hy-
drogen atom implies that the nucleus (the proton) is treated as a point particle. Now
suppose that, on the contrary, the charge of the proton +e is evenly distributed over
a sphere of radius R = 10−13 cm.
a. Derive the modified potential Vm , which corresponds to this distribution of the
nuclear charge.
b. Assume that the wavefunction of the hydrogen atom does not change much due
to the modified potential. Calculate in lowest order in R/aB the average energetic
displacement h∆V i for the state (n = 1, ` = 0, m = 0). How will the energy displace-
ment be in comparison to the states (n = 2, ` = 0, m = 0) and (n = 2, ` = 1, m = 0)?
c. Calculate in the same way h∆V i for muonic hydrogen in the ground state.
We consider a two-level system. Without perturbation the system would have the
(0) (0)
Hamiltonian H (0) , the eigenenergies E1,2 and the eigenfunctions ψ1,2 . Now we switch
on a stationary perturbation of the form H (1) = (|1ih2| + |2ih1|).
a. Calculate the eigenenergies directly solving the perturbed Schrödinger equation.
b. Calculate the perturbed energies using TIPT and compare to the exact calculation
of the eigenenergies.
c. Calculate the eigenstates directly solving the perturbed Schrödinger equation.
d. Calculate the perturbed states using TPIT and compare to the exact calculation
of the eigenfunctions.
Consider an electron in a one-dimensional box, that is, in a well inside the interval
x ∈ [0, a] delimited by infinite walls. When a uniform electric field E is applied, also
in x-direction, the electron experiences a force equal to −eE, being −e the electron
charge, so that the potential energy inside the box becomes eEx.
a. What is the energy of the ground state of the electron (in first order approximation)?
We can assume that eEa is much smaller than the ground state energy the electron
would have in the absence of electric fields.
b. Use first-order TIPT to get an approximation for the ground state wavefunction
by calculating the first term of the correction.
Calculate the corrections for the eigenvalues and eigenfunctions up to first order.
ground state is not degenerate while the first excited state is three times degenerate.
Consider that the particle in this box is subject to a perturbation of the shape H (1) =
V0 for 0 < x < a/2 and 0 < y < a/2 and H (1) = 0 outside this region.
a. Obtain the first-order correction for the ground state energy.
b. Obtain the first-order correction for the (degenerate) energy of the first excited
state, in addition to the optimal base (which follows from the linear combinations of
degenerate states) which most closely approximates the perturbed states.
4 See [146], Complement A XI.
6.2. VARIATIONAL METHOD 201
hψ|Ĥ|ψi
Eg ≤ ≡E , (6.26)
hψ|ψi
not only when ψ is the wavefunction of an excited state, but even when it represents
a (imperfect) trial to the ground state. Assuming normalized wavefunctions we can
discard the denominator hψ|ψi = 1. To verify the theorem,
P we expand the function ψ
into orthonormal (unknown) eigenfunctions, |ψi = n cn |ψn i. Since ψ is normalized,
X X
1 = hψ|ψi = hψm |c∗m cn |ψn i = |cn |2 . (6.27)
m,n n
As the ground state is that of the lowest energy, Eg ≤ En , we have demonstrated the
relationship (6.26)
X X
Eg = Eg |cn |2 ≤ En |cn |2 = hĤi . (6.29)
n n
In practice, the ansatz ψα for the ground state allows us to calculate an energy
that must be minimized via
In the Excs. 6.2.3.1 and 6.2.3.2 we will approach the fundamental state of a quartic
potential and a harmonic oscillator, respectively, by trying several trial wavefunctions
and optimizing their free parameters.
using the definition of E (6.31). For the expression ∂E/∂c∗q we get analogous results.
Hence,
X
0= cm (hq|Ĥ|mi − Ehq|mi) . (6.33)
m
The condition for the existence of solutions is that the secular determinant disappears,
0 = det(hq|Ĥ|mi − Ehq|mi) . (6.34)
The solution of this equation leads to a set of values E, and the lowest value, Emin ,
is the best approximation for the ground state energy. The coefficients of the wave-
function are obtained by solving the eigenvalue equation (6.33) with Emin .
In Exc. 6.2.3.3 we will use the Rayleigh-Ritz method to estimate the effect of a
finite nuclear mass of the hydrogen atom on the energy levels. In Exc. 6.2.3.4 we will
use the Rayleigh-Ritz method to find the maximum number of atoms allowing for a
stable Bose-Einstein condensate made of atoms subject to an attractive interatomic
force.
6.2.3 Exercises
6.2.3.1 Ex: Variational method applied to a quartic potential
Determine the ground state energy of the quartic potential V (x) = bx4 making the
2
variational ansatz ψα (x) = (α/π)1/4 e−αx /2 . Formulae:
Z ∞ Z ∞ Z ∞
2 √ 2 √ 2 √
e−x dx = π , x2 e−x dx = 21 π , x4 e−x dx = 34 π
−∞ −∞ −∞
6.3. WKB APPROXIMATION 203
~2 d2
− ψ(x) + V (x)ψ(x) = Eψ(x) , (6.35)
2m dx2
and rewrite it as follows,
d2 ψ
= k(x)ψ(x) . (6.36)
dx2
204 CHAPTER 6. APPROXIMATION METHODS
p
with k(x) = 2m[E − V (x)]/~2 . For now, we will restrict ourselves to energies
E > V (x). In this scheme, the wavefunctions are usually complex functions, so that
we can write them in polar coordinates, containing an amplitude A(x) and a phase
φ(x), which are both real numbers:
Substituting this function into the Schrödinger equation we obtain a system of coupled
equations in terms of A(x) and φ(x),
h i
A00 = A (φ0 )2 − k 2 and (A2 φ0 )0 = 0 . (6.38)
The equations (6.39) and (6.37) are completely equivalent to the Schrödinger equation.
The second Eq. (6.38) is easy to solve,
C
A= √ 0 , (6.39)
φ
being C a real constant. We can not say the same thing about the solution of the first
Eq. (6.38). In order to solve it we are going to use the WKB approach, assuming that
A varies slowly, so the term A00 → 0. By doing this approximation we can rewrite
Eq. (6.39) in this way:
(φ0 )2 = k 2 . (6.40)
Solving this last expression we obtain two linearly independent solutions, φ0 = ±k.
So we get the expression for the phase:
Z
φ(x) = ± k(x)dx . (6.41)
We write this indefinite integral, because the constant term can be absorbed by the
constant C. Finally, we obtain the expression for the wavefunction in the WKB
approximation:
C R
ψ(x) = p e± |k(x)|dx . (6.42)
|k(x)|
Here, taking the absolute value of the wavevector, we have already generalized for the
case that the energy E of the particle is lower than the potential V (x) (classically
forbidden region).
ψ(x) = AeiS(x)/~ ,
The WKB approximation can be used to describe continuous potentials (or barri-
ers) by stepwise constant potentials. The transmission |T |2 through these parts can
be obtained by multiplying the individual tunneling probabilities,
Z
2
ln |T | ' −2 κ(x)dx ,
barrier
1
p
with κ(x) = ~ 2m[V (x) − E].
Figure 6.1: (a) Potential for which we want to obtain the connection formula. (b) Turning
point.
Let us start with the right turning point Fig. 6.1(a)]. First, we shift the coordinate
system so that the turning point coincides with zero, as shown in Fig. 6.1(b). As seen
above, the WKB solutions will be given by the following equations:
h R0 0 0
R0 0 0
i
√1 Bei x k(x )dx + Ce−i x k(x )dx if x≤0
k(x)
ψ(x) ≈ Rx 0 0 . (6.43)
√ 1 De− 0 |k(x )|dx if x≥0
k(x)
206 CHAPTER 6. APPROXIMATION METHODS
In the vicinity of the turning point we approximate the potential by a straight line
(Taylor series expansion up to first order) with the following functional dependence,
V (x) ≈ E + V 0 (0)x . (6.44)
The Schrödinger equation for this potential acquires the following format,
d2 ψt
= α3 xψt , (6.45)
dx2
with α = [ 2m 0
~2 V (0)]
1/3
. Through a change of variables, z = αx, we fall back on Airy’s
equation,
d2 ψt
= zψt , (6.46)
dz 2
having as solution a linear combination of the two solutions of the Airy equation,
ψt (x) = aAi(αx) + bBi(αx) . (6.47)
Now let’s have a look at the WKB solutions in the two regions in the √ vicinity of
3/2
the turning
Rx point. In the classically forbidden region we have k(x) = α −x, thus
being 0 |k(x0 )|dx0 = 32 (αx)3/2 . Thus, the WKB solution in the classically forbidden
region near the turning point will be given by:
D 2 3/2
ψ(x) ≈ √ e− 3 (αx) . (6.48)
~α3/4 x1/4
Using the asymptotic forms of Airy functions in the solution (6.47) we obtain the
following expression for ψt (x),
a 2 3/2 b 2 3/2
ψp (x) ≈ √ e− 3 (αx) + √ e 3 (αx) , (6.49)
2 π(αx)1/4 π(αx)1/4
q
4π
which when compared to equation (6.48) shows us that a = α~ D and b = 0.
Repeating the previous steps in the negative region we see that the WKB solution in
the asymptotic forms of the Airy solutions for approximately linear potentials takes
the following format (with b = 0):
1 h 2 3/2 2 3/2
i
ψ(x) ≈ √ Bei 3 (−αx) + Ce−i 3 (−αx) , (6.50)
~α3/4 (−x)1/4
and
a 1 h iπ/4 i 2 (−αx)3/2 −iπ/4 −i 23 (−αx)3/2
i
ψp (x) ≈ √ e e 3 − e e . (6.51)
π(−αx)1/4 2i
a
When compared, 2i√ π
eiπ/4 = √B~α and − 2i√
a
π
e−iπ/4 = √C~α . Having all this infor-
mation we can rewrite the WKB solutions for all positions in the potential, including
the turning points 5 :
hR i
√2D sin xx2 k(x0 )dx0 + π4 , if x ≤ x2
k(x)
ψ(x) ≈ Rx 0 0 . (6.52)
√D e− x2 |k(x )|dx , if x ≥ x2
k(x)
Repeating the process for a decreasing turning point [left turning point of the
potential of Fig. 6.1(a)], we obtain the following expression:
D0 Rx 0 0
√ e− x |k(x )|dx ,
1
if x ≤ x1
k(x) hR i
ψ(x) ≈ . (6.53)
√2D sin x k(x0 )dx0 + π , if x ≥ x1
0
k(x) x1 4
and hence the zeros of these functions, so the arguments of those sines must be
equal (except for a multiple of π),
Z x2 Z x1
π π
kdx0 + = − kdx0 − + nπ (6.54)
x 4 x1 4
Z x Z x2 1
+ kdx0 = n − π
x1 x 2
Z x2 1
kdx0 = n − π,
x1 2
with n = 1, 2... 6 . With this information we take a harmonic potential of the type
V (x) = 21 κx2 . In this case, the turning points for a given energy E will be at
q q q
− 2E κ
and 2E κ
. For this potential we will have that k(x) = 2m~2
(E − 21 κx2 ).
Calculating the integral of k(x) between these two turning points we get,
Z x2 Z √2E/k r
2m 1 m 1
k(x)dx = ~ √ E − κx2 dx = πE = n− π~ ,
x1 − 2E/k 2 κ 2
p
isolating E and taking ω = κ/m we have,
1
E = n− ~ω ,
2
with n = 1, 2..., the exact spectrum of the harmonic oscillator, but this is just a
coincidence.
Eigenstates: Now we will calculate the eigenstates of the harmonic oscil-
lator. The eigenfunctions were calculated on a computer. The first graph
(Fig. 6.2) compares the first exact excited state with that obtained using the
WKB method.
Note that the WKB approach is very good when x → 0 and x → ∞, regions
where the difference between the oscillator energy and the potential are large
(E V (x → 0) and E V (x → ∞)), because in these regions p the wave-
length λ(x) acquires the lowest values, since it is proportional to |1/ E − V (x)|.
6 Note that n 6= 0, because the integral (6.54) has to be greater than zero.
208 CHAPTER 6. APPROXIMATION METHODS
Figure 6.2: First excited state calculated accurately and through the WKB approximation.
Hence, the spatial region in which the potential needs to be practically constant
is smaller, which explains why the approximation is closer to the exact solu-
tion. In the intermediate regions the difference between E and V (x) begins to
decrease, and the WKB approximation delivers its worst results.
As we increase the energy of the harmonic oscillator, the approximation be-
comes better (for the same reason as discussed in the previous paragraph). The
following graph illustrates this effect for n = 10.
Figure 6.3: Wavefunction of the vibrational state n = 10 calculated exactly and using the
WKB approximation.
Notice that E < 0. Let us make the following substitution to facilitate algebraic
manipulations,
e2 ~2
B=− and C=− l(l + 1) .
4π0 E 2mE
The turning points r1 and r2 are given by the following expressions,
√ √
B − B 2 − 4C B + B 2 − 4C
r1 = and r2 = .
2 2
Thus, returning to the integral we will have the following:
√ √
Z r2 Z r2 p
1 π √ √
p(r)dr = −2mE (r − r1 )(r2 − r)dr = −2mE ( r2 − r1 )2
r1 r1 r 2
√ π √ π √
= −2mE (r1 + r2 − 2r1 r2 ) = −2mE (B − 2 C)
2 2
√
π e2 2m p 1
= − √ − 2~ l(l + 1) = n − π~ .
2 4π0 −E 2
Isolating E we obtain the energy spectrum of the hydrogen atom in the WKB
approximation:
m e2 2 1 13.6
E=− i2 = − h i2 eV .
2~ 4π0
h
n − 1/2 + l(l + 1) n − 1/2 + l(l + 1)
6.3.3 Exercises
6.3.3.1 Ex: Energy levels of hydrogen via WKB
Use the WKB approach to calculate the energy levels of the hydrogen atom.
Recalling that this stationary Schrödinger equation was obtained from the time-
dependent Schrödinger equation via a separation ansatz (2.56), the temporal evolution
of these eigenfunctions is given by,
be good. Note that not only do eigenfunctions oscillate, but the coefficients also
depend on time, because the composition of the states can change. The instantaneous
probability of finding the system in state n is |an (t)|2 . Importing the above linear
combination into the Schrödinger equation,
h i ∂
Ĥ (0) + Ĥ (1) (t) |ψ (1) (t)i = ı~ |ψ (1) (t)i , (6.59)
∂t
we find,
(0) (0) (0) (0)
a1 Ĥ (0) |ψ1 i + a2 Ĥ (0) |ψ2 i + a1 Ĥ (1) |ψ1 i + a2 Ĥ (1) |ψ2 i
" #
(0) (0)
∂a1 (0) ∂a2 (0) ∂|ψ1 i ∂|ψ2 i
= ı~ |ψ i + |ψ i + a1 + a2 (6.60)
∂t 1 ∂t 2 ∂t ∂t
(0) (0) (0) (0)
=⇒ a1 Ĥ (1) |ψ1 i + a2 Ĥ (1) |ψ2 i = ı~ȧ1 |ψ1 i + ı~ȧ2 |ψ2 i ,
because the other terms satisfy the Schrödinger equation of zero order. Replacing the
stationary eigenfunctions,
a1 e−ıE1 t/~ Ĥ (1) |1i + a2 e−ıE2 t/~ Ĥ (1) |2i = ı~ȧ1 e−ıE1 t/~ |1i + ı~ȧ2 e−ıE2 t/~ |2i , (6.61)
and multiplying this equation with h1|× and h2|×, we find with the abbreviation
~ω0 ≡ E2 − E1 ,
Frequently, the perturbation induces only a coupling, but does not directly influence
the energies very much, hn|Ĥ (1) |ni ' 0,
e−ıω0 t eıω0 t
ȧ1 = a2 h1|Ĥ (1) |2i and ȧ2 = a1 h2|Ĥ (1) |1i . (6.63)
ı~ ı~
∂
ı~ |ψ(t)i = Ĥ(t)|ψ(t)i . (6.65)
∂t
(0)
Now, we do a unitary transformation into the interaction picture with S(t) = e−ıĤ t/~
substituting |ψ(t)i ≡ S(t)|ψI (t)i and Ĥ (1) (t) ≡ S(t)Ŵ (t)S −1 (t) in the Schrödinger
equation. This procedure removes the stationary part, as shown in Sec. 2.4.4,
∂
i~ |ψI (t)i = λŴ (t)|ψI (t)i . (6.66)
∂t
If Ŵ (t) were also independent of time, the solution would simply be |ψI (t)i =
e−ıŴ t/~ |ψI (0)i. Otherwise, we integrate the equation,
Z t
λ
|ψI (t)i = |ψI (0)i + Ŵ (τ )|ψI (τ )idτ . (6.67)
ı~ 0
This is called the Dyson series. For N = 1, we get the first order of the perturbation
series 9 ,
Z
λ t
|ψI (t)i = 1 + Ŵ (τ )dτ |ψI (0)i . (6.69)
ı~ 0
The stationary states of the unperturbed Hamiltonian are given byPĤ (0) |f i =
Ef |f i. Now, the perturbed states are expanded on this basis, |ψI (t)i = f |f iaf (t).
7 See Becker-Sauter II, p.118ff and [699], p.104ff. An alternative treatment is found in [456], p.191ff
or in Blochinzew, p.332ff.
8 Note that by substituting Ŵ by Ĥ (1) , the equation (6.66), i~∂ |ψ (t)i = Ĥ (1) (t)|ψ (t)i, cor-
t I I
responds to a first-order perturbative approximation, i.e., the perturbation eigenvalues Ĥ(1) are
calculated with the eigenvectors of the unperturbed system. Thus, in first order TDPT we can
substitute Ŵ for Ĥ (1) .
9 For higher orders,
X n Z t n Z t
N λ λ
|ψI (t)i ≈ Ŵ (τ )dτ |ψI (0)i = T exp Ŵ (τ )dτ |ψI (0)i .
n=1 ı~ 0 ı~ 0
212 CHAPTER 6. APPROXIMATION METHODS
10
The expansion coefficients are ,
Z t
λ
af (t) = hf |ψI (t)i = hf |ψI (0)i + hf | S −1 (τ )Ĥ (1) (τ )S(τ )|ψI (0)idτ . (6.70)
ı~ 0
Now, we assume that the system be initially in the eigenstate |ψI (0)i = |ii. The
amplitudes then are,
Z
λ t ıEf τ /~
ai→f (t) = hf |ii + e hf |Ĥ (1) (τ )|iie−ıEi τ /~ dτ (6.71)
ı~ 0
Z t
λ
= δif + hf |Ĥ (1) (τ )|iieıωif τ dτ .
ı~ 0
The varying potential is considered a perturbation, and a variation of the system’s
state is observed. As the energy is not conserved, [∂t , Ĥ(t)] 6= 0, the time-dependence
is not separable and the system exchanges energy with the potential. In first-order
perturbation theory we only consider weak perturbations, i.e. the initial state is grad-
ually emptied, ai→i (dt) ' ai→i (0) = 1. For an initially empty state the growth is
obviously considerable. For i 6= f we have,
λ
dai→f (t) = ai→f (t + dt) − ai→f (t) = hf |Ĥ (1) (t)|iieıωif t dt . (6.72)
ı~
This formula is nothing more than a generalization of the formula (6.63) obtained
for a two-level system assuming that the initial state does not deplete considerably.
In Exc. 6.4.5.1 we calculate the dynamics of a harmonic oscillator perturbed by a
decaying force.
The δ-function in (6.75) ensures that, for infinitely sharp steps, transitions are im-
possible, unless the energy of the final state is the same as the one of the initial state.
This points to the fact that infinitely sharp steps are not a realistic physical concept.
In practice, the changes applied to a system are often slow and the observation
times are long, because the frequencies of the transitions are high ωf i /2π ' THz.
Let us assume that the perturbation be switched on within a time constant γ −1 . In
Exc. 6.4.5.2 we will study how the rapidity of a perturbation influences the transition
rate. We will see via a temporal analysis of |ai→f (t)|2 , that for slow variations,
γ ωf i , the system adiabatically approaches the final situation. For γ ' ωf i , the
system receives a shock and exhibits oscillating transients. For γ > ωf i , we observe
violent oscillations with largest amplitudes.
The first term being small, we neglect it doing a so-called rotating wave approximation
(RWA). We obtain,
sin2 1 (ωf i − ω)t
|ai→f (t)|2 = |Ωf i |2 1 2 2
. (6.79)
4 (ωf i − ω)
This result coincides with the formula (6.74), except that the energy difference be-
tween the states ωf i is shifted by the frequency of the perturbation ω. The quantity
∆f i ≡ ω − ωf i is called a detuning. The transition probability is maximal, when we
are at resonance, that is ∆f i = 0. In this case,
This can be seen by expanding the numerator in a Taylor series for small (ωf i − ω)t.
Note, that the probability exceeds 1 for long times, which can not be. In fact, the
restriction to the first order in the Taylor expansion used in the derivation of the last
214 CHAPTER 6. APPROXIMATION METHODS
equation is no longer valid for long times, when (ωf i − ω)t > 1, and we need to take
into account higher orders.
Then,
where we call
−eE0 h2|r|1i
Ω≡ (6.83)
~
the Rabi frequency. With this abbreviation the Eqs. (6.63) become,
ȧ1 = −ıΩa2 e−ıω0 t cos(kz − ωt) and ȧ2 = −ıΩ∗ a1 eıω0 t cos(kz − ωt) . (6.84)
ıΩ ıΩ∗
ȧ1 ' − a2 eı∆t and ȧ2 ' − a1 e−ı∆t . (6.85)
2 2
With the equations of motion we can, starting from initial values for a1 (0) and
a2 (0), calculate the temporal evolution.
We solve this system of differential equations by differentiating one and substi-
tuting the other,
Ω ∗ −ı∆t Ω∗ |Ω|2
ä2 = −ıȧ1 e − a1 ∆ e−ı∆t = − a2 − ı∆ȧ2 . (6.86)
2 2 4
We find solutions via the ansatz a2 = e−ı∆t/2 (AeıGt/2 +Be−ıGt/2 ). The equation
for a2 yields,
h i
− ı∆ ( 2ı G − 2ı ∆)Aeı(G−∆)t/2 + (− 2ı G − 2ı ∆)Beı(−G−∆)t/2 .
Separating the parts in A and in B we obtain two equations with the same
result,
G2 = |Ω|2 + ∆2 . (6.88)
G is called the generalized Rabi frequency. Using the initial conditions, a1 (0) = 1
and a2 (0) = 0, we can fix one of the coefficients A and B, since a2 (0) = A + B =
0,
a2 = 2ıAe−ı∆t/2 sin G 2
t. (6.89)
We now import this solution into the differential equation for a1 ,
Ω
ȧ1 = −ı a2 eı∆t = ΩAeı∆t/2 sin G
2
t . (6.90)
2
6.4. TIME-DEPENDENT PERTURBATIONS 215
1
|a2 (t)|2
0.5
0
0 1 2 3 4
Ωt (π)
Figure 6.4: (code) Probability |a2 (t)|2 for the atom to be in the excited state for Ω = Γ and
∆ = 0 (blue), ∆ = Γ (green), and ∆ = 2.5Γ (red). Time is in units of 1/Γ.
Note, that the Rabi formula (6.93) can be derived exactly without perturbative
calculus. In Exc. 6.4.5.6 we show a derivation using the Laplace transformation
method.
corresponds to the probability of the initial state |ii to be depleted. When the final
state lies within a continuum, the sum in (6.94) must be replaced by an integral.
With the density of states written in the form ρ(E), where ρ(E)dE is the number of
states found in the energy range between E and E + dE, the transition probability
is 12 ,
Z Emax
Pi→F (t) = |ai→f (t)|2 ρ(Ef )dEf , (6.95)
Emin
where E ∈ [Emin , Emax ] is the regime of energies within reach of the periodic pertur-
bation. Now, plugging in the expressions (6.79),
Z Emax
sin2 1 (Ef i − E)t
Pi→F (t) = |Ωf i |2 1 2~ 2
ρ(Ef )dEf . (6.96)
Emin 4~2 (Ef i − E)
with the substitution x ≡ (Ef i − E)/2~, we obtain after sufficiently long times 13 ,
Z Emax
π E i −E
Pi→F (t) = |Ωf i |2 t2 δ( f2~ )ρ(Ef )dEf = 2π~t|Ωf i |2 ρ(Ei + E) . (6.98)
Emin t
The transition rate is,
dPi→F (t)
= π|Ωf i |2 ρ(Ei + E) . (6.99)
dt
For narrow distributions we may substitute the density of states ρ by a δ-distribution,
dPi→F 2π
= |hf |Ĥ (1) |ii|2 δ(Ef i − E) , (6.100)
dt ~
where we went back to the definition of the perturbation Hamiltonian (6.77). This
expression is called Fermi’s Golden rule. In Exc. 6.4.5.7 we will calculate the photo-
electric effect.
again using the representation (6.97) of the Dirac function. The approximation
%(ω) = %(ωf i ) can be used if the width of the sinc function is much narrower than
the frequency distribution, which is the case for sufficiently long times, t π/2∆f i .
12 With
P
ρ(E) ≡ f ∈F δ(Ef − E) the integral is converted back into a sum.
13 Remember δ(ax) = 1 δ(x).
a
6.4. TIME-DEPENDENT PERTURBATIONS 217
The probability to encounter |ψi in a given state |f i is |hf |ψi|2 , and the transition
rate is simply the derivative of this. The transition rate out of the initial state into
any other final state is, consequently
1 X d
= |hf |e−ıĤ(t−t0 )/~ |ii|2 . (6.103)
τ dt
f
From this formula we can already see, that at short times, when we can expand the
exponential to first order, we will recover the results of (6.72).
Let us now consider a time-independent perturbation in the Schrödinger picture,
where the time dependence is entirely left to the wavefunction. Now, we expand the
propagator e−ıĤ(t−t0 ) in a perturbative series 14 . Unfortunately, generally [Ĥ (0) , Ĥ (1) ] 6=
(0) (1)
0, so that e−ıĤt 6= e−ıĤ t e−ıĤ t . But we can calculate,
d ıĤ (0) t −ıĤt (0) (0) (0)
ı e e = −Ĥ (0) eıĤ t e−ıĤt + eıĤ t Ĥe−ıĤt = eıĤ t Ĥ (1) e−ıĤt . (6.105)
dt
We want to analyze a steady-state situation, that is, we assume that the interaction
as gradually built up in an infinitely remote past. Hence, we can set Ĥ = Ĥ (0) for
t0 → −∞,
Z t
(0) (0)
e−ıĤt = e−ıĤ t 1 − ı eıĤ t1 Ĥ (1) eεt1 e−ıĤt1 dt1 , (6.107)
−∞
We calculate the matrix elements up to second order substituting Ĥ with Ĥ (0) in the
last integral,
hf |e−ıĤt |ii (6.109)
(0)
Z t (0) (0)
Z t1 (0) (0)
' hf |e−ıĤ t
1−ı eıĤ t1
Ĥ (1) eεt1 e−ıĤ t1
1−ı eıĤ t2
Ĥ (1) eεt2 e−ıĤ t2
dt2 dt1 |ii
−∞ −∞
Z t
= e−ıωf i t hf |ii − ıe−ıωf t eıωf t1 hf |Ĥ (1) |iieεt1 e−ıωi t1 dt1
−∞
Z t X Z t1 (0)
− e−ıωf t1 eiωf t1 hf |Ĥ (1) eεt1 e−ıωi t1 |mihm| eıĤ t2
Ĥ (1) |iieεt2 e−ıωi t2 dt2 dt1 .
−∞ m −∞
P
where we separated the perturbation orders and inserted m |mihm| = 1 using the
closure relation. Using the time-independence of Ĥ (1) ,
Z t
hf |e−ıĤt |ii ' e−iωf i t δf i − ıe−ıωf t hf |Ĥ (1) |ii e(ıωf −ıωi +ε)t1 dt1 (6.110)
−∞
X Z t Z t1
− e−ıωf t hf |Ĥ (1) |mihm|Ĥ (1) |ii e(ıωf −ıωi +ε)t1 e(ıωm −ıωi +ε)t2 dt2 dt1
m −∞ −∞
The first term is Fermi’s Golden rule, the second order corresponds to the Kramers-
Heisenberg formula, which serves to describe Thomson, Rayleigh and Raman scat-
tering. The generalization of this transition rate to all perturbation orders can be
written,
1 2π X 1 X Ĥ (1) |mihm|Ĥ (1)
= 2 hf | Ĥ (1) + + ... (6.113)
τ ~ ~ m ωi − ωm
f
2
X
1 Ĥ |m1 ihm1 |...|mn−1 ihmn−1 |Ĥ
(1) (1)
+ n−1 |ii δ(ωf i ) .
~ m ,...,m
(ωi − ωm1 )...(ωi − ωmn−1 )
1 n−1
Figure 6.5: Graphical illustration of the various transitions orders. From left to right:
absorption, Raman transition via an intermediate virtual state |m1 i, three-photon process
via two intermediate virtual states |m1 i and |m2 i.
6.4.5 Exercises
6.4.5.1 Ex: Perturbed harmonic oscillator
Consider a one-dimensional harmonic oscillator (HO) initially prepared (t = −∞)
in the ground state |0i of the unperturbed Hamiltonian H (0) = ~ω↠â, such that
H (0) |ni = En |ni with En = n~ω.
1
R tf
a. Through the expression, af (t) ≈ i~ ti
Wf i eiωf i t dt, and the perturbative Hamil-
2 2
tonian W (t) = −eExe−t /τ (x is the position operator of the HO), applied between
t = −∞ and t = +∞, calculate the probability of the system to be in the excited
state |ni, specifying n, at t = +∞. Analyze the result.
2 2
b. Do the same for a shape-changing perturbation, W (t) = Λx2 e−t /τ .
with γ ωf i .
a. Calculate the transition rate for long times, t γ −1 .
b. Analyze the transition rate at a given time as a function of γ.
The population of a degenerate two-level system be initially in state |1i. What should
be the duration of a perturbation to transfer the population to state |2i?
Solve the problem of a two-level atom interacting with a laser using the Laplace
transformation method.
A hydrogen atom ground state in the ground state 1s is placed in an electric field
E(t) = E0 cos ωt, such that W (t) = −er·E(t) = W0 e−ıωt +W0† eıωt with W0 = er·E0 /2.
Find, via Fermi’s Golden rule,
2π
R= |hf |W (t)|ii|2 ρ(Ef − Ei ∓ ~ω) ,
~
using the density of states ρ(Ek )dEk = V /(2π)3 k 2 dkdΩ, the probability per unit
of time for the atom to be ionized, by exciting from the ground state ψ100 (r) =
e−r/aB /(πa3B )1/2 to the state described by the plane wave ψk (r) = e−ık·r /V 1/2 . Sim-
plify the calculation by assuming E0 = E0 êz and k = kêz .
6.5. FURTHER READING 221
Appendices to ’Quantum
Mechanics’
This chapter compiles some fundamental functions and polynomials, which are rele-
vant in quantum theory.
With this fixing we give the energy in terms of e2cgs /aB , the wavevector in terms of
1/aB , the distance in terms of aB and the mass in terms of me , such that,
223
224 CHAPTER 7. APPENDICES TO ’QUANTUM MECHANICS’
7.2.2 {3j}-symbols
The Clebsch-Gordans are related to Wigner’s {3j}-symbols,
j1 j2 j −j1 +j2 −m
p j1 j2 j
= (−1) 2j + 1 , (7.5)
m1 m2 m m1 m2 −m
and
0 j j (−1)j
=√ . (7.8)
0 0 0 2j + 1
7.2.3 {6j}-symbols
{6j}-symbols describe the recoupling of two spins. They can be evaluated by,
X (−)t (t + 1)!
j1 j2 j3 p
= ∆(j1 j2 j3 )∆(j1 J2 J3 )∆(J1 j2 J3 )∆(J1 J2 j3 ) ,
J1 J2 J3 t
f (t)
(7.9)
where,
f (t) = (t − j1 − j2 − j3 )!(t − j1 − J2 − J3 )!(t − J1 − j2 − J3 )!(t − J1 − J2 − j3 )! (7.10)
(j1 + j2 + J1 + J2 − t)!(j2 + j3 + J2 + J3 − t)!(j3 + j1 + J3 + J1 − t)! .
7.2.4 {9j}-symbols
{3j}-symbols describe the recoupling of three spins. They can be evaluated by,
j1 j2 J12 X
j1 j2 J12 j3 j4 J34 J13 J24 J
j j4 J34 = (−)2g (2g+1) .
3 J 34 J g j2 g J24 g j1 j3
J13 J24 J g
(7.11)
{9j}-symbols satisfy the following orthogonality relation,
X j1 j2 J12 j1 j2 J12
Jˆ12 Jˆ34 Ĥ13 Ĥ24 j j4 J34 j j4 J34 = δJ13 H13 δJ24 H24 .
3 3
J12 ,J34 H13 H24 J J13 J24 J
(7.12)
Higher momenta:
Z x1 Z x1
2 2 x1 2
e−ax dx = xe−ax + 2a x2 e−ax dx . (7.14)
x0 x0 x0
The derivative,
dJk (x)
2 = Jk−1 (x) − Jk+1 (x) . (7.16)
dx
Sum rules (empirically found),
∞
X ∞
X ∞
X
Jk (x) = Jk (x)2 = 1 6= |Jk (x)|2 . (7.17)
k=−∞ k=−∞ k=−∞
226 CHAPTER 7. APPENDICES TO ’QUANTUM MECHANICS’
The spherical Bessel function, the spherical von Neumann function, and the spher-
ical Hankel function of the first and second kind are defined by:
r
π
jk (x) = Jk+1/2 (x) (7.18)
2x
r r
π k+1 π
nk (x) = Nk+1/2 (x) = (−1) J−k−1/2 (x)
2x 2x
(1,2)
hk = jk ± ink .
The series,
k k
1 d sin x 1 d cos x
jk (x) = (−x)k , nk (x) = −(−x)k . (7.19)
x dx x x dx x
The derivative,
k
jk0 (x) = jk (x) − jk+1 (x) . (7.20)
x
Recursion:
d
Hn (x) = 2Hn−1 (x) (7.23)
dx
d −x2 2
e Hn (x) = e−x Hn+1 (x)
dx
Hn+1 (x) = 2xHn (x) − 2nHn−1 (x) .
Particular values:
Series:
∞
X (2n)! (−1)n−k
H2n (x) = (2x)2k (7.25)
(2k)! (n − k)!
k=0
X∞
(2n + 1)! (−1)n−k
H2n−1 (x) = (2x)2k+1
(2k + 1)! (n − k)!
k=0
int(n/2)
X 1 (−1)k
Hn (x) = n! (2x)n−2k .
k! (n − 2k)!
k=0
ex x−m dn −x n+m
L(m)
n (x) ≡ (e x ) (7.26)
m! dxn
Ln ≡ L(0)
n (x) .
Series:
n
X
n + m (−x)k
L(m)
n (x) = . (7.27)
n−k k!
k=0
Recursion:
d (m) (m+1)
L (x) = −Ln−1 (x) . (7.28)
dx n
Related functions:
q
2
umn (ε) ≡ e−ε · (iε)n−m · m! n−m 2
n! · Lm (ε ) (7.29)
q
umn (0) ≈ (ıε)n−m · m!(n−m)!
n!
2
√
un+1,n (0) ≈ ıε · n + 1 .
Fourier transforms:
Z ∞ √ √
2 (−1)k π 2
e−ax x2k cos xp · dx = k
√
2k+1
· ep /4a
· He2k (p/ 2a) (7.30)
2 a
−∞
Z ∞ √
2 (−1)m π 2
e−x /2 2m
x L2m 2
n (x ) cos xp · dx =
√
2n!
· e−p /2
· Hen (p)Hen+2m (p)
−∞
Z ∞ √
2 (−1)m π 2
e−x /2 2m+1
x L2m+1
n (x2 ) sin xp · dx = √
2n!
· e−p /2
· Hen (p)Hen+2m+1 (p)
−∞
Z ∞ Z ∞ Z ∞
e−ax−bp f (|x − p|) · dxdp = 1
a+b e−ax f (x)dx + e−bp f (p)dp .
−∞ −∞ −∞
228 CHAPTER 7. APPENDICES TO ’QUANTUM MECHANICS’
1 dn 2
Pn (x) ≡ (x − 1)n (7.31)
2n n! dxn
dm
Pn(−m) (x) ≡ (1 − x2 )m/2 m Pl (x) .
dx
Series:
" #
(−1)m (n + m)! 1 − (n−m)!(m+n+1)!
1!(m+1)!
1−x
2 +
Pn(m) (x) = m 2 m/2
(1−x )
1−x 2
.
2 m!(n − m)! + (n−m)!(n−m+1)!(m+n+1)!(m+n+2)!
2!(m+1)!(m+2)! 2 − ...
(7.32)
p X ` 1 j
1−`−m
Yj`m (r) ≡ (−1) 2j + 1 Y` m−q (ϑ, ϕ)êq . (7.35)
q
m−q q −m
where,
∞
X Z ∞
(±z)t 1 xη−1 dx
gη (z) = ± = , (7.38)
t=1
tη Γ(η) 0 z −1 ex
±1
is also called the Bose/Fermi function. The upper sign holds for bosons, the lower for
fermions. The Sommerfeld expansion,
Z ∞ Z y Z ∞ Z x
g(x)dx g(y + x)η−1 dx g(y − x)η−1 dx
= g(x)dx + − (7.39)
0 ex−y + 1 0 0 ex + 1 0 ex + 1
Z y
π2 0
≈ g(x)dx + g (x) + ...
0 6
or,
Z ∞ Z ∞ η−1 −2 x
∂fη (Z) 1 ∂ xη−1 dx 1 x Z e dx
= = (7.43)
∂Z Γ(η) ∂Z 0 Z −1 ex + 1 Γ(η) 0 (Z −1 ex + 1)2
Z ∞ Z ∞
1 d −1 η−1 xη−2 fη−1 (Z)
= xη−1 −1 x
dx = −1
dx = .
ZΓ(η) 0 dx Z e + 1 ZΓ(η) 0 Z ex + 1 Z
231
Chapter 8
e2 1
α≡ ' , (8.1)
4π0 ~c 137
we obtain, r
2E1 ~ e2
v= = = = αc , (8.2)
me me aB 4πε0 ~
which shows that the electron velocity is very high and that relativistic effects may
indeed be not negligible.
233
234 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
As already discussed in Sec. 2.1.2 we can, in order to find a relativistic wave equation,
try the approach of inserting the quantum operators into the relativistic energy-
momentum relation 1 .
E 2 = c2 p2 + m2e c4 . (8.5)
We obtain,
m c 2
1 ∂2 2 e
−∇ + ψ=0 . (8.6)
c2 ∂t2 ~
This is the Klein-Gordon equation. The stationary solution of this equation is a
spherical wave,
1
ψ = ψ0 e−2πr/λC , (8.7)
r
where λC = h/me c is the Compton wavelength. We show this in Exc. 8.1.5.1. For
example, in the case of heavy bosonic particles, such as a field of π-mesons, ψ is the
Yukawa potential.
In the framework of the standard model, it is believed that matter is composed
of two fundamental types of particles, bosons and fermions. Bosons are exchanged
between fermions conveying the interaction between them. A typical example is
the one of two electrons whose Coulomb interaction is mediated by the exchange
of photons. Bosons obey the Klein-Gordon equation, fermions the Dirac equation
derived in the following section.
The matrices αi must satisfy the condition (8.10). It is possible to show that this
requires at least four-dimensional matrices of the following form:
I 0 0 σj
α0 = and αj = , (8.11)
0 −I σj 0
where j = x, y, z = 1, 2, 3. In this notation the components of the matrices are also
matrices:
1 0 0 1 0 −ı 1 0
I= , σx = , σy = , σz = , (8.12)
0 1 1 0 ı 0 0 −1
called Pauli spin matrices. The state vector must also have four dimensions,
!
~ ~ t)
φ(r, φ1 (r, t) χ1 (r, t)
~
Φ(r, t) = ~
with φ(r, t) = and χ~ (r, t) = . (8.13)
χ
~ (r, t) φ2 (r, t) χ2 (r, t)
φj are called large components, χj are called small components. This designation is
explained later. Combining the matrices αj to a three-dimensional vector α
~ , we can
now write the Dirac equation (8.9) like,
~ ~
~ t) = me c2 α0 + c~
ı~∂t Φ(r, ~ t) .
α · p Φ(r, (8.14)
236 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
Or, using the notation (8.11) and combining the Pauli matrices σj to a three-dimensional
vector ~σ , we can write the Dirac equation as:
! !
∂ ~
φ 2 I 0 0 ~σ · p ~
φ
ı~ = me c +c . (8.15)
∂t χ
~ 0 −I ~σ · p 0 χ
~
γ 0 ≡ α0 and γ k = γ 0 αk . (8.17)
We obtain, ! !
0 I 0 k 0 σk
γ = and γ = . (8.18)
0 −I −σk 0
We also define another important matrix by,
!
0 I
γ5 ≡ ıγ0 γ1 γ2 γ3 = . (8.19)
I 0
With this, using Einstein’s notation 3 , the Dirac equation (8.13) adopts the
form,
ı~γ µ ∂µ ψ − mcψ = 0 . (8.20)
The complete system is summarized in the Minkowski metrics of time-space in
the form,
[γ µ , γ ν ]+ = 2η µν , (8.21)
for µ, ν = 0, .., 5, that is, all matrices γk anticommute.
The Dirac equation can now be interpreted as an eigenvalue equation, where
the rest mass is proportional to the eigenvalue of a momentum quadrivector,
the proportionality constant being the speed of light:
γ 0µ = S −1 γ µ S . (8.24)
3∂ 1
0 ≡ ∂
c t
8.1. THE DIRAC EQUATION 237
If, in addition, the matrices are all unitary, as is the case of Dirac’s set, then S
is unitary,
γ 0µ = U † γ µ U . (8.25)
8.1.2.1 Anti-particles
Disregarding for a moment the non-diagonal matrix, the Dirac equation separates
into two independent equations,
~
∂φ ∂~
χ
ı~ ~
= me c2 φ and ı~ = −me c2 χ
~ . (8.26)
∂t ∂t
These are eigenenergy equations with the eigenvalues me c2 and −me c2 . The state
with negative energy is interpreted as anti-particle. Therefore, the non-diagonal ma-
trix mixes particles and anti-particles. We will study in Exc. 8.1.5.2 the so-called
Zitterbewegung as a solution of the Dirac equation.
finally obtaining,
Since u and v only vary slowly in time, the derivatives on the left-hand side are small
quantities. However, the condition that both derivatives must zero is too strong,
because it leads to the trivial solution u = 0 and v = 0. We find the first non-trivial
solution by the condition v̇ = 0. The second equation then becomes,
1
v= (~σ · p)u . (8.31)
2me c
238 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
(~σ · p)2
ı~u̇ = c u. (8.32)
2me c
We need, therefore, to evaluate the expression (~σ · p)2 ,
pz px − ıpy 2 2 1 0
~σ · p = and (~σ · p) = p . (8.33)
px + ıpy −pz 0 1
Inserted into the differential equation (8.32) for u we obtain precisely the Schrödinger
equation for a free particle,
p2
ı~u̇ = u . (8.34)
2me
Let us return to the question, why we call u the strong component. We have from
the equation (8.31),
1 1 p2 †
v† v = 2
(~σ · p)2 u† u = 2
u u, (8.35)
(2me c) 2me c 2me
2
p
and since 2m e
me c2 follows immediately v† v u† u.
In this non-relativistic approximation the components u are much larger than the
components v. The mixture between particles and antiparticles only matters when
p2 2 1 2 2
2me ' me c , resp., 2 me v ' me c or |v| ' c. The electron only receives small
positronic contributions as it approaches the speed of light. In the ground state of
the hydrogen atom the electron has a velocity of of v = αc ' c/137. That is, the
contribution of the weak components is small, but present.
Example 42 (Vanishing rest mass): Let us note that for the case of vanishing
rest mass, me = 0, the Dirac equation (8.15) dramatically simplifies. Taking the
time derivative of the upper equation (8.15) and inserting the lower equation
(8.15), we find,
ŝ ≡ ~2 ~σ , (8.37)
and we calculate the commutation relations between its components. From the defi-
nitions (8.12) we obtain the rule,
~2 0 1 0 −ı ~2 0 −ı 0 1 ~2 2ı 0
[ŝx , ŝy ] = 4 − 4 = = ı~ŝz .
1 0 ı 0 ı 0 1 0 4 0 −2ı
(8.38)
8.1. THE DIRAC EQUATION 239
In general terms the following holds true: [si , sj ] = ijk ı~sk . It is interesting to com-
pare this with the commutation relation for the orbital angular momentum [ˆli , ˆlj ] =
ijk ı~ˆlk . The coincidence suggests a generalization of the concept of angular mo-
mentum: We now call angular momentum operator every three-dimensional vector
operator satisfying this commutation relation 4 . We consider the eigenvalue equation
for ŝz , which is incorporated in the Dirac equation,
~=~ 1 0
ŝz φ
φ1
= m s ~
φ1
. (8.39)
2 0 −1 φ2 φ2
we obtain coupled stationary equations for the large and small components,
~
(E − me c2 )φ(r) = cσ · p~
χ(r) and (E + me c2 )~ ~
χ(r) = cσ · pφ(r) . (8.42)
p̂ −→ p̂ − qA ≡ ~π and Ĥ −→ Ĥ + qU . (8.44)
called the minimal coupling. We briefly mentioned this already in Sec. 2.5.3, and the
rules will be derived in Sec. 9.1. In addition to the substitution of the momentum,
we must add the scalar potential qU , and we obtain the Dirac equation for a particle
inside an applied electromagnetic field,
˙
~ ~
~ = me c2 α0 + c~
ı~Φ ~ ,
α · ~π + qU Φ (8.45)
Ĥ ≡ me c2 α0 + c~
α · [p − qA(r)] + qU (r) (8.49)
e2
with A=0 and U (r) = − .
4πε0 r
We adopt the standard procedure from non-relativistic physics, which consists
in rewriting the Hamiltonian in terms of observables, which commute with the
Hamiltonian 7 .
5 In E/c U/c
quadrivetorial notation: πµ = pµ − qAµ with pµ = and Aµ = .
p A
6 See also http://einstein.drexel.edu/∼bob/Term Reports/Whitehead 3.pdf
7 Typical examples are the Hamiltonian of the harmonic oscillator (3.76) written in terms of
α · p = (~
rε~ α · r)(~
α · p) = r · p + ıγ5 α
~ · (r × p) (8.53)
~ · L = rpr + ıα0 ~j 0 − ı~ ,
= r · p + ıγ5 α
ı~(α0 j 0 − 1)
~ · p = ε pr +
α , (8.54)
r
The final radial Hamiltonian is,
~ ıcεα0 ~j 0 e2
Ĥ = me c2 α0 + cε pr − ı + − . (8.55)
r r 4πε0 r
For now the choice of the quantities ε and j 0 must seem arbitrary, so we will
have to discover their properties. We will see that j 0 is a non-zero integer related
to the total angular momentum j. The following properties will be proven in
Exc. 8.1.5.5(a) and (b),
~2
(~j 0 )2 = J2 + 4
(8.56)
0
[~j , Ĥ]− = [ε, Ĥ]− = [α0 , Ĥ]− = 0 .
into 2 by 2 matrices. Using the fact that ε commutes with all other terms and
anti-commutes with α0 and that ε−1 = ε† is a unitary transformation, we may
go to a new basis via,
0!
e2
! !
me c2 − 4πε −ıcpr − c ~r − c ~jr φ~ φ~
0r =E . (8.57)
0
e2
icpr + c ~r − c ~jr −me c2 − 4πε r
ε~
0
χ ε~
χ
2
Substituting α = e
4πε0 ~c
and a± ≡ ~
me c∓E/c
(that is, ±me c2 − E ≡ ± a~c
±
), we
find,
j 0 +1
1
! !
a+
− αr d
− dr − r φ~
0 =0. (8.58)
d
dr
− j r−1 − a1− − αr ε~
χ
√
E2
−1
d e−r/a e−r/a
a+ a− = ~ m2e c2 − − a1 − 1
where a ≡ c2
[that is, dr r
= r r
],
we find,
j0
1 α d 1
! !
a+
− r
− dr − a
− r f~
1 j0
=0. (8.60)
d
dr
− a
− r
− a1− − α
r
~g
These are then substituted into our system of equations. In order for the equa-
tion to go to zero as required, each term in the resulting series must separately
go to zero. The coefficient of the rs terms are,
fs gs
− αfs+1 − (s + 1 + j 0 )gs+1 + =0 (8.62)
a+ a
gs fs
− αgs+1 − (s + 1 − j 0 )fs+1 + =0.
a− a
These can be combined by multiplying the first equation (8.62) by a and and
the second by a− and then subtracting the former from the latter. Exploiting
a a
a+
= a− , this gives us an expression directly relating the fs coefficients with
the gs coefficients,
Combining these equations we can write the value s0 in a very simple form,
p
s0 = j 02 − α2 . (8.66)
This places a lower bound on the series. Note that this bound becomes imaginary
if α > j 0 . This will be discussed in more detail shortly.
The upper bound of the series is also useful. It can be shown that the series must
terminate if the energy eigenvalue is to be less than me c2 [208]. The implication
of this result is that if the series terminates at index s1 such that,
E2 E2
s1 me c − 2 = α2 2 .
2 2
(8.69)
c c
This can be solved for the energy eigenvalues,
−1/2
α2
E = ±me c2 1 + 2 . (8.70)
s1
Note that the ’negative energy’ solution corresponds to positron energy levels.
From here forward, we drop the negative root and look only at the electron
solution.
The two end points of the series, the indices s0 and s1 are separated by an
integer number of steps. Calling this integer n0 we can write,
s1 = n0 + s0 = n0 + j 02 − α2 .
p
(8.71)
Plugging this into (8.70) gives a result for the energy eigenvalues in terms of
only the two quantum numbers n0 and j 0 ,
−1/2
α2
En0 ,j 0 = me c2 1 + .
p 2
0 02
n + j −α 2
This is the final result quoted for the energy eigenvalues of the hydrogenic atom
by Dirac [208]. It turns out that later developments in the field [296, 301] prefer
to use an equivalent set of quantum numbers that maps more closely to the
familiar ones. The number j 0 is closely related to the total angular momentum
quantum number j. j 0 has the range 1, 2, 3, while j has the range 21 , 23 , 52 . It
is natural, and in fact correct, to make the identifications,
j0 = j + 1
2
, (8.72)
244 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
and
n0 = n − j 0 = n − j − 1
2
, (8.73)
for the principal quantum number n [296]. Combining these two adjustments
with equation (8.72), we get the Sommerfeld fine-structure formula,
−1/2
2
α
En,j = me c2 1 + , (8.74)
q 2
1
n−j− 2
+ (j + 12 )2 − α2
with j = ` ± 12 and ` = 0, 1, ... The derivation of the form of the actual wave
~
functions φ(r) and χ~ (r) is very tedious [296] and will not be reproduced here.
The second term reproduces the energy of Bohr’s model, but there are correction
terms proportional to α4 . We will show in Secs. 8.1.4, that the energy levels, called
fine structure, result from several relativistic corrections of different origins.
In the expression (8.75) for the electron energy in the Coulomb potential, the last
term is positive and proportional to 1/n4 . It describes relaxation of the binding due
to the contribution of weak components. The term containing the quantum number
j is called the spin-orbit coupling. To better understand this contribution we must
first analyze more deeply the matrices ~σ .
Now, ~σ ·p is an operator entity, which acts on the subsequent operators and wavefunc-
tions. We thus have to apply the product rule, (~σ · p)V ψ = V (~σ · p)ψ + [(~σ · p)V ]ψ,
to the first occurrence of operator this operator in equation (8.77),
~ ' qU (r)φ
~+ E 0 −qU (r) 2~ q ~ . (8.80)
E0φ 1
2me 1− 2me c2 (~σ · p) φ + 4m2e c2 [(~
σ · p)U (r)](~σ · p)φ
In the following we will make use of a general relationship which is similar to (8.51),
It holds for [~σ , B] = 0 = [~σ , C] and will be demonstrated in Exc. 8.1.5.4. The
relationship yields,
(~σ · p)2 = p2 and [~σ · pU (r)](~σ · p) = pU (r) · p + ı~σ · [pU (r) × p] , (8.82)
∂U ∂U r ∂U ∂U ∂
∇U (r) = ∇r = and ∇U (r) · ∇ = êr · ∇ = . (8.84)
∂r ∂r r ∂r ∂r ∂r
We get,
E 0 − qU (r) ~− ~
2
~q
~+ 1
~ = qU (r)φ
E0φ 1− p2 φ
∂qU ∂ ~
φ+ ~
σ ·
1 ∂U
r × pφ~
2me 2me c 2 4me c2 ∂r
2 ∂r 4m2e c2 r ∂r
2
p4 ~2 q
p q 1 ∂U ∂U ∂ ~
' + qU (r) − + s · l − φ. (8.85)
2me 8m3e c2 2m2e c2 r ∂r 4m2e c2 ∂r ∂r
where we again applied the non-relativistic approximation (8.78) in the second line
and made use of the definitions s = ~2 ~σ and l = r × p. The term in the bracket can
be used as the Hamiltonian allowing to calculate the fine structure as first-order per-
turbations to the non-relativistic energy levels obtained from non-relativistic theory,
p2 1 e2 p4 e 1 ∂U ~2 e ∂U ∂
Ĥ ' − − 3 2
− 2 2
s·l− . (8.86)
2me 4πε0 r 8me c 2me c r ∂r 4m2e c2 ∂r ∂r
The first two terms are those arising from Bohr’s atom model, the third one is a
correction due to the relativistic velocity of the electron, the forth comes from the
electron’s spin-orbit coupling, and the fifth is called the Darwin term. All contri-
butions represent perturbations to the non-relativistic Schrödinger theory of Bohr’s
atom and will be discussed extensively in Secs. 8.2. We will show in 8.1.5.6 that l̂2 ,
ŝ2 , and ĵ2 are constants of motion of the above Hamiltonian.
246 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
~ = qU (r)φ
~+ (~σ · ~π )2 ~
E0φ φ . (8.87)
2me
p2 1 e2 ı~e e ~ ,
Ĥ ' − − (∇ · A + A · ∇) + s·B (8.90)
2me 4πε0 r 2me me
neglecting terms in A2 . Note however, that the kinetic energy is calculated with the
momentum projected onto the spin, ~σ · ~π . The third term can be simplified within the
Coulomb gauge ∇ · A = 0 yielding,
e
Ĥint = (A · p) . (8.91)
me
The Pauli equation serves for a classical (non-relativistic) approach to the elec-
tron’s spin-orbit coupling, as we will see below and in the discussion of the fine
structure in Sec. 8.2.2.
8 The spin of the electron does not generate a magnetic field, in contrast to the angular momentum
caused by its orbital motion. It only interacts with the environment through the requirement of
symmetrization for being a fermion.
8.1. THE DIRAC EQUATION 247
µ
~` l
= −gl . (8.94)
µB ~
−~ ~ = e~ ~σ · B
µs · B ~ = e s·B
~. (8.96)
2me me
We conclude, that the electron carries, besides mass, charge and spin, also a magnetic
dipole moment,
µ
~s e s
=− s = −2 , (8.97)
µB me µB ~
For the g-factor of the electron, we obtain ge = 2 9 . Neutron and proton are also
fermions with spin 21 , but they do not obey the Dirac equation! Their g-factors are
µs 2
9 The exact value is ge ≡ sµ = 2.002319314... The deviation ge − 2 ' απ
− 0.164 α
π2
is due to
B
the coupling of the spin to the fluctuations of the electromagnetic vacuum. We need to use quantum
electrodynamical methods to calculate the corrections.
248 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
gproton = 5.5858 and gneutron = −3.8261. The large deviation from g = 2 points to
the existence of an internal structure.
The rapid motion of the electron within the electrostatic field E~ of the nucleus pro-
duces, following the theory of relativity, in the electron’s reference frame a magnetic
~ 0 with which the electronic spin can interact. As we will show in Exc. 8.1.5.7,
field B
the field seen by the electron can be approximated in first order in v/c by,
~ 0 ' v × E~ .
B (8.98)
c2
With this the interaction energy (8.95) becomes,
Ĥ`s = −~ ~ 0 = e s · (v × E)
µs · B ~ = − e s · (p × ∇U ) (8.99)
me c 2 m2e c2
e r ∂U 1 ∂V (r)
= − 2 2s · p × =− 2 2 s·l ,
me c r ∂r me c r ∂r
8.1.5 Exercises
8.1.5.1 Ex: Yukawa potential
Show, that Yukawa’s potential satisfies the Klein-Gordon equation.
non-collinear velocities can not simply be concatenated, but must be rotated, too [260, 374].
8.1. THE DIRAC EQUATION 249
an environment such that the non-relativistic Schrödinger equation for the ion has
the same mathematical form as the Dirac equation (although the physical situation
is different) [276]. Then, in 2013, it was simulated in a setup with Bose-Einstein
condensates [439].
α · B)(~
(~ α · C) = B · C+ıγ5 α
~ · (B × C) . (8.102)
S = ~2 ~ς where ~ = I ⊗ ~σ
~ς ≡ γ5 α (8.103)
8.1.5.7 Ex: Magnetic field generated by the orbiting proton at the lo-
cation of the electron
Calculate the magnetic field generated by the orbiting proton as it is perceived by the
electron.
We will discuss the various terms in the following sections. Note that the expression
for the Darwin term differs from that of (8.86). We will see in Exc. 8.2.6.1, that they
are, in fact, equivalent.
p p2 p4
Ekin = p2 c2 + m2e c4 ' me c2 + − + ... . (8.110)
2me 8m3e c2
8.2. FINE STRUCTURE OF HYDROGEN-LIKE ATOMS VIA TIPT 251
using the eigenvalues calculated in (4.53). Finally, we obtain the following relativistic
correction,
2 1 3
∆Erl = En (Zα) − . (8.115)
n(` + 21 ) 4n2
Obviously, the degeneracy with respect to the angular momentum ` is lifted by this
correction.
−j(r0 )d3 r0
Z
µ0
A(x) = , (8.117)
4π V |x − r0 |
respectively,
(x − r0 ) × j(r0 ) 3 0
Z
~ µ0
B(x) = ∇x × A(x) = d r (8.118)
4π V |x − r0 |3
µ0 ∞ 0 ∞ 0 0 2π (x − r0 ) × v Ze
Z Z Z
=− dz r dr dφ δ(r − r0 )δ(z 0 )
4π −∞ 0 0 |x − r0 |3 2πr
Zeµ0 (x − r) × v
= ,
4π |x − r|3
where we replaced the expression for the current density (8.92). With the ex-
pression for the Coulomb potential between the electron and the proton and its
radial derivative,
~ Zeµ0 −r × v ε0 µ0 r × v dV (r)
B(0) = =− (8.120)
4π r3 e r dr
1 r × v dV (r) 1 dV (r)
=− 2 =− l.
ec r dr eme c2 r dr
The advantage of introducing the potential V is, that this expression also holds
for more complicated atoms with many electrons, where the potential may devi-
ate considerably from the Coulombian potential. Note, that the magnetic field
is very strong, B ' ξ(aB )~/µB ≈ 5 T. Inserting the magnetic field into Pauli’s
expression (8.95) together with the magnetic moment of the spin (8.97),
~ 1 1 dV (r)
Ĥ`s = −~
µs · B(0) = s·l . (8.121)
m2e c2 r dr
1
where we still have to apply the corrective Thomas factor of 2
.
−1 dV Ze2 1 En Z 2 α2 n2 1
ξ(r) ≡ = − = , (8.123)
2m2e c2 r dr 8πε0 m2e c2 r3 ~2 r̃3
After the introduction of the spin, the Hilbert space of the particles’ wavefunctions
must be extended. The wavefunctions are now products of spatial wavefunctions and
spin eigenvectors:
s
|n, `, m` , ms i = Rn` (r)Y`m (θ, φ) 1 . (8.124)
s2
The new Hilbert space is the tensorial product of position space and spin space. The
radial Hamiltonian for the hydrogen atom including the centrifugal term and the
spin-orbit coupling now takes the form:
p2 l2
Ĥ = + V (r) + + ξ(r)l · s . (8.125)
2m 2me r2
We may again consider the energy term V`s as a small perturbation, and calculate
it using unperturbed wavefunctions,
∆E`s = hn, `, s, m` , ms |V`s |n, `, s, m` , ms i = hn, `|ξ(r)|n, `ih`, s, m` , ms |s·l|`, s, m` , ms i .
(8.126)
Assuming a Coulombian potential, we first look at the radial part (8.123), which can
easily be calculated using the formulae (4.53),
E n Z 2 α 2 n2 1
hn, `|ξ(r)|n, `i = . (8.127)
~2 n3 `(` + 21 )(` + 1)
To diagonalize the angular part of the Hamiltonian, we need the common wavefunc-
tions of l2 and l · s. We can rewrite the coupling term as:
l · s = 12 (j2 − l2 − s2 ) . (8.128)
In the common eigensystem of j2 , l2 , and s2 the Hamiltonian, therefore, is diagonal.
We know the basis of this system from the theory of the addition of angular momenta.
The states of the basis are linear combinations of the functions |n, `, m` , ms i. Since
the spins precess around each other, `z and sz are not good observables, the non-
coupled base is not appropriate. But s2 , l2 , and j2 are good observables. In the
coupled basis {n, (`, s)j, mj },
~2
hn, (`, s)j, mj |s · l|n, (`, s)j, mj i = 2 [j(j + 1) − `(` + 1) − s(s + 1)] . (8.129)
Since j = ` ± 1/2, we find that every level splits into two levels, one with the energy
En` +`ζn` and the degeneracy 2`+2 and the other with the degeneracy En` −(`+1)ζn`
with the degeneracy 2`, where we introduced the abbreviation,
~2
ζn` ≡ 2 hξ(r)i . (8.130)
All in all, we get an energy correction due to the spin-orbit interaction of,
j(j + 1) − `(` + 1) − 43
∆E`s = −En (Zα)2 . (8.131)
2n`(` + 1/2)(` + 1)
Note, that the coupling l · s lifts the degeneracy with respect to l, but not with respect
to `z (see Fig. 8.1). As we have already seen in Exc. 8.1.5.6, in the presence of an
energy associated with the coupling l · s, only the total angular momentum l + s is a
constant of motion.
254 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
where f (r0 ) is a radially symmetric and normalized density-type function that takes
significant values only in the vicinity of r within a volume (λC /2π)3 centered at r0 = 0.
Expanding the potential V (r + r0 ) near the origin,
The second term is null due to the parity of f (r0 ) and the third produces the Darwin
correction using V (r) = V (r). Letting the function be constant within the volume,
f (r) ' f0 , and with the normalization,
Z ~/2me c Z ~/2me c Z ~/2me c 3
~
1= f (r)dxdydz = f0 , (8.135)
−~/2me c −~/2me c −~/2me c me c
Also,
Ze %(r) Ze2 δ 3 (r)
∇2 V (r) = −e∇2 = −e =− . (8.137)
4πε0 r ε0 ε0
Hence,
Z
Ze2 π~2 Ze2 3
f (r0 )V (r + r0 )d3 r0 = − + δ (r) + ... , (8.138)
4πε0 r 2m2e c2 4πε0
11 The smearing out of the electron’s position is also known as Zitterbewegung. See Exc. 8.1.5.2.
8.2. FINE STRUCTURE OF HYDROGEN-LIKE ATOMS VIA TIPT 255
which is precisely the electrostatic energy with the Darwin correction in the expres-
sions (8.86) and (8.109).
To estimate the importance of this effect we inserting the wavefunctions (4.52)
evaluated at the origin,
Z
∗ π~2 Ze2 3
hĤdw i = d3 rψn`m (r) δ (r)ψn`m (r) (8.139)
2m2e c2 4πε0
π~2 Ze2 2 π~2 Ze2 Z 3
= |ψ n00 (0)| δ `0 δ m0 = δ`0 .
2m2e c2 4πε0 2m2e c2 4πε0 πn3 a3B
We obtain,
π~2 Ze2 1
Hdw 2m2e c2 4πε0 πa3B ~2 Z α2
= = = ≈ 0.01% . (8.140)
H0 e2
4πε0 aB
2m2e c2 a2B 2
(Zα)2
∆Edw = hĤdw i = −En δ`0 . (8.141)
n
That is, the levels are now degenerate in j (see Fig. 8.1) 12 . Obviously the levels
which are most affected by relativistic corrections are those with low values of n and
`.
The levels are labeled by n`j . For example, the state 3d5/2 has the main quantum
number n = 3, the orbital angular momentum ` = 2, and the total angular momentum
j = 5/2. For large n or j the fine structure disappears. The new energy scheme is
shown in Fig. 8.1. We note that, taking into account all relativistic corrections (but
without the Lamb shift), we still have a partial degeneracy of the quantum number j.
For example, the states 2 s1/2 and 2 p1/2 have the same energy. This is a particularity
of the hydrogen atom.
12 It is interesting that the quantum treatment presented here, including relativistic corrections,
8.2.6 Exercises
8.2.6.1 Ex: The Darwin term
Show that the expressions for the Darwin correction (8.86) and (8.109) are equivalent.
where gp ≡ µI /I is once again a factor taking into account possible corrections be-
tween the classical derivation and quantum mechanics 14 .
~,
(~σ · ~π )(~σ · ~π ) = [p − qA]2 − ~q~σ · B (8.145)
[p − qA(r)]2 1 e2 p4 e 1 ∂U ~2 e ∂U ∂ ~q ~
Ĥ ' − − 3 2
− 2 2
S·L− − ~σ · B(r) ,
2me 4πε0 r 8me c 2me c r ∂r 4m2e c2 ∂r ∂r 2me
(8.146)
and expanding the bracket
in the Coulomb gauge, we see that two new terms are added to the energy called the
hyperfine structure,
with µ~ S = − mee S.
Up to now we did not say anything about the origin of the magnetic field. We
only notice that any magnetic field will interact with the electron’s orbit and with its
spin. We now make use of our knowledge that the proton has a spin of its own which
produces, at the position of the electrons, a magnetic vector potential,
~I × r
µ0 µ
A(r) = , (8.149)
4π r3
interacting with the angular momentum of the electron L in the form,
e e µ0
ĤLI = A · p̂ = µI × r) · p
(~ (8.150)
me me 4πr3
e µ0 µN µ0 µB µN
= gp (I × r) · p = gp L · I ,
me 4πr3 ~ 2πr3 ~ ~
using the definition of Bohr’s magneton.
In addition, the potential vector (8.149) generated by the nuclear spin produces a
magnetic field [374],
~ = ∇ × A = µ0 [3(~
B µI · r̂)r̂ − µ ~ I δ 3 (r) ,
~ I ] + 32 µ0 µ (8.151)
4πr3
as will be shown in Exc. 8.3.3.1. This field interacts with the spin of the electron S
in the form,
ĤSI = −~ ~ = − µ0 [3(~
µS · B µI · r̂)(~µS · r̂) − (~ ~ I )] − 32 µ0 µ
µS · µ ~ I δ 3 (r) (8.152)
~S · µ
4πr3
µ0 µB µN S I
= ge gp [3(I · r̂)(S · r̂) − (S · I)] + 32 µ0 ge µB · gp µN δ 3 (r) ,
4πr3 ~ ~ ~ ~
inserting the expressions (8.97) and (8.143). The first term gives the energy of the
nuclear dipole in the field due to the electronic orbital angular momentum. The
second term gives the energy of the ’finite distance’ interaction of the nuclear dipole
with the field due to the electron spin magnetic moments. The final term, often known
as the Fermi contact term relates to the direct interaction of the nuclear dipole with
the spin dipoles and is only non-zero for states with a finite electron spin density at
the position of the nucleus (those with unpaired electrons in s-subshells).
We now discuss the two cases in which L = 0 or L 6= 0 separately in the following
subsections [146, 614, 374].
because the s-orbitals have a high probability at the nuclear region, but fall off quickly
at larger distances. Hence,
S I
ĤSI ' 23 µ0 ge µB · gp µN δ 3 (r) . (8.153)
~ ~
Defining the complete total angular momentum of the atom,
F≡I+S , (8.154)
`=0 2µ0 ge gp µB µN
∆Ehf s = h(S, I)F, mF |ĤSI |(S, I)F, mF i = hS · Iihδ 3 (r)i (8.155)
3~2 Z
∗
= 32 µ0 ge gp µB µN [F (F + 1) − I(I + 1) − S(S + 1)] ψn00 (r)δ 3 (r)ψn00 (r)d3 r
1 Z 3/2 2
2
= 3 µ0 ge gp µB µN [F (F + 1) − I(I + 1) − S(S + 1)] π naB
√ .
As an example consider the hyperfine structure of the state 1s1/2 of the hydrogen
atom. With J = I = 12 and Z = n = 1 we obtain (see Exc. 8.3.3.2),
3
L=0 `=0 2 1 Z
∆Ehf s (F = 1) − ∆Ehf s (F = 0) = 3 µ0 ge µB gp µN 2 π naB (8.156)
2ge gp m2e c2
= α4 ≈ (2π~) · 1.420 GHz .
3mp
F≡I+J , (8.159)
260 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
N·J J I·J J
N −→ , I −→ . (8.160)
|J| |J| |J| |J|
We get for the coupling between the projected spins (8.160) of the electronic layer
and the nucleus,
(N · J)(I · J) (N · J)(F2 − I2 − J2 )
N · I −→ = . (8.161)
|J|2 2|J|2
We calculate
L6=0
∆Ehf s = h((L, S)J, I)F, mF |ĤJI |((L, S)J, I)F, mF i (8.162)
µ0 µB µN N·I
= ge gp
4π ~ ~ r3
3
µ0 µB µN N · J[F (F + 1) − I(I + 1) − J(J + 1)] Z n
−→ ge gp .
4π ~ ~ 2J(J + 1) aB n4 L(L + 12 )(L + 1)
L6=0 AJ
∆Ehf s = 2 [F (F + 1) − J(J + 1) − I(I + 1)] . (8.164)
Note, that the J · I-coupling breaks the degeneracy of J in the hydrogen atom, but
not of Jz . We can derive the following interval rule,
Besides the magnetic interaction between the angular momenta of the nucleus
and the electronic shell there is an interaction between the nucleus, when it is not
spherically symmetric, and the shell. This interaction causes deviations from the
interval rule and an additional splitting of the hyperfine states.
1 e2 1 e2
Ĥqud = − − , (8.166)
4π0 |re − rN | 4π0 |re |
8.3. HYPERFINE STRUCTURE 261
Table 8.1: List of atomic data [678] showing the natural linewidth of the D2 line,
frequencies of the D1 and D2 lines, and the hyperfine splitting.
where re is the electronic coordinate and rN is the nuclear coordinate, both having
their origin in the center mass of the nucleus. For re > rN this interaction can be
obtained after several mathematical steps as [482],
where the constants AJ and BJ depend on the atom and the total electronic angular
momentum.
262 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
In Excs. 8.3.3.3 and 8.3.3.4 we determine the hyperfine structures of sodium and
rubidium atoms.
8.3.3 Exercises
8.3.3.1 Ex: Field of a magnetic moment
a. Calculate the vector potential A(r) and the magnetic dipole moment µ ~ produced by
an orbiting electron by Biot-Savart’s law using the expansion of |r − r0 |−1 in Legendre
polynomials.
~
b. Calculate the magnetic field B(r).
8.3.3.2 Ex: Probability for finding the electron near the nucleus
Calculate the expectation value hδ (3) (r)in`m for encountering the electron of a hydro-
gen atom close to the nucleus.
2~
τ= = 124.4 ps (8.170)
me c2 α5
and decays preferably in two gamma rays with energy of 511 keV each (in the center-
of-mass). The triplet state with parallel spins (S = 1, Ms = −1, 0, 1) is known as
ortho-positronium (o-Ps) and denoted as 3 S1 . It has an average life of 138.6 ns, and
the most common form of decay produces three photons. Other forms of decay are
negligible. For example, the decay channel producing five photons is 10−6 times less
likely. Measurements of these lifetimes and the positronium energy levels have been
used in precision tests of quantum electrodynamics.
While the precise calculation of the positronium energy levels is based on the
Bethe-Salpeter equation, the similarity between positronium and hydrogen allows for
an approximate estimate. In this approach, the energy levels are suppodsed to be
264 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
different from those of hydrogen because of the difference in the value of the reduced
mass µ, used in the energy equation. Since µ = me /2 for positronium, we have
and the binding energy and the energies of excitation are greater for the same reason.
F.ex. while for H = p+ e− the transition 2S − 2P1/2 is at 10 eV , 121 nm, for p+ µ−
it is at 1900 eV. Muonic atoms are interesting because they have amplified Lamb
shifts, hyperfine interactions, and quantum electrodynamical corrections. Therefore,
the displacement due to the finite distribution of charges in the proton rp = 0.8 fm
should influence the spectrum. While in p+ e− the 2S level is shifted upward by the
Lamb shift by a value of 4.4×10−6 eV, in p+ µ− it is shifted down by a value of 0.14 eV.
In Exc. 8.4.5.2 we calculate the spectrum of the muonic hydrogen and in Exc. 6.1.3.3
we compare the energy corrections due to the finite extension of the nuclei for muonic
and for standard hydrogen in first order TIPT.
leading to less and less spaced levels. These Rydberg states form the Rydberg series.
e2
mωn2 = and mωr2 = n~ . (8.174)
4π0 r
Eliminating r,
me4
ωn = . (8.175)
(4π0 )2 n3 ~3
266 CHAPTER 8. ELECTRON SPIN AND THE ATOMIC FINE STRUCTURE
Radiation of this frequency will be emitted by an atomic antenna. On the other hand,
the Bohr model predicts frequencies between orbitals,
En+1 − En me4 1 1 n→∞ me4 2
ωn = = − −→ . (8.176)
~ 2(4π0 ) ~
2 2 (n + 1) 2 n 2 2(4π0 ) ~ n3
2 2
e2
Ucou = − . (8.177)
4πε0 r
The similarity of the effective potential ’seen’ by the outer electron and the authentic
hydrogen potential suggests a classical treatment within the planetary model. There
are three notable exceptions:
An atom can have two (or more) electrons in highly excited states with com-
parable orbital radii. In this case, the electron-electron interaction gives rise to
a significant deviation from the hydrogen potential. For an atom in a multiple
Rydberg state the additional term Uee includes a sum over each pair of highly
excited electrons:
e2 X 1
Uee = . (8.178)
4πε0 i<j |ri − rj |
If the valence electron has very low angular momentum (interpreted classically
as an extremely eccentric elliptical orbit), it can pass close enough to the nucleus
to polarize it, giving rise to an additional term,
e2 αd
Upol = − . (8.179)
(4πε0 )2 r4
If the outer electron penetrates the inner electronic shells, it sees more of the
charge of the nucleus and therefore feels a larger force. In general, the modifi-
cation of the potential energy is not simple to calculate and should be based on
some knowledge of the nucleus’ geometry.
In hydrogen the binding energy is given by:
E1
EB = − . (8.180)
n2
8.5. FURTHER READING 267
The binding energy is weak at high values of n, which explains the fragility of the
Rydberg states that can easily be ionized, e.g. by collisions.
Additional terms modifying the potential energy of a Rydberg state require the
introduction of a quantum defect, δ`, in the expression for the binding energy:
E1
EB = − . (8.181)
(n − δ` )2
The long lifetimes of Rydberg states with high orbital angular momentum can be
explained in terms of overlapping wavefunctions. The wavefunction of an electron in
a state with high ` (large angular momentum, ’circular orbit’) has little overlap with
the wavefunctions of the internal electrons and therefore stays relatively unperturbed.
Also, the small energy difference between adjacent Rydberg states decreased the decay
rate according to the result (13.37).
8.4.5 Exercises
8.4.5.1 Ex: Positronium
Calculate and compare the fine and hyperfine structure of positronium.
F = q E~ + q ṙ × B
~, (9.1)
where
∂A
E~ = −∇Φ − and ~ =∇×A ,
B (9.2)
∂t
where Φ and A are called scalar and vector potential, respectively.
It is important to realize here, that the momentum p not only involves the mo-
mentum of the particle mv, but the field also carries a momentum qA(r). As we
learned in electrodynamics it is possible to derive this force from a Lagrangian for the
electronic motion,
m
L(ri , ṙi ) = ṙ2 − qΦ(r) + q ṙ · A(r) . (9.3)
2
With this aim we first determine the momentum by,
∂L
pi = = mṙi + qAi , (9.4)
∂ ṙi
269
270 CHAPTER 9. ATOMS WITH SPIN IN EXTERNAL FIELDS
That is,
1
H(ri , pi ) = (p − qA)2 + qΦ . (9.6)
2m
The following equations hold,
∂H ∂H
ṙi = and ṗi = − . (9.7)
∂pi ∂ri
The first equation is easily verified by inserting the Hamiltonian (9.5). The second
leads to the Lorentz force,
∂H ~ i,
Fi = mv̇i = ṗi − q Ȧi = − − q Ȧi = qEi + q(v × B) (9.8)
∂ri
where the last step of the derivation will be shown in the Exc. 9.1.3.1 using the
Coulomb gauge ∇ · A = 0.
mv −→ p − qA and H −→ H + qΦ . (9.9)
This substitution rule, called minimal coupling, can be applied in quantum mechanics,
The fourth term called diamagnetic term is quadratic in A and usually so small
that it can be neglected. The second and third terms describe the interaction of the
electron through its momentum p̂ with the potential vector A produced by magnetic
moments inside the atom or outer magnetic fields. Within the Coulomb gauge we
have (∇ · A)ψ = (A · ∇)ψ + ψ(∇ · A) = (A · ∇)ψ, such that,
e
Ĥint = me A · p̂ . (9.12)
χ(r, t) ≡ −A(r, t) · r ,
assuming that the potential only weakly varies in space, such that,
∂χ(r, t) ∂A(r, t) ~ t) ,
∇χ(r, t) ' −A(r, t) and = −r · = −r · E(r,
∂t ∂t
9.2. INTERACTION WITH MAGNETIC FIELDS 271
1 ∂χ(r, t)
Ĥ = (p̂ − qA + ∇χ)2 + qΦ + q (9.13)
2me ∂t
p̂2 ~ t) ,
' + qΦ − d · E(r,
2me
9.1.3 Exercises
9.1.3.1 Ex: Lagrangian of an electron in the electromagnetic field
a. Show that the Lagrangian (9.3) reproduces the Lorentz force (9.1).
b. Show that the Hamiltonian (9.5) reproduces the Lorentz force (9.1).
Thus the interaction energy between the electron and the field is given by the Hamil-
tonian (9.12),
µB
∆Ezee (B) = − Bhn, L, mL |L̂z |n, L, mL i = −µB mL B . (9.16)
~
In the Excs. 9.2.8.1 and 9.2.8.2 we represent the interaction between an atomic an-
gular momentum and a magnetic field in different bases characterized by different
quantization axes.
272 CHAPTER 9. ATOMS WITH SPIN IN EXTERNAL FIELDS
~
before projecting the result onto the B-field. The potential is,
~ = − µB (L̂ + 2Ŝ) · B ~ −→ − µB L̂ · Ĵ
h i
V̂zee (B) = −~
µJ · B Ĵ
·B ~ + 2 Ŝ · Ĵ Ĵ
·B ~
~ ~ |J| |J| |J| |J|
µB h i
~ µB 1 h 2 2 2 2 2 2
i
~
=− L̂ · Ĵ + 2 Ŝ · Ĵ Ĵ · B = − Ĵ + L̂ − Ŝ + 2( Ĵ + Ŝ − L̂ ) Ĵ · B
~|J|2 ~|J|2 2
µB 1 3Ĵ2 − L̂2 + Ŝ2 ~.
=− Ĵ · B (9.19)
~ 2 |Ĵ|2
Therefore,
µB h i
V̂pb (B) = − ~ −→ − µB L̂ ·
(L̂ + 2Ŝ) · B
~
B ~
B
+ 2 Ŝ ·
~
B ~
B ~ ,
·B (9.24)
~ ~ ~
|B| ~
|B| ~
|B| ~
|B|
9.2. INTERACTION WITH MAGNETIC FIELDS 273
Figure 9.1: Coupling angular moments for the effect (a) Normal Zeeman effect, (b) anoma-
lous Zeeman effect, (c) Paschen-Back effect, (d) Zeeman effect of the hyperfine structure,
and (e) Paschen-Goudsmith effect.
such that
∆Epb (B) = −µB (mL + 2mS )B . (9.25)
The derivations we have made so far have focused on simple situations well de-
scribed by CSCOs in various coupling schemes. The projections on the different
quantization axes [the total spin (9.18) in the Zeeman case or the applied magnetic
field (9.23) in the Paschen-Back case] ensure that the Hamiltonians V̂ls and V̂zee (B) in
these CSCOs are described by diagonal matrices. However, in regimes intermediate
between Zeeman and Paschen-Back, V̂ls ' V̂zee (B), it is generally not possible to find
a diagonal representation.
In order to calculate the energy spectrum in intermediate regimes we must, there-
fore, determine all the components of the matrix,
µB
V̂ls + V̂zee (B) = ξ(r)L̂ · Ŝ + (L̂ + 2Ŝ) . (9.26)
~
Using L̂± ≡ L̂x ± ıL̂y and Ŝ± ≡ Ŝx ± ıŜy , we can easily rewrite the energy in the
following way,
µ
B ~.
V̂ls + V̂zee (B) = ξ(r) L̂z Ŝz + 21 L̂+ Ŝ− + 21 L̂− Ŝ+ + (L̂ + 2Ŝ) · B (9.27)
~
This operator acts on the uncoupled states,
Figure 9.2: Transition between the Zeeman regime and the Paschen-Back regime for the
case L = 1 and S = 1/2.
where µ
~ F is the total magnetic momentum,
µB µN
µ
~F = µ ~I = −
~J + µ gJ Ĵ + gp Î . (9.30)
~ ~
we project the nuclear spin and the total electronic momentum separately in the
direction of F̂,
F̂ F̂ F̂ F̂
Ĵ −→ Ĵ · |F| |F| and Î −→ Î · |F| |F| . (9.31)
Note the negative sign in (9.30) due to the negative charge of the electron. The
Landé factor gJ [see (9.21)] is the one caused by the coupling of the orbital angular
momentum L̂ and the electron spin Ŝ and depends on the state under consideration.
Thereby,
h µ µN i
B F̂ F̂ F̂ F̂ ~
V̂zee (B) = − gJ Ĵ · |F| + gp Î · |F| |F| B (9.32)
~ |F| ~
µB µN
= − gJ Ĵ · F̂ + gp Î · F̂ (B · F̂z ) .
~|F̂|2 ~|F̂|2
such that
∆Ehf s + ∆Ezee (B) ' ∆Ehf s + µB gF mF B , (9.34)
The Zeeman effect of the hyperfine structure becomes a hyperfine structure of the
Zeeman effect, also called Paschen-Back effect of the hyperfine structure or Paschen-
Goudsmith effect. We can diagonalize the potential on a basis, where I, mI , J, and
mJ are good quantum numbers. Using the expression (8.169) but disregarding the
quadrupolar contribution to the hyperfine interaction, BJ ' 0, we obtain,
~ ' AJ Ĵ · Î − µ
'0
V̂hf s + V̂zee (B) = V̂hf s − (~
µJ + µ~I )·B ~
~J · B (9.37)
! ~
2
!
AJ ~ B
B ~ ~ B
B ~ AJ µB ˆ
−→ 2 Ĵ · · Î · − µJz B = 2 Jˆz Iˆz + gJ Jz B ,
~ B B B B ~ ~
where we neglect the interaction of the dipole moment of the nucleus with the external
~ I ' 0. We obtain for strong magnetic fields,
magnetic field, µ
where ge = 2.002.. is the g-factor of the electron and gp = 5.586.. of the proton.
Now, in large molecules the most electrons are paired, such that Ĵ = 0. In that
case, we are left with the interaction between the nuclear spin and the applied
magnetic field,
∆Enmr (B) = µN gI mI B , (9.42)
where the g-factor of the nucleus gI must be looked up in data tables. This
is the regime where nuclear magnetic resonances (NMR) can be excited with a
large variety of applications in spectroscopy and imaging.
where
2
hm0I m0J | ηĴ
~ |mI mJ i = (mI mJ )2 + 14 I−
2 2 2 2
J+ + 41 I+ J− δm0I mI δm0J mJ + (9.45)
1 0 0
+ 2 (mI mJ + mI mJ ) I+ J− δmI
0 mI +1 δm0J mJ −1 +
1 0 0
+ 2 (mI mJ + mI mJ ) I− J+ δmI
0 mI −1 δm0J mJ +1 +
1
+ 4 I+ J− I++ J−− δmI mI +2 δmJ
0 0 mJ −2 +
1
+ 4 I− J+ I−− J++ δmI mI −2 δmJ
0 0 mJ +2 .
9.2. INTERACTION WITH MAGNETIC FIELDS 277
The matrix hm0I m0J |Hhf s + HB |mI mJ i is divided into 2F + 1 diagonal blocks, each
labeled mF . The total number of levels is,
X X X
2F +1 = (2I+1)(2J+1) = 1 .
F =|I−J|,..,I+J mF =|−F,..,F | mI =|−I,..,I|, mJ =|−J,..,J|, mI +mJ =mF
In this form the matrix can be programmed, e.g. using computational software
such as MATLAB, and all eigenvalues of the Hamiltonian for any state 2S+1 XJ and
nuclear spin I can be calculated numerically. Obviously, the eigenvalues follow from a
diagonalization of the matrix and do not depend on the chosen base. Fig. 9.3 shows the
result obtained for 6 Li (I = 23 ) in the state 2s 2 P3/2 knowing that AJ /h = −1.17 MHz
and BJ = 0.
400 6
20
Li, 2 S1/2 6
Li, 2 P3/2
200 10
(MHz)
(MHz)
0 0
ν
-200 -10
-400 -20
0 100 200 0 2 4
B (G) B (G)
Figure 9.3: (code) Hyperfine and Zeeman structure of the state 2 P3/2 of 6 Li.
That is, the hyperfine structure does not depend on the F quantum number
at any B-field amplitude, which means that the mJ , mI , and mF will be good
quantum numbers at arbitrary B-field strengths. So see this, we calculate the
Landé-factor (9.35),
J µN I
gF = gJ − gJ ,
F µB F
and consider the particular Zeeman state mF = F , mJ = J, and mI = I,
The energy displacement of the fully stretched states is always linear in√the
magnetic field. We can also look at the matrix elements I+ = 0 and I− = 2I
and note that all non-diagonal terms vanish.
When one of the spins, J or I, is equal to 1/2 only two possible hyperfine states
exist: F = I ± J. For this case there is an approximate analytic formula called the
Breit-Rabi formula [44], which will be derived in Exc. 9.2.8.4,
~ · Ĵ − gI µN B
∆Ehf s + ∆Ezee (B) = h A~2J Î · Ĵ + gJ µB B ~ · Îi (9.46)
1
r
AJ AJ (I + 2 ) 4mF
=− + µN gN mF B ± 1+ x + x2 ,
4 2 2I + 1
~ = Bêz
B (9.47)
Ĥ = 1
2m (p̂ − q Â)2 , (9.48)
where p̂ is the operator of the canonical momentum and  is the potential vector,
~ = ∇ × Â. The vector potential,
related to the magnetic field by B
−By
A= 1 Bx (9.49)
2
0
reproduces the field (9.47). However, we have the freedom in choosing the potential
vector, given by the gauge transformation, to add the gradient of a scalar field, for
example,
By
χ ≡ 12 Bxy =⇒ ∇χ = 1 Bx =⇒ A0 ≡ A + ∇χ = Bxêy . (9.50)
2
0
The potential vector A0 gives the same magnetic field and only changes the general
phase of the wavefunction, but the physical properties do not change. In this gauge,
9.2. INTERACTION WITH MAGNETIC FIELDS 279
The operator p̂y commutes with this Hamiltonian, since the ŷ operator is absent due
to the choice of the gauge. Thus, the operator p̂y can be replaced by its eigenvalue
~ky . Hence, by introducing the cyclotron frequency,
qB
ωc ≡ , (9.52)
mc
we obtain,
2
p̂2x mωc2 ~ky
Ĥ = + x̂ − . (9.53)
2m 2 mωc
This is exactly the Hamiltonian of the quantum harmonic oscillator, except that the
minimum of the potential is displaced in position space by the value,
~ky
x0 ≡ . (9.54)
mωc
To find the energies, we note that the translation of the potential of the harmonic
oscillator does not affect the energies. The energies of this system are therefore
identical to those of the standard quantum harmonic oscillator,
En = ~ωc n + 12 , (9.55)
for n ≥ 0. Since the energy does not depend on the quantum number ky , we will
have degeneracy. To derive the wavefunctions, we remember that p̂y commutes with
the Hamiltonian. Then the wavefunction splits into a product of eigenstates of the
momentum in y-direction and eigenstates of the harmonic oscillator |φn i shifted by a
value x0 in x-direction:
Ψ(x, y) = eıky y φn (x − x0 ) . (9.56)
That is, the state of the electron is characterized by two quantum numbers, n and ky .
Each set of wavefunctions with the same n is called Landau level. Effects due
to Landau levels are only observed, when the average thermal energy is lower than
the separation of the energy levels, which means that low temperatures and strong
magnetic fields are required. Each Landau level is degenerate, because of the second
quantum number ky , which can adopt the values,
2πN
ky = , (9.57)
Ly
with N ∈ N. The allowed values of N are further restricted by the condition that the
center of mass of the oscillator, x0 , must be physically inside the system, 0 ≤ x0 < Lx .
Using (9.54) this gives the following range for N ,
mωc Lx Ly
0≤N < . (9.58)
2π~
280 CHAPTER 9. ATOMS WITH SPIN IN EXTERNAL FIELDS
For particles with charge q = Ze, the upper limit in N can simply be written as
a ratio of fluxes,
Φ ZBLx Ly
Z = = Nmax , (9.59)
Φ0 (hc/e)
where Φ0 = h/2e is the fundamental flux quantum and Φ = BA the flux through the
system (with area A = Lx Ly ). Thus, for particles with spin S, the maximum number
of particles per Landau level is,
Φ
Nmax = Z(2S + 1) . (9.60)
Φ0
of the Landau levels is done via the quantum Hall effect. To discuss this effect let
us briefly recapitulate the Hall effect. In the scheme of Fig. 9.4, charges are deviated
by the Lorentz force exerted by an applied magnetic field B ~ from a driven current
density jx into a current density jy until a sufficient amount of surface charge density
has accumulated to generate an electric field exerting a Coulomb force on the charges
which neutralizes the Lorentz force, F = q(E~ + v × B) ~ = 0. Resolving this condition
by v, we obtain for the current density,
E
j = %qv = %q êx = ς E~ , (9.61)
B
where the last equation is Ohm’s law and
ςxx ςxy 1 ςxx −ςxy
ς= and % = ς −1 = 2 2
(9.62)
−ςxy ςyy ςxx + ςyy ςxy ςyy
the conductivity and the resistivity, respectively. The Hall resistivity does therefore
depend linearly on the magnetic field,
E B
%xy = = . (9.63)
j %q
9.2. INTERACTION WITH MAGNETIC FIELDS 281
2π~ 1
%xy = with ν∈Z. (9.64)
e2 ν
(h/e2 ) 1
0.5
xy
0
0 5 10
B (T)
9.2.8 Exercises
9.2.8.1 Ex: Zeeman effect with different quantization axes
The Zeeman effect can be described in several ways depending on the choice of the
quantization axis. Consider a magnetic field B ~ = Bx êx and calculate the interaction
Hamiltonian V (B) = −~ ~
µJ · B
a. choosing the quantization axis êx in the direction of the magnetic field,
b. choosing the quantization axis êz perpendicular to the direction of the magnetic
field.
more adequate.
c. Show that in the regime ~A µB B, it is convenient to use the representation that
privileges the individual spins of the total momentum.
d. Analyze the intermediate regime ~A ' µB B.
For the level P3/2 the interval factor is less. In particular for 6 Li it is so small
that we are immediately in the Paschen-Back scheme. This means that the matrix
p p p p
∞ hm̃J m̃I |mJ mI iB = δmp pδ p p is diagonal. The element of the transition matrix
J ,m̃J mI ,m̃I
is then,
p p (Eκ)
B hmJ mI |Tq |msJ msI iB
X X s s s s p p (Eκ)
= p p p p
∞ hm̃J m̃I |mJ mI iB ∞ hm̃J m̃I |mJ mI iB ∞ hm̃J m̃I |Tq |m̃sJ m̃sI i∞
m̃s
J
m̃s
I
p p
m̃J m̃I
X s s s s p p (Eκ)
= ∞ hm̃J m̃I |mJ mI iB ∞ hmJ mI |Tq |m̃sJ m̃sI i∞ .
m̃s
J
m̃s
I
The matrix elements in the pure Zeeman regime can be expressed by [Deh07, un-
published],
p p (Eκ)
Discuss the pure Paschen-Back regime via ∞ hmJ mI |Tq |msJ msI i∞ .
9.3. INTERACTION WITH ELECTRIC FIELDS 283
This is the Stark effect. This effect is usually weak, and its observation requires strong
fields or high spectral resolution. Stationary perturbation theory TIPT gives,
Z
(1) (0) (0)
En = hψn | − d̂ · E|ψn i = eEz z|ψn(0) |2 d3 r = 0 , (9.66)
R3
with d = −er and E~z = Ez êz . This only applies when the states have well-defined
parity and are NOT degenerate in `. When they ARE degenerate in `, which is the
case of hydrogen, the states have no defined parity (−1)` . For example, the states s
and p contributing to the same state |ψn,j i have different parities. In this case, the
condition (9.66) is not automatically satisfied, and the first perturbation order yields
a value. This is the case of the linear Stark effect. In the Excs. 9.3.2.1 and 9.3.2.2 we
explicitly calculate the Stark energy shift for a hydrogen atom subject to an electric
field.
Other atoms do not have this degeneracy, and we must calculate the quadratic
Stark effect in second order TIPT,
X (0) (0)
(0) hψn0 |ẑ|ψn i
|ψn(1) i = eEz |ψn0 i . (9.67)
En − En0
n0 6=n
and
X |hψ (0) (0) 2
n0 |ẑ|ψn i|
En(2) = e2 Ez2 . (9.68)
0
En − En0
n 6=n
To simplify the matrix elements, we separate the radial part from the angular part,
Z ∞ Z
(0)
hψn0 |ẑ|ψn(0) i = hn0 J 0 m0J |ẑ|nJmJ i = r3 Rn0 J 0 RnJ dr YJ∗0 m0 zr YJmJ dΩ . (9.69)
J
0
and called the irreducible matrix element, no longer depends on the magnetic quantum
number. On the other hand, the angular part may be expressed by Clebsch-Gordan
coefficients, as will be discussed more extensively in Sec. 13.2.3. The result is called
Wigner-Eckart theorem,
Z
hn0J J 0 m0J |ẑ|nJ JmJ i 1 J 1 J0
= YJ∗0 m0 zr YJmJ dΩ = . (9.71)
hn0J J 0 ||ẑ||nJ Ji J 0
2J + 1 mJ 0 −m0J
With [ẑ, L̂z ] = 0, which was shown in Exc. 4.3.4.2, and obviously [ẑ, Ŝz ] = 0 we
find,
0 = hJ 0 m0J |[ẑ, Jˆz ]|JmJ i = (mJ − m0J )hJ 0 m0J |ẑ|JmJ i . (9.72)
This means that for mJ 6= m0J , the matrix elements hJ 0 m0J |ẑ|JmJ i should disappear.
Therefore, the matrix is diagonal in mJ . We consider dipole transitions with |J −J 0 | ≤
1 1,
J 1 J +1 (J + 1)2 − m2J
= , (9.73)
mJ 0 −mJ (2J + 1)(J + 1)
J 1 J m2J
= ,
mJ 0 −mJ J(J + 1)
J 1 J −1 J 2 − m2J
= .
mJ 0 −mJ J(2J + 1)
States with the same |mJ | lead to the same quadratic Stark effect,
∆E ∼ A + B|mJ |2 . (9.74)
The factors A and B depend on the main quantum number n and also on L, S, J.
Moreover, they depend on the energy distance of all contributing levels, because of
the denominator in the perturbation equation (9.67). Only levels with different parity
(−1)L contribute. The formulae (9.73) will be derived explicitly in Exc. 9.3.2.3.
9.3.2 Exercises
9.3.2.1 Ex: Stark effect in hydrogen
Consider the hydrogen atom immersed in a uniform electric field E~ applied along the
êz -direction. The term corresponding to this interaction in the total Hamiltonian
is Ĥ (1) = −eE~ẑ. For typical electric fields produced in laboratory, the condition
Ĥ (1) Ĥ0 , which allows the use of TIPT, is satisfied. The effect of the perturbation
Ĥ (1) , called Stark effect, is the removal of the degeneracy of some of the hydrogen
atom states. Calculate the Stark effect for the state n = 2.
2 Ultimately, all this is simply a consequence of the uncertainty principle, which forbids us to
287
288 CHAPTER 10. ATOMS WITH MANY ELECTRONS
However, the fact that this state is never observed with two electrons shows,
that theory must be corrected to allow a true description of reality. Will deepen
this argument in Exc. 10.1.3.1.
We need to construct the total wavefunction in another way. Let us consider linear
combinations of Ψ(1,2) ,
ΨS,A ≡ √1 (Ψ(1,2)
2
± Ψ(2,1) ) = √1
2
[ψa (x1 )ψb (x2 ) ± ψa (x2 )ψb (x1 )] . (10.2)
That is, for a symmetric system, the probability of finding two particles at the same
location is doubled, whereas for an anti-symmetric system, this probability is zero.
Wolfgang Pauli showed that the (anti-)symmetric character is related to the spin
of the particles. Particles with integer spin called bosons must be symmetric. Par-
ticles with semi-integer spin called fermions must be anti-symmetric. Electrons are
fermions. Therefore, in an atom, they can not be in the same state (location), but
must be distributed over a complicated shell of orbitals. We note, that this applies
not only to elementary particles, but also to composed particles such as, for exam-
ple, atoms. We will determine in Exc. 10.1.3.2 the bosonic or fermionic character of
several atomic species.
and the wavefunction need not be anti-symmetric. But if we exchange the spin along
with the position, the particles must be indistinguishable. This must be taken into
account in the wavefunction by assigning a spin coordinate, ψa (x1 , s1 ). The exchange
operator should now be generalized,
Px,s Ψ(1,2) ≡ Px,s ψa (x1 , s1 )ψb (x2 , s2 ) = ψa (x2 , s2 )ψb (x1 , s1 ) = Ψ(2,1) . (10.6)
We now assume that the electrons not only do not interact with each other, but
there is also no interaction between the position and the spin of each electron. That is,
for a while we will discard L · S-coupling 4 . We can then write the total wavefunction
of an electron as the product of a spatial function, ψ(x), and a spin function, χ(s) =
α ↑ +β ↓, where α and β are probability amplitudes of finding the electron in the
respective spin state, such that,
ψ(x, s) = ψ(x)χ(s) . (10.7)
For two particles, the total spin function is,
X (1,2) = χa (s1 )χb (s2 ) . (10.8)
The (anti-)symmetrized version is
X S,A = √1 (X (1,2)
2
± X (2,1) ) = √1 [χa (s1 )χb (s2 )
2
± χa (s2 )χb (s1 )] , (10.9)
as we have already seen in Sec. 4.4.1. Since there are only two spin directions, there
are four possibilities to attribute the spins ↑ and ↓ to the functions χm (sn ),
↑↑ = χ1,1
S 1
X = √
2
(↑↓ + ↓↑) = χ1,0 and X A = √12 (↑↓ − ↓↑) = χ0,0 (10.10)
↓↓ = χ1,−1
For the total wavefunction, which must be anti-symmetric for electrons, there are two
possibilities,
Ψ X = 12 (Ψ(1,2) + Ψ(2,1) )(X (1,2) − X (2,1) ) = √1 [ψa (x1 )ψb (x2 )
S A
2
+ ψa (x2 )ψb (x1 )]χ0,0
χ1,1
A
Θ = A S 1 (1,2) (2,1) (1,2) (2,1) 1 .
Ψ X = 2 (Ψ −Ψ )(X +X ) = 2 [ψa (x1 )ψb (x2 ) − ψa (x2 )ψb (x1 )] χ1,0
√
χ
1,−1
(10.11)
That is, the two electrons may be in a triplet state with the anti-symmetric spatial
wavefunction, or in a singlet state with the symmetric spatial wavefunction 5 .
How to generalize these considerations to N particles? The symmetric wavefunc-
tions contain all permutations of the label ak , where we understand by ak the set of
quantum numbers unambiguously specifying the state of the particle k,
X
ΘS = N ψa1 (x1 )...ψaN (xN ) , (10.12)
Px,s ak
4 In the case of L · S-coupling, the total wavefunction can not be written as a product of spatial
and spin functions, but it must be anti-symmetric anyway.
5 In the coupled image, the total spin S = s +s can have the following values S = |s −s |, .., s +
1 2 1 2 1
s2 = 0, 1. In the case S = 0 the magnetic quantum number can only have one value (singlet), mS = 0.
In the case S = 1 it can have three values mS = −1.0, +1 (triplet) (see Exc. 9.2.8.3).
290 CHAPTER 10. ATOMS WITH MANY ELECTRONS
The Slater determinant is zero, when two sets of quantum numbers are identical,
ai = aj . For example, for two electrons in an electronic shell, |ni , li , mi , si i =
|nj , lj , mj , sj i. This is Pauli’s strong exclusion principle:
The total wavefunction must be antisymmetric with respect to the ex-
change of any pair of identical fermions and symmetrical with respect to
exchange of any pair of identical bosons.
Pauli’s weak exclusion principle (usually sufficient for qualitative considerations) says
that two fermions in identical states can not occupy the same region in space. That
is, their Broglie waves interfere destructively, as if Pauli’s principle exerted a repulsive
interaction on the particles. This ’force’ has a great impact on the phenomenology of
the bonds between atoms, as we will discuss in the following sections.
Ψ = {ψa (x1 )ψa (x2 ), ψa (x1 )ψb (x2 ), ψb (x1 )ψa (x2 ), ψb (x1 )ψb (x2 )} (10.15)
with the same probability of p = 1/4. When the quantum particles approach each
other, x1 ' x2 , they must become indistinguishable. Bosonic indistinguishable parti-
cles can stay in one of the following three states,
Ψ = {ψa (x1 )ψa (x2 ), √12 [ψa (x1 )ψb (x2 ) + ψb (x1 )ψa (x2 )], ψb (x1 )ψb (x2 )} (10.16)
Ψ = { √12 [ψa (x1 )ψb (x2 ) − ψb (x1 )ψa (x2 )]} (10.17)
different physics as we deal with systems of large numbers of particles, as we can see
in the cases of the free electron gas and the Bose-Einstein condensate.
We finally note a result of the standard model of particle physics assigning a
fermionic character to all fundamental constituent particles of matter while the me-
diators of fundamental forces are always bosons.
10.1.3 Exercises
10.1.3.1 Ex: Indistinguishability of particles
R L/2 R L/4
Consider the observable quantity 0 0
|Ψ(x1 , x2 )|2 dx1 dx2 for the case of the
wavefunction defined in (10.1) and show, that it does depend on particle exchange.
10.2 Helium
The simplest atom to discuss Pauli’s principle is helium. The helium atom has a
charged nucleus Z = +2e and mass mHe ≈ 4mH .
and we get two Schrödinger equations, the Hamiltonian being equal to the one of
hydrogen-like atoms:
2
~ 2 e2 Z
− ∇i − Ψi (ri ) = En(i) Ψi (ri ) , (10.19)
2µ 4πε0 ri
with i = 1, 2. For hydrogen-like atoms we have,
1 1
E= En(1) + En(2) = EB Z 2
+ 2 , (10.20)
n21 n2
292 CHAPTER 10. ATOMS WITH MANY ELECTRONS
with EB = −13.6 eV. With this, we get the energy for the ground state:
The value predicted by Bohr’s model is far from experimental reality: The ioniza-
tion energy measured for the first electron is 24.6 eV, for the second 54.4 eV, totalizing
a binding energy for two electrons of −78.983 eV . This corresponds to an error of
about 38%. The lower energy of the first electron is due to the shielding of the nucleus
by the second.
e2
∆E = hn1 , `1 , m`1 ; n2 , `2 , m`2 | |n1 , `1 , m`1 ; n2 , `2 , m`2 i . (10.22)
4πε0 r12
This correction is called the Coulomb integral and has the value:
Z
e2 2 1
∆E = |Ψn1 ,`1 ,m`1 (r1 )| |Ψn2 ,`2 ,m`2 (r2 )|2 dV1 dV2 . (10.23)
4πε0 r12
This integral is always positive. The term |Ψn1 ,`1 ,m`1 (r1 )|2 dV1 is the probability of
finding the electron inside the volume element dV1 and, when multiplied by −e, gives
the charge associated with that region. Thus, the integral represents the Coulombian
interaction energy of the confined charges within the two volume elements dV1 and
dV2 . ∆E is the total contribution to the potential energy. Calculating the Coulomb
integral for the ground state, which will be done in Exc. 10.2.3.1, we obtain,
5Z e2 5Z
∆E = = EB , (10.24)
4 4πε0 2aB 4
with aB the Bohr radius. ∆E corresponds to 34 eV. Thus, the ground state energy is
EHe (1s) = −108.8 eV + 34 eV = −74.8 eV. Comparing with the experimental value
of −78.983 eV we have an error around 5.3%.
EB
En = −ζ , (10.25)
n2
10.2. HELIUM 293
and the energy for the ground state becomes, assuming total shielding, B = 1,
Comparing with the experimental value of −78.983 eV we have an error around 15%.
For a shielding constant of around B = 0.656 the experimental value is reproduced.
This means that the effective nuclear charge felt by the second electron is only partly
shielded by the former. The TPIT method (10.22) and the shielding concept (10.25)
can be combined in a variational calculation, where the effective charge ζ is the vari-
ational parameter. In Exc. 10.2.3.2 we study the reciprocal shielding of the electrons
at the example of the helium-type ion H− .
Figure 10.1: Helium levels for the excitation of the first electron and allowed singlet and
triplet transitions. Note that the state (1s)↑↑ does not exist.
As we have seen in the discussion of the fine structure of hydrogen, the energy of
the l · s-coupling given by (8.131) is ∝ En (Zα)2 ∝ Z 4 . For helium which still has
a small Z, the energy of the coupling is weak (∼ 10−4 eV), so that we can count
on a direct coupling of the spins of the two electrons. If in an excited state the
294 CHAPTER 10. ATOMS WITH MANY ELECTRONS
ΘA
± =
√1
2
[ψ100 (r1 )ψ200 (r2 ) ± ψ100 (r2 )ψ200 (r1 )] · χA,S , (10.27)
where the (+) sign holds for χA (S = 0) and the (−) sign for χS (S = 1). The energies
are,
Z Z
e2
∆E S,A = 12 dr13 dr23 Θ∗A ± Θp mA (10.28)
4πε0 |r1 − r2 |
Z Z
e2
= 12 dr13 dr23 [|ψ100 (r1 )|2 |ψ200 (r2 )|2 + |ψ100 (r2 )|2 |ψ200 (r1 )|2 ]
4πε0 |r1 − r2 |
Z Z
e2
± 21 dr13 dr23 2ψ ∗ (r1 )ψ200∗
(r2 )ψ100 (r2 )ψ200 (r1 )
4πε0 |r1 − r2 | 100
≡ ∆Ecoulomb ± ∆Eexchange .
is the Coulomb energy (10.23) between the electronic orbitals. We note that this part
can be calculated from the Hamiltonian using non-symmetrized orbitals. The second
integral,
Z Z
e2
∆Eexchange = dr1 dr23
3
ψ ∗ (r1 )ψ200
∗
(r2 )ψ100 (r2 )ψ200 (r1 ) , (10.30)
4πε0 |r1 − r2 | 100
Hamiltonian,
p21 p2
Ĥ S,A = + 2 + V (r1 ) + V (r2 ) + V (|r1 − r2 |) ± ∆Eexchange , (10.31)
2m 2m
but only through the symmetry character of the spatial wavefunction. On the other
hand, on a much smaller energy scale, the spin enters through the L · S-interaction.
The potential is not spherically symmetric, the term r12 depends on the angle
between r1 and r2 . Thus, the total wavefunction Ψ(r1 , r2 ) is not separable into a
radial and an angular part. By consequence, unlike for hydrogen, the Schrödinger
equation with the Hamiltonian (10.21) has no analytical solution.
Example 50 (TIPT for excited helium states): We consider the two elec-
trons of a helium atom occupying different orbits described by wavefunctions
denoted by ψa (1) ≡ ψn1 ,`1 ,m`1 (r1 ) and ψb (2) ≡ ψn2 ,`2 ,m`2 (r2 ). Applying
the Hamiltonian without the interelectronic interaction term, the total states
Θ = ψa (1)ψb (2) and ψa (2)ψb (1) have the same energy Ea + Eb . To calculate
the energy correction, we use TIPT for degenerate states. We have to calcu-
(1)
late the secular determinant det(hn, ν|H (1) |n, µi − En,µ δµ,ν ). The terms of the
(1)
perturbation matrix H are:
(1) e2
H11 = hψa (1)ψ( 2)| |ψa (1)ψb (2)i
4πε0 r12
(1) e2
H22 = hψa (2)ψb (1)| |ψa (2)ψb (1)i
4πε0 r12
2
(1) e (1)
H12 = hψa (1)ψb (2)| |ψa (2)ψb (1)i = H21 .
4πε0 r12
(1) (1) (1)
The terms J ≡ H11 = H22 are Coulomb integrals. The term K ≡ H12 is
called exchange integral:
e2 1
K= hψa (1)ψb (2)| |ψa (2)ψb (1)i .
4πε0 r12
Hence, as J and K are positive, the determinant is:
J − E K
=0,
K J − E
yielding,
E (1) = J ± K .
That is, the states that were previously degenerate with energy E = Ea + Eb
are now split into two states with energies E = Ea + Eb + J ± K. And the
corresponding eigenfunctions are:
1
ΨS,A (1, 2) = √ [ψa (1)ψb (2) ± ψb (1)ψa (2)] .
2
This result shows that the repulsion between the two electrons breaks the degen-
eracy (of separable functions written in product form) into states with an energy
difference 2K. Note that the eigenfunctions are symmetric, which is discussed
in the next section.
296 CHAPTER 10. ATOMS WITH MANY ELECTRONS
If we exchange the electrons, the above integral changes sign, because ΨA (r1 , r2 )
changes sign. But the integral can not depend on the nomenclature of the integration
variables, so it must be zero. The transition between a symmetric and an antisymmet-
ric state can not occur. Looking at the spin wavefunction in Θ = ΨS χA or ΨA χS , we
find that transitions are only allowed between singlet states or between triplet states.
That is, there is a selection rule for the spin postulating ∆S = 0 7,8 .
Because of the differences observed in the singlet and the triplet spectrum of
helium, illustrated in Fig. 10.1, it was first believed that they belong to different
atomic species, called para-helium and ortho-helium. A chemical analysis showed
later that it was the same element.
10.2.3 Exercises
10.2.3.1 Ex: Helium atom
Compare the measured binding energy with the prediction of Bohr’s model considering
the inter-electronic interaction up to first order TIPT.
a product, Θ = Ψ(x)χ(s), the symmetry character is preserved for the two functions separately. The
eigenvalues of the operators Px and Ps are then good quantum numbers. But this only holds for
weak L · S-coupling. The electric dipole operator for the transition does not act on the spin (which
prevents the recoupling S = 1 ↔ S = 0 via E1-radiation) and also does not act on the symmetry
character of the orbitals (which prevents transitions ΨS ↔ ΨA ).
In principle, this holds for any species of atoms with two valence electrons. In reality however, the
influence of the L · S-coupling grows with Z, which weakens the interdiction of the intercombination
transition. In this case, only the operator Px,s yields good eigenvalues.
10.3. ELECTRONIC SHELL STRUCTURE 297
calculate the fundamental state of a given problem through a test function that is not
a solution of the original problem. This method, when applied to a helium atom, uses
as test function the solution of the problem without coulombian interaction between
the electrons, which only feel the interaction with the original charge of the nucleus.
However, this method could be further improved if we considered an effective nuclear
charge, due to its interaction with the electrons themselves, and then obtaining the
test function. Apply this correction to the case of helium. Interpret the result. Help:
Z p D1E
sin θ2 r12 + r22 − 2r12 r22 cos θ2 Z
p dθ 2 = and = .
2 2 2 2
r1 + r2 − 2r1 r2 cos θ2 r r
1 2 r aB
XZ Z Z
p2i X Ze2 X e2
Ĥ = Ekin +Vncl−ele +Vele−ele = − + , (10.33)
i=1
2m i=1 4πε0 |ri | i<j=1 4πε0 |ri − rj |
is extremely complicated to solve, even for the simplest case (Z = 2) we must use
approximation methods.
Z 2 e2 Z(Z − 1)e2
Vncl−ele ' and Vele−ele ' , (10.34)
aB 2aB
we realize that Vele−ele /Vncl−ele varies between 41 for Z = 2 and 12 for Z 1/2.
For this reason the use of alternative methods to describe multielectronic atoms is
298 CHAPTER 10. ATOMS WITH MANY ELECTRONS
necessary. Nevertheless, the set of quantum numbers derived from Bohr’s atomic
model are still the same as those used for many-electron atoms, and the orbitals are
used as starting points for more sophisticated methods.
To calculate most of the atomic properties we need reasonably realistic potentials.
The most important terms of the Hamiltonian are the Coulombian potential between
the nucleus and the electrons, Vne , being naturally spherical, and the interaction
potentials between the electrons, Vee , which we will try to approximate by a spherical
potential and treat the deviations caused by the approximation afterward. Knowing
the effect of the shielding of the nucleus by electronic charges, we already know the
asymptotes (see Fig. 10.2),
Ze2 e2
Vef f = − for r→0 and Vef f = − for r → ∞. (10.35)
4πε0 r 4πε0 r
0
V (r)/(e2 /aB )
-5
-10
0 1 2 3
r/aB
2
Figure 10.2: (code) External potential (shielded Coulombian) Vcl ∝ er (blue, upper curve),
2
interior potential (non-shielded Coulombian for Z = 4) Vbl ∝ Zer (green, lower curve), and
effective potential (red, middle curve).
complex atoms are similar to the wavefunctions of hydrogen. So, we can use these
quantum numbers n, `, m` , and ms for every electron.
However, the effective radial potential depends very much on the species and is
quite different from the Coulomb potential. So, the degeneracy in ` is lifted. In
general, electrons with small ` are more strongly bound, because they have a higher
probability of being near the nucleus, where the potential is deeper (see Fig. 10.2).
The same argument explains why electrons with small n are more strongly bound.
We will discuss these effects in more depth in Sec. 10.4.1 by comparing the excitation
levels of the valence electron in different alkalis.
We note that, with (10.50), we can identify the Lagrange multiplier µ as a chemical
potential. In particular, for non-interacting neutral atoms, we have µ = 0. In addition,
since for an atom both the potential and the electronic density must have spherical
symmetry, we can write for r 6= 0,
3/2
1 ∂2 e2 3
2
[rVef f (r)] = [−Vef f (r)]3/2 . (10.56)
r ∂r 2ε0 5C
Z 3e4/3
Vef f (r) ≡ − χ(αr) setting α≡ 2/3
Z 1/3 . (10.57)
r 5Cε0
2. There is no electronic shell structure assumed; apart from the fact that the
kinetic energy was derived in a way as to respect Pauli’s principle, the whole
calculation was done within the laws of classical electromagnetism; no quantum
mechanics was involved and, hence, no set of quantum numbers has been found.
3. Apart from the Pauli principle used to calculate the density of states (10.40),
quantum statistical effects of identical particles (such as wavefunction anti-
symmetrization) are not taken into account.
A more refined model which deals with third criticism and, in addition, is closer to
density functional theory (DFT) is the Thomas-Fermi-Dirac model.
10.3. ELECTRONIC SHELL STRUCTURE 303
With this we calculate all energies and eigenfunctions (only the radial parts are of
interest) minimizing the total energy and respecting the weak Pauli principle, that is,
we classify the states in the order of increasing energies ei and fill them successively
with electrons. For the total wavefunction we obtain,
N
! N
X X
Ĥi ΨN = En ΨN with ΨN = ψ1 · .. · ψN and En = ei . (10.62)
i=1 i=1
With the eigenfunctions we calculate the charge densities e|ψj (rj )|2 . We integrate the
interaction energy between these charge densities and the potentials exerted by the
nucleus and all other electrons j 6= i to obtain an effective potential that represents
an improved estimation for the electronic mean field,
X Z
Ze2 e2
Vef f ←− − + d3 rj |ψj (rj )|2 . (10.63)
4πε0 ri 4πε0 |ri − rj |
j6=i
We replace that potential in the SchrödingerPequation, and repeat the whole process
from the beginning, until the total energy i ei does not get any lower. This self-
consistent method is called Hartree method. Fock improved these calculations using
antisymmetric wavefunctions for the valence electrons. This method is called Hartree-
Fock method.
Z
X X
1
Ĥ = ĥi + 2 V̂ij , (10.64)
i=1 i6=j
where ĥi is the Hamiltonian only of the electron i, and V̂ij is the interaction term
between the electrons i and j. To implement the method we must suppose that the
multi-electronic state can be written as the product of the individual states of each
electron:
Ψ0 (1, ..., Z) = ψ1 (1)ψ2 (2)...ψZ (Z) , (10.65)
where ψi (1) = φi (r1 )χ(α) = ψiα (r1 ) represents the spin-orbital state of electron 1, that
is, the spatial wavefunction of the electron in the state i and with spin α. However, due
to the symmetrization postulate, the physical state of the system must be expressed
by a Slater determinant,
1
Ψ(1, ..., Z) = √ det [ψ1 (1)ψ2 (2)...ψZ (Z)] . (10.66)
Z!
Now, we use the variational principle to minimize the expectation value of the ground
state energy by varying the functions ψk (n). In this way, the correct orbitals are those
that minimize the energy. The expectation value is written as,
Z
X X
E = hΨ|Ĥ|Ψi = hΨ| ĥi |Ψi + hΨ| 21 V̂ij |Ψi . (10.67)
i=1 i6=j
Hence,
Z
X Z
X
1
E= hψi |ĥi |ψi i + 2 [hψi ψj |V̂ij |ψi ψj i − hψj ψi |V̂ij |ψj ψi i] . (10.69)
i=1 i,j
The above expression can be minimized via Lagrange multipliers under the constraint
that the states are orthogonal hψi |ψj i = δij ,
X
δ hΨ|Ĥ|Ψi − ij [hψi ψj |V̂ij |ψi ψj i − hψj ψi |V̂ij |ψj ψi i] (10.70)
i,j
10.3. ELECTRONIC SHELL STRUCTURE 305
is the Fock operator and k is the energy associated with the spin-orbital ψk . The
operator Jˆi , called Coulomb operator, represents the mean potential sensed by electron
1 in the orbital k due to the presence of electron 2 in the orbital i:
Z
Jˆi ψk (1) = ψi∗ (2)V12 ψi (2)dr2 ψk (1) . (10.72)
Once we know all wavefunctions, the energies of the orbitals can be obtained in the
following way:
Z ( ) Z
X
∗
dr1 ψk (1) ĥ1 + (2Jˆi − K̂i ) ψk∗ (1) = k dr1 ψk∗ (1)ψk (1) = k , (10.74)
i
that is, Z X
k = dr1 ψk∗ (1)ĥ1 ψk∗ (1) + (2Jˆki − K̂ki ) , (10.75)
i
where,
Z
Jˆki = dr1 ψk∗ (1)Jˆi ψk (1) is the Coulomb integral (10.76)
Z
K̂ki = dr1 ψk∗ (1)K̂i ψk (1) is the exchange integral .
Furthermore, if assuming that, taking an electron away from the orbital ψk the elec-
tronic distribution remains unchanged, it is possible to associate the energy k with
the ionization energy of the electron in this orbital, Ik ' k . This equality is known
as Koopman’s theorem.
10.3.5 Exercises
10.3.5.1 Ex: Effective potential in the Thomas-Fermi model
Calculate numerically and plot the effective potential in the Thomas-Fermi model
from the differential equation (10.59) for Z = 40 [770].
306 CHAPTER 10. ATOMS WITH MANY ELECTRONS
This only works for atoms with up to 18 electrons. When the layer 3p is completely
filled, the next to be occupied is not 3d but the 4s. The new scheme is illustrated
in Fig. 10.4. The anomalies beginning at Z = 18 arise due to electron-electron
interaction. The real potential evolves from one to the other Coulombian potential,
as distance from the nucleus is increased, as illustrated in Fig. 10.2. Near the nucleus,
the electrons shield the positive charge less than for large distances r. Thus, those
states that have a high probability near the nucleus are energetically lowered. That
is,
E2s < E2p and E6s < E6p < E6d . (10.79)
The degeneracy of the orbital angular momentum in the Schrödinger model is thus
lifted. The shielding is, as can be seen in the example of the excited states of lithium,
a large effect in the range of some eV.
Figure 10.4: Illustration of the sequence of filling the orbitals with electrons.
The shielding also accounts for the anomalies in the periodic system, such as in
K or Ca. Since E4s < E3d , the 4s state is filled before the 3d. Similar anomalies
308 CHAPTER 10. ATOMS WITH MANY ELECTRONS
also occur in Rb (5s), Cs (6s), and Fr (7s). In rare earths the shielding effect is even
more pronounced. Here, the energy of the state 6s is even below the energy of the
4f , which means that the shells 6s, 5s, 5p, and 5d protect the 4f shell very well 9 .
Resolve Exc. 10.4.4.2.
10.4.1.1 Alkalines
The electronic shell structure of alkalines consists of a completed noble gas shell and
an additional valence electron. Their spectrum is therefore very similar to hydrogen.
An empirical approach can be stressed to describe this feature,
µEG c2 Z 2 α2
En,` = − , (10.80)
2 (n − ∆n,` )2
where µEG is the reduced mass relative to the noble gas shell and ∆n,` is called
quantum defect. The quantum defect is tabulated for most alkaline states and is
particularly important for low energy states. For sodium, for example, the values are:
For states with a large angular momentum, the quantum defect disappears. In
these states, the electron is far from the nucleus and the potential is similar to that of
hydrogen. Alkalines are currently widely studied in quantum optics laboratories, for
being comparatively simple, but having a sufficiently rich structure to be interesting.
The fundamental electronic transitions typically lie in the visible and near-infrared
spectral range and can be excited with comparatively simple laser sources. The
lifetime of excited states is typically longer than 20 ns, which corresponds natural
linewidths of approximately (2π) 10 MHz.
optical transitions can be excited within the 4f shell of the Nd. However, these transitions are only
allowed due to perturbations of the crystalline field. The very strong shielding ensures a long life of
the excited state. For this reason this crystal is an excellent laser material.
10.4. THE PERIODIC SYSTEM OF ELEMENTS 309
Figure 10.5: Comparison of the excitation energies of the valence electron for several alkaline
atoms. Grey line are meant to guide the eye.
total spin S adopts in the field generated by the total orbital angular momentum L 10 .
ji = li + si . (10.81)
HF S ∝ ji · jj . (10.82)
10 In addition, there are the small contributions due to l · l -coupling and to s · s -coupling, where
i j i j
i 6= j.
10.4. THE PERIODIC SYSTEM OF ELEMENTS 311
Figure 10.7: Hierarchy of coupling energies at the example of two electrons out of closed
shells, (2p)(3d), under the presumption of perfect jj-coupling (left) or LS-coupling (right).
Pure jj-coupling only exists for very heavy nuclei. Normally, we have a so-called
intermediate coupling, which is a mixture of LS and jj-coupling. This can consider-
able relax the intercombination prohibition. When the coupling is pure, we have the
following dipolar selection rules:
LS-coupling: ∆S = 0, ∆L = ±1, ∆` = ±1
jj-coupling: ∆j = 0, ±1 for one e− , ∆j = 0 for all others
In addition we have for the two couplings: ∆J = 0, ±1, but J, J 0 = 0 is forbidden,
∆mJ = 0, ±1 when ∆J = 0 but mJ , mJ 0 = 0 is forbidden.
Of course, with the presence of other atoms, other interactions may generate other
relevant contributions to the Hamiltonian.
The following interactions contribute to the potential V : The Coulomb interac-
tions,
Z
X Z
X
Ze2 e2
Vncl−ele = − and Vele−ele = , (10.84)
i=1
4πε0 |r − ri | i<j=1
4πε0 |ri − rj |
Vsym , (10.85)
312 CHAPTER 10. ATOMS WITH MANY ELECTRONS
Z
X 1 1 dVcl
Vls = − (li · si ) , (10.86)
i=1
e2m2 c2 |r − ri | dri
Z
X
e2 σi · σj [σi · (ri − rj )][σj · (ri − rj )]
Vss = − 3 , (10.87)
m2 |ri − rj |3 (ri − rj )5
i6=j=1
Z
X
Vll = cij (li · lj ) , (10.88)
i6=j=1
interactions between the spin of the electrons and the nuclear spin and between the
orbital angular momentum of the electrons and the nuclear spin,
A
Vhf s = J·I , (10.89)
~2
In addition, static external fields may displace energy levels and can influence the
internal coupling of angular momenta and spins,
Vext = −d · E~ , − µ ~.
~ ·B (10.91)
What quantum numbers are good depends on the relative amplitudes of intra-
atomic interactions:
Case 1: fine structure with L·S-coupling plus Zeeman splitting of hyperfine structure:
r a
Vncl−ele , Vele−ele Vele−ele , Vsym V`s Vhf s VB the quantum number are
|ni , `i , L, S, J, F, mF i.
Case 2: fine structure with j · j-coupling plus Zeeman splitting of hyperfine structure:
r a
Vncl−ele , Vele−ele V`s Vele−ele , Vsym Vhf s VB the quantum number are
|ni , `i , ji , J, F, mF i.
Case 3: fine structure with L·S-coupling plus hyperfine structure of Zeeman splitting:
r a
Vncl−ele , Vele−ele Vele−ele , Vsym V`s VB Vhf s the quantum number are
|ni , `i , L, S, J,mJ , mI i.
Case 4: fine structure with L·S-coupling plus Paschen-Back splitting of fine structure:
r a
Vncl−ele , Vele−ele Vele−ele , Vsym VB V`s Vhf s the quantum number are
|ni , `i , L, S,mL , mS , mI i.
10.5. FURTHER READING 313
Vncl−ele splitting in n
coarse structure ↓
Vele−ele splitting in `
↓
Vsym splitting in S &
↓ Vls splitting in ji
fine structure Vee splitting in L ↓
↓ Vsym , Vele−ele splitting in J
VLS splitting in J .
↓
hyperfine structure Vhf s splitting in F
↓
Zeeman effect VLS splitting in mF
10.4.4 Exercises
10.4.4.1 Ex: Filled electronic shells
Show at the example of hydrogen that completely filled electronic layer are isotropic.
I.N. Levine, Allyn and Bacon, 7th ed. Pearson (1983), Quantum Chemistry [ISBN]
J.I. Steinfeld, The MIT Press, Cambridge (2005), Molecules and Radiation [ISBN]
B.H. Bransden, C.J. Joachain, John Wiley & Sons (1983), Physics of Atoms and
Molecules [ISBN]
Physics Standard Reference
Group P E R I O D I C T A B L E Laboratory Data Program
IA physics.nist.gov www.nist.gov www.nist.gov/srd VIII
1
Atomic Properties of the Elements
314
1 2S1/2 2 S0
U.S. DEPARTMENT OF COMMERCE
H Frequently used fundamental physical constants Technology Administration He
1 Hydrogen For the most accurate values of these and other constants, visit physics.nist.gov/constants National Institute of Standards and Technology Helium
1.00794 1 second = 9 192 631 770 periods of radiation corresponding to the transition 4.00260
133 2
1s between the two hyperfine levels of the ground state of Cs 1s
13.5984 IIA !1 IIIB IVB VB VIB VIIB 24.5874
speed of light in vacuum c 299 792 458 m s (exact)
2 1 !34 2 3 4 3 2 1
3 S1/2 4 S0 Planck constant h 6.6261 H 10 Js (h = h/2p) 5 P°1/2 6 P0 7 S°3/2 8 P2 9 P°3/2 10 S0
!19
elementary charge e 1.6022 H 10 C
Li Be !31 B C N O F Ne
2 electron mass me 9.1094 H 10 kg
Lithium Beryllium 2 Boron Carbon Nitrogen Oxygen Fluorine Neon
mec 0.5110 MeV
6.941 9.01218 !27 10.811 12.0107 14.00674 15.9994 18.99840 20.1797
2 2 2 proton mass mp 1.6726 H 10 kg 2 2 2 2 2 2 2 3 2 2 4 2 2 5 2 2 6
1s 2s 1s 2s 1s 2s 2p 1s 2s 2p 1s 2s 2p 1s 2s 2p 1s 2s 2p 1s 2s 2p
5.3917 9.3227 fine-structure constant a 1/137.036 8.2980 11.2603 14.5341 13.6181 17.4228 21.5646
!1
2 1 Rydberg constant R4 10 973 732 m 2 3 4 3 2 1
11 S1/2 12 S0 15 13 P°1/2 14 P0 15 S°3/2 16 P2 17 P°3/2 18 S0
R4c 3.289 84 H 10 Hz
Na Mg R4hc 13.6057 eV Al Si P S Cl Ar
!23 !1
3 Sodium Magnesium Boltzmann constant k 1.3807 H 10 JK Aluminum Silicon Phosphorus Sulfur Chlorine Argon
22.98977 24.3050 26.98154 28.0855 30.97376 32.066 35.4527 39.948
2 2 2 2 2 3 2 4 2 5 2 6
[Ne]3s [Ne]3s VIIIA [Ne]3s 3p [Ne]3s 3p [Ne]3s 3p [Ne]3s 3p [Ne]3s 3p [Ne]3s 3p
5.1391 7.6462 IIIA IVA VA VIA VIIA IB IIB 5.9858 8.1517 10.4867 10.3600 12.9676 15.7596
2 1 5 4 3 2 2 3 4 3 2 1
19 S1/2 20 S0 21 2D3/2 22 3F2 23 4F3/2 24 7S3 25 6S5/2 26 D4 27 F9/2 28 F4 29 S1/2 30 1S0 31 P°1/2 32 P0 33 S°3/2 34 P2 35 P°3/2 36 S0
4
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
Potassium Calcium Scandium Titanium Vanadium Chromium Manganese Iron Cobalt Nickel Copper Zinc Gallium Germanium Arsenic Selenium Bromine Krypton
39.0983 40.078 44.95591 47.867 50.9415 51.9961 54.93805 55.845 58.93320 58.6934 63.546 65.39 69.723 72.61 74.92160 78.96 79.904 83.80
Period
2 2 2 2 3 2 5 5 2 6 2 7 2 8 2 10 10 2 10 2 10 2 2 10 2 3 10 2 4 10 2 5 10 2 6
[Ar]4s [Ar]4s [Ar]3d4s [Ar]3d 4s [Ar]3d 4s [Ar]3d 4s [Ar]3d 4s [Ar]3d 4s [Ar]3d 4s [Ar]3d 4s [Ar]3d 4s [Ar]3d 4s [Ar]3d 4s 4p [Ar]3d 4s 4p [Ar]3d 4s 4p [Ar]3d 4s 4p [Ar]3d 4s 4p [Ar]3d 4s 4p
4.3407 6.1132 6.5615 6.8281 6.7462 6.7665 7.4340 7.9024 7.8810 7.6398 7.7264 9.3942 5.9993 7.8994 9.7886 9.7524 11.8138 13.9996
2 1 2 3 6 7 6 5 4 1 2 1 2 3 4 3 2 1
37 S1/2 38 S0 39 D3/2 40 F2 41 D1/2 42 S3 43 S5/2 44 F5 45 F9/2 46 S0 47 S1/2 48 S0 49 P°1/2 50 P0 51 S°3/2 52 P2 53 P°3/2 54 S0
5
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
Rubidium Strontium Yttrium Zirconium Niobium Molybdenum Technetium Ruthenium Rhodium Palladium Silver Cadmium Indium Tin Antimony Tellurium Iodine Xenon
85.4678 87.62 88.90585 91.224 92.90638 95.94 (98) 101.07 102.90550 106.42 107.8682 112.411 114.818 118.710 121.760 127.60 126.90447 131.29
2 2 2 2 4 5 5 2 7 8 10 10 10 2 10 2 10 2 2 10 2 3 10 2 4 10 2 5 10 2 6
[Kr]5s [Kr]5s [Kr]4d5s [Kr]4d 5s [Kr]4d 5s [Kr]4d 5s [Kr]4d 5s [Kr]4d 5s [Kr]4d 5s [Kr]4d [Kr]4d 5s [Kr]4d 5s [Kr]4d 5s 5p [Kr]4d 5s 5p [Kr]4d 5s 5p [Kr]4d 5s 5p [Kr]4d 5s 5p [Kr]4d 5s 5p
4.1771 5.6949 6.2171 6.6339 6.7589 7.0924 7.28 7.3605 7.4589 8.3369 7.5762 8.9938 5.7864 7.3439 8.6084 9.0096 10.4513 12.1298
2 1 3 4 5 6 5 4 3 2 1 2 3 4 3 2 1
55 S1/2 56 S0 72 F2 73 F3/2 74 D0 75 S5/2 76 D4 77 F9/2 78 D3 79 S1/2 80 S0 81 P°1/2 82 P0 83 S°3/2 84 P2 85 P°3/2 86 S0
6
Cs Ba Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
Cesium Barium Hafnium Tantalum Tungsten Rhenium Osmium Iridium Platinum Gold Mercury Thallium Lead Bismuth Polonium Astatine Radon
132.90545 137.327 178.49 180.9479 183.84 186.207 190.23 192.217 195.078 196.96655 200.59 204.3833 207.2 208.98038 (209) (210) (222)
Atomic Ground-state 2 1 4- 5- 6 7 8 9 6 5- 4- 3 2 1 2
Number Level
57 D3/2 58 G°4 59 -I °9/2 60 -I 4 61 H°5/2 62 F0 63 S°7/2 64 D°2 65 H°15/2 66 -I 8 67 -I °15/2 68 H6 69 F°7/2 70 S0 71 D3/2
1
La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
Symbol 58 G°4 Lanthanum Cerium Praseodymium Neodymium Promethium Samarium Europium Gadolinium Terbium Dysprosium Holmium Erbium Thulium Ytterbium Lutetium
138.9055 140.116 140.90765 144.24 (145) 150.36 151.964 157.25 158.92534 162.50 164.93032 167.26 168.93421 173.04 174.967
2 2 3 2 4 2 5 2 6 2 7 2 7 2 9 2 10 2 11 2 12 2 13 2 14 2 14 2
Ce [Xe]5d6s [Xe]4f5d6s [Xe]4f 6s [Xe]4f 6s [Xe]4f 6s [Xe]4f 6s [Xe]4f 6s [Xe]4f 5d6s [Xe]4f 6s [Xe]4f 6s [Xe]4f 6s [Xe]4f 6s [Xe]4f 6s [Xe]4f 6s [Xe]4f 5d6s
Name 5.5769 5.5387 5.473 5.5250 5.582 5.6436 5.6704 6.1501 5.8638 5.9389 6.0215 6.1077 6.1843 6.2542 5.4259
Cerium 2 3 4 5 6 7 8 9 6 5- 4- 3 2 1 2
140.116 89 D3/2 90 F2 91 K11/2 92 L°6 93 L11/2 94 F0 95 S°7/2 96 D°2 97 H°15/2 98 -I 8 99 -I °15/2 100 H6 101 F°7/2 102 S0 103 P°1/2?
Atomic
†
Weight [Xe]4f5d6s 2 Ac Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr
5.5387 Actinium Thorium Protactinium Uranium Neptunium Plutonium Americium Curium Berkelium Californium Einsteinium Fermium Mendelevium Nobelium Lawrencium
(227) 232.0381 231.03588 238.0289 (237) (244) (243) (247) (247) (251) (252) (257) (258) (259) (262)
CHAPTER 10. ATOMS WITH MANY ELECTRONS
Ground-state Ionization 2 2 2 2 2 3 2 4 2 6 2 7 2 7 2 9 2 10 2 11 2 12 2 13 2 14 2 14 2
[Rn]6d7s [Rn]6d 7s [Rn]5f 6d7s [Rn]5f 6d7s [Rn]5f 6d7s [Rn]5f 7s [Rn]5f 7s [Rn]5f 6d7s [Rn]5f 7s [Rn]5f 7s [Rn]5f 7s [Rn]5f 7s [Rn]5f 7s [Rn]5f 7s [Rn]5f 7s 7p?
Configuration Energy (eV) 5.17 6.3067 5.89 6.1941 6.2657 6.0262 5.9738 5.9915 6.1979 6.2817 6.42 6.50 6.58 6.65 4.9 ?
† 12
Based upon C. ( ) indicates the mass number of the most stable isotope. For a description and the most accurate values and uncertainties, see J. Phys. Chem. Ref. Data, 26 (5), 1239 (1997). March 1999
Chapter 11
Molecular dimers
In systems of many particles (gases, fluids, or solids) interatomic interactions must
be considered. These interactions usually have electrostatic origins, but generally can
not be given in the form of closed expressions. For example, the collision of two atoms
i and j can occur in a multitude of channels, that is, interaction potentials V (ri − rj ).
Interatomic forces do not only govern collisions, but can sustain molecular bound
states. This introduces new degrees of freedom in the systems of many particles
through possible excitations of vibration or rotation movements.
600
1
Σ+ 3 +
g , Σu
400 S1/2 + P3/2
(cm−1 )
1 3
Πg , Πu
200
1
Πu ,3 Πg
E
0
1
Σ+ 3 +
u , Σg S1/2 + P1/2
-200
10 20 30 40 50 60
R (aB )
Figure 11.1: (code) Example of an interatomic potential spaghetti: The lowest states of the
molecule 85 Rb2 .
In this course, we will not go beyond homo- or heteronuclear dimers, that is,
molecules consisting of two identical or different atoms.
315
316 CHAPTER 11. MOLECULAR DIMERS
At short distances, the exchange of an electron between atoms can decrease energy.
The so-called ionic bond is then sustained by the Coulombian attraction between two
ions, and the binding energy can be estimated through electrostatic interaction.
(11.1)
Na + 5.1 eV → Na+ + e−
Cl + e− → Cl− + 3.8 eV
Na+ + Cl− → NaCl + 4.9eV
− − − − −− − − − − − − − − − − − −−
Na + Cl → NaCl + (−5.1 + 3.8 + 4.9) eV .
1 We are not considering metallic bonds nor hydrogen bridge bonds, here.
11.1. MOLECULAR BINDING 317
The molecules are polar and, therefore, have a permanent electric dipole moment.
The bond has no preferential direction, since each atom is perfectly isotropic. There-
fore, this type is well suited for the construction of crystalline lattices.
Figure 11.3: Scheme for (a) ionic binding of NaCl and (b) covalent binding of H2 .
To understand covalent bond, we consider the example H+ 2 and estimate the in-
teraction energy for each distance R between the nuclei. In this case, in contrast to
atoms, the spherical symmetry is broken, and therefore the energy degeneracy with
respect to parity is abolished, that is, for wavefunctions ψ(−x) = ±ψ(x) the energies
vary differently with R. The even wavefunction, which has an increased probability of
the electron of being between the nuclei, is binding, which means that the interaction
potential exhibits a minimum at a certain distance. The odd wavefunction, which
disappears between the nuclei, is anti-binding, which means that the interaction po-
tential is repulsive at all distances. In fact, an electron located at the center between
two positive charges can overcome the Coulomb repulsion between the nuclei, whose
mutual distance is twice. Obviously, the energy can not fall below that of the fun-
damental state of He+ , being approximately −4 × 13.6 eV. With two electrons, as
in the case of the neutral molecule H2 , the anti-parallel orientation of the spins, ↑↓,
allows us to place the two electrons in the same orbital, while for parallel orientation,
↑↑, we expect anti-binding. Each electron without a partner in an orbital can form a
covalent bond, for example, phosphorus [P]=[Ne]3s2 3p↑↑↑ has three available orbitals
corresponding to different magnetic quantum numbers. The covalent bond is direc-
tional (sp1 , sp2 , or sp3 hybridization), which is essential for the molecular structure
such as in CH4 . Do the Exc. 11.1.6.1.
Figure 11.4: System with two heavy and one light mass.
X ≡ Ma RaM +Mb Rb
= 21 (Ra + Rb ). The distance of the heavies is R ≡ Ra − Rb , and
the coordinate of the light mass counting from the center-of-mass is r = Ra − 21 R−Re .
Introducing the reduced mass of the heavies Mr = M 2 ,
2
−~ 2 −~2 2 −~2 2
∇X + ∇R + Vnn (R) + ∇r + Vne (|r + R2 |) + Vne (|r − R
2 |) Θ(X)Ψ(R, Re )
2M 2Mr 2me
= Etot Θ(X)Ψ(R, Re ) , (11.3)
Here, we made the ansatz for the total wavefunction Ψ = Θ(X)Ψ(R, Re ), assuming
that the center-of-mass is only determined by the heavy masses,
−~2 2
∇X Θ(X) = Ecm Θ(X) (11.4)
2M 2
−~ 2 −~2 2
∇R + Vnn (R) + ∇ + Vne (|r + 12 R|) + Vne (|r − 12 R|) Ψ(R, Re ) = EΨ(R, Re ) ,
2Mr 2me r
where Etot = Ecm + E.
∇2R [ψ(R, r)φ(R)] = φ(R)∇2R ψ(R, r) + 2[∇R φ(R)] · [∇R ψ(R, r)] + ψ(R, r)∇2R φ(R)
' ψ(R, r)∇2R φ(R) , (11.5)
11.1. MOLECULAR BINDING 319
postulating that the first two terms are negligible compared to the third. We can now
separate the second equation (11.4) in two parts, the first being,
−~2 2
∇r + Vne (|r + 12 R|) + Vne (|r − 12 R|) ψ(r, R) = ε(R)ψ(r, R) . (11.6)
2me
We solve this equation for the electronic degree of freedom r by choosing a fixed
internuclear distance R, and we substitute in the second expression (11.4), which
gives,
−~2 2
∇R + Vnn (R) + ε(R) φ(R) = Eφ(R) . (11.7)
2Mr
Treating the interatomic distance R as a fixed parameter, the solution of Eq. (11.6)
provides the electronic orbitals and their energies ε(R). The Born-Oppenheimer po-
tential is composed of the electrostatic repulsive potential of the nuclei and the kinetic
energy of the electron, Vnn (R) + ε(R). In other words, the presence of the electrons
generates an additional interaction energy between the nuclei. By inserting this com-
plete interatomic potential into Eq. (11.7), we can determine its vibrational structure
φvib (R).
Fig. 11.5 shows an example of binary effective nuclear potentials, as a function of
the distance R between the two nuclei. Each curve corresponds to a different solution
of the electronic equation, that is, to a different electronic state. In many cases, such
nuclear potentials have a stable equilibrium region. The bound states located in these
regions are molecular bound states of two atoms.
Figure 11.5: Effective nuclear potentials for the interaction between two rubidium atoms
(Rb2 ).
2 Note that the Born-Oppenheimer approximation is no longer good in the presence of relativistic
defining the average one-electron energy ≡ h1sa |ĥ|1sa i and the non-diagonal cou-
pling (often called resonance integral) hab ≡ h1sa |ĥ|1sb i = h1sa |ĥ|1sb i. We can im-
mediately diagonalize this matrix, the eigenvalues and the eigenvalues being:
The one-electron effective Hamiltonian eigenstates are called molecular orbital (MO).
They are one-electron wavefunctions delocalized over the spatial regions of the molecule.
ha hb
We need to first normalize the MOs, which is more complicated than it might
seem, because the AOs are not orthogonal. For example, when the atoms approach
each other, their AOs may have the shape shown in Fig. 11.6. However, by defining
the overlap integral by S ≡ h1sa |1sb i, we can normalize as follows:
1
|φ± i = √ (|1sa i ± |1sb i) , (11.12)
2(1±S)
since,
1
hφ± |φ± i = 2(1±S) (h1sa |1sa i ± h1sa |1sb i ± h1sb |1sa i + h1sb |1sb i) = 1 . (11.13)
These eigenfunctions merely show the symmetry of the molecule. The two hydrogen
atoms are equivalent and, therefore, the eigenorbital must give equal weight to each
1s orbital. Thus, our choice of the one-electron Hamiltonian does not really mat-
ter that much, because every one-electron Hamiltonian exhibiting the symmetry of
the molecule would give the same molecular orbitals. For historical reasons, |φ+ i is
denoted by |σi and |φ− i by |σ ∗ i.
The second step in MO theory consists of constructing the determinant from the
MOs corresponding to the wanted states. For illustration we will look at the lowest
singlet state constructed from molecular orbitals. We note that hab < 0, such that
|σi has an energy inferior to |σ ∗ i. Neglecting the interaction, the lowest singlet state,
e2
hΦM O |Ĥ|ΦM O i = hΦM O |ĥ1 |ΦM O i + hΦM O |ĥ2 |ΦM O i + hΦM O |V̂12 |ΦM O i +
4πε0 Rab
e2
= 2σ + Jσσ + . (11.17)
4πε0 Rab
Each of the first two terms represents the energy of a single electron (either 1 or 2)
in the field produced by both the nuclei (ĥ), while the third is the average repulsion
between the two electrons. Note that the first and second term are both positive,
such that the bond must come from the one-electron part. This is the MO energy
for the ground state of the H2 . We can try a more reasonable ansatz for the 1s-
type basis functions 3 by determining the unknown quantities from above (σ and
Jσσ ) numerically and plot the total energy as a function of Rab (blue dotted curve
in Fig. 11.7). The exact adiabatic energy function determined from experimental
data (solid black curve) agrees well at low energies. Summarizing the results with
some key numbers, we note that MO theory predicts a bond distance of 0.072 nm
in reasonable agreement with the exact value of 0.074 nm. We can also compare the
3 It turns out that it is more convenient to adjust the exponential decomposition of the hydrogenic
binding energies,
De = EH2 (Requil ) − 2EH . (11.18)
MOs theory predicts a binding energy of 5.0 eV compared to the experimental value
of 4.75 eV. In view of the simplicity of the wavefunction and the absence of adjustable
parameters the agreement is not so bad. Unfortunately, far away from the equilibrium
distance, we have an unpleasant surprise: the molecule does not dissociate into two
hydrogen atoms!
1s 1s s1s* E
ya yb y
|y|2
1s 1s s1s
ya yb y
|y|2
H H H2
Figure 11.8: Illustration of MO theory for a dimer of two equal atoms, each one with a
valence electron in the atomic 1s orbital: When the atoms approach, the atomic orbitals
form new molecular binding and anti-biding orbitals.
The two terms in the middle of the last line, called covalent configurations, are exactly
what we expect near dissociation: one electron in each hydrogen atom. However, the
first and last term (which are called ionic configurations) correspond to having two
electrons in one atom and none in the other, which gives us H + and H − at dissoci-
ation! Since the weight of these terms is fixed, it is obvious that we got the wrong
wavefunction (and therefore the wrong energy) when dissociating the molecule. Near
the equilibrium point, the ionic terms contribute significantly to the true wavefunc-
tion, such that the MO theory is good at this point. But it is always terrible at
dissociation.
324 CHAPTER 11. MOLECULAR DIMERS
|1sa (1)i|1sa (2)i + |1sb (1)i|1sa (2)i | ↑ (1)i| ↓ (2)i + | ↓ (1)i| ↑ (2)i
|Ψi ∝ √ √
2 2
≡ |Ψspace i|Ψspin i . (11.20)
VB theory assumes that this wavefunction is a good approximation to the true wave-
function at all binding distances and not only at large distances Rab . To verify this
approximation, we can calculate the average energy for this VB state. First, we
normalize the VB wavefunction,
1
|ΨV B i = √ (|1sa (1)i|1sa (2)i + |1sb (1)i|1sa (2)i) (| ↑ (1)i| ↓ (2)i − | ↓ (1)i| ↑ (2)i) .
2 1 + S2
(11.22)
Now we want to calculate hĤel i for this state. We note that the spin part does not
matter, since the Hamiltonian is independent of spin:
The only remnant of the spin state is the fact that the spatial wavefunction is sym-
metric, which is only possible when the spin part is antisymmetric. Treating each
term in hĤi separately,
1
hΨ|ĥ1 |Ψi = (h1sa (1)|h1sb (2)| + h1sb (1)|1sa (2)|) ĥ1 (|1sa (1)i|1sa (2)i + |1sb (1)i|1sa (2)i)
2
1
= h1sa (1)|1sb (2)|ĥ1 |1sa (1)i|1sb (2)i + h1sa (1)|1sb (2)|ĥ1 |1sb (1)i|1sa (2)i
2
+h1sb (1)|1sa (2)|ĥ1 |1sa (1)i|1sb (2)i + h1sb (1)|1sa (2)|ĥ1 |1sb (1)i|1sa (2)i
= 21 ( + Shab + + Shab ) , (11.24)
or,
+ Shab
hΨV B |ĥ1 |ΨV B i = . (11.25)
1 + S2
11.1. MOLECULAR BINDING 325
Since the two electrons are identical, the elements of the ĥ2 are the same as those of
ĥ1 . The only remaining term is the average value of the interaction:
1
hΨ|V̂12 |Ψi = 2 (h1sa (1)|h1sb (2)| + h1sb (1)|1sa (2)|) V̂12 (|1sa (1)i|1sa (2)i + |1sb (1)i|1sa (2)i)
(11.26)
1
= 2 h1sa (1)|1sb (2)|V̂12 |1sa (1)i|1sb (2)i + h1sa (1)|1sb (2)|V̂12 |1sb (1)i|1sa (2)i
+ h1sb (1)|1sa (2)|V̂12 |1sa (1)i|1sb (2)i + h1sb (1)|1sa (2)|V̂12 |1sb (1)i|1sa (2)i .
The second and third terms are the same. They are called exchange integrals, because
the ’bra’ orbitals have switched order as compared to the ’kets’:
K = h1sa (1)|1sb (2)|V̂12 |1sb (1)i|1sa (2)i = h1sb (1)|1sa (2)|V̂12 |1sa (1)i|1sb (2)i .
(11.27)
The first and forth terms are also the same. They are called Coulomb integrals,
because they seem to be due to the Coulomb interaction between two charge densities:
J = h1sb (1)|1sa (2)|V̂12 |1sa (1)i|1sb (2)i = h1sa (1)|1sb (2)|V̂12 |1sb (1)i|1sa (2)i . (11.28)
J +K
hΨV B |V̂12 |ΨV B i = . (11.29)
1 + S2
|Ψspace i ' |1sHa i|1sHb i|1sO i|1sO i|2sO i|2sO i|2pxO i|2pxO i|2pyO i|2pyO i . (11.32)
However, there are two things wrong with this wavefunction. First, we know that
atomic orbitals hybridize in a molecule. Therefore, we need to make appropriate
linear combinations of AOs (in this case sp3 hybrids) to obtain the hybridized AOs.
In this case, the four sp3 hybrids can be written symbolically as:
|sp3 i = cs,i |2si + cx,i |2px i + cy,i |2py i + cz,i |2pz i . (11.33)
|Ψspace i ' |1sHa i|1sHb i|1sO i|1sO i|sp31O i|sp31O i|sp32O i|sp32O i|sp33O i|sp34O i . (11.34)
The other problem with this state is that it lacks the adequate symmetry to describe
fermions; the general state must be antisymmetric. In the case of two electrons this
concept is easy to apply - singlets have symmetric space parts and triplets antisym-
metric ones. However, in the case of many electrons, the rules are not so simple;
in fact, the time of numerical computation grows exponentially with the number of
electrons.
Formally, we will leave the derivation at this point to defining an operator A which
’antisymmetrizes’ the wavefunction. In this case,
|Ψspace i ' A |1sHa i|1sHb i|1sO i|1sO i|sp31O i|sp31O i|sp32O i|sp32O i|sp33O i|sp34O i .
(11.35)
In general, the results of VB theory are very accurate for small systems, where it can
be applied. The predicted bond lengths are rather short, and the binding energies
tend to be too small, but the results are nevertheless qualitatively excellent. In ad-
dition, the correct hybridized atomic orbitals fall directly off the calculation, giving a
good qualitative insight. Also, note that the atomic configurations should not change
(or very little) when the geometry of the molecule changes (since the orbitals depend
on the atom and not on the molecular structure). Therefore, these VB wavefunctions
have a strong connection to the diabatic states discussed above. However, the ex-
ponential amount of time that one must invest to perform these calculations makes
them impractical for most molecules of interest.
11.1.6 Exercises
11.1.6.1 Ex: Classical model of the covalent binding
Consider the molecule H+ 2 with the two nuclei separated by 1 nm and an electron
located in the middle between the nuclei. Calculate the electrostatic force acting on
the nuclei.
11.1. MOLECULAR BINDING 327
∇2R X 1 X ∇2
i
− − ∇i ∇j − +V
2µAB i,j
2M i
2
where that the Coulombian interactions are included in the fifth term. Write down
the time-independent Schrödinger equation for this molecule.
c. The Born-Huang approximation consists in assuming that the total wavefunction
can be expanded on a basis of wavefunctions of the nuclei and the electrons, that is,
X
Ψ(r, R) = |χk (R)i|φk (r, R)i ,
k
where χ and φ are the wavefunctions of the nuclei and the electrons, respectively.
For the Schrödinger equation calculated in the previous item, use the Born-Huang
approximation and obtain the set of coupled equations
(
X 1
− ∇2R + hφk |∇2R |φk i + 2hφk |∇R |φk i.∇R | −
2µAB
k
X 1 X X X
hφi |∇i .∇j |φk i + 1 hφi |∇2i |φk i − hφi |V |φk i| |χk i = E |χk i
2M 2
k i,j i k
which includes, although approximately, the kinetic energy of nuclei and electrons.
Help: Use ∇2 (αβ) = α∇2 β + β∇2 α + 2∇α · ∇β.
d. Make a brief comparison between the Born-Huang approximation (and the coupled
equations obtained in the previous equation) and the Born-Oppenheimer approxima-
tion.
P2 e2
Ĥ = + Vmol (R) with Vmol (R) = + VBO (R) . (11.36)
2Mr 4π0 R
The wavefunction can be separated into an angular part and a radial part, φ(R) =
Rv (R)Y`m (θ, φ). The angular part, which was discussed in Sec. 4.1.3, describes a
rigid rotation of the homonuclear atoms around their center-of-mass with the rotation
energy,
L2 ~2 `(` + 1)
V` (R) = 2
= , (11.38)
2Mr R 2Mr R2
where uv (R) = rRv (R) is the radial wavefunction of nuclear motion. The interatomic
potential causes a motion of vibration. The vibrational states of the adiabatic poten-
tial are quantized and characterized by a well-defined vibrational energy. We will
discuss the ro-vibrational structure in the following sections.
4 We note here that at great distances other forces called van der Waals forces dominate the
Molecules have much more degrees of freedom than atoms. For example, the atoms of
a molecular dimer may vibrate inside the mutual interaction potential. In the center-
of-mass system we can imagine these vibrations as oscillations of an atom with reduced
mass and quantized energy. The molecule can rotate and have a momentum of inertia.
These degrees of freedom contribute energies to the molecule’s Hamiltonian, either
directly or through interactions with other degrees of freedom. Therefore, molecular
spectra are characterized by a much greater complexity.
However, the energy regimes of the strongest excitations are quite different. A
typical range for binding energies (depth of the interatomic potential) is ∆Ep '
20..200 THz (0.1..1 eV) 5 . Electronic excitations occur in the regime ∆Ee ' 100..1000 THz
(1..10 eV). The spacing between vibrational excitations typically is Ev+1 −Ev ' THz
(0.01 eV). Finally, the rotational excitations are on the scale of E`+1 − E` ' 100 MHz
(10−6 eV). Since at room temperature (a gas of molecules in thermal equilibrium
at T = 300 K) the energy is on a scale of 2.5 × 10−2 eV, the degree of freedom of
the electronic excitation is frozen, while a wide distribution of vibrational and rota-
tional states can be excited (e.g. by intermolecular collisions). The large difference of
scales facilitates their separation and, therefore, the identification of the origin of the
observed states in experimental measurements.
are more weakly bound, because the electrons are not in the most binding orbital.
330 CHAPTER 11. MOLECULAR DIMERS
The equilibrium energy is not of interest here, and the first derivative disappears in
equilibrium. Therefore,
d2 Vmol (0)
Vmol (x) ' 12 k 2 x2 with k≡ . (11.41)
dx2
Using the effective mass we can write the Hamiltonian,
~2 d2 ~2 d2 ~2 d 2
Ĥmol = − 2 − 2 + 12 kx2 = − + 21 kx2 . (11.42)
2m1 dx1 2m2 dx2 2Mr dx2
Ev = ~ω(v + 12 ) . (11.43)
p
with ω = k/Mr . That is, at the bottom of deep potentials, the energy levels are
equidistant.
0
E
-5
-10
0 2 4 6
R/Re
Figure 11.10: (code) Many potentials are approximately harmonic at the center such as,
for instance, the Morse potential (blue). The red curve shows the approximate harmonic
potential.
where r is the interatomic distance, re the equilibrium bond distance, De the depth
of the potential measuring from the dissociation limit, and a a parameter controlling
the range of the potential. At the bottom of the potential we can make the harmonic
11.2. ROVIBRATIONAL STRUCTURE OF MOLECULAR POTENTIALS 331
D0 = De − E0 . (11.47)
The number of vibrational states is limited v = 0, 1, .., vmax . With E < 0, we find,
1 1
vmax < xe − 2 . (11.48)
~2 ∂ 2
− + V (R) Ψ(v) = Ev Ψ(v) ,
2m ∂R2
it is convenient to introduce new variables,
√
2mDe 2m
x ≡ aR , λ≡ , εv ≡ Ev ,
a~ a2 ~2
such that,
∂2
− 2 + V (x) Ψn (x) = εn Ψn (x) with V (x) = λ2 e−2(x−xe ) − 2e−(x−xe ) .
∂x
The eigenvalues and eigenfunctions are [169]:
1 1
1 2
εn = 1− λ12 λ − n − 2
(n+ 21 )2 − λ12 (n+ 12 )2 Ψn (z) = Nn z λ−n− 2 e− 2 z L(2λ−2n−1)
2
= λ
and n (z) ,
h i1
n!(2λ−2n−1) 2
where z = 2λe−(x−xe ) e Nn = Γ(2λ−n)
and,
z −α ez dn Γ(α + n + 1)/Γ(α + 1)
L(α) z n+α e−z =
n (z) = 1 F1 (−n, α + 1, z) ,
n! dz n Γ(n + 1)
is the generalized Laguerre polynomial. The matrix elements of the spatial
operator x̂ are (assuming m > n and N = λ − 21 ),
s
2(−1)m−n+1 (N − n)(N − m)Γ(2N − m + 1)m!
hΨm |x|Ψn i = .
(m − n)(2N − n − m) Γ(2N − n + 1)n!
[~ω(v + 1/2)]2
Ev = ~ω(v + 1/2) − ,
4De
332 CHAPTER 11. MOLECULAR DIMERS
q
where v is the vibrational quantum number and ω = a 2D m
e
. The energy
difference between adjacent levels decreases with v,
v+1
Ev+1 − Ev = ~ω − (~ω)2 .
2De
This fact describes well the vibrational structure of non-rotating molecules.
However, the equation fails above some value of v > vmax , where Evmax +1 −
Evmax is zero or negative,
2De − ~ω
vmax = .
~ω
This failure is due to the finite number vmax of bound states in the Morse
potential. For energies above vmax all energies are possible, and the equation
for Ev is no longer valid.
The dipole moment, d = h|d̂|i, of the molecule depends on the distance of the
nuclei, since the electronic orbitals |i depend on distance. Therefore, we can expand,
dd̂0 1 d2 d̂0 2
d̂ = d̂0 + x̂ + x̂ + .. . (11.50)
dx 2 dx2
Therefore, the transition matrix is,
dd̂0 0 d2 d̂0 0 2
h0 , v 0 |d̂|, vi = d̂0 δv,v0 + hv |x̂|vi + hv |x̂ |vi + .. . (11.51)
dx dx2
The first term disappears, that is, transitions can only occur, when the dipole moment
varies with the distance. Therefore, homonuclear dimers do not undergo vibrational
transitions.
11.2. ROVIBRATIONAL STRUCTURE OF MOLECULAR POTENTIALS 333
For heteronuclear molecules with electronic charges that do not depend on the
interatomic distance, the dipole moment varies linearly with small displacements. In
this case, we only need the second term of the expansion. Within the harmonic
approximation, the position operator can be expressed by, x̂ ∝ â + ↠. Therefore,
only transitions ∆v = ±1 are possible. However, due to anharmonicities, higher
order terms, x̂n ∝ (â + ↠)n become influential, and transitions with ∆v = ±2, ±3, ..
become possible.
Thus, in anharmonic potentials, the vibrational selection rules are replaced by the
concept the overlapping wavefunctions called Franck-Condon factor.
Raman spectroscopy is a very useful tool to analyze ro-vibrational spectra. In this
method, inelastic Raman scattering gives rise to Stokes and anti-Stokes lines in the
spectrum at frequencies corresponding to ∆v = ±1, ±2. The ground state spectrum
is asymmetric, because of the absence of the lower state. In homonuclear dimers, the
nuclear spins have a major impact on the Raman spectra. Parity considerations show
that there can only be odd or even lines.
For this reason, the transitions are drawn vertically in the scheme of potentials shown
in Fig. 11.11(right). To yield considerable rates, transitions must occur when the
nuclear velocities in the two coupled states are similar, which is the case at the
classical turning points. At these points, the wavefunctions are maximal 6 .
With this Franck-Condon principle, we can determine which are the strongest
transitions between vibrational levels of a molecule, as represented in Fig. 11.11(left).
In particular, we are interested in transitions between vibrational levels of different
electronic states.
The exact transition probabilities are calculated via the square module of the
transition dipole moment (TDM). The TDM is an out-of-diagonal matrix element of
the electric dipole operator M, given by:
×10−4
0
(a.u.)
-5
V (R)
-10
10 15 20 25 30
R (aB )
Figure 11.11: (Left, code) Molecular wavefunctions in a potential for three different
vibrational states. (Right) Pictorial representation of the classical statement of the
Franck-Condon principle. Transition (a) has high intensity (or probability), because
here both, the position and the relative velocity of the nuclei do not change. Transi-
tions (b) and (c) are unlikely, because they necessitate either a change in the position
of the nuclei (case b) or in velocity (case c).
Since the electronic wavefunctions of different states are orthogonal, it follows that
R (A)∗ (B)
ψe ψe dr = 0, canceling the second term.
Looking at the first term, we note that the electronic dipole moment M̂e (r, R)
also depends on the nuclear coordinates as a parameter. The quantum formulation
of the Franck-Condon principle consists in stating that, in a molecular state, the
electronic dipole moment varies little with the nuclear coordinates. Thus, along with
the condition of the Born-Oppenheimer approximation, we can split the first TDM
term into electronic and nuclear integrals:
Z Z
MAB = M̂e ψe(A)∗ ψe(B) dr ψn(A)∗ ψn(B) dR . (11.55)
Thus, we have a comparative expression for the transition probability given by:
Z 2 Z 2
PAB ∝ |MAB |2 = M̂e ψe(A)∗ ψe(B) dr ψn(A)∗ ψn(B) dR .
(11.56)
L2 ~2 `(` + 1)
V` (R) = = . (11.57)
2Mr R2 2Mr R2
As we shall now see, this energy creates a substructure of the vibrational levels.
The moments of inertia in the three axes of space are,
X
Iqq = mi ri2 (q) . (11.58)
i
X X L2x L2y L2
Erot = 1
2 mq vq2 = 1
2 Iqq ωq2 = + + z , (11.59)
q=1,2,3 q=1,2,3
2Ixx 2Iyy 2Izz
Figura
Figure 3: Esquema
11.12 da fotoassociação
: Photoassociation para formar
scheme moléculas
to form Rb2demolecules.
Rb2 . No processo a), o par (a),
In process de átomos
a pair of
free livres
atomsabsorve um afóton
absorbs da radiação
photon of the incidente,
incident formando
radiation,umforming
estado ligado no potencial
a bound state inexcitado.
the excited
Em seguida, em b), a molécula recém-formada decai por emissão espontânea para estados ligados do
potential.
potencialThen, in (b),
fundamental, ou the
aindanewly formeda um
pode retornar molecule decays
estado de by spontaneous
dois átomos livres. emission to a
bound state of the fundamental potential, or it can return to a state of two free atoms.
4 Detecção molecular & alguns resultados
O método
Many mais direto
molecules havequeautilizamos
symmetry paraaxis,
se detectar
suchmoléculas se baseia
that there are na
twoionização molecular.
different moments
Aplica-seIum ≡
of inertia, campo
I elétrico
= I na amostra
and I ≡ molecular,
I . e coloca-se um
Interpreting detector momenta
angular de íons na direção
as do
quantum
⊥ xx yy k zz
campo. Partículas que se ionizem adquirem uma carga líquida, sendo atraídas pelo campo e caindo
operators,
no detector de íons. Para promover a ionização de moléculas, aplica-se a luz de um laser pulsado
na amostra, cuja frequência pode ionizar
de vários fótons (razão pela qual
L̂2as moléculas
1 via processo
1
a técnica recebeu o nome de "REMPI-
Ĥ = resonantly
+ −
enhanced L̂2z . ionization ). As moléculas
multiphoton (11.60)
2I⊥ em 2I
ionizadas pelo laser pulsado caem no detector 2I⊥ de tempo
umk intervalo bem denido após os pulsos
de luz, de forma que conhecendo os instantes em que os pulsos foram emitidos, é possível olhar apenas
We
para must first econsider
as moléculas the rotation
ignorar a maioria of the molecule
dos íons esporádicos que caemrelative to the symmetry axis
no detector.
of the Outra vantagemForgetting
molecule. desse métodoexternal
de detecçãofields
é que we
a frequência de ionização
calculate the energydepende
of do nível
the vi-
molecule
bracional das moléculas que se deseja ionizar.
2 Ou seja, variando-se a frequência do laser ionizante,
associated with
encontramos the picos
diversos observables L̂ de
de intensidade with the quantum
ionização, referente aosnumber ` andvibracionais
diversos estados L̂z with the
quantum
em que number K.daWe
as moléculas findsethe
amostra eigenvalues,
encontram. Por essa razão, uma varredura do laser de ionização
pode ser chamada de "espectro vibracional", sendo este essencial para o nosso trabalho, pois de fato
desejamos estimar
~2a`(`
população
+ 1) de moléculas
1 que
1 está (ou não) no estado fundamental de vibração.
Na M
E(`, K, gura
`) =5, encontra-se um +espectro vibracional
− ~2 Kjanela
numa 2
= espectral
B`(` +relativamente
1) + (A − curta, 2
B)K feito em
, (11.61)
nosso laboratório. A2Iesta 2Ikcapazes
⊥ fase, éramos 2Ide
⊥ promover um considerável resfriamento vibracional
das moléculas, sem contudo alcançar uma população macroscópica no nível fundamental de vibração.
withNeste
` =espectro, foi K
0, 1, .., capaz=de−`,
se identicar um M
.., `, and pico de ionização partindo do nível vibracional ν = 4 do
` = −`, .., ` and introducing the rotational
estado eletrônico 2fundamental.
constants, A ≡ ~ /2Ik and B ≡ ~ /2I⊥ . We then analyze this equation in the context
2
of applying an external field that defines both, the direction ê0z in the laboratory as
well as the projection of the angular motion L̂2 on this direction, m` . That is, we
have two axes, the internuclear axis êz and the rotation axis of the molecule ê0z .
Each level |`, m` i is 2(2` + 1) times degenerate, 5 because K = −`, .., ` and K can
be positive or negative. Each level ` contains 2` + 1 states. Note that for spherical
molecules, A = B, and the degree of freedom K disappears. In Exc. 11.2.6.2 we
calculate the rotational spectrum of a homonuclear diatomic molecule.
The rotational constant can be approximated by,
~2 `(` + 1)
Erot = , (11.62)
2Mr hR2 i
11.2. ROVIBRATIONAL STRUCTURE OF MOLECULAR POTENTIALS 337
p
where hR2 i is the expectation value for the outer turning point of the vibrational
level. As an example, the rotational constant for the vibrational state of 87 Rb2 which
is 5.9 cm−1 below the dissociation limit is Bv = νrot
`=1 `=0
− νrot = 81 MHz. To be more
2 2
precise, we would need to calculate hR iv = hψv |R |ψv i.
Transitions between vibrational levels occur together with rotational transitions
∆` = ±1. Therefore, the frequencies of transitions depend on the rotational constant
Bv , which depends on the vibrational state. The energies of the molecule are,
Ev,` = ~ω(v + 1/2) − ~ωxe (v + 1/2)2 + .. + hcBv `(` + 1) − hcDv `2 (` + 1)2 + .. . (11.63)
Under the influence of a rapid rotation, the atoms of the molecule are subjected to
centrifugal force and, hence, are further away from each other 7 .
Since at ambient temperatures many rotational levels are populated, we experi-
mentally observe many lines known as P -branch, when ∆` = −1, as Q-branch, when
∆` = 0, and as R-branch, when ∆` = 1. See Exc. 11.2.6.3.
The space available for the particle is limited between the classical turning point,
1/α
C
which for a given energy is rt = |E| . The momentum corresponding to this
1/2
energy is kt = 2m|E|
~2 . Heisenberg’s uncertainty relation requires kt rt > 2, that
is, at least half of the wavelength must fit within the potential (between 0 and rt ) at
the height of the bound state. Therefore,
2~2
|E|1−2/α > . (11.67)
mC 2/α
For a Coulomb potential, with α = 1 and C = e2 /4πε0 , we obtain the energy of
the ground state of the hydrogen atom,
e2
E > E1 = − , (11.68)
4πε0 2aB
but there is no state higher than all others. That is, all energies En = E1 /n2 exist.
For α = 2, we do not get a condition for the energy. For the Casimir-Polder
potential, α = 3, we obtain,
8~6
E<− 3 2 . (11.69)
m C
This means that, in contrast to the Coulomb potential, the binding energy must be
lower than a certain limit.
∗ (n − 2)Γ 1 + n1 ∗ h2n
E(v ) = De − (v + vD ) n , (11.72)
2Γ 12 + n1 (2πMr ) C 2
where v ∗ is a number counting the vibrational levels from the top to the bottom
starting at the dissociation limit.
-0.1
(μK)
-0.2
V (R)
-0.3
-0.4
-0.5
0 1 2 3 4
R (μm)
Figure 11.13: (code) Highest vibrational states obtained by the LeRoy-Bernstein method.
Separating the radial and angular contributions, ψ(R) ≡ Y (ϑ, ϕ)f (R)/R, we obtain,
1 ∂2 l(l + 1)
− + V (R) + f (R) = Ef (R) . (11.74)
2µ ∂R2 2µR2
p
Now, we introduce the local wavevector, k(R) = 2µ[E − V (R)] − l(l + 1)/R2 and
write,
f 00 = −k 2 f . (11.75)
Figure 11.14: (code) (a) Numerical computation of the relative wavefunction for a low energy
collision, E & 0 and ` = 0. The blue curve shows the interatomic Li-Rb potential (a)3 Σ,
the red curve illustrates the relative Broglie wavefunction of the molecule. The asymptote
of this extrapolated wavefunction (green curve) cuts the abscissa at as = −120aB , which
is just the scattering length for collisions in this channel. (b) Numerical computation of
the wavefunction for a vibrational state. The potential is an interpolation (black) between
a short-range Morse potential (green), and a long range potential (blue). The red curve
illustrates the relative wavefunction for the ninth vibrational state (counting from ground
state).
340 CHAPTER 11. MOLECULAR DIMERS
and put it in a matrix form using the set of functions of the basis φi (Rj ) = δ(Ri − Rj )
with i = 1, .., N , where Ri = R0 +i(RN −R0 )/N . This problem has N eigenvalues Ei .
The Fourier grid method now evaluates the kinetic energy at each point in the grid.
We insert the local terms Hii = H(Ri ) and the non-local terms Hij = H(Ri , Rj ) into
the Hamiltonian, as well as the potential energies Vij = V (Ri )δij . The kinetic energy
is the inverse Fourier transform from momentum space of Trs = T (kr )δrs = (kr2 /2µ)δrs
and becomes [495, 472, 223],
!
π2 i−j 1 1
Hij = (−1) − for i 6= j (11.77)
4µ(RN − R1 )2 sin2 π(i−j) sin2 π(i+j)
2N 2N
!
π2 2N 2 + 1 1
Hij = − + V (Ri ) for i=j .
4µ(RN − R1 )2 3 sin2 2N
πi
×10−3
0 0
(cm−1 )
(a.u.)
-500 -2
-4
V (R)
-1000
V
-6
0 10 20 30 5 10 15 20 25 30
R (aB ) R (aB )
×104
(a.u.)
30 3 1
(M Hz)
(aB )
20 2
FC with N = 10
0.5
10 1
Rt
Bv
0 0 0
0 20 40 0 20 40 0 20 40
N N N
Figure 11.15: (code) Numerical computation of the wavefunction using the Fourier grid
method at example of the interatomic potential Li-Rb (1)1 Π. (a) Short range Morse potential
(green), long range (blue), interpolation (black) and [419]. (b) Vibrational wavefunctions,
(c) external turning point (red) and center of mass (green), (d) rotational progression, and
(e) Franck-Condon overlap with the tenth vibrational state.
11.2.6 Exercises
11.2.6.1 Ex: Transitions between vibrational states
Calculate the dipole moment between two arbitrary vibrational states of (a) a har-
monic potential and (b) a Morse potential.
interact, couple and generate new energy terms, which need to be taken into account
when calculating the various potentials of interatomic interaction,
P2 X (k)
(k)
Ĥ = +Vcoulomb (R)+ Vhf s + Vzeeman +Vdipole,spin−spin (R)+Vdipole,spin−orbit (R) .
2Mr
k=1,2
(11.81)
The Coulomb interaction for interacting alkaline gases can be expressed as:
S=0 S=1
Vcoulomb (R) = Vcoulomb PS=0 + Vcoulomb PS=1 . (11.82)
The projectors PS=0,1 will be required to expand the Hilbert space for the degrees of
freedom of the spins.
The van der Waals forces include all intermolecular forces. These are long-range
forces that occur between permanent and induced atomic dipoles ∼ 1/r6 9
Here, Bm is the width of the minimum, Rm the position of the minimum, Dm the
length. A long-range potential can be written,
C6 C8 C10
Vvdw (R) = De − − 8 − 10 . (11.84)
R6 R R
De is the energy of dissociation. The van der Waals coefficients Ck , which determine
the potential shape at large distances, can be calculated using other methods with
higher precision. To obtain a closed formula, the short and long range parts can be
joined by,
V = Vmorse F + Vvdw (1 − F ) , (11.85)
10
where F ≡ e−(R/Rt ) .
The situation is different for collisions of identical atoms in excited states, which
has a much larger range because of the resonant interaction between dipoles. In this
case, an additional Movre-Pichler potential dominated by a coefficient C3 arises,
e e e
Vvdw = Vmovre + Vdispersion . (11.86)
In contrast, excited state collisions of different species are purely short ranged.
9 They also occur, in a pure form, in optical resonators such as in the Casimir effect. Since the
√
lowest frequency in a cavity is ω = 2πc/L, the zero point energies inside and outside the cavity are
different. This causes an attractive force between the cavity mirrors ∼ 1/r3 , 1/r4 .
11.3. VAN DER WAALS FORCES AND SPIN COUPLING 343
When the atoms approach each other, at intermediate distances, they couple their
spins:
At short distances, they form a molecular dimer described by the quantum numbers:
The molecular interaction is so strong that L and S couple to the z-axis instead of
coupling to each other. This case is analogous to the Paschen-Back effect,
A common notation is to label the states Λ = Σ, Π, ∆, .... That is, in the symbol
X(2S+1 ΛΩ )±σ , where σ = g, u is the inversion symmetry, X, A, B, .. and a, b, .. are
the singlet and triplet series starting from the lowest energy levels. An alternative
notation is to assign labels ordered by energy X = (1), (2), ... Finally, ± is the
symmetry upon reflection. For example, X 1 Σ+ g.
L is projected onto the z-axis before coupling to `. The resulting angular momentum
afterward directly couples to S.
L→Λ (11.92)
((L, `)k, S)J .
11.3. VAN DER WAALS FORCES AND SPIN COUPLING 345
A common notation is to label the states by Ω = 0, 1, 2, .... That is, in the symbol
X(Ω)± −
s , the letter X = 1, 2, .. is a label ordered by energy. For example 2(0g ).
The last bracket is dropped for unsymmetrized recoupling. We will study examples of
spin recoupling in the ground state channels in Exc. 11.3.5.1, 11.3.5.2, and 11.3.5.3.
11.3.5 Exercises
87
11.3.5.1 Ex: Spin recoupling of identical Rb ground state channels
a. Unravel the molecular hyperfine structure of identical 87 Rb ground state channels.
b. Project the collisional channels |f1 mf 1 f2 mf 2 i = |1 − 1 1 − 1i and |22 22i on the
singlet and triplet potentials, S = 0, 1.
346 CHAPTER 11. MOLECULAR DIMERS
Collisions
Until now, we mostly restricted our studies to individual atoms or molecules. In
practice, however, we investigate atomic or molecular gases by spectroscopic methods.
The constituents of thermal gases are constantly in motion. They are subject to
Doppler-shifts of their spectral lines and they collide with each other. Colliding
atoms often get so close, that the electronic orbitals affect each other, which leads to
distortions of the interaction potentials, spectral lineshifts and modifications of the
interatomic forces. These are the topics to be addressed in the present chapter.
Figure 12.1: Atoms may interact via exchange of (a) virtual photons or (b) real photons.
is the thermal de Broglie wavelength and r0 the range of the interaction potential.
As the Van der Waals interaction gives rise to elastic collisions, the total energy
of the relative motion is conserved in time. As the potential energy vanishes at
large interatomic separation the total energy is usually expressed in the form E =
~2 k 2 /2mr . This implies that also the wavenumbers for the relative motion before
and after the collision must be the same and shows that, far from the potential
center, the collision can only affect the phase of the wavefunction - not its wavelength.
Apparently, the appearance of a shift in phase relative to the free atomic motion
347
348 CHAPTER 12. COLLISIONS
provides the key to the quantum mechanical description of elastic collisions. This
being said, we postpone the discussion of the actual collisional behavior to Sec. 12.2.
First we prepare ourselves for this discussion by analyzing the stationary states for
the motion in the presence of an interaction potential.
An important simplifying factor in the description of ultracold collisions is the
emergence of universal behavior in the relative motion of the atoms. The latter
applies to low-energy collisional states as well as to weakly bound states. Universal
means in this context that, asymptotically (for r r0 ), the wavefunctions become
independent of the details of the interaction potential but can be characterized in
terms of a few parameters, each representing some characteristic length scale of the
collisional system. In other words, very different short-range physics can give rise to
the same scattering behavior. From a theory point of view this universality has the
enormous advantage that the essential features of ultracold collisions can be described
with the aid of simple model potentials for which analytical solutions can be obtained.
In our analysis of the collisional motion three characteristic length scales will
appear, the interaction range r0 the scattering length a and the effective range re , each
expressing a different aspect of the interaction. The range r0 is the distance beyond
which the interaction may be neglected even for k → 0. The second characteristic
length, the s-wave scattering length a, acts as an effective hard-sphere diameter. It
is a measure for the interaction strength and determines the collision cross section in
the limit k → 0 as will be elaborated on in Sec. 12.2. The third characteristic length,
the effective range re expresses how the potential affects the energy dependence of
the cross section and determines when the k → 0 limit is reached.
The s-wave scattering length is the central parameter for the theoretical descrip-
tion of bosonic quantum gases. It determines both the thermodynamic and the col-
lisional properties of these gases. In single-component fermionic gases the s-wave
scattering length plays no role because the wavefunction for the relative motion of
the atoms has to be antisymmetric. In two-component fermionic gases this restriction
is absent for collisions between atoms of different components. As a consequence, in
these systems the inter-component s-wave scattering length determines the collision
related properties - for instance the thermalization rate.
In Sec. 12.1.1 we show how the phase shift appears as a result of interatomic
interaction in the wavefunction for the relative motion of two atoms. For free particles
the phase shift is zero. An integral expression for the phase shift is derived. In
Sec. 12.1.2 and beyond we specialize to the case of low-energy collisions (kr0 1).
The basic phenomenology is introduced and analyzed for simple model potentials like
the hard-sphere (Sec. 12.1.2) and the spherical well (Sec. 12.1.3), where the existence
of a short range is manifest. For the discussion of arbitrarily shaped potentials, we
refer to the script [729].
The starting point for the description of the relative motion of two atoms at energy
E is the Schrödinger equation (4.33),
1 L2
p2r + + V (r) ψ(r, ϑ, ϕ) = Eψ(r, ϑ, ϕ) . (12.2)
2mr r2
Here mr is the reduced mass of the atom pair and V (r) the interaction potential. As
discussed in Sec. 4.1.3 the eigenfunctions ψ(r, ϑ, ϕ) can be separated in a radial and
a angular part, ψ = R` (r)Y`m (ϑ, ϕ), where the functions Y`m are spherical harmonics
and the functions R` (r) satisfy the radial wave equation,
~2 d2 2 d `(` + 1)
− 2− + + V (r) R` (r) = ER` (r) . (12.3)
2mr dr r dr r2
Figure 12.2: Example showing the high-lying bound states near the continuum of the singlet
potential 1 Σ+
g (the bonding potential) of the hydrogen molecule; v and J are the vibrational
and rotational quantum numbers, respectively. The dashed line shows the effect of the
J = 3 centrifugal barrier. The presence of a rotational barrier gives rise to an exponential
suppression of the radial wavefunction for r < rtp and is negligible at distances where the
interaction becomes noticeable r r0 .
350 CHAPTER 12. COLLISIONS
choosing k and κ as real positive number. This puts Eq. (12.3) in the form,
2 `(` + 1)
R`00 + R`0 + ε − Ṽ (r) − R` = 0 . (12.6)
r r2
represent the amplitude c` and the asymptotic phase η` of the wavefunction. Note
that this equation is singular in the origin except for the case of vanishing phase shifts.
Therefore, in the case of free particles we require η` = 0 for all angular momentum
values `. This implies that the general solution reduces to the regular one,
For r 1/k the spherical Bessel and Neumann functions assume their asymptotic
form and we find,
c`
R` (k, r) −→ [cos η` sin(kr − 12 `π) + sin η` cos(kr − 21 `π)] (12.15)
kr→∞ kr
c`
= sin(kr + η` − 12 `π) .
kr
where we introduced a the constant η` representing the asymptotic phase shift. For a
given value of k this phase shift fixes the general solution of the radial wavefunction
R` (k, r) up to an ` dependent normalization constant c` . Note that in view of the
k dependence of the phase shift, R` is a function of k and r rather than a function
of the product kr. Whereas in the case of free particles the phase shifts must all
vanish, in the presence of the interaction they provide the proper asymptotic form
of the distorted waves. The non-zero asymptotic phase shift is the signature of the
interaction at short distance; the motion becomes free-particle like (undistorted) only
at large distance from the scattering center. In elastic scattering the relative energy
~2 k 2 /2m is conserved; hence, asymptotically also k and the de Broglie wavelength.
This leaves only the asymptotic phase of the wave to be affected.
Example 54 (Scattering matrix ): Rewriting Eq. (12.15) in complex notation,
1 1
c` −iη` e−i(kr− 2 `π) iη` e
i(kr− `π)
2
R` (k, r) ' i e −e , (12.16)
r→∞ 2k r r
352 CHAPTER 12. COLLISIONS
R (r)
0.5
0 5 10 15 20 25
r/r0
Figure 12.3: (code) Radial wavefunctions (` = 0) for various values of k (down to the
k → 0 limit) in the case of a hard sphere potential. The boundary condition is fixed by the
requirement that the wavefunction vanishes at the edge of the hard sphere, R0 (ka) = 0.
The phase shift follows from the boundary condition R0 (k, a) = 0, which can be
written in the form,
cos η0 sin ka + sin η0 cos ka = 0 . (12.23)
Hence, the phase shift is
η0 = −ka . (12.24)
With this expression Eq. (12.22) reduces to
c0
R0 (k, r) = sin[k(r − a)] . (12.25)
kr
This expression is exact for any value of k, as announced above. The linear k de-
pendence of η0 simply expresses its definition in which the shift of the wave (by a) is
compared to the de Broglie wavelength λdB , η0 = −2πa/λdB . As a consequence the
phase shift vanishes for k → 0,
This result is obvious when comparing the finite shift a to the diverging wavelength
λtherm . Interestingly, in the limit k → 0 the expression (12.25) becomes k indepen-
dent,
a
R0 (r) ∼ 1 − for a ≤ r 1/k . (12.27)
k→0 r
This important result is illustrated in Fig. 12.3. In the limit k → 0 the wavefunction
is essentially constant throughout space (up to a distance 1/k → ∞ at which it starts
to oscillate), except for a small region of radius a around the potential center.
Here |U0 | = κ20 is called the well depth (κ0 is chosen to be real and positive, κ0 > 0).
The energy of the continuum states is given by ε = k 2 . In analogy, the energy of the
bound states is written as,
εb = −κ2 . (12.29)
We now have to solve the radial wave equation (12.6) with the spherical well potential
(12.28). Since the potential is constant inside the well (r ≤ r0 ) the wavefunction has
to be free-particle like with the wave number given by,
p q
K+ = 2mr (E − V0 )/~2 = κ20 + k 2 . (12.30)
As the wavefunction has to be regular in the origin, inside the well it is given by,
R` (r) = C` j` (K+ r) for r ≤ r0 , (12.31)
where C` is a normalization constant. This expression holds for E > V0 (both E > 0
and E ≤ 0).
Outside the well (r > r0 ) we have for E > 0 free atoms, Ṽ (r) = 0, with relative
wavevector k = [2mr E/~2 ]1/2 . Thus, for r > r0 the general solution for the radial
wave functions of angular momentum ` is given by the free atom expression (12.11),
R` (k, r) = c` [cos η` j` (kr) + sin η` n` (kr)] for r > r0 . (12.32)
1.5
R (r)
0.5
0 1 2 3 4 5
r/r0
Figure 12.4: (code) (Left) Scheme of the flat bottom potential. (Right) (a) Radial
wavefunctions for square wells: (a) continuum state (ε = k 2 > 0); (b) Zero energy
state (ε = k 2 = 0) in the presence of an asymptotically bound level (ε = −κ2 = 0);
(c) bound state (ε = −κ2 < 0). Note the continuity of R0 (r) and R00 (r) at r = r0 .
The wavefunctions are not normalized and are shifted relative to each other only for
reasons of visibility.
The full solution [see Fig. 12.4(a)] is obtained by the continuity conditions for
R` (r) and R`0 (r) at the boundary r = r0 . These imply continuity of the logarithmic
derivative with respect to r,
j 0 (%i ) R0 (r) cos η` j`0 (%e ) + sin η` n0` (%e )
K+ ` = ` =k , (12.33)
j` (%i ) R` (r) r=r0 cos η` j` (%e ) + sin η` n` (%e )
12.1. MOTION OF INTERACTING NEUTRAL ATOMS 355
(1) (1)
h` (%e ) K+ ∂% ln j` (%i ) − k∂% ln h` (%e )
e2iη` = − (2) (2)
, (12.35)
h` (%e ) K+ ∂% ln j` (%i ) − k∂% ln h` (%e )
where the expression ∂% ln stands for the logarithmic derivative. This expression may
look a bit heavy, but is valuable as it represents the exact result for arbitrary `.
In Exc. 12.1.5.2 we simplify this formula for ` = 0. As the formula (12.35) lacks
transparency from the physical point of view, we analyze in the coming sections the
case ` = 0 directly discussing the radial wavefunctions u` .
10
5
a/r0
-5
-10
0 1 2 3 4
κ0 r0 /π
Figure 12.5: (code) The s-wave scattering length a normalized on r0 as a function of the
depth of a spherical square potential well (blue curve). Note that, typically, a ' r0 (green
line), except near the resonances at κ0 r0 = (n + 21 )π being an integer.
Note that this expression coincides with the general result given by Eq. (12.33) for
the case ` = 0; i.e. the boundary condition of continuity for u00 /u0 coincides with that
for R00 /R0 , as we know from a calculation left to Exc. 12.1.5.3. Furthermore, for a
vanishing potential (κ0 → 0) we have K+ → k and the boundary condition properly
yields a zero phase shift (η0 = 0).
At this point we introduce the effective hard-sphere diameter a(k) to describe, in
analogy with Eq. (12.24), the behavior of the phase shift,
By this procedure we extract the linear k dependence as well as the negative sign from
the phase shift. This is a good idea because the linear k-dependence does not arise
from the potential but simply from the definition of the phase in which, as discussed
earlier, the shift of the wave is compared to the de Broglie wavelength. In the limit
k → 0, we have K+ r0 → κ0 r0 and with the definition,
γ ≡ κ0 r0 (12.43)
is called the well parameter. As shown in Fig. 12.5, the value of a can be positive,
negative or zero depending on the value of γ. Therefore, rather than using the pictorial
term effective hard-sphere diameter the name scattering length is used for a. Next to
the range, the scattering length represents the second characteristic length that can be
associated with the interaction potential. As the name suggest, it is a measure for the
scattering behavior of atoms, and we elaborate on this in Sec. 12.2. Also, in Sec. ??
we will show that a is also a measure for the effective strength of the interaction.
Fig. 12.5 and Eq. (12.42) show that a is typically a quantity of the size of r0 ,
although for γ = tan γ it is zero and for γ = (ν + 21 )π, with ν being an integer,
12.1. MOTION OF INTERACTING NEUTRAL ATOMS 357
na3 1 . (12.46)
Otherwise, the interaction with neighboring atoms will distort the relative motion
of the colliding pair. This violates the binary scattering approximation on which
Eq. (12.45) is based. The dimensionless quantity na3 is called the gas parameter.
When its value is small, the gas is called nearly ideal or weakly interacting 2 .
where ε = −κ2 is the discrete energy eigenvalue of a bound state with ` = 0. The
solutions are of the type (see Fig. 12.30),
C0 sin K− r for r ≤ r0
u0 (k, r) = (12.48)
c0 e−κr for r ≤ r0
2 Note that weakly interacting does not mean that that the potential is ’shallow’. Any gas can be
Figure 12.6: Reduced radial wavefunctions u0 (r) for continuum states (ε > 0) in the k → 0
limit for increasing well depth near the threshold value κ0 r0 = (n + 21 )π: (a) presence of
an almost bound state (a < 0); (b) presence of zero-energy resonance (κvb = 0, a → ±∞);
(c) presence of a weakly bound state (a > 0); (d) deeper binding of the least bound state.
For r > r0 the wavefunction is given by u0 (r) = c0 (r − a); hence, the value of a is given
by the intercept with the horizontal axis. This gives rise to a characteristic node at r = a,
which is real for a > 0 (just as for hard spheres of diameter a), but virtual for a < 0. The
wavefunctions are not normalized.
where κ > 0 because the bound state wavefunction has to be normalized. The bound
state energy is obtained by requiring the continuity of the logarithmic derivative when
connecting the inner part of the wavefunction to the outer part,
u00 (r)
= K− cot K− r0 = −κ , (12.49)
u0 (r) r=r0
γ = (v + 21 )π , (12.52)
where v = 0, 1, .., rmax is the vibrational quantum number. This shows that a mini-
mum well parameter (γ = π/2) is required to bind the first state 3 . For the least-bound
level, vmax , we have,
(vmax + 12 )π = Int ( πγ − 12 ) (12.53)
and the total number of bound s levels follows with,
The relation between κ and v for a given vibrational level depends on the ratio κ/K−
and is given by
κ
cot K− r0 = − . (12.55)
K−
Note that this relation corresponds to K− r0 ' (vmax + 21 )π for the least-bound state
-0.5
-1.5
−(γ/π)2
-2
0 2 4 6 8
(γ/π)2
Figure 12.7: (code) Appearance and increase of binding of the first three bound levels for
increasing well depth. The quadratic dependence near threshold is universal (i.e. independent
of the well shape). The full crossover curve is obtained by numerical solution of Eq. (12.49)
and corresponds to a π/2 phase shift of K+ r0 near threshold. The dashed line shows the
increase in well depth.
(κ/K− 1) and to K− r0 ' (vmax + 1)π for deeply bound levels (K− /κ 1), as is
illustrated in Fig. 12.8(a).
potential center like a halo. This behavior holds for arbitrary short-range potentials.
Figure 12.8: Bound states oscillate inside the well and decay exponentially outside the well:
(a) the boundary condition depends on the ratio κ/K− ; (b) a slight reduction of the well
depth can turn the least bound state into a halo state.
With Eqs. (12.56) and (12.57) we have obtained two expressions for κ0 cot κ0 r0 and
arrive at the conclusion that in the presence of a weakly bound state the scattering
length is given by,
a ' 1/κ . (12.58)
κ→0
This expression reveals the tight relation between the binding energy of the least-
bound state, given by Eq. (12.29), and the scattering length,
~2 κ2 ~2
Eb = − −→ − . (12.59)
2mr κ→0 2mr a2
(a) 4
2 ṽ = 31 32 33 34 35
ηres (k)
π/2
−π/2
-2
(b) 0
ares (k)/r0
-0.1
-0.2
-0.3
0 10 20 30 40 50 60
kr0
Figure 12.9: (code) (a) Resonance contribution to the s-wave phase shift η0 (k) for a large
well parameter slightly detuned from the threshold value (at γ = 31.5π), such that the scat-
tering length is negative (∆γ = −0.5): The linear shift of the background contribution is
not included in the plot. Note that the π phase jumps arise from the modulo-π representa-
tion of the arctangent and do not represent an observable phenomenon; the physical phase
increases monotonically and equals ηres = 21 (modulo π) at the center of the resonances; (b)
contribution of the resonances to the effective hard sphere diameter a(k) = −η0 (k)/k. As
(in this example) the lowest resonance is not close to threshold the resonant enhancement is
small, |ares (k)| = r0 1.
ṽ = 29 30 31 32 33 34
2πγ
1 Δε =
r02
sin2 ηres
4kres
Γ=
0.5 r0
0
-1000 -500 0 500 1000 1500 2000 2500
k2 r02
Figure 12.10: (code) (a) Transition from bound states to Breit-Wigner s-wave resonances
plotted for ∆γ = −0.5 with respect to the threshold at γ = 31.5π (same conditions as
Fig. 12.9). The bound states are indicated as zero-width spikes at energies ε = −κ2 , with
κ following from Eq. (12.49). For ε > 0 the plot is based on Eqs. (12.60). The width of
the resonances increases with the square root of the energy. Note that the band of energies
typical for the quantum gases (kr0 1) corresponds to a narrow zone, unresolved on the
energy scale of the plot.
is,
2 2πγ
∆εres = ε(ṽ+1)
res − ε(ṽ) π
res = 2(ṽ + 1) r 2 ' r02
. (12.65)
0
To analyze a given resonance we expand K+ cot K+ r0 about the point of zero crossing.
For this purpose we introduce the notation,
q
δk kres
K+ = κ20 + (kres + δk)2 = Kres + , (12.66)
Kres
where δk = k − kres is called the detuning from resonance. Thus, restricting ourselves
to the low-energy (but not zero energy) s-wave resonances (1 < kres r0 Kres r0 ' γ),
we may approximate K+ cot K+ r0 ' Kres cot K+ r0 . Expanding cot K+ r0 about the
zero crossing at K+ r0 = (ṽ + 1/2)π and retaining only the linear term we obtain (see
Problem 3.4),
Kres cot K+ r0 = −δk kres r0 . (12.67)
Hence, the diverging argument of the arctangent becomes,
k 1 −(k + kres ) −2kres /r0
tan ηres = '− = 2 2 )r
' . (12.68)
K+ cot K+ r0 δk r0 (k − kres 0 ε − ε0
The expansion (12.66) is valid over the full range of the resonant change in phase
provided the following condition holds,
Kres γ
δk r0 ' kres r0 , (12.69)
kres
which is satisfied for the lowest resonances as long as the well parameter is sufficiently
large (γ kres r0 ). As long as δk kres we may further approximate k ' kres . With
these approximations and after restoring the dimensions, Eq. (12.68) can be written
as a function of the energy E = ~2 2k 2 /2mr ,
k −Γ2
tan ηres = ' , (12.70)
K+ cot K+ r0 ε − εres
12.1. MOTION OF INTERACTING NEUTRAL ATOMS 363
where
Γ/2 = 2kres /r0 (12.71)
is called the spectral width of the resonance. Comparing the expressions for Γ and
∆Eres we find that for given r0 the width Γ is independent of γ, whereas the resonance
spacing is proportional to γ . Thus, only for sufficiently large well parameters (γ 1)
the spectral width becomes smaller than the resonance spacing,
Γ ∆εres ⇐⇒ kres r0 πγ . (12.72)
Knowing the tangent of ηres , we readily obtain the sine and Eq. (12.70) is replaced
by the Breit-Wigner formula,
(Γ/2)2
sin2 ηres = . (12.73)
(ε − ε0 )2 + (Γ/2)2
For optical resonances this energy dependence is known as the Lorentz lineshape. Note
that Γ corresponds to the full-width-at-half-maximum (FWHM) of this line shape.
The lowest energy resonances are plotted in Fig. 12.10 along with the highest-energy
bound states.
The resonance near threshold (almost bound level) deserves special attention, as
this type of resonance is the only one that can play an important role within the band
of energies relevant for the quantum gases (kr0 1). Fig. 12.10 shows that near
the threshold (at γ = 31.5π) the resonance narrows down and becomes asymmetric,
which means that the Breit-Wigner lineshape is lost. Using Eq. (12.54) we calculate
ṽ = 31. The narrow line is reminiscent of a bound level but the scattering length is
negative (see Fig. 12.10). Under these conditions the wavefunction has a virtual node
at r = −|a| (see Fig. 12.30). Accordingly, the level is called a virtual level and the
wavefunction is said to represent a virtual bound state. In analogy with the bound
states its energy is written as ε = κ2res , where κ is to be defined later.
Figure 12.11: Wavefunctions corresponding to the same binding energy (ε = −κ2 ) plotted
for three different values of r0 . Outside r0 the wavefunctions fall off exponentially, always
with the same decay exponent κ; this is the essence of the Bethe-Peierls boundary condition.
The dashed lines show the extrapolation for r → 0. (a) reference case; (b) for κr0 1 most
of the probability density of a bound state is found outside the well (halo state); (c) for
zero-range potentials (κr0 = 0) the oscillating part of the wavefunction is compressed into
a delta function and only the decaying exponent remains (universal limit). Note that these
wavefunctions do not share the same normalization.
e−κr
R0 (r) = c0 for r>0, (12.76)
r
R √
and with κ > 0. Unit normalization, r2 R02 (r)dr = 1, is obtained for c0 = 2κ. For
E > 0 we can arrive at the same conclusion. The boundary condition for k →0 and
given value of r0 is given by Eq. (12.41), which we write in the form,
1
= κ0 cot κ0 r0 . (12.77)
r0 − a
Reducing the radius r0 , the scattering length a can be conserved by increasing κ0 . In
the limit r0 → 0 the well depth should diverge in accordance with,
1
− = cot κ0 r0 → 0 . (12.78)
κ0 a
This is again satisfied for κ0 r0 ' π/2. In the zero-range limit the radial wavefunction
12.1. MOTION OF INTERACTING NEUTRAL ATOMS 365
(a) 50
(a.u.)
0
V
-50
0 20 40 60 80 100
R (aB )
(b) 200 (c)
0
(a.u.)
(aB )
-100
0
as
V -200
-200 -300
0.8 0.9 1 10 20 30
η R (aB )
Figure 12.12: (code) Dependence of the scattering length on the potential depth.
Because this equation can be solved analytically in the limit k → 0 it is ideally suited
to analyze the conditions under which the potential V (r) may be neglected and thus
to determine r0 . To solve Eq. (12.85) we look for a clever substitution of the variable
r and the function R` (r) to optimally exploit the known r dependence of the potential
in order to bring the differential equation in a well-known form. To leave exibility in
the transformation we search for functions of the type,
~2
Ev = − . (12.87)
2mr a2
C6
V =− , (12.88)
R6
we get,
Ev = −[(vD − v)H(mr , C6 )]−1/3 , (12.89)
12.1. MOTION OF INTERACTING NEUTRAL ATOMS 367
where vD is the vibrational quantum number at the dissociation limit and H a con-
stant. Also,
1/4
Γ(3/4) 2µC6
a= √ 1 − tan π(vD + 21 ) . (12.90)
2 2Γ(5/4) ~ 2
Z R Z ∞
(k+1) (k+1) (k) 2mr C6 (k)
δγ = 2mr V δγ dR = δγ (∞) − − δ dR (12.98)
0 R R6 γ
Z ∞ (k)
δγ
= 2mr C6 dR .
R R6
In particular,
−1 −1
δα(1) = 2mr C6 , δα(2) = (2mr C6 )2 (12.99)
4R4 4 · 9R9
(1) 1 (2) 1
δβ = 2mr C6 5 , δβ = (2mr C6 )2 .
5R 5 · 10R10
368 CHAPTER 12. COLLISIONS
where we used the expression for the s-wave phase shift, η0 = −ka. This is the alter-
native boundary condition we were looking for. Substituting this into Eq. (12.106)
we obtain the inhomogeneous equation
(∇2 + k 2 )ψk (r) ' −4πac0 δ 3 (r) . (12.108)
k→0
This inhomogeneous equation has the solution (12.104), as demonstrated in Exc. 12.1.5.4.
For functions f (r) with regular behavior in the origin we have 4 ,
∂ ∂
rf (r) = f (0) + r f (r) = f (0) , (12.109)
∂r r=0 ∂r r=0
and the pseudo potential takes the form of a delta function potential 5 ,
4π
Ṽ (r) = − δ 3 (r) ' 4πaδ 3 (r) , (12.110)
k cot η0 k→0
g 3 4π~2
V (r) = δ (r) with g= . (12.111)
2 mr
4 Note that the wavefunction ψk is irregular,
∂ ∂ c0 sin(kr + η0 )
δ (r) [rψk (r)] = δ 3 (r)
3
r = δ 3 (r)c0 cos[k(r − a)] = c0 δ 3 (r) .
∂r ∂r kr k→0
5 Note that the dependence on the relative position vector r rather than its modulus r is purely
formal as the delta function restricts the integration to only zero-lengthR vectors. This notation is used
to indicate that normalization involves a 3-dimensional integration, δ 3 (r)d3 r. Pseudo potentials
do not carry physical significance but are mathematical constructions that can chosen such that they
provide wavefunctions with the proper phase shift.
370 CHAPTER 12. COLLISIONS
This expression, for na3 1 valid in the zero energy limit, is convenient for calculating
the interaction energy, as will be shown in Sec. ??.
Figure 12.13: (code) The coupling of molecular potentials (solid lines) generates new adia-
batic potentials (dash-dotted lines).
12.1.5 Exercises
12.1.5.1 Ex: Asymptotic radial function for hard-sphere potentials
Using asymptotic expressions for the Bessel and von Neumann functions derive the
radial function R` (k, r) for the two limiting cases ka 1 and ka 1.
with Ĥ = p2 /2m + V (r) and Ek = ~2 k 2 /2m. The boundary conditions are given
by the scattering geometry in such a way that at long distances the wavefunction
behaves as (see Fig. 12.14),
eiks r
ψk (r) ∼ eik·r + fk (Ω) . (12.114)
r
For elastic scattering processes we have ks = k. The scattering amplitude fk (Ω)
depends on the energy Ek and on the scattering solid angle. Experimentally, we
scatter individual particles described by wave packets. Since the scattering theory is
linear, we can describe the packets by superpositions of stationary solutions ψk 6 .
(∆ + k 2 )ψk (r) = 2m
~2 V (r)ψk (r) . (12.115)
This equation is not a common eigenvalue problem, since any energy Ek generates a
solution. The equation (12.115) is a partial inhomogeneous differential equation with
the left side describing free propagation and the right side describing a source that
depends on the solution. Such differential equations are usually solved using Green’s
functions. We choose a point source and we solve,
along with the boundary conditions. The solution takes the form [81],
1 eik|r|
G(r, k) = − , (12.117)
4π |r|
such that,
Z
ψk (r) = e ik·r
+ G? 2m
~2 V ψk (r) = e ik·r
+ 2m
~2 d3 r0 G(r − r0 , k)V (r0 )ψk (r0 ) .
V
(12.118)
The equation (12.118) is called Lippmann-Schwinger equation. Of course, this equa-
tion does not solve, but only reformulate the problem taking into account the bound-
ary conditions. It is more appropriate for an implementation of approximations. See
Exc. 12.2.7.1 and 12.2.7.2.
Now let us consider the far field, r → ∞, to verify the asymptotic behavior and
find an expression for fk (Ω) as a function of V (r). For r → ∞ we can approximate
p p
k|r − r0 | = kr (êr − r0 /r)2 = kr 1 − 2êr · r0 /r + (r0 /r)2 ' kr − k0 · r0 ' kr ,
(12.119)
12.2. SCATTERING THEORY 373
Starting from the wavefunctions ψin ≡ eik·r and ψs ≡ fk (Ω)eikr /r we can calculate
the current densities,
~ † ~k
Jin = (ψin ∇ψin − c.c.) = (12.122)
2mi m
~ † ~k0 1
Js = (ψs ∂r ψs − c.c.)êr = |fk (Ω)|2 + O(r−3 ) .
2mi m r2
The number dI(Ω) of particles scattered per second into the solid angle dΩ is
simply dI(Ω) = |Js |r2 dΩ. With this we can calculate the differential effective cross
section defined by the ratio between dI(Ω) and the number |Jin | of incident particles
per second,
dσ dI(Ω)
≡ = |fk (Ω)|2 . (12.123)
dΩ |Jin |dΩ
At time t0 the results (12.125) and (12.126) must match. To verify this, we write
(12.125) replacing the plane wave eik·r using the Lippmann-Schwinger equation (12.118)
with Green’s function (12.117), and then we compare the coefficients,
Z " Z #
ik|r−r0 |
d3 k m 3 0e 0 0
ψ(r, t0 ) = ak ψk (r) + d r V (r )ψk (r ) . (12.127)
(2π)3 2π~2 |r − r0 |
The scattering process is illustrated in Fig. 12.15. To simplify the calculation of
the second term in this equation, we assume that ψk is smooth, that is, there are no
resonances, such that we can approximate, ψk ' ψk0 . With k ' k · êk0 we obtain,
Z Z
d3 k ik|r−r0 | 0 d3 k 0
3
ak e ψ k (r ) = ak eik·(êk0 |r−r |) ψk0 (r0 ) (12.128)
(2π) (2π)3
(12.125)
= ψ(êk0 |r − r0 |, t0 )ψk0 (r0 ) .
Here, ψ(êk0 |r − r0 |, t0 ) is the incident wave package evaluated to the right, where by
definition it is ' 0. The expression (12.127) therefore has the form,
Z
d3 k
ψ(r, t0 ) = ak ψk (r) , (12.129)
(2π)3
and a comparison of the coefficients with (12.126) gives, Ak = ak . Finally, we evaluate
ψ(r, t) at the time of detection t > t0 to understand, that the above stationary analysis
is actually physically correct. According to (12.126) we have,
Z
d3 k
ψ(r, t) = Ak ψk (r)e−iEk (t−t0 )/~ (12.130)
(2π)3
Z
(12.120)(12.125) d3 k eikr
' ψ0 (r, t) + ak fk (Ω)e−iEk (t−t0 )/~ .
(2π)3 r
Hence, ψ0 (r, t) describes the evolution of the wave packet without scatterer,
Z
d3 k
ψ0 (r, t) = ak eik·r e−iEk (t−t0 )/~ . (12.131)
(2π)3
| {z }
ψ(r,t0 )
If fk it’s smooth around k = k0 , which allows us to place this amplitude (fk ' fk0 )
in front of the integral, and with k ' k · k̂0 we obtain,
t large fk0 (Ω)
ψ(r, t) =⇒ ψ0 (r, t) + ψ0 (k̂0 r, t) . (12.132)
| {z } | r {z }
packet not scattered
packet scattered
12.2. SCATTERING THEORY 375
The scattering process is illustrated in Fig. 12.15: According to the last equation
the scattering process involves the superposition of the non-scattered packet and a
packet scattered in the direction Ω. The latter involves the amplitude Ψ0 (k̂0 r, t) of
a packet propagating in forward direction, which only needs to be evaluated at the
right time and distance. This packet will then be multiplied with the amplitude
describing the angular dependency fk0 ; the angle, therefore, only appears through
this amplitude and not in the wavefunction ψ0 . In two situations the above analysis
can not be applied :
2m
ψ(r) = ψi (r) + G ? ~2 V ψ (r) (12.133)
2m
2m 2
= ψi (r) + ~2 (G ? V ψi )(r) + ~2 [G ? V (G ? V ψi )](r)
Z
= ψi (r) + 2m
~2 G(r − r0 )V (r0 )ψi (r0 )d3 r0
V
Z
2m 2
+ ~2 G(r − r0 )V (r0 )G(r − r00 )V (r00 )ψi (r00 )d3 r0 d3 r00 .
V
In the so-called Born approximation we consider only the first perturbation order,
and inserting a plane wave, ψi (r) = eikz /(2π)3/2 , we obtain,
Z 0
eikz m eik|r−r | 0
ψ(r) = − V (r0 )eikz d3 r0 . (12.134)
(2π)3/2 (2π)3/2 2π~2 V |r − r0 |
The asymptotic behavior r r0 , it follows with (12.119) using z 0 = r0 ·êz and defining
ks = kêr e ki = kêz ,
Z ik(r−r·r0 /r)
eikz m e 0
ψ(r) ' − V (r0 )eik·r d3 r0 (12.135)
(2π)3/2 (2π)3/2 2π~2 V r
Z
eikz m eikr 0
= + V (r0 )ei(ki −ks )·r d3 r0
(2π)3/2 (2π)3/2 2π~2 r V
1 ikz eikr
≡ e + f (ki , ks ) ,
(2π)3/2 r
with
Z
m 0 m
f (ki , ks ) ≡ V (r0 )ei(ki −ks )·r d3 r0 = − hks |Ṽ |ki i .
2π~2 V 2π~2
376 CHAPTER 12. COLLISIONS
where the factor (2` + 1)i` is a convention facilitating the calculation later on. By
inserting this separation ansatz for the radial and angular variables into the stationary
Schrödinger equation (12.113), we obtain the radial Schrödinger equation,
2
∂ `(` + 1) 2 2m
− + k rR` (r) = 2 V (r)rR` (r) , (12.137)
∂r2 r2 ~
where ψk must satisfy the boundary conditions (12.114). Fortunately, we can also
expand the incident wave by partial waves 7 ,
∞
X
ikz ir cos θ
e =e = (2` + 1)i` j` (kr)P` (cos θ) . (12.138)
`=0
We now use the result (12.138) to find the boundary conditions for the radial
r→∞
waves R` . In the infinity we have rV (r) −→ 0. For this reason,
r→∞ (2) (1)
R` (r) −→ α` [h` (kr) + s` h` (kr)] , (12.139)
(1,2)
where the Hankel functions h` (kr) ∼ e±i(ρ−(`+1)π/2) describe, respectively, incident
(2) (1)
(h` ) and outgoing (h` ) spherical waves.
To determine the coefficients α` and s` we note first that, without potential,
V (r) = 0, the solution of the radial equation (12.137) is known,
(2) (1)
R` (r) = j` (kr) = 21 [h` (kr) + h` (kr)] , (12.140)
(2)
such that α` = 21 and s` = 1. For V (r) 6= 0 the incident wave h` is the same, but
(1)
not the incident h` , which results in s` 6= 1. However, particle number conservation
requires that the number of particles entering the potential is equal to the number of
particles coming out. That is, the total radial flow must be,
~ ~
0 = jr` (r) = [R∗ ∂r R` − R` ∂r R`∗ ] = [|s` |2 − 1] , (12.141)
2im ` 4mkr2
7 For the more general case of arbitrary vectors k e r, we use the addition theorem for Y
`m and
express P` (cos θ) by spherical functions,
∞ X
X `
∗
eik·r = 4π i` j` (kr)Y`m (Ωk )Y`m (Ωr ) .
`=0 m=−`
12.2. SCATTERING THEORY 377
e−i(kr+w` i(kr+w`
approximating 2R` ' kr + s` e kr . Hence, |s` | = 1, that is,
We call
e2iδ` − 1 eiδ` sin δ`
= ≡ f` (12.145)
2ik k
the amplitude of the partial wave [735, 104].
The quantity
4π 4π
σ` = 2
(2` + 1) sin2 δ` = 2 (2` + 1)|f` |2 (12.148)
k k
is called effective partial cross section. Obviously, σ` ≤ 4π
k2 (2` + 1) holds. The phase
shift e2iδ` has a simple physical interpretation: We consider the function,
eiδ` (2) (1)
eiδ` j` (kr + δ` ) = 2 [h` (kr + δ` ) + h` (kr + δ` )] (12.149)
iδ`
` i(kr+δ` ) ` −i(kr+δ` )
e (−i) e (+i) e krδ` 1 (2) (1)
∼ + −→ 2 [h` + e2iδ` h` ] ∼ R` .
2 kr + δ` kr + δ`
(1) ikr
8 Com h` ∼ (−i)`+1 ekr .
378 CHAPTER 12. COLLISIONS
Now comparing the case V = 0 giving R` (r) = j` (kr) with the case V 6= 0 giving
R` (r) ∼ eiδ` j` (kr + δ` ), we notice that a positive displacement, δ` > 0, pulls the
wavefunction into the potential, while a negative displacement, δ` < 0, pushes the
wavefunction out, as illustrated in Fig. 12.16.
Figure 12.16: Phase shift δ` (k) of the scattered wavefunction. (a) An attractive potential
increases the kinetic energy and the wavefunction oscillates faster, which causes a positive
phase shift and a negative scattering length. (b) A repulsive potential slows down the wave-
function oscillation and produces a positive scattering length. (c,d) In an attractive potential
deep enough to support vibrational states, the wavefunction performs several oscillations.
The sign of the scattering length then depends on the distance of the last bound state to
the continuum.
The deeper meaning of the optical theorem is the conservation of particle number:
The flux of scattered particles, (~k/m)σ = Isct , must be extracted from the incident
flux I0 by scattering, and therefore, is missing in the forward direction. It is the
interference of the scattered wave with the incident wave, which diminishes the non-
scattered wave and therefore creates a shadow of the scatterer in the forward direction.
The particles missing in the shadow of the scatterer are precisely those that have been
scattered. This is the message of the optical theorem, which is always valid in the
absence of possible (inelastic) processes leading to trapping or a transformation of the
particles.
12.2. SCATTERING THEORY 379
Element amixed
s atriplet
s atriplet
p BF B,s−wave BF B,p−wave
[aB ] [aB ] [aB ] [ G] [ G]
1
H 1.23
2
H −6.8
6
Li −35 160, 186, 215
7
Li 10 −27.3
23
Na 52 85
39
K 118 81.1
40
K 158 1.7 −100 200 200
41
K 225 286
85
Rb −450 −363 156
87
Rb 105 109.3 685.43, 911.74, 1007.34
133
Cs −240 −350
135
Cs 163 138
while for r < R0 the solution R` must be found by integrating the radial equation
(12.137). The scattering phase s` must be chosen in a way that R` and ∂r R` be
continuous at R0 . The normalization factor vanishes in the logarithmic derivative,
such that,
∂r h` + s` ∂r h`
(2) (1)
1 ∂R`
γ` ≡ ∂r ln R` |R− = = . (12.155)
0 R` ∂r R− (2)
h + s` h
(1) +
0 ` ` R0
9
Now
2(∂r − γ` )j`
s` − 1 = (12.156)
(γ` − ∂r )h
(1)
` R0
2i
or with s` − 1 = cot δ` −i expressing δ` by γ` ,
(∂r − γ` )n`
cot δ` = . (12.157)
(∂r − γ` )j` R0
4π 4π 2` + 1
σ` = (2` + 1) sin2 δ` = 2 . (12.158)
k 2 k 1 + cot2 δ`
The poles of s` are just the bound states: A bound state asymptotically satisfies
(1)
R` (r) ∼ h` (iκr) ∝ e−κr with the binding energy EB = −~2 κ2 /2m. The condition
of continuity is given by,
∂r h`
(1)
γ` = (1)
, (12.159)
h ` R0
and the insertion into the general continuity condition (12.157) gives,
(1) (1)
h` ∂r n` − n` ∂r h`
cot δ` = (1) (1)
=i. (12.160)
h` ∂r j` − j` ∂r h`
9 We (1,2)
have for the spherical Hankel functions: h` (x) = j` (x) ± iy` (x).
und einsetzen in die allgemeine Stetigkeitsbedingung (6.62) ergibt
(1) (1) (1)
hl ∂r nl − nl ∂r hl hl =jl +inl
cot δl = (1) (1)
= i. (6.66)
12.2. SCATTERING THEORY
h ∂r jl − jl ∂r h 381
l l
CE
Er
σl −iΓr /2 δl
(n +1)π
Γr
Γr
nπ
Er E Er E
Abb.12.17
Figure 6.6: :Resonanzpeak der Breite
Peak of the resonance Γr : Links
of width der Wirkungsquerschnitt
Γr : (Left) effective cross sectionσlσ(E)
` (E), (right)
und rechts
scattering dieδ`Streuphase
phase δl (E);scheme
(E), and (center) in derofMitte ist die ofLage
the position des Poles
the pole in the in der energy
complex
komplexen
plane E. E-Ebene skizziert.
that is,
sin δ` ' (R0 k)2`+1 . (12.166)
In other words, the scattering phases decrease rapidly, in the regime of cold collisions,
with increasing `, and ` = 0 type collisions dominate,
α` (E)'α` (0) 1 + R0 α0 (0)
k cot δ0 ' − . (12.167)
R02 α` (E)
then it is the only relevant parameter of the collision. For R0 α0 1 we find a ' R0 .
For example, for a hard sphere we have R` (R0 ) = 0, α` = ∞, a = R0 > 0 and
cot δ0 = −1/kR0 . For small kR0 we obtain δ0 ' −kR0 < 0, which corresponds to
a negative phase shift for the repulsive potential, as expected. The effective cross
section is,
4π 1 4π
σ0 = 2 2 ' 2 . (12.169)
k 1 + cot δ0 k + 1/a2s
2
In comparison to the effective cross section for higher angular momenta, σ` ∝ sink2 δ`
behaves like,
σ` ∝ R02 (R0 k)4` → 0 . (12.170)
12.3. COLD ATOMIC COLLISIONS 383
In contrast, we find that the scattering at low energies has an s-wave character, σ
being dominated by σ0 ,
σ(E = 0) = 4πa2s . (12.171)
For a hard sphere (a = R0 ) we find an effective cross section four times larger than
classically expected (σcl = πR02 ).
V(r)
as
r
Figure 12.18: The cold collision can be described by a hard core potential.
In summary, for kinetic energies below the centrifugal barrier, only s-wave colli-
sions are significant. The higher-order partial waves are frozen behind the centrifugal
barrier. That is, the energy of cold collisions is not enough to excite a rotational
motion, not even the one with the lowest rotational energy allowed by quantum me-
chanics. Such a collision is called cold collision or s-wave collision.
R→∞ −1 T →0
The relative wavefunction tends asymptotically to ψ(R) −→ kdB sin [kdB (R − a)] −→
R − a. This means that for temperatures so low that the length of the Broglie wave
−1
of the relative motion is much longer than the potential range kdB Rturning , the
scattering becomes independent of temperature, and the scattering length a becomes
well defined 10
Generally, a repulsive interaction potential corresponds to a positive scattering
length and an attractive potential to a negative one. However, if the attractive po-
tential supports bound states, the value of the scattering length depends on the energy
of the last bound state with respect to the dissociation threshold [728, 729].
difference between an atom before and after an elastic collision is the phase shift δ0 of this wave.
384 CHAPTER 12. COLLISIONS
smallest of the values 8πa and 8π/k. Even though a(B) can be increased to divergent
values near a Feshbach resonance,
p for finite collision energies it will√never exceed an
effective value aef f = 1/ a2 + 1/k 2 . For a thermal gas, k = ~−1 2πmkB T , while
for a pure condensate k = 2π/L, where L is the size of the condensate. Therefore,
the effective scattering length can not exceed the size of the condensate. Unitarity
2 2
k ~2
also means that the kinetic energy exceeds the binding energy, kB T ≡ ~2m ≤ 2ma 2.
10−6
10−8
(cm2 )
10−10
σ
10−12
10−2 100 102
T (μK)
Figure 12.19: (code) Temperature dependence of the collision cross section for various scat-
tering lengths. For higher scattering lengths, the unitarity limit is reached at lower temper-
atures.
The collision rate depends on the density, temperature and collisional cross section
[402],
√
γcoll.peak = 2σelast v̄n0 , (12.173)
√
γcoll.aver = γcoll.peak /2 2 .
√
Note that at the unitary limit, as v̄ ∝ T , we have that σelas ∝ T and n0 ∝ r̄−3 ∝
T −3/2 , such that the collision rate is independent of temperature.
At low temperatures, k → 0, the delocalization of the colliding particles is greater
than its short-range structure. This is the Wigner threshold law [742, 735]. For
elastic collisions of neutral ground state atoms, the cross section (12.169)√ becomes
temperature-independent. Thus, the rate coefficient decreases as hσv̄i ∝ T , while
the collision rate increases as hσv̄n0 i ∝ T −1 .
For inelastic collisions, σ ∝ T −1 . For three-body collisions see [239].
Figure 12.20: The symmetrization of the collision wavefunction produces two paths with
Für must
angles θ and π − θ, which eine be
zentralsymmetrisches
added coherently. Potential V (r) gehen wir zur Partialwel
darstellung über,
For the isotropic potential V (r) we use the partial wave
* representation,
X f (θ) = il (2l + 1)fl Pl (cos θ)
`
f (θ) = i f` P` (cos θ) . l (12.179)
` Pl (cos θ) = (−1) Pl (cos(π − θ))
↓ l
*
f (θ)
With P` (cos θ) = (−1)` P` (π − cos θ) + (π − θ) = 2
wefget, il (2l + 1)fl Pl (cos θ), (12.
X l gerade
f (θ) − f (π − cos θ) = 2 i` f` P` (cos θ) , (12.180)
und finden, dass nur gerade
`par Drehimpulse l vorkommen (für ungerade l we
selt das Legendre Polynom das Vorzeichen und die Beiträge vernichten
even angular momenta appear 11
gegenseitig).
and we find that only
11 For ` odd the Legendre polynomial changes sign, and the contributions vanish.
kR0
X
σ' 4π
k2 (2` + 1) sin2 δ` (12.188)
`=0
kR0
X
' 4π
k2 (` + 1) cos2 [kR0 − (` + 1)π/2] + ` sin2 (kR0 − `π/2)
`=0
kR0
X kR0 (kR0 + 1)
= 4π
k2 `(cos2 + sin2 ) = 4π
k2 = 2πR02 , (12.189)
2
`=0
Figure 12.21: (a) Bound state for ` = 0. (b) Bound state for ` > 0 in a potential including
the centrifugal barrier ~2 `(` + 1)/2mr2 . (c) Resonances for ` = 0 are broad and possibly not
defined with Γr > Er . A defined resonance with Γr < Er requires that |∂E α0 | be large. (d)
For ` > 0 we obtain narrow resonances called shape resonances, because the decay of the
state is suppressed by the centrifugal barrier.
C6
V1 (r) = − . (12.190)
R6
or long-range van der Waals potential. Two other molecular states arise from the
interaction of an excited atom with the ground state atom. The predominant term of
the interaction is the potential of resonant dipole-dipole interactions,
C3
V2,3 (r) = − ∓ . (12.191)
R3
which gives rise to an attractive and a repulsive potential. The R−3 dependence of the
resonant dipole interaction means that the associated potentials greatly modify the
asymptotic level even at internuclear distances where the ground state van der Waals
potential is still relatively shallow. The photoassociation process involves a slow ap-
proach of two identical ground state atoms. A mono-modal optical field, tuned to the
red of the atomic resonance, is applied. When the two atoms reach an internuclear
distance RC such that the energy of the applied field ~ωC exactly coincides with the
potential difference V2 (Rc ) − V1 (Rc ), the probability of transferring population from
the fundamental molecular state to the excited molecular state is maximal. This
molecular resonance is sometimes called Condon point. The conventional way of cal-
culating this probability follows the procedure elaborated in Sec. 13.1 for a two-level
atom. First, we would solve the time-independent molecular Schrödinger equation
to obtain the molecular wavefunctions. Then, write down the coupled differential
equations by describing the time dependence of the expansion coefficients of the rel-
evant molecular wavefunctions, solve for the coefficients and calculate the square of
their absolute values. Finally, we would need to integrate the transition probability
inside a region ∆R around the Condon point, where the transition probability is not
be negligible. The dressed states picture allows to reduce this rather laborious pro-
gram, essentially, to a problem of a curve crossing of two levels. The bottom panel
of Fig. 12.22 illustrates the photoassociation in the dressed states picture. The basis
states are now product states made up of the field and the molecular levels. Further-
more, we approximate the molecular states themselves as products of atomic states.
This approximation is justified by the long range, weakly perturbative influence of
the van der Waals and resonant dipoles interactions. Calling the ground and excited
states |1i and |2i, respectively, we have
The two molecular curves intersect at the Condon point and optically couple to the
applied field. This optical coupling produces an avoided crossing near RC and mixes
the states of the molecule-field basis. The well-known Landau-Zener formula (LZ)
expresses the probability of traversing from one adiabatic molecular state to another
as a function of the interaction strength, the relative velocity of the collision partners,
12.3. COLD ATOMIC COLLISIONS 389
and the relative slopes of the two curves. The probability of LZ is given by,
2πh1, n| Ω2 |2, n − 1i
exp , (12.194)
v d ∆V12 (RC )
dR
d
where v is the relative radial velocity of the approaching particles and dR ∆V12 (RC )
is the difference of slopes of the two non-interacting potentials at the Condon point.
The dipole-field interaction operator Ω must be taken with the dipole of the molecular
transition. A reasonable approximation is to take the moment of molecular transition
as twice the atomic moment and average over the whole space. The result is,
2π
√
3
h1, n|Ωat |2, n − 1i
exp d , (12.195)
v dR ∆V12 (RC )
where Ωat denotes the atomic operator of the dipole-field interaction. In the case of
3
a crossing of an essentially flat potential V1 and V2 (R) = − CR3 , the absolute value of
the derivative of the slope difference is,
d
∆V12 (RC ) = 3C3 . (12.196)
dR RC 3
Figure 12.22: (Above) Molecular states resonantly coupled by a light field at the Condon
point. (Below) Same coupling of the molecular states represented by an avoided intersection
in the dressed states basis.
At short distances the coupling ((S, I)f1 , f2 )f breaks up and a ((s1 , s2 )S, I)f coupling
emerges. This does not hold for fully stretched spin states f1 + f2 = f , because f
is a good quantum number at all distances (and small enough magnetic fields). In
contrast mf is a good quantum number at all distances and at all fields.
In order to obtain the potentials coupled by hyperfine interaction, we first calculate
for a single vibrational level the relative wavefunctions |ψ(f1 ,f2 )f (R)i from the known
perturbed but uncoupled potentials V (R) + Vhf , where,
The antisymmetric part is negligible [719], so that Vhf ' 21 hahf,1 S · i1 + 12 hahf,2 S · i2 .
Then couple the channels via,
Xq s1 s2 S
|ψ(S,I)f (R)i = A Ŝ Iˆfˆ1 fˆ2 i1 i2 I |ψ(f1 ,f2 )f (R)i (12.199)
f1 ,f2 f1 f2 f
Xq s1 s2 S
|ψ(f1 ,f2 )f (R)i = A−1 Ŝ Iˆfˆ1 fˆ2 i1 i2 I |ψ(S,I)f (R)i .
S,I f1 f2 f
(12.201)
!
s1 s2 S q
1 − (1 − δf1 f2 )(−1)S+I+`
× i1 i2 I Ŝ Iˆfˆ1 fˆ2 p .
2 − δf1 f2
f1 f2 f
The last bracket is dropped for unsymmetrized recoupling. Calculations have been
done for 87 Rb collisions and for 6 Li87 Rb collisions. It turns out that the fully stretched
states have pure triplet character.
Note that in strong magnetic fields the coupling (i1 , i2 )I easily breaks up, and the
recoupling to be considered is rather hSmS mi1 mi2 `0 m`0 |f1 mf 1 f2 mf 2 `m` i.
1
+ 4 hahf,1 (i1,+ S− δmS mS +1 δmi1 mi1 −1 + i1,− S+ δmi1 mi1 −1 δmS mS +1 )δmi2 mi2
0 0 0 0 0
+ 14 hahf,2 (i2,+ S− δm0S mS +1 δm2 mi2 −1 + i2,− S+ δm0i2 mi2 −1 δm0S mS +1 )δm0i1 mi1 .
For an example on how to evaluate the matrix at various magnetic fields and obtain
the hyperfine structure of a bound state in the Zeeman and the Paschen-Back regime,
we propose the systems 6 Li23 Na and 6 Li87 Rb.
S = 0, .., s1 + s2 (12.205)
I = 0, .., i1 + i2
fj = |ij − sj |, .., ij + sj
|I − S| ≤ F ≤ I + S
|f1 − f2 | ≤ F ≤ f1 + f2
(−1)I = (−1)S for all F
(−1)f1 = −(−1)f2 for all odd F .
What does they look like for fermions? What does they look like for heteronuclear
collisions?
392 CHAPTER 12. COLLISIONS
where,
2
X
Mif = 0
(CS=0 CS=0 0
− CS=1 CS=1 ) (12.207)
mS ,I,mI
C = hS, mS ; I, mI |F1 m1 ; F2 m2 i .
See also [655]. The reason for the above dependence can be understood as follows.
as and at determine the energy of the last bounds states of the singlet and triplet
potentials. The more those energies are different, the stronger the coupling ∝ |Et −
Es |−1 . An alternative, physical picture is given in [116].
of a third atom receiving the excess energy as kinetic energy and results in trap loss
of the molecule and of the atom. They are suppressed as the gas gets more and
more dilute since the rate goes like n3 . There might be interesting 3-body resonance
phenomena like the Efimov state predicted for Helium droplets [228], [450] and recently
seen in experiment [426, 687, 104, 461].
3-body recombination is in contrast to the more controlled way of creating molecules
via photoassociation or coherent free-bound coupling.
Ultracold collisions are crucial for BEC. They provide the thermalizing elastic
collisions necessary for evaporative cooling, they are the cause for the condensate
self-energy and give rise to nonlinearities in the condensate dynamics. But they
are also interesting by themselves. The spectrum of two-body collisions exhibits
interesting features like shape- and Feshbach resonances [160, 364]. Their analysis
facilitates detailed conclusions on the nature of the interaction potentials. Three body
collisions are important to investigate, because they constitute a decay mechanism of
the intrinsically metastable system that represents a trapped Bose-gas.
Three body collisions are not a quantum statistical effect and don’t require the
presence of a BEC, but they occur only at very high densities comparable to those
necessary for BEC. It should be possible to detect them in photoassociation (PA)
spectra. Photoassociation provides a tool to explore the level structure of excited
states by shining in a laser with frequency ν tuned between the colliding channel
and a vibrational bound level of the excited state potential. Since the excited state
preferentially decays into the continuum, where the atoms have high kinetic energy,
the transition rate may be monitored via trap losses. Or we can shine in an additional
laser that further excites the excited atoms into the ionized continuum where they
can be registered by an ion detector.
Three-body photoassociation processes should reveal themselves by additional
lines in the two-body photoassociation spectra. Those lines should only appear at
very high densities and their strength should scale as the density cubed,
Rb + Rb + Rb + hν → Rb∗3 . (12.208)
The spin-orbit interaction splits the potential curves. The recoupling is described by,
Xq l1 l2 L
|ψ(L,S)j (R)i ∼ L̂Ŝj1 j2 s1 s2 S |ψ(j1 ,j2 )j (R)i . (12.210)
j1 ,j2 j1 j2 j
From Fig. 12.23 we see that the molecular states are remixed at long range: Every
state 0− − + +
g , 0u , 0g , 0u , 1g , 1u , 2g , 2u has several molecular states to which it connects,
−
e.g. 0g connects within the fine structure to (2)3 Σ+ 3
g and (2) Πg . Those multiplets
12.3. COLD ATOMIC COLLISIONS 395
form a closed interacting subspace. According to Movre and Pichler [523, 706] we get
with ∆ = E(2 P3/2 ) − E(2 P1/2 ) = 23 Aso ,
√ ! √ !
2
E(3 Πσ√) − 13 ∆ 3 ∆ E(3 Πσ√) − 13 ∆ − 32 ∆
0−
σ = , 0+
σ =
− 32 ∆ E(3 Σ+
σ) − 32 ∆ E(1 Σ+
σ)
(12.211)
1
E(3 Πσ ) 3∆ − 13 ∆
1σ = 1
3∆
1
E( Πσ ) − 13 ∆ , 2σ = 3
Πσ .
1 1
−3∆ −3∆ E(3 Σ+σ)
E(1 Σ+ 3 +
g ), E( Σu ) ∝ 2C3 /R
3
(12.212)
1 3 3
E( Πu ), E( Πg ) ∝ C3 /R
E(1 Πg ), E(3 Πu ) ∝ −C3 /R3
E(1 Σ+ 3 + 3
u ), E( Σg ) ∝ −2C3 /R .
Inserting the short range potentials and defining X = σC3 /3∆R3 , Y = E/∆ and
σ = + for g and − for u,
√ ! √ !
1 2 1 2
3X + 3X +
0−
σ =
√ 3
2
3 , 0+σ =
√ 3
2
3 (12.213)
3 −6X + 23 3 6X + 32
3X + 13 1
3 − 13
1σ = 13 −3X + 13 − 13 , 2σ = (3X + 1) .
1 1 2
−3 −3 −6X + 3
Note that the structure looks very much like the transition from the Zeeman to the
Paschen-Back regime, where X plays the role of the magnetic field and Y the level
shift. The characteristic equations are,
We thus expect a fine structure j87 = 12 , 23 at long range and exchange interactions
S = 0, 1 at short range. For the projection onto the internuclear axis |L − S| ≤ Ω ≤
L + S.
Let Vk (R) be the uncoupled potentials and,
∆kl (R) = Dkl + Akl [1 − e−Bkl (R−Rkl ) ]2 − 1 (12.216)
+ √
h3 Σ | V (3 Σ) ∆cb 2
Ω=0− √
h3 Π | ∆cb 2 V (3 Π) − ∆bb
Ω=0−
+
h1 Σ | V (1 Σ)(−2∆bb ) −∆bb
Ω=0+
h3 Π V (3 Π) +∆bb
| −∆bb
Ω=0+
+
h3 Σ | V (3 Σ) ∆cB ∆cb
Ω=1
h1 ΠΩ=1 | ∆cB V (1 Π) −∆bB
h3 ΠΩ=1 | ∆cb −∆bB V (3 Π)
h3 ΠΩ=2 | V (3 Π) + ∆bb
(12.217)
for every fine structure. For example, there is a matrix for the two 6 Li S1/2 87 RbP1,3/2
asymptotes and another one for the 6 Li P1/2,3/2 87 Rb S1 asymptotes. Additional
terms enter via rotational effects [68].
Concretely,
1
(E3/2 − E1/2 )
∆bb (R → ∞) = (12.218)
3
1
∆cB (R → ∞) = (E3/2 − E1/2 )
3
1
∆bB (R → ∞) = (E3/2 − E1/2 )
3
1
∆cb (R → ∞) = (E3/2 − E1/2 ) .
3
Figure 12.24: Movre type spin-orbit recoupling scheme for heteronuclear molecules.
the prefactors depend on the collision velocity (that is, the temperature of the sample)
and atomic parameters (for example, the collision cross section for two-body collisions,
which may itself depend on temperature). Hence, the total number of trapped atoms
evolves according to,
N 2 2
n(r) = e−r /2r̄ , (12.222)
(2π)3/2 r̄3
we calculate,
K L
Ṅ = −BN − 3/2 3
N 2 − 3/2 N3 . (12.223)
(4π) r̄ 3 (2π)3 r̄6
Fig. 12.26 shows a simulation of Eq. (12.223).
For condensates in the Thomas-Fermi limit [666] found the following differential
equations for two- and three-body collisions,
Ṅ 1 Ṅ 1
= −Gc2 N 2/5 − and = −Lc3 N 4/5 − . (12.224)
N τ N τ
398 CHAPTER 12. COLLISIONS
Σ+ S1/2 + P3/2
1 , Π1 , Π1
3 1 3
S+P
12 12 S1/2 + P1/2
1
Π1
(1000 cm−1 )
(1000 cm−1 )
10 1
Σ+
0+ 10 1
Σ+
0+ , Π0
3 +
3
Σ+
0− ,1
3
Σ+
0− , Π0
3 −
8 8
E
V
6 6
3
Π0+ ,0− ,1,2 3
Π2
4 4
10 20 30 10 20 30
R (aB ) R (aB )
106
105
N
104
0 0.05 0.1
t (s)
8π ω 3
Γα
vlm→v 0 l0 m0 = |h|d|i|2 , (12.225)
3 hc3
In electric fields the dipole moments will lead to Stark shifts. For |d| = 1 Debye
in a field of 1 V/m we expect a Stark shift of HStark ≈ 500 kHz. It may be possible
to measure this by photoassociation spectroscopy. Note also that since homonuclear
dimers do not have a permanent electric dipole moment, transitions between vibra-
tional ground states are forbidden.
dipole moment (Debye)
0 0
0
(a.u.)
-2 -2
-0.01
-4 -4
V -0.02
-6 -6
10 20 30 0 50 10 20 3
R (aB ) N R (aB )
Figure 12.27: (code) (a) Dipole moment at the example of LiRb as a function of interatomic
distance. (b) Dipole moment at the example of LiRb as a function of vibrational quantum
number.
12.3.13 Exercises
12.3.13.1 Ex: Photoassociation
Consider a laser focused into a cold, confined cloud of Na atoms at a temperature of
450 µK. For a detuning of 600 MHz, calculate the laser intensity (in W/cm2 ) required
to produce a 25% probability of photoassociation. The transition moment (a.u.) of
Na is 2.55.
that is, the regularization of the interaction potential, since the matrix T by itself does not diverge.
12.4. RESONANCES IN COLD COLLISIONS 401
Figure 12.29: (a) The Feshbach resonance is due to a coincidence of collisional and bound
channels. (b) Feshbach resonance in 85 Rb. The atoms collide in the hyperfine state F = 2,
and the resonant bound state is in the F = 3 multiplet (courtesy [160]).
be manipulated with the help of radiation fields [242, 88, 509] or, near a collisional
Feshbach resonance, by magnetic fields [703, 704]. Feshbach resonances were predicted
in nuclear systems [244]. Its recent revival in the context of cold atomic collisions is due
to the prospect of its use for manipulating the scattering length and thus controlling
the mean-field energy of a Bose-Einstein condensate 15
200
(cm2 )
(aB )
10−15
0
as
-200 10−20
1070 1080 1090 1070 1080 1090
B (G) B (G)
Figure 12.30: (code) (a) Scattering length and (b) collision cross section upon crossing a
Feshbach resonance. The scattering length can be positive (solid red line) or negative (dashed
blue line). A movie on a simulation of a Feshbach resonance can be watched here (watch
movie).
The collisional Feshbach resonance arises when the energy of the state of two
colliding atoms coincides with the energy of a vibrational molecular level belonging
to a higher energy asymptote (see Fig. 12.29,right). This coincidence can strongly
perturb the collisional channel, because the resonance shifts the phase of the relative
de Broglie wavefunction and allows the atoms to tunnel into the molecular state
for a short period of time, the duration of which is determined by the Heisenberg
uncertainty relation. If the sum of the magnetic dipole moments of the atoms is
15 Itis even possible to invert the sign of the scattering length. This was a prerequisite for conden-
sation of the atomic species of 85 Rb [603]. In addition, Feshbach resonances are interesting because
they can install a coherent free-bound coupling between an open-channel of colliding atoms and a
bound molecular state of the same atoms, as shown in Fig. 12.29. This coupling can lead to the
creation of molecular condensates.
402 CHAPTER 12. COLLISIONS
different from the dipole moment of the molecule, the resonance can be tuned via
external magnetic fields exploring the Zeeman effect. When a Feshbach resonance is
crossed, the scattering length crosses a singularity, as shown in Fig. 12.30(left) 16 .
16
Part III
Quantum Optics
403
Chapter 13
405
406 CHAPTER 13. INTERACTION OF LIGHT WITH MONO-ELECTRONIC ATOMS
With this scope we separate the Hamiltonian in a stationary part and a time-
dependent part 1 ,
where Ĥ (0) contains the kinetic energy and the Colombian potential of Eq. (9.11). In
Sec. 6.4.4, inserting the expansion,
X
|ψi = ak (t)|kie−ıEk t/~ , (13.2)
k
along with the Hamiltonian (13.1) in the Schrödinger equation, we obtained the first
order perturbative approximation (6.72). Setting the initial condition to ck (t ≤ 0) =
δki and supposing that the probability of finding the atom initially in the ground state
|ii for short times is 1, we got,
Z t
(1) 1 0
af (t) ' hf |Ĥint |iieıωf i t dt0 . (13.3)
ı~ 0
On the other hand, the energy density is proportional to the number of photons N (ω)
inside the volume V ,
N (ω)~ω
u(ω) = . (13.8)
V
The intensity corresponds to a flow of energy,
that is, its energy commutes with the other observables of the system.
13.1. SEMICLASSICAL THEORY OF THE EXCITATION OF ATOMIC TRANSITIONS407
Which of the two processes occurs, depends on the initial and final energies. For
Ef = Ei + ~ω the first term describing the process of absorption will dominate, for
Ef = Ei − ~ω the second term describing emission prevails.
The fact that, in a coupled atom-radiation system in equilibrium, the radiation field
excites the same number of transitions in absorption i → f as in stimulated emission
f → i is called the principle of detailed balance.
Obviously, the situation is different, if instead of two states we have several states
that can be excited by radiation or decay.
are imposed by the periodic boundary conditions on the spatial part of the plane
wavefunctions e−ık·r ,
kx,y,z = 2π
L nx,y,z , (13.21)
where (nx , ny , nz ) is a set of integers representing the various states of the photon.
In the limit of large L the variation between successive k is very small, such that we
can treat the numbers as continuous variables. That is, the number of states with k
between (kx , ky , kz ) and (kx + dkx , ky + dky , kz + dkz ) is,
L3 V 2 V ω2
dnx dny dnz = (2π)3 dkx dky dkz = (2π)3 k dkdΩ =≡ ρc (ω)dωdΩ ,
(2π)3 c3 dωdΩ
(13.22)
where the third expression is in spherical coordinates, the fourth one uses the relation
ω = ck, and the last one defines the density of states,
V ω2
ρc (ω) = (2π)3 c3 . (13.23)
Thus, the spontaneous emission rate of photons into the solid angle dΩ is,
Z Z
(sp) (sp) π~e2 2 V ω2
Wif dΩ = Wif ρc (ω)dω dΩ = |M f i | δ(ω + ωf i ) dω dΩ
ω ω ε0 m2 ωV (2π)3 c3
(13.24)
~e2
= |Mf i |2 ωf i dΩ ,
8π 2 ε 2 3
0m c
This simplified treatment with only two atomic states considers light as a scalar
field. In fact, light is a vector field and can have two independent orthogonal polar-
izations. The ttransition matrix may depend on polarization, such that,
Z X
(sp) ~e2
Wif = |Mfλi |2 ωf i dΩ . (13.25)
8π 2 ε0 m2 c3
λ=1,2
13.1.4 Exercises
13.1.4.1 Ex: Energy density of plane waves
~ t)2 =
Derive the result (13.7) for the temporal averages of the squares of the fields E(r,
~ t)2 = [∇ × A(r, t)]2 .
[−∂t A(r, t)]2 and B(r,
e−ık·r = 1 − ık · r − 1
2! (k · r)2 + ... . (13.26)
This expansion is justified by the fact that the wavelength (∼ 600 nm in the visible
spectrum) is much larger than the size of the scattering atom, kaB 1. The dipolar
approximation supposes,
e−ık·r ' 1 , (13.27)
such that we can remove the spatial dependence. In this approximation there is only
an interaction of the electric field of the radiation with the atom via an electric dipole
~ Thus,
term d · E.
We can calculate the expectation value of the velocity of the moving charge by the
Heisenberg equation using the unperturbed Hamiltonian,
The interpretation of the last equation is, that the states |ii and |f i are connected
through a displacement of the electronic cloud which, therefore, represents the induc-
tion of an electric dipole during the electronic transition. It is convenient to introduce
the electric dipole moment 4 .
function of the recoil operator dipole, eık·r̂ = |p + ~kihk|. For discussions of photonic recoil the
operator character must be maintained even in the dipole approximation, which can be done by
reinserting the recoil operator in the Hamiltonian
P after the dipole approximation has been applied.
4 In the presence of several atoms d = −e
j rj , where the rj are the radii of the orbits of the
various electrons of the atom.
13.2. DIPOLAR TRANSITIONS 411
where A(r, t) and Φ(r, t) are respectively the vector and scalar potentials of
the external field. In the Coloumb gauge the vector potential satisfies the wave
function,
1 ∂2A
∇2 A − 2 =0,
c ∂t
with solution A(r, t) = A0 eır·k−ωk t − c.c.. For atoms |r| λ, hence over
the extent of an atom the vector potential is spatially uniform A(r, t) = A(t),
the denominated dipole approximation. In this gauge, and after some unitary
transformation, the above Hamiltonian becomes,
1 ∂χ
Ĥ = [−ı~∇ + e(A + ∇χ]2 − e .
2m ∂t
where we can choose the gauge function χ(r, t) = −A(t) · r, which lead us to
the following relations,
∂χ ∂A
∇χ = −A and = −r · = −r · E~ ,
∂t ∂t
and the above Hamiltonian becomes,
~2 2
Ĥ = − ∇ − d̂ · E~ .
2m
(dp) π
Wf i = I(ωf i )|df i |2 cos2 θ δ(ω − ωf i ) . (13.33)
ε 0 ~2 c
In case of unpolarized (or randomly polarized) radiation we can replace the angular
distribution cos2 θ by its average value,
Z 2π Z π
1
cos2 θ = cos2 θ sin θdθdφ = 1
3 , (13.34)
4π 0 0
such that,
(dp,no−pol) π
Wf i = I(ωf i )|df i |2 δ(ω − ωf i ) . (13.35)
3ε0 ~2 c
This expression also represents the stimulated emission rate in the electric dipole
approximation.
412 CHAPTER 13. INTERACTION OF LIGHT WITH MONO-ELECTRONIC ATOMS
The total spontaneous emission rate can be obtained from Eq. (13.25) integrating
over all possible orientations,
Z X
(sp) ~e2
Wf i = |Mfλi |2 ωf i dΩ (13.36)
8π 2 ε0 m2 c3
λ=1,2
Z 2π Z π 2
~e2 mωf i
=2 2 2 3 · df i ωf i sin θdθdφ
ˆ
8π ε0 m c 0 0 e~
Z 2π Z π
e2 3 2 2 e2
= ω f i |r f i | cos θ sin θdθdφ = ω 3 |rf i |2 ,
4π 2 ε0 ~c3 0 0 3πε0 ~c3 f i
such that, for non-polarized light,
(sp) 4α 3 4α
Wf i = 2
ωf i |rf i |2 = 2 ωf3 i |rf i |2 = Γ . (13.37)
3c 3c
This is the rate of spontaneous decay of an excited atomic state. It can be measured
experimentally which, in turn, allows the calculation of the dipole moment,
r
3πε0 ~Γ
df i = . (13.38)
k3
In Exc. 13.2.6.1 we calculate the Rabi frequency from the dipole moment of an
atomic transition and the electric field of a radiation field.
with W0 = E0 df i and df i being the transition matrix element between atomic states.
The factor 13 comes from the fact that the vector E~ of the electric field can have any
polarization, but only polarizations along the direction of the oscillation of the dipole
moment contribute.
For a single atom, the result (13.40) is symmetric with respect to an exchange of
the initial and final states, that is, the rates for absorption and induced emission of
light are the same. For a sample of atoms being in thermal equilibrium, the popu-
lations Ni of the ground state and Nf of the excited state are unequal according to
Boltzmann’s law. Therefore, as we have shown in Sec. 1.2.5,
Nf Rf →i 6= Ni Ri→f . (13.41)
Thus, Einstein came to the conclusion that Fermi’s golden rule correctly describes
absorption, but does not contain all contributions of emission. The rates being related
to the Einstein coefficients by the equation (1.65), we find,
Rf →i = Bf i Nf u(ωf i ) (13.42)
and
~ωf3 i ~ωf3 i Ri→f ωf3 i
Sf →i = Af i Nf = B N
if f = = |df i |2 , (13.43)
π 2 c3 π 2 c3 u(ωf i ) 3π0 ~c3
exploiting the relation (1.70). We shall return to the subject of transitions in Sec. 13.1.
Defining,
ê±1 ≡ √1 (∓êx − ıêy ) , ê0 ≡ êz . (13.45)
2
we obtain
q
r±1 ≡ r · ê± = r · √1 (∓êx
2
− ıêy ) = √1 (∓x
2
− ıy) = ∓ √12 r sin ϑe±ıϕ = r 4π
3 Y1,±1 (ϑ, ϕ)
q
4π
r0 ≡ r · ê0 = r · êz = z = r cos ϑ = r 3 Y1,0 (ϑ, ϕ) . (13.46)
Similarly,
±1 ≡ ˆ
· ê± , 0 ≡ ˆ
· ê0 . (13.47)
Applying the expansion into spherical coordinates to the matrix element rf i = hf |r|ii
with êq · êq0 = δqq0 , it is easy to check,
X X X q X
·rf i =
ˆ ·êq )êq ·
(ˆ (rf i ·êq )êq = q hf |rq |ii = 4π
3 q hf |rY1,q |ii .
q=0,±1 q=0,±1 q=0,±1 q=0,±1
(13.48)
The matrix elements are,
q
hf |rq |ii = 4π3 hnf `f mf |rY1,q |ni `i mi i (13.49)
Z ∞ q Z
= 3
r Rnf ,`f Rni ,`i dr 4π 3 Y`∗f ,mf Y1,q Y`i ,mi dΩ .
0
This is the Wigner-Eckart theorem. The electric dipole operator is a simpler example
of a tensor operator Qqκ (r) characterizing the transition between atomic states. In
Excs. 13.2.6.2 and 13.2.6.3 we explicitly calculate, for a hydrogen atom subjected to
a magnetic field, components of the electric dipole operator.
Selection rules may be violated in higher orders, e.g., by multipolar radiation, as
in the cases of magnetic dipole transitions or electric quadrupole transitions. This
also is the case of the phenomenon of phosphorescence, which is a type of fluorescence
emitted by metastable states.
13.2.3.1 Parity
as shown above. That is, states with ` pair (impair) have even (odd) parity. Now
the integral (13.52) only does not vanish, when `i + `f + 1 = even. Therefore, dipole
transitions must change the parity of the states. F.ex., transitions S → P would be
possible, while S → S would be prohibited.
Table 13.1: Allowed transitions: (1-3) rigorous rules, (4-5) LS-coupling, and (6) in-
termediate coupling (https://en.wikipedia.org/wiki/Selection rule).
(E1) (M1) (E2) (M2) (E3) (M3)
∆J = 0, ±1 ∆J = 0, ±1, ±2 ∆J = 0, ±1, ±2, ±3
(1)
(∆J = 0 = 0) (∆J = 0 = 0, 1; 1 = 1 ) (0 = 0, 1, 2; 1 = 1 , 3 ; 1 = 1)
2 2 2 2 2
(2) ∆MJ = 0, ±1 ∆MJ = 0, ±1, ±2 ∆MJ = 0, ±1, ±2, ±3
(3) Pf = −Pi Pf = Pi Pf = −Pi Pf = Pi
one e − jump no e− jump none or one e− jump one e− jump one e− jump one e− jump
(4)
∆L = ±1 ∆L = 0; ∆n = 0 ∆L = 0, ±2 ∆L = ±1 ∆L = ±1, ±3 ∆L = 0, ±2
if ∆S = 0 if ∆S = 0
if ∆S = 0 if ∆S = 0
(5) ∆L = 0, ±1, ±2 ∆L = 0, ±1, ±2, ±3
∆L = 0, ±1 ∆L = 0 (L = 0 = 0, 1) (L = 0 = 0, 1, 2; 1 = 1)
(L = 0 = 0)
if ∆S = ±1 if ∆S = ±1 if ∆S = ±1 if ∆S = ±1 if ∆S = ±1
(6)
∆L = 0, ±1, ±2 ∆L = 0, ±1, ±2, ±3 ∆L = 0, ±1 ∆L = 0, ±1, ±2, ±3, ±4 ∆L = 0, ±1, ±2
(L = 0 = 0) (L = 0 = 0) (L = 0 = 0, 1) (L = 0 = 0)
z
B y
e- s+ s- p
x
k e
s+
s-
Figure 13.1: Selection rules due to polarization ˆ of the incident light. The projection of this
vector onto the axes π = ˆ · ê0 and σ± = ˆ
· ê± is proportional to the excitation probability
(and, obviously, also to the emission probability).
Using the spherical unit vectors ê±1 = √12 (∓êx − ıêy ) and ê0 = êz , we can reduce
Cartesian vector operators to first-order tensor operators:
X
T~ = Tq(1) êq resp. Tq(1) = T~ · êq . (13.65)
q
With this product it is possible to represent scalar, vector or tensor products of higher
ranks,
(0)
T(1) × U(1) = √1 T~ ~
·U (13.67)
0 3
(1)
T(1) × U(1) = ~
√1 (T ~ · êm
× U)
m 2
(2)
T(1) × U(1) ~ · U)
= − √16 (3Tz Uz − T ~
0
(2)
T(1) × U(1) = ± 12 [(Tx Uz + Tz Ux ) ± ı(Ty Uz + Tz Uy )]
±1
(2)
T(1) × U(1) = − 12 [(Tx Ux − Ty Uy ) ± ı(Tx Uy + Ty Ux )] .
±2
13.2. DIPOLAR TRANSITIONS 419
From the possible values for the Clebsch-Gordan coefficients follow directly the selec-
tion rules for multipolar radiation:
0 0
hjm|T(k)
q |j m i = 0 else E 0 − E = ~ω (13.69)
0 0
|j − j| ≤ k ≤ j + j
m0 − m = q
τ 0τ = T .
13.2.6 Exercises
13.2.6.1 Ex: Rabi frequency
From the expression for the dipole moment d and the relationship between the inten-
sity I and the electric field derive the Rabi frequency Ω produced by a laser beam of
intensity I by exciting an atomic dipole transition with the wavelength λ and decay
width Γ.
Since Ni = hψi |ψi i, we have |ψi (t)i = |ψi (0)ieıωif t−Γt/2 . The Fourier transform is,
Z ∞ Z ∞
1 1
|ξ(ω)i = √ |ψi (t)ie−ıωt dt = √ eıωif t−ıωt−Γt/2 dt|ψi (0)i (13.76)
2π 0 2π 0
1 eı(ωif −ω)t−Γt/2 − 1 1 1
=√ lim |ψi (0)i = √ |ψi (0)i .
2π t→∞ ı(ω if − ω) − Γ/2 2π ı(ω − ω if ) + Γ/2
The spectrum,
1 1
|ξ(ω)|2 = , (13.77)
2π (ω − ωif )2 − Γ2 /4
is a Lorentz distribution. Note, that the natural linewidth can be blurred by line
broadening effects, such as the Doppler broadening or collisions between atoms. These
effects will be discussed in the Chp. 14.6.1.
Excited states can sometimes decay into various states of lower energy. In this
case the linewidth is simply given by the sum of the partial decay rates, since the
convolution of LorentzP distributions LΓk with widths Γk is again a Lorentzian with
the total width Γ = k Γk .
The notion of the line strength becomes significant when working with real atoms
characterized by degenerate ground and excited states. In such cases we need to
extend the meaning of d12 and consider transitions between each of the degenerate
sublevels. For a non-degenerate two-level atom, the quantities d12 and A21 are simply
related by,
ω03
A21 = d2 . (13.79)
3π0 ~c3 12
If the lower level would be degenerate, the spontaneous emission rate coefficient would
be given by the sum of all possible deexcitation rates. In this case, d212 is defined as
the sum of the elements of the transition matrix coupling the excited and the lower
states, X
d212 = |hψ1,m1 |d|ψ2 i|2 . (13.80)
m1
Now, it can be shown that the spontaneous emission rate from any sublevel of a
degenerate excited state toward a lower level (that is, the sum over all lower sublevels)
is the same for all excited sublevels 8 This statement reflects the intuitively plausible
idea that spontaneous emission must be isotropic and unpolarized, if the sublevels
of an excited state are uniformly populated. Therefore, the insertion of Eq. (13.80)
into (13.79) should produce correct results, even when the excited state is degenerate.
Comparing the sum over all upper and lower degeneracies with the line strength S,
X
S12 = |hψ1,m1 |d|ψ2,m2 i|2 = g2 d212 . (13.81)
m1 ,m2
Therefore, the insertion of Eq. (13.81) into (13.79) must be accompanied by a factor
of 1/g2 to correct for the fact that all excited sub-levels radiate at the same rate.
Therefore, using the S12 of Eq. (13.81) the correct expression relating the transition
dipole between degenerate levels to the spontaneous emission rate is,
3π0 ~c3
S12 = g2 A21 , (13.82)
ω03
meaning that the line strength is proportional to the sum of the spontaneous emission
rates A21 from each one of the g2 excited levels toward all fundamental levels.
For an atom with two levels separated by an energy ~ω0 the oscillator strength for
emission is defined as a measure for the radiative decay rate A21 as compared to the
radiative decay rate γe of a classical electronic oscillator with frequency ω0 :
1 A21
f21 = − . (13.83)
3 γe
In case of degeneracy the oscillator strength for absorption is consequently defined by,
g2 g2 A21
f12 = − f21 = . (13.84)
g1 3g1 γe
8 This applies to Zeeman sublevels (summing up (3j)-coefficients). Check for other degeneracy
e2 ω02
γe = . (13.85)
6π0 me c3
Therefore, in terms of the A21 coefficient and of fundamental constants, the oscillator
strength for absorption is given by,
2π0 me c3
f12 = A21 . (13.86)
e2 ω02
Oscillator strengths obey certain sum rules that are useful for analyzing the relative
intensities of atomic spectral lines. For example, atoms with single valence electrons
(which are closer to the classical situation) obey the following sum rule,
X
fik = 1 , (13.87)
k
where the sum goes over all the excited states reached from the ground state. Alkaline
atoms are approximately one-electron systems, and the oscillator strength of the first
transition S −→ P is typically of the order of f12 = 0.7 − 0.95. The sum rule tells us
that most of the total transition probability for the excitation of the valence electron is
concentrated in the first transition S −→ P , and that transitions to higher states will
be comparatively weaker. Another sum rule exists for the excitation and spontaneous
emission from excited intermediate states j:
X X
fji + fjk = Z , (13.88)
i<j k>j
which is called the Thomas-Reiche-Kuhn sum rule. In the form of many electrons
[Eq. (13.88)] this rule is very useful, when Z is the number of equivalent electrons,
that is, electrons with the same quantum numbers n, l. Note also, that the numbers
are intrinsically negative. Oscillator strengths are often used in astrophysics and
plasma spectroscopy 9 . They are sometimes tabulated as log gf , where,
important quantities, such as the polarization, the susceptibility, the index of refrac-
tion, the extinction coefficient, and the absorption coefficient. We shall see later that
the polarization can be considered as densities of dipole moments induced into the
dielectric medium by the oscillating light field. But for the moment, we simply begin
by defining the polarization P̃~ with respect to an applied electric field Ẽ~ as,
~ = ε χ̃ Ẽ~ .
P̃ (13.90)
0 e
The tilde ∼ ornamenting the quantities means that they are complex, and that mea-
surable quantities must be taken from the real parts. χe is the linear complex electric
susceptibility, which is an intrinsic property of the medium responding to the light
field. It is related to the medium’s dielectric constant,
ε̃ = ε0 (1 + χ̃e ) . (13.91)
The relationship between the amplitudes of the electric and magnetic fields is,10
(χ) √ √ p 1
B0 = εµE0 = ε0 µ0 1 + χ̃e E0 = (η + ıκ)E0 = (η + ıκ)B0 . (13.97)
c
We use the subscript (χ) to mark quantities within the dielectric medium. The average
energy density is,
~ 2 = η 2 ū .
ū(χ) = 12 ε0 η 2 |E| (13.98)
10 In a dielectric medium, µ ' µ0 .
13.3. SPECTRAL LINES AND LIFETIMES 425
where
(χ)
I¯0 = 21 ε0 cη E~02 (13.100)
is the intensity at the point, where the light enters the medium, and
ωκ ω
K=2 = χ00e (13.101)
c ηc
is called absorption coefficient. Note, that the energy flow I¯(χ) in the dielectric
medium is always the product of energy density and propagation velocity c/η. Note
also, since the frequency ω of the light propagating through the dielectric remains the
same, the wavelength shrinks like λ = c/η ν [456].
1 1 ~ ~∗ Re ε ~ 2 1 η + ıκ ~2
ū(χ) = Re (εE~ · E~∗ + B·B )= |E| + E
4 µ0 4 4µ0 c
ε0 ε0 ε0 ε0
= (1 + χe )E0 + |η + ıκ| E0 = (η − κ + η + κ2 ) E02 = η 2 E02 ,
0 2 2 2 2 2 2
4 4 4 2
where we only consider the real part of the susceptibility χe . On the other hand,
the Poynting vector is,
2
1 ~∗ ] = ε0 c E02 e−2ωκz/c η + ıκ = ε0 c ηE02 e−2ωκz/c .
h i
I¯ = Re [E~ × B
2µ0 2 c 2
Now, we must treat the refractive index of the dielectric medium carefully. The
expression for the energy density must be modified following Eq. (13.98)
u(χ) (ω)
u(ω) = . (13.105)
η2
To use the Einstein coefficients, which suppose propagation at the speed of light in
the vacuum 11 , we must correct the energy density in the dielectric medium before
using it in the equation (13.104) [456]. Therefore, we should express u(ω) in this
equation by u(χ) (ω)/η 2 ,
u(χ) (ω)
N2 A21 = B12 (N1 − N2 gg12 ) . (13.106)
η2
Multiplying the two sides with ~ω0 , we see that the left side describes the rate of energy
scattered out of the beam by spontaneous emission, while the right side describes the
energy loss of the beam, i.e., the difference between energy removed by absorption
and energy returned to the beam by stimulated emission,
the number of photons that traverse an atom per unit of time. This flux can not depend on the
index of refraction. Therefore, we must use the spectral energy density calculated in the vacuum ūω .
13.3. SPECTRAL LINES AND LIFETIMES 427
Then we consider how to write the polarization in terms of a susceptibility when the
field contains two conjugate frequencies, ±ω. Substituting into Eq. (13.90), we have,
~
P(t) = 12 0 E~0 [χ(ω)eıωt + χ(−ω)e−ıωt ] . (13.109)
The polarization can also be expressed in terms of the density of the transition dipoles
in a gas of two-level atoms,
~ N N
P(t) = d12 (t) −→ hd12 (t)i . (13.110)
V V
where d is the transition dipole of only one atom, N = V is the atomic density, and
the quantum expectation value for the transition dipole moment is,
Z X
hd12 i = −e Ψ rj Ψd3 r . (13.111)
j
Making the weak field approach, C1 (t) ' 1, replacing the values obtained for
a1,2 in the transition dipole (13.113) and replacing the average of the orientations,
|hr12 i|2 −→ 13 |hr12 i|2 , in the polarization (13.110), we obtain,
~ N e2 |hr12 i2 E~0 eıωt eıωt e−ıωt e−ıωt
P(t) = + + + .
V 6~ ω0 + ω − ıγ ω0 − ω + ıγ ω0 − ω − ıγ ω0 + ω + ıγ
(13.120)
We compare this result with Eq. (13.109) and identify the susceptibility χ(ω) in terms
of the atomic properties and the frequency of the incident field,
N e2 |hr12 i|2
1 1
χ(ω) = + (13.121)
30 ~V ω0 − ω − ıγ ω0 + ω − iγ
N e2 |hr12 i2
ω0 − ω ω0 + ω 1 1
= + + ıγ + .
30 ~V (ω0 − ω)2 + γ 2 (ω0 + ω)2 + γ 2 (ω0 − ω)2 + γ 2 (ω0 + ω)2 + γ 2
In most practical situations in the laboratory ω will not be tuned more than some
100 GHz away from ω0 , hence |ω0 − ω| . 1011 Hz. With optical frequencies ω '
1015 Hz, it is clear that the second term on the right hand side of Eq. (13.121) will
be negligible compared to the first one. Therefore, we can discard the second term
and write the susceptibility as,
The Lorentzian profile term governs the frequency dependence of the absorption co-
efficient. We see that K exhibits a peak at the resonance frequency ω0 and a width
of Γ. The factor of π inserted in the numerator and denominator of the right term of
Eq. (13.123) allows to normalize the profile. We have also assumed in Eq. (13.123)
that the gas is sufficiently dilute for η ' 1 to hold, and that the line is sufficiently
narrow to be able to replace ω with ω0 , such that,
ω ω0
−→ . (13.124)
cη c
The absorption cross section has the same lineshape, since from Eqs. (1.73) and
(13.123) we have,
πd212 ω0
σ0a = L(ω − ω0 ) , (13.125)
30 ~cV
13.3. SPECTRAL LINES AND LIFETIMES 429
consistent with our previous expression for the frequency dependence of the absorption
cross-section. We can also write the imaginary component of the susceptibility in
terms of the cross section using the Eqs. (1.73) and (13.123)
cN
χ00 = σ0a . (13.126)
ω0 V
Do the Excs. 13.3.7.1 until 13.3.7.3.
13.3.7 Exercises
13.3.7.1 Ex: Solution for the susceptibility
Show that the solution of Eq. (13.114) is given by the result (13.119).
431
432 CHAPTER 14. THE BLOCH EQUATIONS
where {|ψk i} is a complete set of orthonormal states of the system under study. We
have a statistical distribution
P of these states with pk being the probability of finding
|ψk i in the set. Obviously, k pk = 1. That is, the density operator acts on a member
of the set |ψk i such as to extract the probability of finding the system in |ψk i,
X
ρ̂|ψm i = pk |ψk ihψk |ψm i = pm |ψm i . (14.2)
k
If all members of the set are in the same state, for example |ψk i, the density operator
reduces to,
ρ̂ = |ψk ihψk | , (14.3)
and the system is in a pure state with pk = δkk0 . Each time a quantum state can
be expressed by a single wave function, it is a pure state, but it does not have to be
an eigenstate. Starting from the equation (14.2) we find the diagonal elements of the
density matrix, which are the probabilities of finding the system in |ψm i,
Here, m and µ are degenerate quantum numbers, m, n are discrete, and λ, µ are continuous quantum
numbers. The set of quantum numbers is complete, when
X Z
|kmihkm| = 1̂ = |λµihλµ|dλdµ .
k,m
P
The degree of degeneracy of a state |ki is Tr P̂k = m 1. The probability of finding the system in
P
the state |ki is hP̂k i = pn m 1.
14.1. DENSITY MATRIX 433
Assuming that all |ψk i are orthonormal, the non-diagonal elements are necessarily
zero. Besides that, X
hψk |ρ̂|ψk i = 1 . (14.5)
k
So, ρ̂ contains all information about the system. When a state is unknown, ρ̂ describes
the probability of finding the system in each state. When the state is fully known, ρ̂
describes a pure state, that is, a vector in the Hilbert space, which is unequivocally
determined by a complete set of observables with their respective quantum numbers.
The properties of the density operator are,
ρ̂ = ρ̂†
hρ̂i ≥ 0
Tr ρ̂ = 1
. (14.6)
Tr ρ̂2 ≤ 1
det ρ̂ = 0 for a pure state
ρ̂ = ρ̂2 for a pure state
14.1.1.1 Entropy
In a very general sense, the entropy determines in what direction a reversible process
will take place. It is related to the size of the available phase space on both sides of
the reaction. For example, the coupling of discrete and continuous modes is governed
by entropy considerations.
Entropy measures of the lack of information about a system from which we only
know hĤi,
S ≡ −kB hln ρ̂i = −kB Tr (ρ̂ ln ρ̂) . (14.7)
The information entropy (or von Neumann entropy) of statistically independent sys-
tems ρ̂ ≡ ρ̂1 ⊗ ρ̂2 is additive S = S1 + S2 . We can also define absolute temperatures
by T −1 ≡ ∂S/∂hĤi. The entropy of a pure state is 0. Hamiltonian processes conserve
entropy, for they correspond to non-dissipative unitary transformations. On the other
side, relaxation increases the entropy and the phase space volume. Another common
definition is the so-called purity or Renyi entropy,
SR ≡ h1 − ρ̂i = 1 − Tr (ρ̂2 ) . (14.8)
Quantum states can exhibit coherences. For example, if we express a state |ψi on
a basis of eigenstates |1i and |2i:
|hψ|1i|2 h1|ψihψ|2i
ρ̂ = |ψihψ| = . (14.9)
h2|ψihψ|1i |hψ|2i|2
434 CHAPTER 14. THE BLOCH EQUATIONS
The measurement process is not described by this equation. A pure state will always
remain pure. If the eigenstates do not interact, the density operator will remain
diagonal. The von Neumann equation conserves the properties of hermiticity, ρ̂ = ρ̂† ,
completeness, Tr ρ̂ = 1, and purity det ρ̂ = 0.
The density operator for a statistical mixture in a canonical ensemble (where S is
maximum, U is variable, and N is fixed) follows from a variational problem with the
Lagrange parameters δ(S − kB αh1̂i − kB βhĤi) = 0, since Tr ρ̂ and hĤi are fixed by
boundary conditions. We find,
1 −Ĥ/kB T
ρ̂ = e with Z ≡ Tr e−Ĥ/kB T . (14.11)
Z
We also have the expectation values, hHi = −∂ ln Z/∂β and (∆H)2 = −∂hĤi/∂β
with the abbreviation β ≡ (kB T )−1 . All quantities are fixed, except the kinetic
energy, which balances the interaction with a heat bath. T is the only equilibrium
parameter. The density operator satisfies a Boltzmann distribution 2 ,
Z
p2 ∂ 2 kB
U = hĤi = =− ln e−p /2mkB T dp = T . (14.12)
2m ∂(1/kB T ) 2
as the projection of the state vector |ψk i on the basis vector |ni. Now, we can write
the density operator matrix representation within the basis {|ni} using the definition
of ρ̂ in Eq. (14.1) and replacing the expansions of |ψk i and hψk | of Eq. (14.13):
X X X X X
ρ̂ = pk |ψk ihψk | = pk |nihn|ψk ihψk |mihm| = pk cnk c∗mk |nihm| .
k k m,n k m,n
(14.15)
The matrix elements of ρ̂ in this representation are
X
ρnm ≡ hn|ρ̂|mi = pk cnk c∗mk (14.16)
k
2 The von Neumann entropy S of a mixture can be expressed in terms of the eigenvalues or in
terms of the trace and the logarithm ofPthe density operator ρ̂. Since ρ̂ is a semi-definite positive
operator, itsPspectrum λi , given by ρ = i λi |ϕi ihϕi | where
P {|ϕi i} is an orthonormal basis, satisfies
λi > 0 and λi = 1. Then the entropy becomes S = − i λi ln λi = −Tr (ρ ln ρ).
14.1. DENSITY MATRIX 435
P
with the diagonal elements hn|ρ̂|ni = k pk |cnk |2 and,
X X
ρ∗nm = hn|ρ̂|mi∗ = pk c∗nk cmk = pk hm|ψk ihψk |ni = hm|ρ|ni = ρmn , (14.17)
k k
Example 60 (Density operator for a single atom): For a very simple system
such as a single atom with several levels, that without spontaneous emission can
be described by a single wavefunction |ψ1 i, we can let pk = δ1k . That is, the
equations (14.15) and (14.17) reduce to,
X
ρ̂ = cn1 c∗m1 |nihm| and hn|ρ|mi = cn1 c∗m1 . (14.18)
m,n
With the definition of the density operator (14.1) we can write the Eq. (14.19) as,
X
Tr ρ̂ ≡ pk hn|ψk ihψk |ni . (14.20)
n,k
which shows that the trace of the density operator representation is always 1 regardless
of the basis of the matrix representation.
Expectation values of observables are expressed by,
X
hÂi = pk hψk |Â|ψk i . (14.22)
k
The Eq. (14.26) says that the ensemble average of any dynamic observable  can be
calculated from the diagonal elements of the operator matrix ρ̂Â: Since the trace is
independent of the basis (this will be shown in Exc. 14.1.5.1), each unitary transfor-
mation taking the matrix representation from a basis {|ni} to another one {|ti} leaves
the trace invariant. Using the definition of a unitary transformation we can easily
show that the trace of a cyclic permutation of a product is invariant. For example,
and in particular
Tr [ρ̂Â] = Tr [Âρ̂] = hÂi . (14.28)
In the Excs. 14.1.5.2 and 14.1.5.3 we apply the density operator to pure and mixed
states of a two-level system. In Exc. 14.1.5.4 we study thermal mixtures and Exc. 15.1.5.2
Glauber states.
That is, after the measurement, the operator becomes diagonal on the basis of the
eigenvalues of  4 , as explained in Sec. 2.2.7. The expression can be thought of as
the mathematical formulation of von Neumann’s state reduction postulate.
3 In the presence of degeneracy or a continuous part of the spectrum we can generalize the definition
of the expectation, X
hX̂i ≡ Tr ρ̂X̂ = hkm|ρ̂X̂|kmi .
k,m
4A projective measure always increases entropy. The entropy of a pure state is zero, while that
of a mixture is always greater than zero. Therefore, a pure state can be converted into a mixture
by a measurement, but the reverse can not happen. Thus, the action of measuring induces an
irreversible change in the density matrix reminiscent of the collapse of the wavefunction. Strangely,
the measurement reduces the amount of information by quenching the quantum interference of the
compound system in a process called quantum decoherence. A subsystem can be taken from a mixed
state to a pure state only at the price of increasing the von Neumann entropy elsewhere in the global
system.
14.1. DENSITY MATRIX 437
We note that the density operator (14.29) describes the whole ensemble after
the measurement. The sub-ensemble corresponding to a particular result an of the
measurement is described by a different density operator,
P̂n ρ̂P̂n
ρ̂0n = . (14.30)
Tr [ρ̂P̂n ]
This is true, when |an i is the only eigenvector with the eigenvalue an . If not, P̂n in the
expression (14.30) should be replaced by the projection operator onto the sub-space of
an 5 . In Exc. 15.1.5.3 we study the projection of Glauber states and in Exc. 14.1.5.7
of entangled states.
(which we may imagine as the expectation value), we can state the following: If f satisfies some
natural properties (such as the one to produce positive values for positive operators), then there
exists a unique density matrix ρ̂, such that f (Â) = Tr (ρ̂Â) for all Â. That is, every reasonable
’family’ of expectation values’ can be represented by a density matrix, which suggests that the
density matrix provides the most general description of a quantum state.
438 CHAPTER 14. THE BLOCH EQUATIONS
Returning to the density operator definition (14.1), we can express its temporal
dependence in terms of time-dependent quantum states and of the time evolution
operator (2.131),
X X
ρ̂(t) = pk |ψk (t)ihψk (t)| = pk U (t, t0 )|ψk (t0 )ihψk (t0 )|U † (t, t0 ) . (14.31)
k k
Writing, X
ρ̂(t0 ) = pk |ψk (t0 )ihψk (t0 )| , (14.32)
k
we see immediately,
ρ̂(t) = U (t, t0 )ρ̂(t0 )U † (t, t0 ) , (14.33)
where, for the common case of a time-independent Hamiltonian,
Now we find the time derivative of the density operator differentiating the two sides
of (14.33) and substituting the Eqs.
dU 1 dU † 1
= ĤU and = − U † Ĥ (14.35)
dt ı~ dt ı~
for the time derivatives U and U † . The result is
dρ̂(t) ı
= [ρ̂(t), Ĥ] . (14.36)
dt ~
The commutator itself can be considered as a superoperator acting, not any more on
states but on operators, that is, we can write,
ı
Lρ̂(t) ≡ [ρ̂(t), Ĥ] , (14.37)
~
where L is called Liouville operator. The equation (14.36) is called Liouville equation
or von Neumann equation. The Liouville equation describes the time evolution of the
density operator which, in turn, describes the distribution of an ensemble of quantum
states. Even though the form of the Liouville equation resembles the Heisenberg
equation, Eq. (14.31) shows that ρ̂(t) is in the Schrödinger picture.
For a two-level system perturbatively interacting with a light field, the Hamilto-
nian can be decomposed as in (2.138) into a stationary part and a time-dependent
part,
Ĥ = Ĥatom + V̂ (t) , (14.38)
where Ĥa is the part of the Hamiltonian describing the atomic structure and V̂ (t)
the interaction of the dipole transition with the classical oscillating electric field. The
interaction picture is the natural choice for this type of problem. In this case, we can
transform the density operator into the interaction picture defined by (2.139),
where the ’tilde’ decoration (replacing the ’hat’) emphasizes, that we are now in the
interaction picture. We look for the time evolution rate of ρ̃(t) analogously to the
Liouville equation. Calculating the time derivatives on both sides of (14.39) and
substituting Eq. (14.36) for ddtρ̂ results in,
dρ̃(t) ı
= [ρ̃(t), V̂ (t)] . (14.40)
dt ~
This equation shows that the time evolution of the density operator in the interaction
picture depends only on the time-dependent part of the total Hamiltonian.
We always can write the Rabi frequency as Ω = |Ω|eıφ and attribute the phase to
the atomic position if necessary. Locating the atom in the center of the coordinate
system, we finally get,
0 2Ω
~
H= ~ , (14.50)
2 Ω −~∆
dρ(t) ı
= [ρ(t), H] . (14.51)
dt ~
The matrix representation of the Hamiltonian given in this section are given in
the basis of the unperturbed states, but we still need to derive the matrix form of the
Liouville equations (14.36), (14.40), and (14.40) in the various pictures. This will be
the topic of the next section.
14.1.5 Exercises
14.1.5.1 Ex: Trace of an operator
P
The trace of an operator  is defined by Tr  = n hn|Â|ni.
a. Show that the trace is independent of the chosen basis!
b. Show that Tr ÂB̂ = Tr B̂ Â!
14.1. DENSITY MATRIX 441
Calculate ρ̂, ρ̂2 and hĤi for the following two cases:
a. The atom is in a superposition state, |ψi = α|1i + β|2i and
b. The atom is a statistical mixture of eigenstates, ρ̂ = µ|1ih1| + ν|2ih2|.
b. Assume that the spin |ψb i is measured and verify whether the new density operator
describing our knowledge of the system is pure.
c. Now consider the entangled state |ψe i = c11 | ↑i| →i + c22 | ↓i| ←i. Write down
again the density operator for the complete system, measure the spin |ψb i, and verify
whether the new density operator describing our knowledge of the system is pure.
dρ̂(t)
hm| |ni = ~ı hm|[ρ̂(t), Ĥ]|ni = ~ı hm|[ρ̂(t), Ĥatom + V̂ (t)]|ni (14.53)
dt
= ~ı (En − Em )hm|ρ̂(t)|ni + ~ı hm|[ρ̂(t), V̂ (t)]|ni ,
where |mi and |ni are members of a complete set of vectors of a basis {|ki} which are
also eigen-kets of ĤA and span the space of Ĥ. Now, we insert the closing expression
14.2. BLOCH EQUATIONS FOR TWO-LEVEL ATOMS 443
P
k |kihk| in the commutator on the right-hand side of Eq. (14.53):
X
hm|[ρ̂(t), V̂ (t)]|ni = [hm|ρ̂(t)|kihk|V̂ |ni − hm|V̂ |kihk|ρ̂(t)|ni] . (14.54)
k
For our two-level atom the complete set only includes two states: |1(t)i = |1i and
|2(t)i = e−ıω0 t |2i. In addition, the matrix elements of the dipole coupling operator V̂
are only non-diagonal,
V ≡ h1|V̂ |2i = h2|V̂ |1i = V ∗ . (14.55)
Hence, Eq. (14.53) adopts the form,
dρ̂11 ı ∗
= ~ [ρ̂12 V − ρ̂21 V ]
dt
dρ̂22 ı dρ̂11
= ~ [ρ̂21 V − ρ̂12 V ∗ ] = −
dt dt
, (14.56)
dρ̂12
= ıω0 ρ̂12 + ~ı [V (ρ̂11 − ρ̂22 )]
dt
dρ̂21 dρ̂∗12
= −ıω0 ρ̂21 + ~ı [V ∗ (ρ̂22 − ρ̂11 )] =
dt dt
remembering that the dash of the diagonal terms, called populations, must be unitary,
and that the non-diagonal terms, called coherences, must be complex,
The above set of equations constitutes the optical Bloch equations in the Schrödinger
picture. It does not include loss terms due to spontaneous emission. We transform the
Bloch equations to the interaction picture by replacing the Liouville equation (14.36)
by (14.40), and calculating the matrix elements. We obtain,
dρ̃22 ı dρ̃12 ı
= (V ρ̃21 − V ∗ ρ̃12 ) and = V (ρ̃11 − ρ̃22 ) . (14.58)
dt ~ dt ~
We would also have obtained this expression via the substitution, ρ̂12 = ρ̃12 eiω0 t .
The interaction picture simplifies the expressions for the temporal dependence of the
coherences by eliminating the first term on the right-hand side. Transforming to the
interaction picture removes the temporal dependence of the basis vectors spanning
the Hilbert space of the two-level atom.
We have derived the optical Bloch equations from the Liouville equation, which is
the fundamental equation of motion of the density operator, and we have seen how a
unitary transformation can be used to represent these equations in the Schrödinger,
Heisenberg or interaction picture. So far, the Bloch equations do not include the pos-
sibility of spontaneous emission. We will learn later, how to include this phenomenon.
rotating wave approximation (RWA) is good, when the Rabi frequency is sufficiently
small, Ω ω. Otherwise, we observe an energy correction of the levels called Bloch-
Siegert shift. The RWA can be implemented in the time dependence of the coupling
operator,
V (t) = ~Ω cos ωt → ~2 Ωe−ıωt , (14.59)
neglecting the part 21 ~Ωeıωt .
Once the RWA made, we can transform to the rotating system by the prescription,
which, applied to the Bloch equations in the Schrödinger picture Eq. (14.56), yields,
dρ22 ıΩ ıΩ∗ dρ12 ıΩ
= ρ21 − ρ12 , = −ı∆ρ12 + (ρ11 − ρ22 ) . (14.61)
dt 2 2 dt 2
In Exc. 14.2.6.1 we derive the Bloch equations from the equations of motion for the
population amplitudes a1 and a2 .
For arbitrary starting conditions, the solution of these equations is not simple. To
solve the problem we write the equations in a matrix form,
ı ∗
ρ11 0 0 2Ω − 2ı Ω
ρ22 0 0 − 2ı Ω∗ ı
~≡
ρ , M≡ 2Ω , ~˙ = M~
ρ ρ .
ρ12 ıΩ − 2ı Ω −ı∆ 0
2
ρ21 − 2ı Ω∗ ı ∗
2Ω 0 ı∆
(14.62)
To solve this system of differential equations, we calculate the eigenvalues of the
matrix,
det(M − λI4 ) = λ2 (∆2 + |Ω|2 ) + λ4 = 0 (14.63)
λ = 0, ±ıG ,
p
with the generalized Rabi frequency G ≡ ∆2 + |Ω|2 . Therefore, the general solution
is,
(1) (2) (3)
ρ22 (t) = ρ22 + ρ22 eıGt + ρ22 e−ıGt (14.64)
(1) (2) (3)
ρ12 (t) = ρ12 + ρ12 eıGt + ρ12 e−ıGt .
The coefficients follow from the Bloch equations with particular starting conditions.
With a little algebra we get
(1) 2
ρ22 = ρ22 (0) + 2G 1
2 |Ω| (1 − 2ρ22 (0)) − ∆ (Ωρ∗12 (0) + Ω∗ ρ12 (0)) (14.65)
(2)
ρ22 = 4G 1
2 −|Ω|2 (1 − 2ρ22 (0)) + (∆ + G)Ωρ∗12 (0) + (∆ − G)Ω∗ ρ12 (0)
(3)
ρ22 = 4G 1
2 −|Ω|2 (1 − 2ρ22 (0)) + (∆ − G)Ωρ∗12 (0) + (∆ + G)Ω∗ ρ12 (0)
(1)
ρ12 = [∆Ω(1 − 2ρ22 (0)) + Ω (Ωρ∗12 (0) + Ω∗ ρ12 (0))]
1
2G2
(2) Ω ∗
ρ12 = ∆−G
4G2 −Ω(1 − 2ρ22 (0)) + (∆ + G) Ω∗ ρ12 (0) + (∆ − G)ρ12 (0)
(3) Ω ∗
ρ12 = ∆+G
4G2 −Ω(1 − 2ρ22 (0)) + (∆ − G) Ω∗ ρ12 (0) + (∆ + G)ρ12 (0) .
14.2. BLOCH EQUATIONS FOR TWO-LEVEL ATOMS 445
such that,
And comparing with the solutions of the Schrödinger equation for a coupled two-level
atom 2.4.7.1 and (6.93), we verify,
2
−ıt∆/2 Gt 2
ρ22 = ıΩ
G e sin 2 = |a2 | (14.70)
ρ12 = −e−ıt∆/2 cos Gt ∆
2 − ı G sin 2
Gt −ıΩ ıt∆/2
G e sin Gt ∗
2 = a1 a2 .
Many times we will restrict ourselves to atoms of two or three levels. For a two-level
system we obtain the Pauli spin matrix defined in (2.45). Every 2 × 2 matrix can be
446 CHAPTER 14. THE BLOCH EQUATIONS
the length of which is simply the generalized Rabi frequency. Note that here we allow
for the possibility of complex Rabi frequency.
Now, using σ̂z = [σ̂ − , σ̂ + ], we may write the Hamiltonian of an unperturbed two-
level system,
The Bloch equations (14.76) then follow as the Heisenberg equation with the Hamil-
tonian (14.78) using [σ̂k , σ̂m ] = 2ıklm σ̂m ,
ˆ
d~σ ˆ ] = ı [G · ~σ
= ~ı [Ĥ, ~σ ˆ, ~σ
ˆ] (14.78)
dt 2
Gy σ̂z − Gz σ̂y
= ıGx ˆ] ıGy ˆ] ıGz ˆ] ˆ.
= Gz σ̂x − Gx σ̂z = G × ~σ
2 [σx , ~
σ + 2 [σ̂y , ~
σ + 2 [σ̂z , ~
σ
Gx σ̂y − Gy σ̂x
As will be shown in Exc. 14.2.6.4, these equations are identical to the Liouville equa-
tions (14.62). ρ12 describes the polarization and ρ22 − ρ11 the population inversion
of the atom. The equation is analogous to the equation of motion for a rigid rotor or
spinning top (for example, a dipole in a homogeneous field). It displays phenomena
such as precession and nutation. The physical content and usefulness of the Bloch
vector will become clearer when we use the formalism to analyze electric and magnetic
couplings. In Exc. 14.2.6.5 we verify that the Bloch vector is normalized (as long as
spontaneous emission is not considered).
that is,
0
~(t) = sin Ωt .
% (14.81)
− cos Ωt
That is, a resonant pulse rotates a Bloch vector initially pointing in the direction
π
−z within the plane z-y, until it arrives, at time t = 2Ω , at the +y direction and
π
at time t = Ω at the +z direction. This means that the entire population has been
transferred to the excited state. The Bloch vector continues to rotate (the movement
is called nutation) around the torque vector G which, as can be seen from Eq. (14.76),
points at the +x direction when ∆ = 0. The nutation frequency is proportional to the
force Ω of the atom-field interaction. With the Eq. (14.69) we see that the population
oscillates between the ground and excited state with the frequency Ω. This means
that the energy ~ω is periodically exchanged between the atom and the field. A pulse
of resonant light of duration such that τ = π/2Ω is called a π/2-pulse. The nutation
is illustrated in Fig. 14.1(a).
Once the coherence has been excited by a detuned radiation, ∆ 6= 0, the Bloch
vector does not stand still, even after the radiation has been switched off. To see this,
we consider again the general solution (14.65) now entering Ω = 0. If the Bloch vector
is initially at a point in the unitary circle of the plane z-y, it will rotate according to
the formula,
ρ22 (t) = ρ22 (0) , ρ12 (t) = ρ12 (0)e−ı∆t , (14.82)
that is,
ρ12 (0) sin ∆t
~(t) = ρ12 (0) cos ∆t .
% (14.83)
2ρ22 (0) − 1
That is, the Bloch vector performs a motion of precession around the symmetry axis.
The precession is illustrated in Fig. 14.1(b).
The evolution of the Bloch vector on the surface of the Bloch sphere under the in-
fluence of radiation fields can be considered a coherent trajectory of the wavefunction
448 CHAPTER 14. THE BLOCH EQUATIONS
(a) 1 (b) 1
2ρ22 − 1
2ρ22 − 1
0 0
-1 -1
1 1
0 0
1 1
Re ρ12 -1 -1 0 Re ρ12 -1 -1 0
Im ρ12 Im ρ12
Figure 14.1: (code) (a) Nutation of the Bloch vector. The red circles show the evolution of
the Bloch vector on the Bloch sphere for a resonant π-pulse. (b) Precession of the Bloch
vector.
of the atomic state, which is therefore subject to interference phenomena [360]. Inter-
ferometers can be realized by sequences of consecutive pulses splitting populations,
exciting coherences, and remixing populations.
Sensors based on interferometry of atomic excitation are nowadays among the most
accurate and most sensitive. We will discuss the method of radiation pulse sequences
in several exercises: In the Excs. 14.2.6.6, 14.2.6.7, and 14.2.6.8 the Ramsey method.
where, because of |Brf (t)| B0 , we choose the z-axis as the quantization axis and
the rf-field as a perturbation along the x-axis.
The energy of a nuclear magnetic dipole moment placed in a magnetic field is
[205],
Ĥ = −µ~ˆ · B
~ = −γI Î · B
~ , (14.85)
where γI is the gyromagnetic ratio of the specific nucleus and Î the nuclear spin
satisfying the usual commutation relation [Iˆm , Iˆn ] = ı~kmn Iˆk . Hence, as shown in
Exc. 14.2.6.10, we can derive from the Heisenberg equation,
~ˆ
dµ
ı~ ~ˆ, Ĥ] ,
= [µ (14.86)
dt
the Bloch equations,
~ˆ
dµ
~ˆ × B
= γI µ ~ . (14.87)
dt
Now, let us restrict to a two-level system, I = 12 such that Iˆz = ~2 σ̂z . In thermal
equilibrium, without time-dependent perturbation,
~ˆ · B
Ĥ0 = −µ ~0 = −µ̂z B0 = −γI Iˆz B0 ≡ −ωL Iˆz . (14.88)
with m = ± 21 . The energy difference between the two states, ∆E = γI ~B0 , results in
a small population bias favoring the lower energy state in thermal equilibrium and,
450 CHAPTER 14. THE BLOCH EQUATIONS
e−Ĥ0 /kB T
1
e− 2 β~ωL σ̂z X e−β~ωL m
ρ̂ = = = |I, mi hI, m| , (14.90)
Z Z Z
m=±1/2
Now, as shown in (14.39), we can express the time-evolution of the density operator
as,
ˆ ˆ
ρ̂(t) = e−ıĤ0 t ρ̂(0)eıĤ0 t = e−ıβ~ωL Iz t ρ̂(0)eıβ~ωL Iz t . (14.92)
Then, as shown in Exc. 14.2.6.11,
cos ωL t − sin ωL t 0
hÎ(t)i = sin ωL t cos ωL t 0 hÎ(0)i , (14.93)
0 0 1
which corroborates the result (14.87). Apparently, the nuclear magnetic dipole mo-
ment (and hence the spin magnetization) precesses around the magnetic field with
the Larmor frequency leaving the populations of the energy levels unaffected.
A perturbation of nuclear spin orientations from equilibrium will occur when an
oscillating magnetic field is applied whose frequency ωrf sufficiently closely matches
the Larmor precession frequency ωL . The populations of the spin-up and -down
energy levels then undergo Rabi oscillations. The stronger the oscillating field, the
faster the Rabi oscillations or the precession around the effective field in the rotating
frame. After a certain time (typically on the order of 2..1000 µs), a resonant rf-π/2-
pulse flips the spin magnetization to the transverse plane, while after a twice longer
time, the initial magnetization is inverted (π-pulse). It is the transverse magnetization
generated by a resonant oscillating field which is usually detected in NMR.
The most important perturbation of the NMR frequency for applications of NMR
is the ’shielding’ effect of the surrounding shells of electrons. Electrons, similar to
the nucleus, are also charged and rotate with a spin to produce a magnetic field
opposite to the applied magnetic field. In general, this electronic shielding reduces
the magnetic field at the nucleus. The corresponding shift in the NMR frequency due
to the electronic molecular orbital coupling to the external magnetic field is called
chemical shift, and it explains why NMR is able to probe the chemical structure of
molecules, which depends on the electron density distribution in the corresponding
molecular orbitals.
6 Note that at high temperatures we get, Z ' 1 and may approximate,
~ˆ
dµ ~ − µ̂z êz − µ̂x êx .
~ˆ × B
= γI µ (14.94)
dt T1 T2
T1 = Γ−1 is the time constant for ’longitudinal magnetic’ relaxation and refers to
the mean time for an individual nucleus to return to its thermal equilibrium state of
the spins. The precessing nuclei can also fall out of alignment with each other and
gradually stop producing a signal. This is called T2 or transverse relaxation, where
T2 = 2γ −1 . Because of the difference in the actual relaxation mechanisms involved (for
example, intermolecular versus intramolecular magnetic dipole-dipole interactions),
T1 is usually longer than T2 , which, in practice, also depends on significant static
magnetic field inhomogeneities.
14.2.6 Exercises
14.2.6.1 Ex: Derivation of Bloch equations
Derive the Bloch equations explicitly based on the temporal evolutions of the coeffi-
cients a1,2 (6.63) knowing that ρij = a∗i aj .
Verify k~
ρk = 1.
Many atomic clocks work according to the Ramsey spectroscopy method: The two-
level atom is resonantly excited by a microwave π/2-pulse. Then, the phase of atomic
coherence precesses freely over a period of time T accumulating an angle φ. Finally,
a second π/2-pulse is applied and the population of the upper-level is measured.
Calculate this population as a function of the angle φ. Neglect spontaneous emission.
452 CHAPTER 14. THE BLOCH EQUATIONS
b. Repeat the calculation of (a), now numerically solving the Bloch equations, which
allow the occurrence of spontaneous emission (Γ12 = 0.03Ω12 ). Interpret the results.
(a) 1 (b) 1
2ρ22 − 1
0.5 0
ρ22
0 -1
0 -1
5 0
5 1
1 -1 0
0
Δ12 /Γ12 Γ12 t Im ρ12
Re ρ12
Figure 14.2: (code) (a) Rabi oscillations damped by spontaneous emission for Rabi frequen-
cies between Ω/Γ = 0.2, .., 5. (b) Evolution of the Bloch vector subject to spontaneous
emission (Γ12 = 0.05Ω12 ) after application of a resonant π-pulse (red) and after a π-pulse
with detuning ∆12 = Ω12 /2 (green).
Ω2
−3Γt/4 3Γ
ρ22 (t) = 1 − e cos λt + sin λt and ρ12 (t) = 0 ,
2|Ω|2 + Γ2 4λ
√ (14.104)
where λ ≡ Ω2 − Γ2 . This solution (which will be derived in Exc. 14.3.5.3),
describes the optical nutation of the Bloch vector along the ρz axis. We note
here that, due to spontaneous emission, the norm of the Bloch vector is NOT
conserved, i.e., the Bloch vector evolves to the interior of the Bloch sphere.
Another case that can be solved analytically is the weakly excited atom, |Ω| γ,
Ω2
ρ22 ' 2
1 + e−2Γt − 2 cos ∆t (14.105)
4G
−ıΩ
and ρ12 (t) ' e−(ı∆+Γ)t − 1 + ρ12 (0)e−(ı∆+Γ)t .
2ı∆ + 2Γ
~Ω
χ = −n ρ12 eı∆t + c.c. (14.107)
~
2|E| 2
We can insert the new expression (14.99) for ρ12 into our previous expression for
~
hd12 i (13.113) and get new expressions for the polarization P(t), (13.110) and (13.120),
and the susceptibility χ (13.122). The modified expression for the susceptibility is,
N d212 −∆ + 21 Γ
χ= . (14.110)
30 ~V ∆2 + 12 |Ω|2 + 14 Γ2
The important new property is the effective width Γeff , which appears in χ, K, and
σ0a . In Exc. 14.3.5.4 and 14.3.5.5 we calculate the impact of the spontaneous emission
on the determinant of the density operator. Solve the Excs. 14.3.5.6, 14.3.5.7, 14.3.5.8,
and 14.3.5.9.
1 2
ρ22 (∞) ' Ω2 − Ω4 + ... (14.113)
4∆2 +Γ 2 (4∆ + Γ2 )2
2
2∆ − ıΓ 2∆ − ıΓ
ρ12 (∞) ' eı∆t Ω − 2eı∆t Ω3 + ... .
4∆2 + Γ2 (4∆2 + Γ2 )2
The theory of nonlinear optics will be developed in Sec. 15.5.2 and Chp. ??.
14.3. BLOCH EQUATIONS WITH SPONTANEOUS EMISSION 457
˙
ρ̂(t) = (L0 + Lsp )ρ̂(t) with
ı ,
− +
L0 ρ̂(t) ≡ [ρ̂(t), Ĥ] and Lsp = Γ
2 (2σ̂ ρ̂ σ̂ − σ̂ + σ̂ − ρ̂ − ρ̂σ̂ + σ̂ − )
~
(14.114)
where σ̂ ± are the Pauli matrices. We show in Exc. 14.3.5.10, that the known Bloch
equations can be derived from the master equation.
q 2
1
2 ∆ − 2ı Γ ± 12 ∆ − 2ı Γ + Ω2 . (14.116)
The real parts of the eigenvalues Re E describe shifts and/or splittings of the transi-
tion line. The imaginary parts Im E describe broadening effects of the lines.
In the simplest case, ∆ = 0 and Γ > 4Ω, we find the saturation broadening already
discussed in (14.100), and we will deepen it in Exc. 14.3.5.11. For the case Γ < 4Ω,
we observe a splitting of the line called Autler-Townes splitting, which will be studied
in Exc. 14.3.5.12. If Ω 6= 0, the spectrum becomes asymmetrical. In the case of
weak excitation, Γ 4Ω, we observe a shift of the transition line with dispersive
dependence (near the resonance) on the frequency of the incident radiation. This is
the dynamic Stark shift (or light shift). In the case of strong excitation, Γ 4Ω, we
observe again at the split spectrum, but now the two lines exhibit an avoided crossing-
type dependence on the radiation frequency. We study these effects in Exc. 14.3.5.14.
Obviously, these effects can be studied by the Bloch equation formalism containing
the terms of spontaneous relaxation.
14.3.5 Exercises
14.3.5.1 Ex: Stationary solution of the Bloch equations
Derive the stationary solution of the Bloch equations including spontaneous emission.
458 CHAPTER 14. THE BLOCH EQUATIONS
Show that the general form of the master equation: ρ̂˙ = − ~ı [Ĥ, ρ̂]− Γ2 (2σ̂ ρ̂σ̂ + − σ̂ + σ̂ ρ̂−
ρ̂σ̂ + σ̂), reproduces the Bloch equations including spontaneous emission.
In this exercise we study saturation broadening effect in a three-level system |1i, |2i,
and |3i in V -configuration, as shown in Fig. 14.6(b), excited by two resonant lasers
with the Rabi frequencies Ω12 and Ω23 .
a. From the eigenvalues E1,2 of the effective Hamiltonian (14.115) of the system,
describe the behavior of the real part (energy shift) and the imaginary part (linewidth)
as a function of the Rabi frequency. Prepare diagrams Ω12 versus Re E1,2 and versus
Im E1,2 and discuss the limits Ω12 > 12 Γ12 and Ω12 < 21 Γ12 .
Saturation broadening can be measured experimentally in a three-level system in V -
configuration. To reproduce the experiment by numerical simulations of the Bloch
equations (14.164),
b. write down the Liouville matrix L of the system and calculate the time evolution of
the Bloch vector via equation (14.168) varying the Rabi frequency Ω23 . Choosing the
parameters Γ23 = Γ12 , Γ13 = 0.001Γ12 , Ω12 = 0.2Γ12 , and ∆12 = 0 = ∆23 , prepare a
3D curve (similar to Fig. 14.2(a)) of the population ρ33 (t).
c. Interpret the results in terms of broadening by saturation. The broadening can
also be understood in terms of the quantum Zeno effect, where the transition |1i-
|2i plays the role of the ’observed system’ and the transition |2i-|3i the role of the
measuring device or ’meter’ (for example, we can observe the light scattered on the
’meter transition’ to infer the evolution of the ’system transition’).
In this exercise we study the Autler-Townes effect in a two-level system |1i and |2i
resonantly excited (∆12 = 0) by a laser with the Rabi frequency Ω12 :
a. From the eigenvalues E1,2 of the effective Hamiltonian (14.115) of the system,
describe the behavior of the real part (energy shift) and the imaginary part (linewidth)
as a function of the Rabi frequency. Prepare diagrams Ω12 versus Re E1,2 and versus
Im E1,2 and discuss the limits Ω12 > 21 Γ12 and Ω12 < 21 Γ12 .
The Autler-Townes effect can be measured experimentally by probing the population
of level |2i via excitation of a third (higher) level by a second laser with the Rabi
frequency Ω23 . Thus, we obtain a three-level system in cascade configuration, as
shown in Fig. 14.6(c). In order to reproduce the experiment by numerical simulations
of the Bloch equations (14.164),
b. write down the Liouville matrix Mred reduced by the trace condition (14.169) and
c. compute the stationary Bloch vector from equation (14.171) varying the detuning
of the probe laser ∆23 and the Rabi frequency Ω12 of the system under study (|1i and
|2i). Choosing the parameters Γ23 = 0.5Γ12 , Γ13 = 0.01Γ12 , Ω23 = 0.1Γ12 , prepare a
3D curve (similar to Fig. 14.2(a)) of the stationary population ρ22 (∞). Interpret the
results.
460 CHAPTER 14. THE BLOCH EQUATIONS
the displacements generate an effective broadening usually with a Gaussian line pro-
file. It can not be included in the Bloch equations, but only as an average over all
trajectories of all atoms. It does not correspond to an accelerated relaxation. Inhomo-
geneous broadening is often due to external perturbations, e.g., Doppler broadening
and broadening due to temporal fluctuations or spatial inhomogeneities of external
electric or magnetic fields. In Exc. 14.4.6.1 we calculate the optical density of atomic
clouds. In Exc. 14.4.6.2 we present a spectroscopic technique bypassing the Doppler
broadening called Doppler-free spectroscopy and calculate the Lamb-dip profile. Fi-
nally, in Exc. 14.4.6.3, we discuss a cooling technique allowing for the reduction of
Doppler broadening, called Zeeman slower.
Ω= √1 Γ . (14.117)
2
We can use equation (14.117) to define the saturation intensity Isat for an atom with
the transition dipole d12 . From Eq. (1.41) we have,
s
2I¯
E0 = . (14.118)
0 c
Therefore, using the definition of the Rabi frequency, ~Ω = d12 E0 , and the relationship
between d12 and Γ given by Eq. (13.38), we have 7 ,
g1 2π 2 c~
Isat = Γ , (14.119)
g2 3λ30
taking into account the degeneracies gj of the levels. In Excs. 14.4.6.4 and 14.4.6.5
we calculate the saturation intensity of popular atomic transitions.
along which the interaction time is short and the time between collisions is long in
comparison with the radiative lifetime of the excited atomic state. Then, the impact
of a collision on the emission of a radiating atom causes a loss of coherence due to a
phase interruption of the excited state atomic wavefunction. The term ’elastic’ means
that the collision does not disturb the populations of the internal states, so we only
need to consider the off-diagonal elements of the density matrix,
dρ12 Ω0
= ı eı(ω−ω0 )t (ρ11 − ρ22 ) − γ 0 ρ12 , (14.120)
dt 2
where γ 0 is the sum of the spontaneous emission γ and the collision rate γcol ,
γ 0 = γ + γcol . (14.121)
The inverse of the collision rate is simply the time between phase interruptions or the
time between collisions. Now, for collisions between hard cores of atoms of mass m
(with reduced mass mred = m/2) and with radius ρ in a gas with density n consisting
of a single species, a standard analysis based on the kinetic theory of dilute gases
shows that the time between collisions is given by the collision rate,
−1
γcol = τcol = σnv̄ , (14.122)
q
8kB T
where v̄ = is the average collision velocity in a homogeneous gas at the
πmred
√
temperature T and σ = 8πρ2 the collision cross section. Thereby,
8ρ2 n
γcol = p . (14.123)
mred /πkB T
We can now relate this simple result of gas kinetics to the phase interruption
rate by reinterpreting the meaning of the collision radius. When an excited atom
propagating through space suffers a collision, the long-range interaction will produce
a time-dependent perturbation of the energy levels of the radiating atom and a phase
shift in the radiation,
Z ∞ Z ∞
η= [ω(t) − ω0 ]dt = ∆ω(t)dt . (14.124)
−∞ −∞
The integral is easily assessed for the two most frequent cases: non-resonant van der
Waals interactions n = 6 and resonant van der Waals interactions n = 3. The phase
shifts are,
2π C6 4π C3
η6 (b) = and η3 (b) = √ . (14.127)
3~ b56 v 3 3~ b23 v
14.4. LINE BROADENING MECHANISMS 463
When the excitation is sufficiently weak, so that power broadening can be neglected
in comparison to collision broadening, the second term can be discarded,
The equations (14.100) and (14.132) express the linewidths in the limits of dominating
power and collision broadening, respectively. Note that the susceptibility, absorption
coefficient, and absorption cross-section retain their Lorentzian profile, but with a
larger width due to collisions. Since each atom is subject to the same broadening
mechanism, the broadening is homogeneous. Solve Excs. 14.4.6.6.
The Liouville equation (14.51) used to derive the Bloch equations assumes im-
mobile atoms. However, we can easily apply the Galilei transformation to a system,
where the atoms move with the given velocity v,
(∂t + v · ∇)ρ̂(r, t) = − ~ı [Ĥ, ρ̂(r, t)] . (14.138)
ı(ωt−k·r)
Since the light fields propagate as e , the solution of the above equation simply
follows from the immobile solution with the substitution ∆ → ∆ − k · v. For a cloud
2
obeying Maxwell’s velocity distribution, P (v) ∼ e−mv kB T ,
Z
1 2 2
ρ̄(∆) = √ e−(k·v) /2δ ρ̂(∆ − k · v)d(k · v) . (14.139)
2πδ R
The average of the density operator over all velocities, ρ̄, therefore follows as the
convolution of the density operator ρ (obtained as the solution of the Bloch equation)
2 2
and the Gaussian function G(∆) = (2πδ 2 )−1/2 e−∆ /2δ ,
This integral has no analytical solution, but it is easy to solve numerically. Resolve
Excs. 14.4.6.7 and 14.4.6.8.
where the matrix X contains the modulation index of the atomic motion kij v. The
stationary solution of the differential equation, averaged over the time of an oscillation
period, can be expressed as an infinite continuous fraction:
This solution replicates the correct excitation spectra even for a multilevel system.
Let us be more specific for a two-level system. Its Hamiltonian is, Ĥint =
1
2 ~Ωe −i[ωt−k·v/Ωa sin Ωa t]
, such that the Bloch equation is,
ρ22 −Γ − 2ı Ω ı
2Ω 0 0 0 ρ22 0
d ı
ρ12 = −2Ω −Λ 0 + 2 cos Ωa t 0 0 ρ12 + 2ı Ω .
− 2ı kv
dt ı ı
ρ21 2Ω 0 −Λ∗ 0 2 kv 0 ρ21 − 2ı Ω
P∞ (14.145)
−ınΩa t
We look for the stationary solution by expanding ρ̂ = n=−∞ ρ̂ n e , letting
ρ̂˙ n = 0, and projecting on e−ınΩa t via,
0.2
0.15
ρ22
0.1
0.05
0
-10 -5 0 5 10
Δ12 /Γ12
Figure 14.3: (code) Spectral broadening due to the periodic movement of the atom.
14.4.6 Exercises
14.4.6.1 Ex: Optical density of a cold cloud
The cross section of an atom with the resonant frequency ω0 moving with velocity v
and irradiated by a laser beam of frequency ω is,
6π Γ2
σ(v) = .
k 4(ω − ω0 − kv)2 + Γ2
2
a. Calculate the absorption profile of the resonance line at 461 nm (Γ461 = (2π) 30.5 MHz)
of a strontium gas cooled to the Doppler limit (kB TD = ~Γ) of this transition.
b. Calculate the absorption profile of the resonance line at 689 nm (Γ689 = (2π) 7.6 kHz)
of a strontium gas cooled to the Doppler limit of the transition at 461 nm.
c. Compare the optical densities in case of resonance.
Help: To evaluate the convolution integral approximate the narrower distribution
by a δ-function maintaining the integral over the distribution normalized.
below the saturation limit, such that the cross section of an atom moving at velocity
v is,
6π Γ2
σ(v) = 2 .
k 4(ω − ω0 − kv)2 + Γ2
The saturation laser has high intensity. We suppose here, Ω ≡ 10Γ, where Ω is the
frequency of Rabi caused by the saturation beam. In this way it creates a population
Ne of atoms in the excited state. As this population lacks in the ground state,
Ng = N − Ne , the absorption of the probe beam is decreased by the factor,
Ne Ω2
= .
N 4(ω − ω0 + kv)2 + 2Ω2 + Γ2
To obtain theR laser probe transmission spectrum, first calculate the optical density,
∞ N −N
OD(ω) = Ln −∞ gN e σ(v)ρ(v)dv where n is the gas density, and then the intensity
of light transmitted through the cell using the Lambert-Beer law II0 = e−OD .
W0 Γ2
W (v) = ,
2π (ω − ω0 + kv)2 + (Γ/2)2
where Γ is the natural width of the spectral line at ω0 , and W0 is a constant. The
frequency of the light is tuned in order to compensate for the Doppler effect at the
beginning of the tube, δ = ω − ω0 = −kv0 (the light is tuned to the red of the
resonance). As the atoms are decelerated, they cease to be resonant with the light
beam and fail to absorb photons. This can be avoided by employing the so-called
Zeeman-slowing technique, which compensates for the effect using the Zeeman-shift
induced by magnetic fields. In this exercise, we will study what happens if this
technique is not used.
a. For an atom with velocity v, write an expression for the mean travel distance
Calculate the average distance required for the atoms to be slowed down to v = 0
(ignoring the Doppler limit). Write the expression as a function of Γ, v0 , k, and W0 .
Help: Do the following change of variables to simplify the evaluation of the integral:
n → v.
c. Typically, the detuning of the light, |δ| = kv0 , is much larger than the natural
width Γ of the transition. What happens to S in the limit when kv0 Γ? Interpret
this result, justifying the need for the Zeeman-slowing technique.
†
Here, σ̂ij ≡ |iihj| = σ̂ji . The constants Rij are eventual incoherent pump rates due
to optical pumping, βij take account of homogeneous broadening, e.g. due to finite
laser linewidths. The levels have the energy ~ωi above the ground level. The Rabi
frequency Ωij is a measure for the force at which the levels |ii and |ji are coupled
by the resonantly irradiated light field. The master equation can be simplified by
applying the rotating wave approximation and transforming to the coordinate system
which rotates with the light frequencies ωij :
ρij → ρ̂ij eıωij t , Ĥatom:f ield → e−ıĤt/~ Ĥatom:f ield eıĤt/~ . (14.150)
8 That is, the atom is quantized and consists of several levels |ii with energies ~ω , while the light
i
fields are described by factors eıωij t , with frequencies ωij tuned near the transitions |ii − |ji.
470 CHAPTER 14. THE BLOCH EQUATIONS
ρ̂˙ = Lρ̂,
ρ̂ = eLatom t ρ̂0 .
P
The
P relation with the von Neumann equation with Ĥ = i,j |iiHij hj| and ρ̂ =
k,l |kiρkl hl| and σ̂ij = |iihj| is:
X X
Latom ρ̂ = − ~ı [Ĥ, ρ̂] = −ı Hkl ρlj |kihj| + ı Hlj ρkl |kihj| . (14.152)
k,l,j k,l,j
For example, for the two-level system with the definition of the external product
(2.113):
Latom ρ̂ = −ıĤ ⊗ Iρ̂ + ıI ⊗ Ĥ ρ̂ . (14.153)
The relaxation terms are,
X
Ldecay ρ̂ = γij (2σ̂ji ρ̂σ̂ij − σ̂ij σ̂ji ρ̂ − ρ̂σ̂ij σ̂ji ) (14.154)
i,j
X
= (2γji δkj ρii − γij ρkj − γik ρkj ) |kihj| .
i,j,k
Example 63 (Liouville equation for two levels): For example, for the two-
level system,
2γ11 0 0 2γ21 2(γ11 + γ12 ) 0 0 0
P
0 0 0 0 0 (kj) γ(kj) 0 0
Latom = − .
P
0 0 0 0 0 0 (kj) γ (kj) 0
2γ12 0 0 2γ22 0 0 0 2(γ22 + γ21 )
So all in all,
" #
X X
ρ̇km = 2 γmj ρjj δkm − (γik + γim ) + ı(ωk − ωm ) ρkm
j i
X , (14.162)
+ ı
2 Ωij ρki δjm + Ω∗ij ρkj δim − Ωij ρjm δki − Ω∗ij ρim δkj
i<j
where the index j runs over all levels |ji into which a population ρkk can decay and
the index i runs over all levels |ii into which the levels |ki and |mi of a dipole moment
ρkm can decay. In Exc. 14.5.5.1 we will apply this formula to three-level systems.
the coherences between |ii and |ji. Now, we must insert the Hamiltonian (14.149)
and the density operator ρij into the Liouville equation (14.36) in order to derive
the generalized Bloch equations. In practice, these calculations are simple but heavy.
Therefore, we describe in Sec. 14.5.4 a simplified recipe for compiling Bloch equations
for arbitrary level systems for real atoms.
Figure 14.6: Three level system (a) in Λ-configuration, (b) in V -configuration, and (c) in
cascade configuration.
Defining the Bloch vector by ρ~, the Bloch equation matrix for three levels in
Raman configuration (that is, in Λ-configuration) using the labeling of Fig. 14.6(a),
is,
~˙ = M~
ρ ρ= (14.164)
ı
0 Γ12 Γ13 Ω
2 12
− 2ı Ω12 0 0 0 0 ρ11
0 0 − 2ı Ω12 ı
Ω 0 0 ı
Ω − 2ı Ω23 ρ22
−Γ12 − Γ23
2 12 2 23
0 Γ23 −Γ13 0 0 0 0 − 2ı Ω23 ı
Ω ρ33
2 23
ı
Ω − 2ı Ω12 0 −Λ12 0 ı
Ω 0 0 0 ρ12
2 12 2 23
− 2ı Ω12 ı
Ω 0 0 −Λ∗12 0 − 2ı Ω23 0 0 ρ21
2 12
0 0 0 ı
Ω 0 −Λ13 0 − 2ı Ω12 0 ρ13
2 23
0 0 0 0 − 2ı Ω23 0 −Λ∗13 0 ı
Ω ρ31
2 12
0 ı
Ω − 2ı Ω23 0 0 − 2ı Ω12 0 −Λ23 0 ρ23
2 23
0 − 2ı Ω23 ı
Ω
2 23
0 0 0 ı
Ω
2 12
0 −Λ∗23 ρ32
The resulting state vector, ρ ~red , has the length n2 − 1, and from M we obtain the
(trace-)reduced, now invertible matrix Mred and the inhomogeneity vector b. The
differential equation is now,
d
~red = Mred ρ
ρ ~red + b , (14.170)
dt
474 CHAPTER 14. THE BLOCH EQUATIONS
and insert the solution obtained again into equations (14.172) with the Hamiltonian
Ĥ(t + ∆t) (or the Liouvillian M(t + ∆t)) adjusted to the new time.
EIT is usually studied in Λ-type systems, but similar phenomena can be found
in cascade-type systems [768, 757], which will be studied here. Disregarding the
decay rate Γ13 , the Bloch equations (14.164) and (14.167) give the coherences,
ıΩ12
ρ̇12 = −Λ12 ρ12 + 2
(ρ11 − ρ22 ) − ıΩ223 ρ13
ρ̇13 = −Λ∗13 ρ13 − ıΩ12
2
ρ23 − ıΩ223 ρ12
ıΩ23
ρ̇23 = −Λ23 ρ23 + 2
(ρ22 − ρ33 ) − iΩ212 ρ13 .
14.5. MULTI-LEVEL SYSTEMS 475
and finally,
ıΩ12 4Λ∗13 Λ23 + Ω212
ρ12 = .
2 Λ12 (4Λ13 Λ23 + Ω212 ) + Ω223 Λ23
∗
Ω2
N |d12 |2 Γ23 + 2Λ12 −2ı∆12
χ= ıΩ12 23
V ε0 ~Ω12 Ω2
Γ23 + 2Λ −2ı∆12 (Γ12 +2ı∆12 )+Ω2
12
23
23
N |d12 |2
= ıΩ12 (Θ−2ı∆ Θ−2ı∆ 12
= χ0 + ıχ00 .
V ε0 ~Ω12
2
12 )(Γ12 +2ı∆12 )+Ω23
×10−3
2 0.5
ρee
1
Re χ
0
0
-20 0 20
0.2
-0.5
-20 0 20
ρmm
0.1
1
0
-20 0 20
1
Im χ
0.5
ρgg
0.9
0.8 0
-20 0 20 -20 0 20
Δge /Γge Δge /Γge
Figure 14.7: (code) EIT signal for the cascade system of strontium with the transitions at
689 nm and 497 nm with Ω12 = Γ12 , Ω23 = Γ23 and ∆23 = 0. The red lines are calculated by
numerical integration of the Bloch equations. The dotted lines are obtained from analytical
formulas based on the assumptions of weak ground state depletion (which is not really correct
in the chosen parameter regime.
the degrees of freedom of the electromagnetic vacuum and using the Markov and the
Born approximations, one arrives at a master equation of the form [268],
ı
ρ̂˙ = − [H, ρ̂] + LΓ ρ̂ + Lβ ρ̂ . (14.173)
~
Here H is the Hamiltonian of coupled atom-laser system. The Liouville superopera-
tors,
X
† †
LΓ ρ̂ = Γkl [σkl , ρσkl ] + [σkl ρ, σkl ] (14.174)
k,l
X
† † † †
Lβ ρ̂ = βkl [σkl σkl , ρ̂σkl σkl ] + [σkl σkl ρ̂, σkl σkl ] ,
k,l
describe dissipation through spontaneous emission from the various excited states and
decoherence due to phase fluctuations of the laser fields, respectively. The operators
σkl = |kihl| describe the transition between two states. Γkl is the spontaneous decay
of level |ki in to |li, and βkl is the emission bandwidth of the laser coupling the two
levels. Simple but tedious algebraic transformations of the master equation lead, in
the rotating wave approximation, to a set of linear first-order differential equations
in the populations of the atomic excitation levels and the coherences between them.
The equations are called the optical Bloch equations.
Alternatively, the Bloch equations may be found by breaking down the multi-level
scheme into a set of three-level systems. Respecting a few symmetry considerations,
14.5. MULTI-LEVEL SYSTEMS 477
the multi-level Bloch equations can then be reassembled from the three-level Bloch
equations corresponding to every possible combination of three levels. Based on such
considerations, we provide in the following a simple recipe for setting up Bloch ma-
trices for arbitrary level schemes. A movie showing a simulation of multi-level Bloch
equations for Li atoms can be watched under the following link: (watch movie).
Let us regard a n-level atom. Its internal state is fully described by the populations
ρkk and the (complex) coherences ρkl , with k, l = 1, .., n. In this work we describe the
coherences by their real and imaginary parts. The labeling is such that the levels are
sorted according to their excitation energy, Ek < El for k < l. We define the Bloch
vector,
~˙ = M~
ρ ρ, (14.176)
where in the given Bloch vector basis the matrix M has the following structure,
(A) (B1 )
M= (C) (D) . (14.177)
(B2 )
(D) (C)
The different blocks of the matrix have the following significations. Block A han-
dles the transfer of populations by spontaneous decay. Its rank corresponds to the
number of levels n. The diagonal elements of this block are the decay rates Γ of the
excited states. The off-diagonal elements Γkl denote the gain of level k from a decay-
ing level l. Conservation of energy thus requires
P that the sum of the transition rates
cancels for every column of matrix A, Γ = k Γkl , as it is the case for the two-level
Bloch matrix. If the levels are sublevels of a Zeeman and/or hyperfine split multiplet,
the rates have to be weighted with Wigner’s (3j) and {6j} symbols, Γkl = ΓSkl . The
relative oscillator strengths Skl are given in Sec. 14.5.4.2.
The blocks Bk treat the interdependence of the populations and the coherences.
B1 describes how the coherence between any pair of states driven by a light field
generating a Rabi frequency Ωkl influences the populations. The block consists of
convoluted 2 × 2 matrices of the form,
ρ̇kk −Ωkl 0 Im ρkl
∼ . (14.178)
ρ̇ll Ωkl 0 Re ρkl
The Rabi frequencies have to be weighted not only with the relative oscillator strength
Skl , but also with the projection Hkl of the laser polarization onto the orientation
478 CHAPTER 14. THE BLOCH EQUATIONS
of the magnetic field and the laser polarization, Ωkl = Ωx Skl Hkl . Here Ωx is the
Rabi frequency generated by a laser on a transition, whose oscillator strength is 1.
The projection is calculated in Sec. 14.5.4.3 for the three possible laser polarizations,
i.e. for σ ± and for π light.
The matrix C rules the influence of the decays of the coherences, of the detunings
∆kl = ωx − ωatom , and the laser linewidths βkl . Note that the detuning of the laser
frequency ωx is negative for red-detuned light. In the chosen basis it breaks down
into an array 2 × 2 matrices aligned along the diagonal of M. Their shape is,
γkl
Im ρ̇kl − 2 −∆kl Im ρ̇kl
∼ . (14.180)
Re ρkl ∆kl − γ2kl Re ρkl
P
where γkl = m,Em <Ek ,El (Γkm + Γlm ) + 2βkl . Often the levels are sublevels of a
Zeeman and/or hyperfine split multiplets. In this case the frequency shift Zkl of the
level is added to the detuning ∆kl . The shift is calculated in Sec. 14.5.4.4 for the
example of the 6 Li D2 line.
The block D governs the interdependences of all laser-driven coherences of the
atom. The block contains 2 × 2 submatrices at any place of the matrix M, where
the row index pair (mn) and the column index pair (kl) have one index in common
provided the two different indices correspond to the Rabi frequency of an incident
laser:,
Im ρ̇mn 0 ± 21 Ωkl Im ρpq
∼ . (14.181)
Re ρ̇mn ± 12 Ωkl 0 Re ρpq
The submatrix elements indexed by column (pq) and row (mn) are non-zero if one
of the indices p or q is equal to one of the indices m or n and the unequal indices
correspond to a laser-driven transition. In order to find the correct signs of the
submatrix elements, we distinguish four cases: 1. For m = p, n = k, and q = l the
signs are: (− + ); 2. for n = q, m = k, and p = l the signs are: (+ − ); 3. for m = q,
n = k, and p = l the signs are: (+ + ); and 4. for n = p, m = k, and q = l the signs
are: (− − ). A proper parametrization is proposed in the next section.
for k ∈ G and l ∈ E,
for k ∈ {1, 2} and f ∈ {3, .., 6}. The Zeeman shift Zkl is given by Eq. (14.196).
Finally, the block D is filled with the components,
where,
where ĝ is an arbitrarily chosen direction, e.g. gravity. The relative amplitude of the
transitions ∆mJ = 0 is proportional to the projection of the polarization vector on
the magnetic field axis ζ0 = (ε̂ · ê3 )2 for π-polarized light. To estimate the amplitude
of the transitions ∆mJ = ±1, we must project onto the coordinates,
With this generalization the Bloch equations can e.g. be employed to calculate
Hanle resonances quantum mechanically. The Hanle effect occurs when a magnetic
and an optical field compete for the quantization axis.
−2.8 MHz, νhf 2 = 0 MHz, and νhf 3 = 1.7 MHz. Hence, the hyperfine splitting is
inferior to the natural decay rate Γ = (2π) 6 MHz,
~
µB |B|
Zkl = 2π~ (gFk mk − gFl ml ) (14.196)
+ νhf 1 δ7≤l≤8 + νhf 2 δ8≤l≤13 + νhf 3 δ13≤l≤16 ,
14.5.5 Exercises
14.5.5.1 Ex: Derivation of three-level Bloch equations
Derive from the general formula (14.162) the three-level Bloch equations for the sys-
Ω
12 Ω
23
tem |1i ←→ |2i ←→ |3i.
possible procedure is the method of optical pumping, from the initial ground state
to an excited state, which subsequently decays to the final state by spontaneous
emission. The problem with this incoherent procedure is, that one can control into
which ground state level the atom will decay, and that it heats atoms due to the
photonic recoil associated with the scattering of light. In this exercise we studied an
alternative method, called Stimulated Raman Adiabatic Passage (STIRAP), which
allows the coherent transfer of population between two states by counter-intuitive
pulse sequences:
a. Consider a three-level system in Λ-configuration, as shown in Fig. 14.6(a), initially
being in the state |1i. Write the system’s Hamiltonian in the interaction picture.
Now, choose ∆12 = 0 = ∆23 , and a temporal variation of the Rabi frequencies
described by Ω12 (t) = Γ12 ( 12 + π1 arctan Γ12 t) and Ω23 (t) = Γ12 ( 12 − π1 arctan Γ12 t).
With this, solve the Schrödinger equation (14.172) iteratively within the time interval
t ∈ [−40/Γ12 , 40/Γ12 ], while continuously adjusting the Rabi frequencies.
b. The dynamics can also be calculated via a numerical simulations of the Bloch
equations (14.164). Write down the Liouville matrix and prepare a simulation using
the same parameters as in (b) and additionally Γ23 = Γ12 /2, Γ13 = Γ12 /500.
c. Interpret the results.
Figure 14.8: Energy levels of an atom in the ground state with Zeeman structure (for
example, |J = 1, mJ = −1, 0, +1i) as a function of the applied magnetic field.
Theoretically, under certain conditions, gases may exhibit negative permittivity and
permeability, and therefore refraction [649]DOI , [647]DOI , [648]DOI , [542]DOI , [433].
To study this phenomenon we consider a three-level system in Λ-configuration with
an electric dipole transition and another magnetic dipole transition. The objective
is to balance the electrical dipole moment excited by a probe laser and the magnetic
dipole moment excited via a Raman transition by both, the probe laser and a control
laser. The Raman transition simulates an effective magnetic field. Since the magnetic
moment is smaller by a factor of α2 , the electric moment must be reduced by detuning
the probe laser, as shown in Fig. 14.9.
Figure 14.9:
R. Loudon, Oxford Science Publications, Oxford (1982), The quantum theory of light
[ISBN]
M.O. Scully, From lasers and masers to phaseonium and phasers [DOI]
I.I. Sobelman, Springer Verlag, Berlin (1972), Introduction to the Theory of Atomic
Spectra [ISBN]
A. Corney, Clarendon Press, Oxford (1977), Atomic and Laser Spectroscopy [ISBN]
Y. Stalgies et al., Light shift and Fano resonances in a single cold ion [DOI]
L.V. Hau et al., Light Speed Reduction to 17 Metres per Second in an Ultracold
Atomic Gas [DOI]
Jian-Qi Shen et al., How to realize a negative refractive index material at the atomic
level in an optical frequency range [DOI]
Jian-Qi Shen et al., Negative permeability in a Λ-type three-level atomic vapor [DOI]
487
488 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
aware that a photon is not a particle, as it simply disappears when performing the
transition from quantum to classical mechanics [434].
where we already isolated the vectorial character due to the polarization ~k of the light
mode k. Obviously, A− + ∗
0k = (A0k ) . As each amplitude and polarization of the wave
given by the vector potential Ak and A∗k must satisfy the wave equation separately,
we arrive at the dispersion relation,
ωk = ck . (15.2)
With the results (13.7) and (13.8) we know that the energy in each radiative mode
containing nk photons is,
where the bar denotes cycle-averaging. The second quantization now consists in
interpreting the mode as a quantum harmonic oscillator, that is, we understand the
observables as operators satisfying commutation rules, such as [Â0k , Â+
0k0 ] ∝ δk,k ,
0
We introduce normalized field operators following the commutation rule (3.74) via,
q q
† −
âk 4ε0 ~V ωk ≡ Â+
0k and âk 4ε0 V ωk ≡ Â0k ,
~
(15.5)
such that,
Ĥk = ~ωk (â†k âk + 12 ) . (15.6)
The analogy allows us to interpret them as creation operator and annihilation operator
of photons. Finally, we can rewrite (15.1) as,
q h i
Âk (r, t) = 4ε0 V ωk ~
~
k âk eı(k·r−ωk t) + â†k e−ı(k·r−ωk t) . (15.7)
We already know such combinations of operators and their complex conjugates from
the quantum harmonic oscillator (3.83).
In the Coulomb gauge, the electric and magnetic field operators for the cavity
modes can be constructed from,
q
ˆ ∂ Âk
E~k = − =ı ~ωk
2ε0 V âk eı(k·r−ωk t) − â†k e−ı(k·r−ωk t) ~k
∂t . (15.8)
q
~ˆk
B = ∇ × Ak = ı 2ε
~ωk
âk e ı(k·r−ωk t)
− ↠−ı(k·r−ωk t)
e k × ~k
0V k
We can calculate the cycle-averaged energy of the k-th cavity mode from a quantum
version of Eq. (15.3), Z
ˆ ˆ
Ēk = ε20 hnk |E~k · E~k |nk idV . (15.9)
The result (15.6) is exactly Planck’s quantum hypothesis (although strictly speak-
ing, he rather suggested a quantization of oscillators in the conducting walls of the
cavity, not of the field) on the distribution of the spectral intensity radiated by a black
body. We now can see that it follows naturally from the quantization of the cavity
field modes. Solve Excs. 15.1.5.1, Exc. 15.1.5.2, and 15.1.5.3.
Now, that we have a clear picture of the quantized field with the energies in the
modes given by Eq. (15.9) and the photon number states given by the eigenstates |ni
of the quantized harmonic oscillator, we are in a position to consider our two-level
atom interacting with this quantized radiation field. If for the moment, we exclude
spontaneous emission and stimulated processes, the Hamiltonian of the combined
atom-field system is,
Ĥatom = ~ωg |gihg| + ~ωe |eihe| = ~ωg |gihg| + ~(ωg + ω0 )|eihe| , (15.12)
where Ĥf ield is the Hamiltonian of the quantized field, expressed by Eq. (15.6), and
Ĥatom:f ield the atom-field interaction. For the Hamiltonian without interaction, Ĥ =
Ĥatom + Ĥf ield , the eigenstates are simply product states of the atomic states and
the photon number states,
The left side of Fig. 15.1 shows, how the eigenenergies of the product states consist
of two ladders, being displaced by the energy difference ~∆, which corresponds to the
490 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
detuning. We write the Hamiltonian of the atom Eq. (15.12) as the sum of projectors
on unperturbed eigenstates using the completeness relation and the orthogonality
of eigenstates. With the same idea we can rewrite the dipole operator defined in
Eq. (13.30),
X X X
d̂ = |ψi ihψi |d̂| |ψj ihψj | = |iihi|eı(ωi −ωj )t d̂|jihj| (15.14)
i j i,j
X X
ı(ωi −ωj )t ı(ωi −ωj )t
= e dij |iihj| = e dij |iihj| + e−ı(ωi −ωj )t dij |jihi| ≡ d̂(+) + d̂(−) .
i,j i<j
using |ψn (t)i = e−ıωn t |ni. Note that d̂ only has non-diagonal elements.
Figure 15.1: (Left) Photons number states and the two stationary states of the two-levels
atom. (Center) Double ladder showing the basis of products states of photon number and
atomic states. (Right) Dressed states constructed by diagonalization of the full Hamiltonian
in the basis of the product states.
Now, let us use the electric field of Eqs. (15.8) to describe the atom-field interaction
through the Hamiltonian Ĥatom:f ield = −d̂ · E, ~ˆ
XXq h
† −ı(k·r−ωk t)
i
ı(ωi −ωj )t ı(k·r−ωk t)
Ĥatom:f ield = ı ~ωk
2ε0 V d ij e |iihj| · ~
k âk e − âk e .
k i,j
(15.15)
For our two-level atom interacting with a single mode radiation field, we only have,
q h i
ı(ωe −ωg )t ı(ωg −ωe )t
~ωk
Ĥatom:f ield = ı 2ε 0V
dge e |gihe| + e |eihg| · (15.16)
h i
·~k âk eı(k·r−ωk t) − â†k e−ı(k·r−ωk t) .
It is important to note that the first and fourth term can be important in higher
order processes, such as multiphotonic absorption or Raman scattering processes,
where the excited state would be a virtual level. In fact, when the Rabi frequency
is very large, Ω1 ' ω, the excitation and deexcitation processes follow each other so
rapidly, that energy conservation can be violated for short times. The energy shift
caused by terms neglected in the RWA are called Bloch-Siegert shift 4 .
3 Remember that the four processes contained in the Hamiltonian are all coherent (absorption and
stimulated emission), and that spontaneous emission must be treated separately see next Sec. 15.4.
4 The shift is not observed, when the non-rotating terms σ ± a± are forbidden by other conservation
or selection rules. For example, when a resonance is excited by σ ± light, the RWA is accurate.
492 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
Figure 15.3: Illustration of the four processes in the atom-field interaction. Terms (b) and
(c) conserve energy in first-order processes, while (a) and (d) do not conserve.
The atom-field product states offer a natural basis for the Hamiltonian of Eq. (15.11).
The states resulting from the diagonalization of the Hamiltonian on this basis are
called dressed states. As indicated in Fig. 15.1, the neighboring doublets the dou-
ble ladder ’repel’ each other under the influence of the interaction Ĥatom:f ield in
Eq. (15.11). The mixed coefficients form the familiar problem of two levels, now
called |ai and |bi. From Fig. 15.1 we see,
|a, N i cos θ sin θ |g, ni
= . (15.21)
|b, N i − sin θ cos θ |e, n − 1i
with
Ω1
tan 2θ = ∆ . (15.22)
The numbers n denote the amount of photons in the laser beam, the numbers N
denote the amount of energy packets within the system, that is, the photons plus the
possible excitation of the atom.
Figure 15.4: (a) Rabi splitting of the lowest dressed states. (b) Avoided crossing of dressed
states.
Generalizing the expression (15.14) to vectorial transitions, the electric dipole operator
reads,
X
d̂ = |iihi|eı(ωi −ωj )t d̂|jihj| (15.24)
i,j
X
= |gα ihgα |eı(ωgα −ωeβ )t d̂|eβ iheβ | + |eβ iheβ |e−ı(ωgα −ωeβ )t d̂|gα ihgα | .
α,β
setting ωga ≡ 0. For strontium, we may introduce the vectorial lowering operator
[202],
ˆ = σ̂x êx + σ̂y êy + σ̂z êz
~σ with σ̂β = |giheβ | with β = x, y, z . (15.26)
In this case, the electric dipole operator reads,
X
d̂strontium = |gihg|e−ıωeβ t d̂|eβ iheβ | + |eβ iheβ |eıωeβ t d̂|gihg| (15.27)
β
ˆ + c.c. ,
X
=d e−ıωeβ t |giheβ |êβ + c.c. = de−ıωe t ~σ
β
15.1.5 Exercises
15.1.5.1 Ex: Photon statistics
An optical resonator contains on average 10 photons in the mode T EM00q . What
is the probability of finding, at any time, 1 photon resp. 10 photons, when the light
is (a) thermal, (b) coherent? For case (a), what is the temperature of the light for
λ = 633 nm?
494 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
Now imagine a device capable of measuring the speed of a single atom randomly
chosen within the cloud.
a. Express the probability of measuring a specific velocity v 0 for this atom using the
density operator.
b. Express the expectation value of the average velocity by the density operator.
where ω is the frequency of the radiation, ω0 the frequency of the atomic transition,
and Ω the Rabi frequency generated by a single photon. We choose the Fock represen-
tation for the radiation mode, we represent the atomic transitions by Pauli matrices,
and we span the product space ρ̂f ield ⊗ ρ̂atom generalizing the operators â± y â± ⊗ I
and σ̂ ± y I ⊗ σ̂ ± . Explicitly we get,
†
P √ 1 0 +
P 0 0
â = n n + 1|n + 1i hn| and σ̂ = n |ni hn|
0 1 1 0
P √ 1 0 −
P 0 1
â = n n|n − 1i hn| and σ̂ = n |ni hn| .
0 1 0 0
(15.29)
Note that we can simplify the above Hamiltonian Ĥn applying the unitary transform,
−ı(n−1/2)ωt
e 0
U≡ , (15.32)
0 e−ı(n−1/2)ωt
yielding,
∆ Ω1 √
n
H̃n = U † Ĥn U + ıU̇ † U = Ω1
2√ 2 . (15.33)
2 n −∆2
The eigenvectors corresponding to the eigenvalues Ên are obtained by Ĥn |xi = enx |xi,
that is, Ên Un−1 |xi = enx Un−1 |xi. Knowing Ên |yi = eny |yi, we obtain |xi = Un |yi.
The temporal evolution of the Jaynes-Cummings state, |ψ(t)i = e−ıĤt |ψ(0)i, is de-
scribed by the transformation,
For large n̄, a resonant π/2-pulse does (ignoring irrelevant dynamical phases),
! !
|2i|n̄ − 1i π/2 (|2i|n̄ − 1i + ı|1i|n̄i)
y , (15.44)
|1i|n̄i (ı|2i|n̄ − 1i + |1i|n̄i)
Obviously, the structure of the field |αi is not affected, and we recover the dy-
namics of a two-level atom excited by a resonant classical radiation as described
by the Bloch equations (14.62).
Figure 15.5: Atomic level scheme for the implementation of resonant interactions with clas-
sical radiation fields (on the lower transition) and dispersive interactions with quantum fields
(on the upper transition).
(0)
In the unperturbed case we have, Ĥn |ψj,n i0 = Ej,n |ψj,n i0 , where the n-photon
subspace is spanned by the basis |ji = (1 0) and (0 1). In second perturbation order,
(1) (2)
0 X hj|Ĥn(1) |iihi|Ĥn(1) |ji nΩ21
hψj,n |Ĥn |ψj,n i ' hj|Ĥn + Ĥn |ji + (0) (0)
=∓ , (15.47)
Ej,n − Ei,n 4∆
j6=i
where the upper sign holds for |ji = (1 0). This result was already obtained in the
Excs. 6.1.3.4 and 14.3.5.14. In matrix notation,
2
(1) nΩ1 /4∆ 0
Ĥn ' . (15.48)
0 −nΩ21 /4∆
2
!
(1)
−ıĤn t eınΩ1 t/4∆ 0
e = 2 . (15.49)
0 e−ınΩ1 t/4∆
The fact that the ground and excited atomic states evolve with different phase factors
is important, as we will show in the following example.
Example 67 (Dispersive π-pulse): As in the previous case, we consider a
two-level atom subject to a coherent field, but now tuned out of resonance.
Introducing the abbreviation ϕ ≡ Ω21 t/4∆, the Jaynes-Cummings evolution is,
!
(1) eınϕ 0
e−ıĤn t = . (15.50)
0 e−ınϕ
The fact that the phase shift nϕ depends on the number of photons, and that
it goes in opposite directions for the ground and excited states, is crucial. We
have already studied in Exc. 14.5.5.7, that the dispersive interaction of the
atom with a radiation field can phase-shift the Bloch vector. Now, we observe
that in addition, it causes a phase shift of the probability amplitude of having
n photons in the radiation field by a value proportional to n, i.e. (ignoring
irrelevant dynamical phases),
! !
|2i|n − 1i nϕ eınϕ |2i|n − 1i
y . (15.51)
|1i|ni e−ınϕ |1i|ni
15.2. THE JAYNES-CUMMINGS MODEL 499
Apparently, the phase of the radiation field is shifted by a value ϕ, which depends
on the state of the atom.
We note here, that the dynamics studied in the last example provides a method
of transferring coherence from an atomic superposition to a quantum correlation of
a radiation field. All we have to do, is to bring the atom into a superposition of
states |1i + |2i, and the field will automatically evolve toward a Schrödinger cat state
|αeıϕ i + |αe−ıϕ i. The transfer of quantum correlations between coupled degrees of
freedom is one of the characteristics of the Jaynes-Cummings model. As examples
we will study the phenomena of quantum collapse and revival in Exc. 15.2.5.3 and of
vacuum Rabi splitting in 15.2.5.4.
The evolution of the coefficients cjn completely describes the Jaynes-Cummings dy-
namics of the system through the formula (15.37). Obviously, the Jaynes-Cummings
state is normalized because,
X
hψ|ψi = Tr |ψihψ| = (|c1,n |2 + |c2,n |2 ) = 1 . (15.55)
n
ρ̂ = |ψihψ| . (15.56)
degrees of freedom from the state. For example, if our focus is on the optical mode,
we ignore the atomic state,
X X
|γi ≡ hj|ψi = c1,n |ni + c2,n−1 |n − 1i . (15.57)
j=1,2 n
(b) We trace over the degrees of freedom NOT under study by DOING a measurement.
For example, if again our focus is on the optical mode, we trace over the atomic states,
(mix)
X X
ρ̂f ield = Tratom ρ̂ = hj|ρ̂|ji = hj|ψihψ|ji (15.61)
j=1,2 j=1,2
X (pure)
= c∗1,m c1,n |nihm| + c∗2,m−1 c2,n−1 |n − 1ihm − 1| =
6 ρ̂f ield .
n,m
It is clear, that this incomplete measurement converts the reduced density operator
into a statistical mixture, which is free of inneratomic correlations of the type c∗2,m c1,n ,
but this means that we also loose possible field correlations. On the other hand,
tracing over the field mode,
∞
X ∞
X
(mix)
ρ̂atom = Trf ield ρ̂ = hn|ρ̂|ni = hn|ψihψ|ni (15.62)
n=0 n=0
X (pure)
= (c1,n |1i + c2,n |2i) c∗1,n h1| + c∗2,n h2| =6 ρ̂atom .
n
After these preliminary remarks let us have a look a some interesting observables.
To determine the internal state of the atom, we must trace over the light field.
The populations and coherences are, therefore,
X X
ρij = hi|Trf ield ρ̂|ji = hi| hn|ψihψ|ni|ji = ci,n c∗j,n . (15.65)
n n
With this, we can calculate the atomic Bloch vector (14.74), whose norm is interest-
ingly NOT preserved, since,
2 Re ρ12
ρ| =
|~
2
2 Im ρ12
= 2|ρ12 | − 2ρ11 ρ22 = −2 det ρ̂ (15.67)
ρ −ρ
22 11
X X X X
=2 c1,n c∗2,n c∗1,n c2,n − 2 |c2,n |2 |c1,n |2 6= 1 .
n n n n
such that,
X
pn = hn|Tratom ρ̂|ni = hn| hi|ψihψ|ii|ni = |hn|ψi|2 = |c1,n |2 + |c2,n |2 . (15.69)
i=1,2
such that,
2 2
X α n X α n
2
πQ(α) ≡ hα|Tratom ρ̂|αi = e−|α| c √ + c √ . (15.71)
n 1,n n! n 2,n n!
502 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
We will derive this result in Exc. 15.2.5.5. This quantity, called Q-function, allows
the illustration of the state in a coordinate system spanned by Re α and Im α [87].
It is generally easy to calculate, but does not exhibit much information, e.g., on
interference phenomena caused by quantum correlations. In the following section,
we will calculate the Wigner function, which can also be evaluated from the Jaynes-
Cummings coefficients [232].
Figure 15.6: (code) Evolution of the state during a Jaynes-Cummings type interaction:
(a) Bloch vector, (b,c) photon distribution after projection on the ground and excited atomic
state, (d) time evolution of the coherence ρ12 showing the phenomenon of collapse and revival,
and (e) W (α) function.
with
βn
2 2 (eıϕ(t) β)n
c1,n = e−|β| √ /2
and c2,n = e−|β| /2 √ . (15.73)
n! n!
We will now concentrate on the field state by explicitly IGNORING the atomic states
|ji,
X (pure)
X
|γi = hj|ψi = |βi + |eıϕ βi or ρ̂f ield = hj|ψihψ|ii = |γihγ| . (15.74)
j=1,2 i,j=1,2
From the optical cat pure state wavefunction (15.74), we can easily calculate the
Wigner function in the Glauber picture, as done in (3.204) and Exc. 3.7.5.4,
2 −2|α−β|2 −2|α−βeıϕ |2 −|β|2 −2(βeıϕ −α)(β ∗ −α∗ )+|β|2 eıϕ
W (α) = e + e ± 2e Re e .
C 2π
(15.75)
15.2. THE JAYNES-CUMMINGS MODEL 503
Now, what happens if we MEASURE the atomic state before we analyze the field
state? The incomplete measurement corresponds to tracing over the atomic degrees
of freedom, which transforms our initially pure state (15.72) into a statistical mixture.
With |ψi given by (15.72),
(mix)
X
ρ̂f ield = Tratom ρ̂ = hj|ψihψ|ji = |βihβ| + |eıϕ βiheıϕ β| =
6 |γihγ| . (15.76)
j=1,2
Calculating the Wigner function from this density operator consequently must yield
a different result, because any correlations are lost. A movie of the dynamics can be
watched here (watch movie) [125, 182, 8, 713].
Obviously, this formula does not contain inneratomic correlations and, since non-
classical field correlation would be entangled with atomic superposition states, we
expect no interesting field correlations, neither.
If on the other hand we decide to ignore the atomic state, we calculate the normally
ordered characteristic function using the density operator (15.58),
(pure) † ∗ † ∗
χN (λ) = Tr ρ̂f ield eλâ e−λ â = hγ|eλâ e−λ â |γi (15.79)
X † ∗ † ∗
= c∗1,m c1,n hm|eλâ e−λ â |ni + c∗2,m−1 c2,n−1 hm − 1|eλâ e−λ â |n − 1i
n,m
† ∗ † ∗
+c∗1,m c2,n−1 hm|e−λâ eλ â |n − 1i + c∗2,m−1 c1,n hm − 1|eλâ e−λ â |ni ,
which now contains field correlations. Inserting the results derived in (3.195) we can
express the characteristic functions by Laguerre polynomials and obtain the Wigner
function as the two-dimensional FFT according to Eq. (15.75). This will be done in
Exc. 15.2.5.8.
[271],
Ĥef f = ω↠â + ω0 σ̂ + σ̂ − 1
2 + Ω1
2 âσ̂ + + ↠σ̂ − + ıΓ + −
2 σ̂ σ̂ . (15.80)
Or in matrix notation,
− 21 ω0 0
1 1
0
2
ω0 − 2ı Γ Ω
2 1
1
Ω
2 1
ω − 21 ω0 0
√ 1
0 ω + 12 ω0 − 2ı Γ 2 2 Ω1
Hef f = √ 1 .
2 2 Ω1 2ω − 12 ω0 0
2ω + 21 ω0 − 2ı Γ
0
..
.
(15.81)
The simulation flowchart is,
|1i(h1|+h2|)|ψi |ψi
|ψi → |ψi →
P |(h1|+h2|)|ψi| P ||ψi|
|1i
= n√(cP
1n |nii+c2n−1 |n−1i)
2
= n √ |1i|ni+c2n−1 |2i|n−1i
c1n
P 2 2
n |c1n +c2n | | n (|c1n | +|c2n | )
projection & . renormalization
dynamic evolution
|ψ(t + dt)i = e−ıHef f dt |ψi
- ↓ %
?
yes ζ > hψ(t + dt)|ψ(t + dt)i no
random variable
Absorption or scattering of light causes the optical field to decay as α(t) ∝ e−κt/2 .
The projection (in component notation) is implemented by,
1
c0jn ≡ pP cjn , (15.82)
2 2
n (|c1n | + |c2n | )
and the dynamical evolution by c0jn ≡ e−κnt/2 cjn . Note, that the dissipation due to
cavity losses can also be taken into account by a master equation. Do the Exc. 16.2.6.1.
The coupling of the atom to the cavities g(z) will, in general, depend on the atomic
position. For simplicity let us, however, consider uniform and identical coupling,
g+ (z) = g− (z) = g, and furthermore assume that both cavities be on resonance,
∆+ = ∆− = 0. Finally, neglecting spontaneous emission and pumping, the total
number of photons is conserved. With ∆a = ω − ωa we get,
n
X
|ψ(t)i = cj,n+ ,n− |j, n+ , n− i . (15.85)
k=0
∞
X N
X
|ψ(t)i = c̃N,D |N, Di . (15.86)
N =0 D=−N
Obviously, the set of three quantum numbers {j, n+ , n− } is equivalent to the set of
two quantum numbers {N, D} and can be recovered by,
Hence, states with odd N + D are excited atomic states, and states with even N + D
are ground states. We may now write down the Hamiltonian,
The subscripts of the coupling constants g+ = g− are there only to clarify to which
mode the photon belongs. Very far from resonance, where the interaction is dispersive,
506 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
or with g+ = g− ,
1 1 g2 ∆ 1 g2
(ĤN )DD0 = N+ ω− + (−1)N +D + N+ δDD0 .
2 2 4∆ 2 2 4∆
(15.90)
If only the upper level of the atom interacts with both light fields, all terms with
negative light shift disappear,
1 1 N g2 N +D ∆ N g2
(ĤN )DD0 = N+ ω+ + (−1) + δDD0 . (15.91)
2 2 4∆ 2 2 4∆
As shown in Sec. 15.2.1, we propagate the evolution of the coupled state via
|ψ(t)i = e−ıĤt |ψ(0)i. As the initial condition we may choose the atom to be in its
ground state and the two field modes in uncorrelated Glauber states,
n n
2
/2−|α− |2 /2 α +α −
cj,n+ ,n− = hj, n+ , n− |ψi = e−|α+ | p+ − δj,0 (15.92)
n+ !n− !
(N +δ−j)/2 (N −δ−j)/2
2
/2−|α− |2 /2 α+ α−
= c̃N,δ = e−|α+ | q δ1+(−1)N +δ−1 ,0 .
N −δ−j
( N +δ−j
2 )!( 2 )!
1 0.2
|1, n|ψ|2
1
0.5 0.1
ρ22 − ρ11
Re ρ12
0 15
0 0
W (α)
0 1 2 3 4 5 6 7 8 9 10
2
1 0 0
1 2
0 0 -1 0.1 0
-1 -1 t -2 -2 xα
pα
Re ρ12 0
Im ρ12 0 1 2 3 4 5 6 7 8 9
Figure 15.7: (code) Evolution of the coupled atom-ring cavity system. (a) Bloch vector, (b)
photon distributions, and (c) Husimi function.
The observables of interest are the atomic Bloch vector (14.74), whose components
15.2. THE JAYNES-CUMMINGS MODEL 507
and analogously for pn− , and the field distribution functions, such as the Husimi
function,
and analogously for Q− (α), or the Wigner functions W± (α) from the coefficients
c̃j,n+ ,n− , respectively, c̃N,δ , as shown in Secs. 3.7.3 and 15.2.2.
15.2.5 Exercises
15.2.5.1 Ex: Time-evolution in the Jaynes-Cummings model
Derive the equations of motion for σ̂ − , σ̂z , and â in the Jaynes-Cummings model.
Show that the number of photons is not a constant of motion, but the total number
of excitations.
and determine the equations of motion for the amplitudes c2 and c1,k .
508 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
Consider the Jaynes-Cummings Hamiltonian and show that the quantum coherence
between the two atomic levels can disappear altogether for long periods and reappear
later. Explain how this is possible.
Calculate the Autler-Townes splitting for an excited atom interacting with an empty
cavity.
Calculate the Q-function for a Jaynes-Cummings state from its definition (15.71).
We will show in this exercise how, via coherent operations in a three-level system,
we can create Schrödinger-type quantum-type correlations in an optical mode. In the
system shown in Fig. 15.8 we imagine the lower transition excited by π/2-pulses of a
classical resonant microwave radiation (as described by the operation (15.42)). The
upper transition is excited by quantum laser pulses tuned very far out of resonance,
thus creating a dispersive dynamics (as described by the operation (15.42)). At time
√ |1i. Now, we apply the following pulse sequence: (i) a
t = 0 the atom is in state
microwave pulse with n̄Ω12 t = π/2, (ii) an optical pulse with Ω223 t/4∆23 = π,
(iii) another microwave π/2-pulse, and finally (iv) an optical pulse of light which is
resonant with the transition |2i-|3i and projects the population of the atom into one
of the states of the microwave transition. Describe the evolution of the state of the
system during the sequence and determine the final state of the optical mode.
hÊ − (r1 , t1 )..Ê − (rn , tn )Ê + (rn+1 , tn+1 )..Ê + (r2n , t2n )i
g (n) (r1 , t1 , .., r2n , t2n ) ≡ q .
hÊ − (r1 , t1 )Ê + (r1 , t1 )i..hÊ − (r2n , t2n )Ê + (r2n , t2n )i
(15.96)
In the following, we will only consider temporal coherences, e.g. one or more collinear
light beams, km k kn , impinging on a photodetector, (tm − zcm ) − (tn − zcn ) = τ , since
Ê(r, t) = Ê(ωt − k · r), we can define the the 1st and 2nd order correlation functions
g (1) and g (2) , which are particularly important,
averages,
Z t
1
h· · ·it = lim · · · dt . (15.98)
t→∞ t 0
The correlation functions must be calculated from the field operators simultaneously
respecting time order and normal order. These functions are useful quantities to de-
scribe phenomena such as photon bunching or to understand the fluorescence spectra
or the scattering of light from correlated atoms. g (1) measures the coherence of a
light field (how much it resembles a sine wave). g (2) measures, for a given degree of
coherence, the deviation of the light field from the quantum state that most closely
approximates a classical light (how much it resembles a laser).
The correlation functions g (1) and g (2) are experimentally measured in Young’s
experiment and in the Hanbury-Brown-Twiss experiment. The experimental schemes
are explained in Figs. 15.9.
Figure 15.9: (a) Scheme of Young’s experiment. (b) Scheme of the experiment of Han-
bury, Brown, and Twiss. Young’s experiment reveals the coherence of a field, that is, its
ability to interfere. In contrast, the Hanbury-Brown-Twist experiment reveals correlations
between the (quasi-)particles constituting the field, that is, effects due to quantum statistics
or interactions.
Coherence and chaos are contrary properties of light. They leave their imprint
in the spectrum of the light or in the autocorrelation functions. The emission spec-
trum of a light source generally emerges as a combination of various physical ef-
fects: The active medium gives rise resonances and broadenings, a resonator con-
taining the active medium imprints a modal structure, and the coupling to a ther-
mal bath gives rise to P
a thermal distribution of the radiation energy according to
Pn = e−~ω(n+1/2)/kB T / n e−~ω(n+1/2)/kB T .
RE (τ )
g (1) (τ ) = , (15.102)
RE (0)
and,
Z ∞
SE (ω) 1
FE (ω) ≡ (Fg (1) )(ω) = = g (1) (τ )eıωτ dτ . (15.103)
RE (0) 2π −∞
The quantity FE (ω) is called line profile or spectrum. Note that, since,
laser
E(t)eıω0 t=
R
t−1 e−ıω0 τ dt
=⇒ (1)
g (τ ) = R = e−ıω0 τ
t−1 dt
Z ∞ . (15.106)
1
=⇒ F g (1) (τ ) = eı(ω−ω0 )τ dτ = δ(∆)
2π −∞
g (2) (τ ) = 1
We see that the absolute values of the first and second-order coherences are constant,
and that the spectrum is narrow like a δ-function.
512 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
For a laser subject to white phase noise (ζ be a normally distributed random number)
we have ,
noisy laser
We see, that the first-order coherence decays exponentially, |g (1) (τ )| = e−γ|τ | 6 , such
that the spectrum has a Lorentzian profile. This result has already been derived in
Sec. 13.3.1 for the natural linewidth of a transition subject to spontaneous emission.
We understand the connection by interpreting spontaneous emission as being induced
by vacuum fluctuations, which do have a white noise spectrum, indeed.
where γ is the spontaneous decay rate of the dipole moment. From E(t) =
eıω0 t+ıζ(t) we calculate,
Z t
1
RE (τ ) = heıω0 t+ıζ(t) e−ıω0 (t+τ )−ıζ(t+τ ) i = e−ıω0 τ lim eı[ζ(t)−ζ(t+τ )] dt
t→∞ t 0
Z ∞
= e−ıω0 τ p(τ 0 )dτ 0 = e−ıω0 τ e−γ|τ | ,
τ
and,
Z ∞ Z ∞
1 RE (τ ) ıωτ 1 γ/π
FE (ω) = e dτ = e−ıω0 τ e−γ|τ | eı(ω−ω0 )τ dτ = .
2π −∞ RE (0) 2π −∞ ∆2 + γ 2
6 Note that, in order to satisfy (15.104), we must take the absolute value of the delay |τ |.
15.3. CORRELATION FUNCTIONS 513
The calculation for the 2nd order correlation function (15.99) is analogous,
X
hE ∗ (t)E ∗ (t + τ )E(t + τ )E(t)i = ∗
hEm ∗
(t)En (t)Em 0 (t + τ )En0 (t + τ )i (15.112)
n,m,n0 ,m0
X
= hEn∗ (t)En (t)En∗ (t + τ )En (t + τ )i
n
X
+ hEn∗ (t)En (t)Em
∗ ∗
(t + τ )Em (t + τ )i + hEm (t)En (t)En∗ (t + τ )Em (t + τ )i ,
n6=m
neglecting all terms which do not satisfy either En∗ (t)En (t) or En∗ (t)En (t + τ ). Now,
assuming a large amount of identical atoms,
Finally, exploiting the result (15.109) and the definitions of the 1st and 2nd order
correlations functions,
hEn∗ (t)En (t)i2 + |hEn (t)En∗ (t + τ )i|2
g (2) (τ ) = = 1 + |g (1) (τ )|2 . (15.114)
hEn∗ (t)En (t)i2
514 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
In summary,
pressure broadening
P
E(t) = n eıω0 t+ıφn (t)
=⇒ g (1) (τ ) = eıω0 τ −γc |τ |
γ/π . (15.115)
=⇒ F g (1) (τ ) =
∆2 + γc2
g (2) (τ ) = 1 + e−γc |τ |
(a) 1 (b) 2
1.5
g (1) (τ )
g (2) (τ )
0.5 1
0 0
0 5 10 0 5 10
ωτ ωτ
Figure 15.10: (code) First and second-order correlation functions for (red) a laser, (cyan) a
laser subject to phase noise, (blue) a laser subject to collision broadening, (magenta) chaotic
light, and (green) spectrally filtered thermal light.
Z X X
hE ∗ (t)E(t + τ )i = e−ıωn t−ıφn eıωm t+ıωm τ +ıφm dt (15.116)
n m
XZ X
= e−ıωn t−ıφn +ıωm t+ıωm τ +ıφm dt = eıωn τ .
n,m n
2
/2δ 2
a Gaussian, p(ω)dω = (2πδ)−1/2 e−(ωn −ω0 ) dω, such that,
thermal ensemble
Z
2
/2δ 2
hE ∗ (t)E(t + τ )i = N (2πδ)−1/2 eıωn τ e−(ωn −ω0 ) dωn
2 2
=⇒ g (1) (τ )= eıωτ −δ τ /2
r . (15.117)
ln 2 − ln 2·ω2 /δ2
=⇒ F[g (1) (τ )] = e
πδ 2
2 2
g (2) (τ ) = 1 + e−δ τ /2
chaotic line
g (1) (τ ) = δ(0)
F[g (1) (τ )] = 1 . (15.118)
(2) (1)
g (τ ) = 1 + |g (τ )|
The last expression is known as Bloch-Siegert relation. Mono-mode chaotic light can
be seen as incoherent multi-mode light, where all modes except a single mode are
filtered by a Fabry-Pérot etalon. This light is characterized by |g (1) (τ )| = 1 and
g (2) (τ ) = 2, despite the coherence length being τ → ∞. Do the Excs. 15.3.5.1 to
15.3.5.3.
a. Calculate |g (1) (τ )|, FE (ω), and |g (2) (τ )| for two interfering and non-interfering
modes neglecting fluctuations.
b. What changes when one mode is broadened by random noise, e.g. induced by
spontaneous emission?
a. Calculate g (1) (τ ), SE (ω), and g (2) (τ ) for amplitude-modulated light: Eam (t) =
eıω0 t (1 + M cos Ωt).
b. Calculate g (1) (τ ), SE (ω), and g (2) (τ ) for phase-modulated light: Epm (t) = eıω0 t+ıM cos Ωt .
c. Repeat the calculation (a) for exponentially decaying amplitude-modulated light:
Edam (t) = e−γt eıω0 t (1 + M cos Ωt).
516 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
Denoting the frequency of the atomic resonance by ωa , the interaction part of the
Hamiltonian is,
Ĥ = ~gk0 (σ̂ − e−ıωa t + σ̂ + eıωa t ) â†k0 eıω0 t−ık0 ·r + âk0 e−ıω0 t+ık0 ·r
X † ıω t−ık·r . (15.123)
+ ~gk σ̂ − e−ıωa t + σ̂ + eıωa t âk e k + âk e−ıωk t+ık·r
k
gk0 is the coupling strength of the interaction between the atom and the pump mode,
σ̂ − is the atomic
p deexcitation operator, âk is the annihilation operator of a photon,
and gk = d ω/(~ε0 V ) describes the coupling between the atom and a vacuum mode
whose volume is V . The atom has two states, the ground state |gi and excited state
|ei. Since we are considering only one atom fixed in space 7 , we can as well locate it
at the origin r = 0. In addition, considering a high power incident laser,
√ √
âk0 |n0 ik0 = n0 |n0 − 1ik0 ' n0 |n0 ik0 , (15.124)
where |jia denotes the atomic state and |nik the number of photons in the scattering
mode.
The temporal evolution of the amplitudes is obtained by inserting the Hamiltonian
(15.125) and the ansatz (15.127) into the Schrödinger equation,
∂ ı
|Ψ(t)i = − Ĥ|Ψ(t)i . (15.128)
∂t ~
As verified in Exc. 15.4.4.1, we obtain,
α̇(t) = −ı Ω20 eı∆0 t β(t) (15.129)
X
β̇(t) = −ı Ω20 α(t)e−ı∆0 t − ıgk γk (t)e−ı∆k t
k
γ̇k (t) = −ıgk eı∆k t β(t) .
Now, we chose the initial conditions,
α(0) = 1 and β(0) = 0 and γk (0) = 0 , (15.130)
we integrate the third equation,
Z t
0
γk (t) = −ıgk eı∆k t β(t0 )dt0 , (15.131)
0
8 The approximation does not necessarily hold for large clouds of atoms.
15.4. SPONTANEOUS EMISSION AND LIGHT SCATTERING 519
and implement the Markov approximation by setting β(t − t00 ) ' β(t), and with
R t −ı(ω −ω )t0 0 P V
R 3
lim e k a
dt = πδ(ωk − ωa ), and replacing k −→ (2π)3 d k, we arrive
t→∞ 0
at,
d X
β(t) ' −ı Ω20 α(t) − gk2 β(t)πδ(ωk − ωa ) (15.134)
dt
k
Z
Ω0 V
= −ı 2 α(t) − (2π) 3 β(t) gk2 πδ(ωk − ωa )d3 k
2 21
= −ı Ω20 α(t) − V
(2π)3 β(t)4πgka πka c = −ı Ω20 α(t) − Γ2 β(t) .
Finally,
d Ω0 d Ω0 Γ
α(t) = −ı β(t) and β(t) = −ı α(t) − β(t) . (15.136)
dt 2 dt 2 2
These are exactly the equations for the probability amplitudes (14.95) derived
from the Schrödinger equation, only that now, the spontaneous emission term has
been derived explicitly. Solve the Exc. 15.4.4.2.
Example 69 (Emission stimulated by vacuum fluctuations): Spontaneous
emission can be regarded as an emission stimulated by vacuum fluctuations.
To see this, we write down the resonant optical cross section (ω0 = ωa ) of a
driven two-level atom without degeneracies, σ0 = λ2 /2π, and the intensity of an
incident laser field, I¯ = cN ~ω0 /V , generating the Rabi frequency (see (1.107)),
I¯ 2πc N
Ω20 = σ0 Γ= 2 Γ. (15.137)
~ωa ka V
Now, we assume that the field is, in fact, a vacuum mode containing only half
a photon, N = 1/2, which corresponds to vacuum fluctuations in the mode k0 .
Then,
πc
Ω21/2 = 2 Γ = g 2 . (15.138)
ka V
Figure 15.12: (a) Geometry of a scattering experiment. (b) Spectral contributions of light
scattered elastically and inelastically by a three-level atom.
where r is now the observation point of the electric field. We restrict to the far-field
and substitute the annihilation operator with the integral of Eq. (15.140) using the
9 Classical theories of light scattering through the excitation of an electronic motion based on the
models of Lorentz or Drude can be found in the script Electrodynamics. Although being classical,
these model are useful for a deeper understanding of many aspects of Compton scattering, Thomson
scattering, and Rayleigh scattering.
15.4. SPONTANEOUS EMISSION AND LIGHT SCATTERING 521
+ d12 ka2 −
Êsct (r, t) ' −ı σ̂ (t − rc ) . (15.145)
4πε0 r
with the definition of the Sommerfeld constant α = e2 /4πε0 ~c. We calculate the total
flux of emitted photons,
Z ¯ 2 Z
(sp) Is r 1 α~ωa4 |r12 |2 cos2 θ
Wf i = dΩ = ρ22 (t − rc ) sin θdθdφ (15.148)
~ωa ~ωa 4πc2
8π α~ωa4 |r12 |2 2α
= 2
ρ22 (t − rc ) = 2 ωa3 |r12 |2 ρ22 (t − rc ) .
3~ωa 4πc 3c
The result coincides with the spontaneous emission rate Γ calculated in (13.37).
A differential scattering cross section can be defined by,
dσ ω I¯s r2
≡ . (15.149)
dΩ ωs I¯0
10 The classical version of this formula is can be found in the script Electrodynamics, Sec. 9.1.3,
Eq. (8.41).
522 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
s/2 s2 /2
Scoh = |ρ21 (∞)|2 = and Sincoh = ρ22 (∞) − |ρ21 (∞)|2 = .
(1 + s)2 (1 + s)2
(15.152)
Hence, Sincoh = sScoh .
0.5
Scoh , Sincoh
0.25
0
0 5 10
s
2 2 2
d∆ = √ −→ . (15.153)
−∞ 4∆ + 2Ω + Γ 2 2Ω2 + Γ2 2Γ
which in turn are related to the field operators (15.8). The field operators follow the
solutions of the Bloch equation, which, being linear, have the following generic form,
X
ρij (t + τ ) = αijkl (τ )ρij (t) + βij (τ ) . (15.154)
k,l
The trace condition is satisfied, when (i, j), (k, l) 6= (1, 1).
To be able to explore this relationship (15.103), needed to calculate the correlation
functions, we have to invoke the so-called quantum regression theorem,
X X
hÂ(t + τ )i = ξi (τ )hÂi (t)i =⇒ hB̂(t)Â(t + τ )Ĉ(t)i = ξi (τ )hB̂(t)Âi (t)Ĉ(t)i .
i i
(15.155)
Example 70 (Quantum regression applied to the Langevin equation): We
have,
We know,
hAµ (t)Fν (t)i = hDµν i and hFµ (t)Aν (t)i = hDµν i (15.157)
~(t + τ ) = eMτ ρ
with the solution (14.171), that is, ρ ~(t) + (1 − eMt )~ ρ(∞) with ρ(∞) =
−1
−M b. This solution can be cast in the following form,
X
ρkl (t + τ ) = α(kl)(mn) (τ )ρmn (t) + β(kl) (τ ) , (15.161)
(mn)
524 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
hσ̂kl (t + τ )i (15.165)
X
(l−k)ıω0 τ (l−k+m−n)ıω0 t (l−k)ıω0 τ (l−k)ıω0 t
= e α(kl)(mn) (τ )he σ̂mn (t)i + e β(mn) (τ )he i.
(mn)
Applying the quantum regression theorem to the Bloch equations, we get for an
arbitrary time-dependent function f (t),
X
hf (t)σ̂kl (t + τ )i = e(l−k)ıω0 τ α(kl)(mn) (τ )he(l−k+m−n)ıω0 t f (t)σ̂mn (t)i
(mn) .
+e(l−k)ıω0 τ β(mn) (τ )he(l−k)ıω0 t f (t)i
(15.166)
Now we can calculate, letting ξi (τ ) ≡ α(12)(mn) , Ĉ(t) ≡ 1, and B̂(t) ≡ σ̂21 (t),
P ıω0 (t+τ ) (m−n)ıω0 t
(1) (mn) e e α(12)(mn) (τ )hσ̂21 (t)σ̂mn (t)i + eıω0 (t+τ ) β(12) (τ )hσ̂21 (t)i
g (τ ) =
hσ̂22 (t)i
(15.169)
P (m−n)ıω0 t
(mn) e α(22)(mn) (τ )hσ̂22 (t)σ̂mn (t)i + β(22) (τ )hσ̂22 (t)i
g (2) (τ ) = .
hσ̂22 (t)i2
Using σ̂21 σ̂mn = σ̂2n δm1 ,
α(12)(12) (τ )hσ̂21 (t)σ̂12 (t)i + β(12) (τ )heıω0 t σ̂21 (t)i
g (1) (τ ) = eıω0 τ (15.170)
hσ̂22 (t)i
α(22)(21) (τ )heıω0 t σ̂22 (t)σ̂21 (t)i + α(22)(22) (τ )hσ̂22 (t)σ̂22 (t)i + β(22) (τ )hσ̂22 (t)i
g (2) (τ ) = .
hσ̂22 (t)i2
Returning to the density operator and letting t → ∞,
(1) ıω0 τ ρ21 (∞) β(22) (τ )
g (τ ) = e α(12)(12) (τ ) + β(12) (τ ) and g (2) (τ ) = eıω0 τ ,
ρ22 (∞) ρ22 (∞)
(15.171)
that is,
(1) ıω0 τ
Mτ Mτ −1
[M−1 b](21)
g (τ ) = e e (12)(12)
− (I − e )M b (12) [M−1 b]
(22)
. (15.172)
Mτ −1
(2) ıω0 τ
[(I−e )M b]
(22)
g (τ ) = e [M−1 b](22)
These correlation functions can easily be calculated via a numerical resolution of the
Bloch equations (15.160). Fig. 15.14 shows the correlation functions and the fluores-
cence spectrum derived by Fourier transform of the first-order correlation function
(15.102). Assuming resonant excitation, ∆ = 0, analytic formulas can be derived, as
will be exercised in Excs. 15.4.4.3 and Exc. 15.4.4.4.
The spectrum 15.14(d) exhibits three lines known as the Mollow triplet. Note that
the spontaneous emission triplet is only observed in the presence of a driving laser,
because it is the laser excitation which causes the splitting. Indeed, the splitting and
the position of the lines are easily understood in the dressed states picture visualized
in Fig. 15.1: The coupling of the two-level atom to a light field splits up the levels |ni
and |n + 1i by an amount corresponding to the Rabi frequency Ω. Now, the transition
from the two excited state |n + 1i levels to the ground state |ni levels can occur on
three different frequencies. In Exc. 15.4.4.5 we calculate the Mollow spectrum for a
transition between one ground and three excited Zeeman states. Fig. 15.15 illustrates
the various methods to analyze scattered light, but not all of them yield information
on the Mollow triplet.
1.5
200
g (1) (τ )
g (2) (τ )
SE (ω)
0.5
ρkk (t)
0.5 1
100
0.5
0
0 0 0
0 5 -10 0 10 -10 0 10 -10 0 10
Γt Γτ Γτ Δ/Γ
Figure 15.14: (code) (a) Temporal evolution of the excited state population ρ22 (t) (cyan) and
coherence ρ12 (t) (magenta) of a laser-driven two-level atom with Ω = 5Γ. (b) Correlation
function g (1) (τ ) and (c) g (2) (τ ). The blue curves in (b-c) are obtained by numerical simula-
tions of the Bloch equations and subsequent application of the quantum regression theorem
(15.172). The green dots are obtained from an analytic solution derived in Exc. 15.4.4.3. (d)
Mollow spectrum obtained by numerical FFT of g (1) (τ ).
(its just a unitary transform of the dressed states picture). Developing a physical
picture the Mollow triplet in the semi-classical framework may give us a deeper insight.
Generally, the Mollow spectrum is anyhow calculated using the semi-classical Bloch
equations, via the correlation function g (1) (τ ) and the Wiener-Khintchine theorem,
but the on the way the physical intuition is lost.
For example, looking at the stationary solution of the Bloch equations (14.99), we
see that ρ
~(∞) is time-independent, so that we might be surprised to see correlations
in time domain (and consequently a structured spectrum) at all. This surprise re-
sults from a common misconception that may arise considering the damping of ρ ~(t)
predicted by the Bloch model, as illustrated e.g. in Fig. 14.2: It seems that the atoms
eventually cease oscillating between the ground and excited states. In most experi-
ments, measurement are made on a large number of atoms and indeed the oscillations
are damped.
Figure 15.15: Schemes for analyzing resonance fluorescence: (a) Heterodyning [545], (b) tem-
poral correlation, (c) spectrum [670], (d) demodulation. Only the schemes (b) and (c) yield
information on the Mollow triplet.
These functions show that the spectrum is essentially composed of Rayleigh scattering
at the frequency of the incident light. The δ-shaped fluorescence spectrum shows,
that the contribution of elastically scattered light dominates below saturation, which
confirms the results (15.152) illustrated in Fig. 15.13. The light is ’antibunched’ and,
at higher τ exhibits a damped oscillation around the value 1.
If pressure broadening is taken into account, the two-level Bloch equations are
given by (14.99), where γ 0 = γ + γcoll is the width of the collision-broadened line.
Within this model and in the limit Ω Γ, the resonance fluorescence spectrum is
given by [456],
γ0 − Γ γ 0 /π Γ
F (ωs ) = + 0 δ(ωs − ω) . (15.174)
γ 0 (ω0 − ωs )2 + γ 02 γ
So, we find that, even at low intensities, a continuous spectrum due to inelastic scat-
tering appears around the resonance frequency ω0 additionally to the elastic Rayleigh
peak.
528 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
(a)
ψ(t)|ψ(t) 1 (b) 1 (c) 0 (d) 0.6
|k|ψ(t)|2
0.4
g (1) (τ )
SE (ω)
0.5 0.5 -0.5
0.2
0 0 -1 0
0 5 10 0 5 10 -50 0 50 -10 -5 0 5 10
Γt Γt Γτ f
Figure 15.16: (code) (a) Time evolution of the norm of a decaying driven two-level atom.
(b) Time evolution of the ground and excited state amplitudes. (c) First-order auto-
correlation function, and (d) spectrum.
15.4.4 Exercises
Inserting the ansatz (15.127) into the Schrödinger equation, derive the equations of
motion (15.129) for the wavefunction amplitudes.
Study the time evolution |ψ(t)i = e−ıĤef f t/~ |ψ(0)i with the effective Hamiltonian,
0 2Ω
~
Ĥef f =
2Ω
~
− ı~
2 Γ
starting from the initial condition h2|ψ(0)i = 1. Calculate the evolution of |ψ(t)i and
the norm hψ(t)|ψ(t)i. Plot the time evolution of the norm for various ratios Ω/Γ and
interpret the curves.
a. Derive the analytic solution of the Bloch equations for a resonantly driven two-level
atom.
b. Calculate the 1st -order correlation function g (1) (τ ) from the formula (15.172).
c. Derive the spectrum of resonance fluorescence [511].
d. Derive the 2nd -order correlation function g (2) (τ ) from the formula (15.172).
15.5. LIGHT SCATTERING FROM MULTI-LEVEL ATOMS 529
Calculate the first-order correlation function and the fluorescence spectrum of this
transition.
Writing the non-vanishing dipole matrix elements as d12 = eh1|r|2i, d23 = eh2|r|3i a
three-level atom has two microscopic oscillating dipoles,
In the semi-classical picture, the radiated field will be a sum of these two terms,
regardless of whether state |2i decays simultaneously to |1i and |3i or vice versa.
for the Λ and the V -system, respectively. The state vector for each type of system is,
X
|ψV (t)i = ci |i, 0i + c1 |c, 112 i + c2 |c, 123 i (15.180)
i=1,2,3
and, X
|ψΛ (t)i = c0i |i, 0i + c01 |b, 112 i + c02 |c, 123 i . (15.181)
i=1,2,3
hψV (t)|Ê1− (t)Ê2+ (t)|ψV (t)i = κh112 023 |a†1 a2 |012 123 ie[ı(ω12 −ω23 )t] h3|3i (15.182)
[ı(ω12 −ω23 )t]
= κe h3|3i
hψΛ (t)|Ê1− (t)Ê2+ (t)|ψΛ (t)i = κ0 h112 023 |a†1 a2 |012 123 ie[ı(ω12 −ω23 )t] h2|3i (15.183)
0 [ı(ω12 −ω23 )t]
=κe h2|3i
15.5. LIGHT SCATTERING FROM MULTI-LEVEL ATOMS 531
for the Λ-system. However, orthogonality of the eigenstates requires h3|3i = 1 and
h2|3i = 0. Therefore, there is a quantum beat note term for V-type atoms, but not
for Λ-type atoms.
This difference originates in quantum mechanical uncertainty. A V-type atom
decays to state |3i via the emission with ω12 and ω23 . Since both transitions decayed
to the same state, one cannot determine along which path each decayed, similar to
Young’s double-slit experiment. However, Λ-type atoms decay to two different states.
Therefore, in this case we can identify the path by the end product.
Quantum beat spectroscopy is a technique which allows for Doppler-free resolution
provided the separation of the adjacent levels is less than the Doppler width. It
consists in generating a coherently distributed population of two upper states, e.g. via
a short laser pulse, and detecting the beat frequency.
Evaluated far from resonance, ω ωm , this result leads to Thomson and Compton
scattering. Close to resonance the second term of the sum may neglected.
The matrix elements can be evaluated by the electric dipole Hamiltonian in second
quantization, 2
1 2π X X Ωf m Ωmi
= δ(ωf − ω + ωs ) . (15.186)
τ ~ ω − ωm
m
ks
We convert the transition rate into a cross section via,
X Z Z Z Z
V 2 V
→ k s dks dΩ = ωs2 dωs dΩ . (15.187)
(2π)3 (2πc)3
ks
532 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
We now insert for the matrix element (first term in ()) the cartesian multipole expan-
sion () and obtain,
r 2
1 2π X ~ω
X
= 2 hf | − ie [ε̂ · dE + ...]|ii δ(ω − ωf ) (15.189)
τ ~ 2ε0 V
f λ
Z
2π V e2 ~ω 2 X 2
−→ 2 ω |ε̂ · hg|dE |ei + ıε̂ · hg|qE |ei · k...] − ...| dΩ .
~ (2πc)3 2ε0 V
λ
P
Finally, letting λ = 2,
Z
1 αω 3 2
= |ε̂ · hg|multipole-tensor|eiwavenumber-tensor| dΩ . (15.190)
τ πc2
For example, for dipole radiation, letting ε̂ = êz ,
Z
1 αω 3 2
= |ε̂ · hg|dE |ei| d cos θdφ (15.191)
τ πc2
Z
αω 3 2 4αω 3
= 2
|hg|d E |ei| | cos θ|2 d cos θdφ = |hg|dE |ei|2 .
πc 3c2
15.5.2.2 Absorption
In first order perturbation theory we have Fermi’s Golden rule,
1 2π X
= 2 |hf |Ĥint |ii|2 δ(ωf − ωi ) . (15.192)
τ ~
f
In the dipolar approximation, Ĥint = −d · E, ~ and separating the field and atomic
degrees of freedom, |f i = |Nf i|Af i, we get for absorption processes,
1 2πe2 X 2
= 2 hNf |Ê − |Ni i · hAf |ε̂ · d̂|Ai i δ(ωf − ωi ) . (15.193)
τ ~
Nf
P
Generalizing to a statistical mixture of Fock states via Nf ,Ni hNi |Ê
+
|Nf ipi hNf |Ê − |Ni i =
Tr (ρ̂Ê + Ê − ), we may also write,
1 2πe2
= 2 |Mabs |2 δ(ωf − ωi )Tr (ρ̂Ê + Ê − ) , (15.194)
τ ~
where
|Mabs |2 ≡ hAf |ε̂ · d̂|Ai i . (15.195)
For an incoming photon ω, we get ωf = Ef and ωi = Ei + ω. See Fig. 15.18(a).
15.5. LIGHT SCATTERING FROM MULTI-LEVEL ATOMS 533
Figure 15.18: Feynman graphs for (a) absorption, (b) spontaneous emission, (c) stimu-
lated emission, (d) two-photon absorption, (e) spontaneous Raman, (f) stimulated Raman,
(g) spontaneous second harmonic generation, (h) and stimulated second harmonic genera-
tion.
ˆ ˆ
In the dipolar approximation, Ĥint = −d · E~1 − d · E~2 , and separating the field and
atomic degrees of freedom, |f i = |Nf i|Af i, we get for two-photon absorption pro-
cesses,
4 X X
1 2πe hNf |E2− |Nm ihNm |E1− |Ni ihAf |ε̂2 · d̂|Am ihAm |ε̂1 · d̂|Ai i
= 4
τ ~ ω1 − ωm
Nf Nm ,Am
(15.199)
2
hNf |E1− |Nm ihNm |E2− |Ni ihAf |ε̂1 · d̂|Am ihAm |ε̂2 · d̂|Ai i
+ δ(ωf − ω1 − ω2 ) .
ω2 − ωm
534 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
P
Generalizing to a statistical mixture of Fock states via Nf ,Ni pi |hNf |Ê2− Ê1− |Ni i =
Tr (ρ̂Ê1+ Ê2+ Ê2− Ê1− ), we may also write,
1 2πe4
= 4 |Mtpa |2 δ(ωf − ωi )Tr (ρ̂Ê1+ Ê2+ Ê2− Ê1− ) , (15.200)
τ ~
where
Mtpa ≡ hAf |ε̂2 · d̂G(Ei + ω1 )ε̂1 · d̂|Ai i + hAf |ε̂1 · d̂G(Ei + ω2 )ε̂2 · d̂|Ai i . (15.201)
For two incoming photons ω, we get ωf = Ef and ωi = Ei + ω1 + ω2 . Here, G denotes
the photon propagator,
X |Am ihAm |
G(ω) = . (15.202)
m
ω − Em
Figure 15.19: Feynman graphs for (a) absorption, (b) spontaneous emission, (c) stimulated
emission, (e)(i-iii) spontaneous Raman.
where
Mf i (ωn , ..., ω1 ) ≡ Sp hAf |ε̂n d̂G(Ei ± ω1 ± ... ± ωn−2 ± ωn−1 ) ... (15.206)
... ε̂n−1 d̂G(Ei ± ω1 ± ... ± ωn−2 ) ...
... ε̂1 · d̂G(Ei ± ω1 )|Ai i .
The upper signs hold for absorbed photons (up to n), the lower for emitted photons,
ωf = Ef + ω1 + ... + ωn and ωi = Ei + ω1 + ... + ωn .
For n-photon processes, there are theoretically up to n + 1 different types, each
one with n! possible temporal sequences, which can be illustrated in Feynman graphs.
Additionally, for every emitted photon can be either spontaneous or stimulated. The
number of possible time sequences is reduced, if some absorbed or emitted photons
have the same frequency by m! (if m is the number og identical photons).
For example there are (3+1) different types of three-photon processes: 3 photons
in, 2 photons in 1 out, 1 photon in 2 out, 3 photons out.
For example second harmonic generation: 3! possible time sequences, 2 processes,
2 equivalent photons = 3!22! possible Feynman graphs.
15.5.3 Exercises
15.5.3.1 Ex: Two-photon transitions in rubidium
Considering the following transitions of rubidium, 5S1/2 − 5P1/2 at 795 nm with
Γ795 = (2π) 6 MHz linewidth 5S1/2 − 5P3/2 at 780 nm with Γ780 = (2π) 6MHz, and
5P3/2 − 5D5/2 at 776 nm with Γ776 = (2π) 700 kHz linewidth. Calculate the rate
for resonant two-photon transitions from the ground state to the 5D5/2 level without
2
bothering about hyperfine splitting induced by a laser intensity of I = 1 mW/cm .
where â and b̂ are the field operators of two oscillator modes. A device providing such
a coupling is called beam splitter. It mixes two modes according to the Hamiltonian,
The beam splitter can be described in the Schrödinger or the Heisenberg picture
exploiting the formalism introduced in Sec. 3.5.
†
+↠b̂)
|ψ(t)i = eıΩt/2(âb̂ |ψ0 i . (15.208)
Figure 15.20: (a) Coupling of two cavity modes by insertion of a beam splitter. (b) Beam
splitter mixing two propagating modes.
we calculate,
¨ = − 1 Ω2 â ¨
â 4 and b̂ = − 41 Ω2 b̂ , (15.211)
15.6. BEAM SPLITTING AND QUANTUM AMPLIFICATION 537
Introducing the abbreviation η ≡ cos2 (Ωt/2), we can describe the evolution as,
! √ √ !
â(t) η −ı 1 − η â0
= √ √ . (15.213)
b̂(t) −ı 1 − η η b̂0
that each the reflected beam suffers a phase shift of π/2 11 . Resolve the Excs. 15.6.8.1
to 15.6.8.4.
where,
√
0 1nb p
√1n 0 2(nb − 1)
b
p ..
~ .
Ĥa+b = 2Ω 2(nb − 1) . (15.219)
.. √
. na 1
√
na 1 0
It is easily verified that the matrix of eigenvectors and the matrix of eigenvalues,
1 0 0 0 0 0 0 0 0 0 0 0
0
−1 1 0 0 0
0
−1 0 0 0 0
0 1 1 0 0 0 0 0 1 0 0 0
U =
√ √ respectively E = 2 Ω
~
0 0 0 − 21 2 −1 1
2 0 0 0 −2 0 0
2
0 0 0 1 0 1 0 0 0 0 0 0
√ √
0 0 0 − 12 2 1 1
2
2 0 0 0 0 0 2
13 A more in-depth discussion is found in Ref. [444].
15.6. BEAM SPLITTING AND QUANTUM AMPLIFICATION 539
satisfy U −1 HU = E. Hence,
The result (15.220) shows that, ignoring (tracing over) one of the modes, the other
mode automatically becomes a Glauber state 14 .
14 This does not mean that the output modes are uncorrelated. For instance, if we measure n
photons in one mode, we know that the other must contain exactly N − n photons. However, the
correlations are so dense in phase space that they are not resolvable or decohere rapidly. In fact, it
can be shown that coherent states are the only pure states that produce uncorrelated outputs when
mixed by a passive linear-optics device [8, 7].
540 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
Besides that,
n N −1
1 X N 2 N X N N (N + 1)
hn̂21 i = n = (n + 1) = (15.223)
2n n=0 n 2n n=0 n 4
s
n
1 X N N (N − 1)
hn̂1 n̂2 i = n n(N − n) = N hn̂1 i − hn̂21 i =
2 n=0 n 4
The squeezing parameter is,
σ 2 (n̂1 − n̂2 ) hn̂2 − 2n̂1 n̂2 + n̂22 i 2hn̂21 i − 2hn̂1 n̂2 i
ξ12 = = 1 = =1. (15.224)
hn̂1 i + hn̂2 i N N
The correlation functions at equal times are,
D E
â†1 â†1 â1 â1 n
1 1 X N hn̂2 i − hn̂1 i
g11 = 2
= 2 n
n(n − 1) = 1 =1 (15.225)
hn̂1 i hn̂1 i 2 n=2 n hn̂1 i2
D E
â†1 â†2 â2 â1 N
1 1 X N N hn̂1 i − hn̂21 i N −1
g12 = = 2 n
n(N − n) = 2
=
hn̂1 ihn̂2 i hn̂1 i 2 n=0 n hn̂1 i N
The Cauchy-Schwarz inequality and the quantum inequality are both met,
√
g12 ≤ g11 g22 (15.226)
s
1 1
g12 ≤ g11 + g22 +
hn̂1 i hn̂2 i
In comparison, a Glauber state is normally divided,
2 X αn 2 X αm
|ψi = |α1 i|α2 i = e−|α1 | /2 √ 1 |nie−|α2 | /2 √ 2 |mi (15.227)
n n! m m!
2 2 X αn αm
= e−|α1 | /2−|α2 | /2 √ 1 √2 |ni|mi .
n,m n! m!
ĉ = √1 (â
2
+ b̂) , dˆ = √1 (â
2
− b̂) . (15.231)
Detected by photodetectors with the gain coefficient g, these two beams produce
currents,
Iˆc = gĉ† ĉ , Iˆd = g dˆ† dˆ . (15.232)
Iˆ+ ≡ Iˆc + Iˆd = g(↠â + b̂† b̂) , Iˆ− ≡ Iˆc − Iˆd = g(↠b̂ + b̂† â) . (15.233)
and,
Iˆ+
2
= g 2 [n̂2a +n̂2b +2n̂a n̂b ] , Iˆ−
2
= g 2 [(↠b̂)2 +(b̂† â)2 +↠â+b̂b̂† +â↠+b̂† b̂] . (15.234)
ˆ 2 i ≡ hIˆ2 i − hIi
Now, with the definition, h(∆I) ˆ 2 , we get the intensity noise of the field,
h∆Iˆ+
2
i = h(∆Iˆ+ )2 i = g 2 h(∆n̂a )2 i , (15.236)
The amplification depends on the mean value hâi of the annihilation operator and
its dispersion h↠âi − h↠ihâi. A coherent state is a state with minimal uncertainty;
when the state is transformed, the uncertainty may increase. This increase can be
interpreted as noise in the amplifier. The gain G can be defined as follows:
hâiout
G= . (15.239)
hâiin
The quantum amplifier can also be described in the Heisenberg picture; the
changes are attributed to the amplification of the field operator. Thus, the evolu-
tion of the operator âout is given by
while the state vector remains unchanged. The gain is then given by,
hâout iin
G= . (15.241)
hâiin
In general, the gain G may be complex, and it may depend on the initial state. For
laser applications, the amplification of coherent states is important. Therefore, it is
15.6. BEAM SPLITTING AND QUANTUM AMPLIFICATION 543
usually assumed that the initial state is a coherent state characterized by a complex-
valued initial parameter α, such that |ini = |αi. Even with such a restriction, the
gain may depend on the amplitude or phase of the initial field.
In the following, the Heisenberg representation is used; all brackets are assumed
to be evaluated with respect to the initial coherent state,
noise = hâ†out âout i − hâ†out ihâout i − (h↠âi − h↠ihâi) . (15.242)
The expectation values are assumed to be evaluated with respect to the initial coherent
state. This quantity characterizes the increase of the uncertainty of the field due
to amplification. As the uncertainty of the field operator does not depend on its
parameter, the quantity above shows how much output field differs from a coherent
state.
where c and s are c-numbers and b̂† is a creation operator characterizing the amplifier.
Without loss of generality, it may be assumed that c and s are real. The commutator
of the field operators is invariant under unitary transformation U :
[âout , â†out ] = 1 = [â, ↠] , [âin , â†out ] = 0 = [âout , â†in ] . (15.244)
From the unitarity of U , it follows that b̂ satisfies the same commutation relations.
The c-numbers are then c2 − s2 = 1. Hence, the phase-invariant amplifier acts by
introducing an additional mode to the field, with a large amount of stored energy,
behaving as a boson. Calculating the gain and the noise of this amplifier, one finds
G = c, and
noise = c2 − 1 . (15.245)
The coefficient g = |G|2 is sometimes called the intensity amplification coefficient.
The noise of the linear phase-invariant amplifier is given by g − 1. The gain can be
dropped by splitting the beam; the estimate above gives the minimal possible noise of
the linear phase-invariant amplifier. The linear amplifier has an advantage over the
multi-mode amplifier: if several modes of a linear amplifier are amplified by the same
factor, the noise in each mode is determined independently;that is, modes in a linear
quantum amplifier are independent.
To obtain a large amplification coefficient with minimal noise, one may use ho-
modyne detection, constructing a field state with known amplitude and phase, corre-
sponding to the linear phase-invariant amplifier. The uncertainty principle sets the
lower bound of quantum noise in an amplifier. In particular, the output of a laser
system and the output of an optical generator are not coherent states.
The multiplicative amplifier D also adds additive noise F . We have DD† = 1,
! !
â†out â†in F1
=D + . (15.246)
âout âin F2
544 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
where the field mode is expressed by the Hermitian quadrature components â0 =
2−1/2 · (x̂θ + ıŷθ ). The expectation value of ∆Jˆ is afflicted with the Heisenberg
uncertainty and can be expressed as the first moment of the Wigner function W (α):
Z
ˆ
hψ|∆J|ψi = W|ψi (α)∆Jd ˆ 2α (15.249)
√ Z √ Z ∞
= 2|αLO | W|ψi (xθ , pθ )xθ dxθ dpθ = 2|αLO | wθ (xθ )xθ dxθ .
−∞
This is called the radon transform. The distribution function wθ (pθ ) as well as the
Wigner function are normalized to 1. Multiple measurements of the expectation value
xθ = hψ|x̂θR|ψi now yields a histogram H|ψi (xθ ) reflecting, if normalized, w|ψi (xθ ) =
H|ψi (xθ )/ H|ψi (xθ ) exactly the distribution function.
Considering the finite detector efficiency [444] wθ (xθ ) must be generalized to a
convolution with an apparatus function ζ(x):
√ 1 2
wθreal ( ηxθ ) = (wθideal ? ζ) where ζ(x) = p e−ηα /(1−η)
. (15.251)
π(1 − η)
A finite detector efficiency degrades the contrast of the quantum interference struc-
tures.
With the procedure of optical homodyne tomography or quantum state endoscopy
the Wigner function for e.g. a Schrödinger
R cat state can be reconstructed from a set
of distribution functions wθ (xθ ) = W (αeıθ )dpθ measured for various phases θ [444].
To do this the data set is exposed to an inverse radon transform:
Z ∞ Z πZ ∞
1 −1/2
αe−ıθ )−xθ ]
W (α) = wθ (xθ )|ζ|eıζ [Re (2 dxdθdζ . (15.252)
4π 2 −∞ 0 −∞
15.6. BEAM SPLITTING AND QUANTUM AMPLIFICATION 545
if the local oscillator can be considered as classical, α = |α|eıθ . I.e. the phase of the
local oscillator permits us to select either one of the quadrature components.
σ 2 (ni − hnj i)
ξi,j = . (15.262)
hni i + hnj i
as
Ĥ = 12 (p̂2a + ω 2 x̂2a ) + 12 (p̂2b + ω 2 x̂2b ) + Ωωx̂a x̂b . (15.265)
√
Setting ω± = ω 2 ± Ωω allows us to rewrite the Hamiltonian as,
Ĥ = 41 (p̂a + p̂b )2 + 14 ω+
2
(x̂a + x̂b )2 + 14 (p̂a − p̂b )2 + 41 ω−
2
(x̂a − x̂b )2 . (15.266)
−1
In this approximation, we find TAB MTab ' M and the transformation defined in
(15.267),
 â
† â†
=M , (15.276)
B̂ b̂
B̂ † b̂†
15
takes us directly from the Hamiltonian (15.275) to ,
† †
Ĥ = (~ω + 2 )(Â Â
~Ω
+ 21 ) + (~ω − 2 )(B̂ B̂
~Ω
+ 21 ) . (15.277)
15.6.8 Exercises
15.6.8.1 Ex: Conservation law at a beam splitter 1
Show that the beam-split transformation
! !
ĉ t −r â
=
dˆ r t b̂
= ~ω(†  + 21 ) + ~ω(B̂ † B̂ + 21 ) + ~Ω
2
(−ÂB̂ † + † B̂) .
15.6. BEAM SPLITTING AND QUANTUM AMPLIFICATION 549
derive the Hamiltonian in terms of the quadrature components x̂a,b and p̂a,b .
b. Set up the Heisenberg equations for the field operators â, ↠, b̂, and b̂† and determine
the eigenvalues of the matrix X defined by,
â â
d â
† â†
=X .
dt b̂ b̂
b̂† b̂†
c. Extend the Heisenberg equations by including loss terms for the cavity modes κa
and κb and determine the corresponding matrix Xκ , as well as its eigenvalues for the
simplifying case Ω0 ≡ Ω.
550 CHAPTER 15. ATOMS IN QUANTIZED RADIATION FIELDS
J.P. Bartolotta et al., Entropy transfer from a quantum particle to a classical coherent
light field [DOI]
R.G. DeVoe et al., Observation and Superradiant and Subradiant Spontaneous Emis-
sion of Two Trapped Ions [DOI]
R. Dum et al., Monte Carlo simulation of master equations in quantum optics for
vacuum, thermal, and squeezed reservoirs [DOI]
J. Eschner et al., Light interference from single atoms and their mirror images [DOI]
D.J. Heinzen et al., Enhanced and Inhibited Visible Spontaneous Emission by Atoms
in a Confocal Resonator [DOI]
D.J. Heinzen et al., Vacuum Radiative Level Shift and Spontaneous-Emission Linewidth
of an Atom in an Optical Resonator [DOI]
B.R. Mollow et al., Power spectrum of light scattered by two-level systems [DOI]
15.7. FURTHER READING 551
Quantum measurement
Since its foundation, the theory of quantum mechanics was driven by the urge to
clarify the relationship between the world and what we can learn about it, that is, be-
tween reality and the observer. Scientists such as Bohr, Heisenberg, Schrödinger, and
Einstein defended controversial positions and struggled for the correct interpretation
of quantum mechanics. The measurement process is supposed to provide informa-
tion about the world out there, but it is not clear whether this information can be
complete and accurate, or whether there are limitations or hidden variables. Also, it
was unclear, to what extend a measurement can be non-invasive or whether it would
always perturb the phenomenon under investigation. The most important step in
this question was the Copenhagen interpretation formulated by Bohr, Heisenberg and
Born in 1927 and elaborated later by von Neumann and Dirac. Although contested
many times in the past, it’s essence still remains valid today.
In this chapter we will study the measurement process from the viewpoint quan-
tum mechanics and discuss some seemingly paradoxical effects, that will allow us to
deepen our understanding. Among them are the quantum jump, Schrödinger’s cat,
the quantum Zeno effect, and the Einstein-Podolski-Rosen paradox.
to predict specific events, without having to assume that these events occur in a random manner.
553
554 CHAPTER 16. QUANTUM MEASUREMENT
through our classical concept of the world. It is, thus, not surprising that one of
the most fascinating areas of investigations is the interface between the classical
and the quantum, the macroscopic and the microscopic worlds. For the pioneers of
quantum mechanics the most important questions were of the type: ’How is it possible
that a microscopic particle flies simultaneously through two slits?’ Nowadays, we
are accustomed to such paradoxes, and we simply accept the fact that we have to
consider a particle as a wave. Nevertheless, we still do not understand very well,
why the classical and the quantum world behave so differently. ’Why does quantum
mechanics allow for quantum superpositions states, which are absolutely forbidden in
classical physics?’, ’Why are the fundamental laws of the quantum world invariant to
the arrow of time, while the macroscopic world always evolves from the past to the
future?, ’How can it be that quantum mechanics allows for effects having no cause,
like spontaneous emission, while the everyday world seems to be deterministic?’
Quantum mechanics must, in some limit, clearly encompass classical physics. But
in spite of Ehrenfest’s correspondence principle, this fact is far from being trivial.
Some predictions of classical and quantum physics are fundamentally different and,
in some cases, even contradictory. The famous Schrödinger cat states are the epitome
of this fact: In one version of this paradox, a particle crosses a double slit. Behind
one of the slits is a detector which, as soon as it registers a particle, actuates a device
killing a cat. We know that in quantum reality the particle crosses both slits in a
superposition state, so that the cat should be in a superposition state as well. Hence,
quantum cats can be in a superposition of ’dead’ and ’alive’.
We believe nowadays that the answers to the above questions are somehow buried
in processes that destroy the quantum superposition of Schrödinger cats during the
transition from the microscopic to the macroscopic world. However, the details of
these quantum coherence destruction processes, called decoherence, are very compli-
cated and the subject of serious efforts in contemporary research. It is one of the mo-
tivations for trying to create in laboratories the largest possible (quasi-macroscopic)
quantum systems, bring them in Schrödinger cat-like superposition states and study
16.1. THE OBSERVER AND THE REALITY 555
Figure 16.2: The measurement of a quantum system presupposes the interaction of the
system with a reservoir, which disturbs its dynamics.
their decoherence 2 .
allows to calculate the time evolution, that is, the trajectory of the wavefunction. The
evolution is coherent and reversible in time.
Now, the process of measuring a pure quantum state includes –according the
Copenhagen interpretation and as discussed in Sec. 2.2.7– two consecutive steps: In
the first step, the interaction of the quantum sample with the measuring device (which
from now on we will call meter) destroys all coherences and projects the pure state
into a statistical mixture of eigenstates of the meter. Following von Neumann, the
impact of the meter on the quantum system is so strong, that its coherent evolution
is interrupted and it is projected onto the degree of freedom that the apparatus
wants to measure, e.g. its position or its momentum, but not both in the same time.
The projection transforms a pure quantum state |ψi into a statistical mixture ρ of
eigenstates, X
ρ̂sample = |ψ(t)ihψ(t)| y ρ̂proj = |hψ|ki|2 |kihk| . (16.2)
k
This process is irreversible, that is, it separates the past from the future. The projec-
tion is not described by the Schrödinger equation. Instead, the sudden reduction of
the state must be postulated, as done by von Neumann’s famous axiom.
In a second step, the observer looks at the measuring device and confirms one of
the possible results. Thus, he transforms the state into a eigenstate of the device 3 :
From this moment, we can again leave the quantum system alone until the next
measure.
2 There are attempts to introduce the concept of the time arrow also in the microscopic world:
’In an isolated system, spontaneous processes occur in the direction of increasing entropy.’ [502].
3 We note that, only if all commuting observables of the system are measured and acknowledged,
From the viewpoint of the quantum system, the evolution of the measurement
process appears discontinuous, because it destroys all possible coherences between its
states. In fact, the problem comes from the non-ideal behavior of the measuring device
(symbolized by | ↑i before the measurement). An ideal non-invasive measurement 4
would leave the quantum state |ψi unchanged:
H
|ψi| ↑i −→ |ψi| %i , (16.4)
while the measuring device changes to a state (| %i after the measurement) indicating
the current state of the system. However, this is normally impossible without previ-
ously established correlation between |ψi and | ↑i. In a real meter device, the coupling
between |ψi and | ↑i requires that the meter and the system to be non-orthogonal.
Ideally, the system evolves independently without being disturbed by the meter. Un-
fortunately, this also means that the meter evolves independently, that is, it is not
influenced by the system and thus does not provide information about the system.
To allow a transfer of information, we need to couple the respective spaces by an
interaction Ω, such that,
sample Ω
Ĥ = . (16.7)
Ω meter
Tracing over all degrees of freedom of the universe except those of the quantum
system, the von Neumann equation (16.5) turns into a master equation,
Figure 16.3: Spontaneous emission can be seen as a coupling of the system under investiga-
tion to an external meter, because it delivers information to the meter, even if only to tell
us: ’The system was in an excited state, but now it’s in a ground state.’
Ĥ = | ↓ih↓ | ⊗ | ↑ih↑ | .
The discussion about the correct interpretation of the measurement process is still
ongoing. Modern theories describe the state reduction in terms of quantum decoher-
ence due to interactions of the system with the environment. Other interpretations
involve decoherent histories or assume multiple worlds [553]. On the practical side,
the current interest in quantum decoherence is motivated by the fact that this phe-
nomenon may turn out to be the fundamental factor limiting the useful operation
of quantum computers. Another interesting area where quantum mechanics meets
classical physics is the phenomenon of quantum chaos.
the following is, that this three-level system, called quantum amplifier, gives a deep
insight into what happens during the process of state reduction and, therefore, can
be considered as paradigmatic for theories on quantum measurement.
To be able to discuss the dynamics of this system on a firmer ground, we will first
introduce the quantum Monte Carlo wavefunction simulation method (MCWF).
aiming at including energy dissipation processes. The problem with this Hamiltonian
†
is that, for being non-hermitian, [Ĥef f , Ĥef f ] 6= 0, it also generates a non-unitary
†
dynamics, e−ıĤef f t 6= eıĤef f t . This means that the mere possibility of spontaneous
emission prevents the reversibility of the dynamics. We observe a temporal decrease
of the norm hψ(t)|ψ(t)i indicating a loss of energy,
The loss of normalization during the evolution, until the next quantum jump occurs,
is due to the dissipation of energy toward the reservoir,
ρsample 0
Tr ρsample → 0 while Tr =1, (16.11)
0 ρreserv
and represents a measure of the probability that an irreversible process has occurred
during the evolution time.
Dissipative processes can be simulated by playing dices with random numbers ζ.
We divide time into small intervals dt and propagate the wavefunction from ψ(t)
to ψ(t + dt). After each interval we evaluate the probability p = 1 − hψ(t)|ψ(t)i
16.1. THE OBSERVER AND THE REALITY 559
accumulated during the time period [0, t + dt] that a dissipative process (such as
spontaneous emission) has occurred. Now, we generate a random number ζ, uniformly
distributed between 0 and 1, which we compare to probability the probability p. In
case, ζ > 1 − hψ(t)|ψ(t)i, we conclude that there was no dissipative process, and we
let the system proceed in peace, only renormalizing the wavefunction to compensate
for the losses [500, ?]. Otherwise, if ζ < 1 − hψ(t)|ψ(t)i, we conclude that there
was a dissipative process, and the system is projected into the eigenstate ψ0 . This
projection is abrupt and called quantum jump. Now, the evolution restarts from zero,
ruled by the effective Hamiltonian. The simulation implemented via,
(1−ıĤdt)|ψ(t+dt)i
!
√ if ζ > 1 − hψ(t)|ψ(t)i
|ψ(t)i y |ψ(t + dt)i ≡ hψ(t)|ψ(t)i . (16.12)
|ψ0 i if ζ < 1 − hψ(t)|ψ(t)i
1 1
ρ11
ρ11
0.5 0.5
0 0
0 100 200 0 100 200
t (ns) t (ns)
As said above, the inclusion of the meter in a global theory is, in general, difficult.
For this reason, to perform the comparison, we choose the simplest imaginable system:
the three-level system with two transitions connecting to a common ground state and
excited by radiation fields. As shown in Fig. 16.6(a) this three-level system can be
an atom with a strong transition and a weak transition, for example, the dipolar
transition S1/2 − P1/2 and the forbidden quadrupolar transition S1/2 − D5/2 in a
single Ba+ ion. We will now name the ’strong transition’ as meter and the ’weak
transition’ as sample and show that this system allows to study the von Neumann
measurement process including the direct observation of quantum jumps 5 . At the
same time, the system is simple enough for a complete theoretical description. In this
sense, the three-level system becomes the epitome of a quantum measurement device.
We turn our attention to the three-level atom: Obviously, the atom will prefer-
entially scatters photons on its strong dipolar transition. However, at times when
the valence electron is ’shelved’ in the metastable state excited by the quadrupole
transition, no fluorescence can be observed on the strong transition.
Figure 16.6: (a) Quantum measurement at the example of a three-level atom incorporating
a weak (sample) transition and a strong (meter) transition. (b) Random Telegraph signal in
the resonance fluorescence due to quantum jumps.
Quantum jumps were experimentally observed in single trapped ions, whose lowest
energies form a three-level systems [532, 624, 625, 69].
object of long-standing debates: ’If we have to go on with these damned quantum jumps, then I’m
sorry that I ever got involved with quantum mechanics.’ [637].
6 The absence of the ’meter’ laser during the ’sample’ stage avoids saturation broadening and
light-shifts of the ground state. Since the ground state is shared by both transitions, its broadening
would reduce the spectral overlap between the ’sample’ transition and the driving laser and therefore
the probability to excite the metastable level. This inhibition of the coherent dynamic by too strong
or too frequent measurements is known as quantum Zeno effect: The more an observer tries to
16.1. THE OBSERVER AND THE REALITY 561
where the Rabi frequencies Ωsd and Ωsd are switched on alternately.
In the simulation 16.7 the quantum jumps to the shelved metastable state D5/2
appear as long periods without population in the P1/2 level (first period S − P , where
the population of S1/2 , illustrated by the red curve, gradually tends to 0 for long
times). The reduction of the system to the shelved state actually occurs by non-
observation of fluorescence on the strong transition. The projection needs a finite
time, simply because we can not be sure whether the non-observation is actually due
to shelving or the incidental absence of scattering events on the S − P transition:
After all, it is not predictable, when the next photon will be spontaneously emitted,
even though the lifetime of the excited state is short. But for longer observation times
it becomes increasingly unlikely that the absence of photons is not due to shelving. It
is this unlikeliness, which lets the population rapidly converge towards the metastable
state. In the second S − P period, Fig. 16.7 shows fast transitions to the P1/2 fol-
lowed by sudden decays to the ground state. These processes correspond to photon
absorption and spontaneous reemission by the strong transition. The succession of
the photon scattering events is so fast, that the signal recorded by photodetectors
appears as a continuous fluorescence. The sudden transitions between bright and
dark periods shown in Fig. 16.6(b), which occur totally randomly, are interpreted as
quantum jumps.
D5/2
0.5
1
S1/2
0.5
1
P1/2
0.5
0
0 200 400 600 800 1000
t (ns)
Figure 16.7: (code) Quantum amplifier comparing Monte-Carlo quantum simulations (red
lines) and Bloch equations (black lines), which coincide well with the averages over 100
Monte-Carlo trajectories (green lines). The periods S − D (white background) represent the
free evolution of the quantum system, the periods S − P (yellow background) represent the
measurement periods.
vacuum modes receiving the spontaneously emitted photons by tracing over the latter
ones.
We will not be able to handle all degrees of freedom. However, we can generalize
the Bloch equations in the following way [511, 774]. We project the total density
operator ρAF R of the atom plus the driving field plus the reservoir of vacuum modes
into the subspace of the atom and the driving field consisting of exactly n photons,
and derive, from the von Neumann equation, the master equation for the atomic state
ρ(n) under the constraint of a fixed number of photons n in the field. The master
equation differs only in one term from the usual Bloch equations: The expression
(n)
Γ12 ρ22 , which describes the spontaneous decay of the population of the excited state
(n−1)
of the meter transition, is replaced by the expression Γ12 ρ22 :
d (n)
ρ = (L|1iΓ12 h2|)ρ(n) + |1iΓ12 h2|ρ(n−1) , (16.15)
dt
P (n)
The substate with of n photons violates the trace condition, j ρjj 6= 1. The
physical explanation for this is the following: While induced emission and absorption
maintain the number of photons in the combined light-atom system (like in the Jaynes-
Cummings model), spontaneous emission decreases the number of photons, leading
to an irreversible loss of energy. The quantum jump observed in MCWF model
16.1. THE OBSERVER AND THE REALITY 563
X4
(n) d (n)
c(t) = ηΓ12 ρ̃22 (t) = η ρ̃ (t) , (16.16)
j=1
dt jj
The second step immediately follows from the homogeneous part of equation (16.15).
As a canonical example of a Kraus operator [48, 129] we take Ĥ = x̂ ⊗ p̂, where the
position and the momentum satisfy the commutation relation, [x̂, p̂] = i. The initial
state of the ancilla be a Gaussian distribution,
Z
1 02
/4σ 2
|φi = dq 0 e−q |q 0 i . (16.17)
(2πσ 2 )1/4
1 2 2
φ(q) = hq|φi = 2 1/4
e−q /4σ . (16.18)
(2πσ )
The Kraus operators are (compared to the previous discussion, we now let λ = 1),
1 2 2
M (q) = hq|e−ıx̂⊗p̂ |φi = 2 1/4
e−(q−x) /4σ , (16.19)
(2πσ )
since the operator e−ıx̂⊗p̂ makes a spatial translation when applied to the degree of
freedom of the position. The corresponding POVM elements are,
1 2
/2σ 2
E(q) = Mq† Mq = √ e−(q−x) , (16.20)
2πσ 2
R
which obey dqE(q) = I.
Calculate hψq |ψq i = hψ 0 |M (q)† M (q)|ψ 0 i.
Note that limσ→0 E(q) = |x = qihx = q|. That is, in a particular limit, these
operators converge to a strong measurement of position. For σ → ∞, we speak of
weak measurement.
Another example would be the three-level atom of Dehmelt’s quantum amplifier.
564 CHAPTER 16. QUANTUM MEASUREMENT
16.1.4 Exercises
16.1.4.1 Ex: Dispersive quantum jumps
Consider a three-level system in V-configuration, as depicted in Fig. 14.6(b), with an
unstable state |1i, a ground state 2i, and a metastable state |3i. Discuss whether the
atom shelved in state |3i is sensitive to light-shift and power broadened induced by
a laser resonant to the |1i − |2i transition. E.g. will power-broadening only effect the
transition rate |1i −→ |3i or also |3i −→ |1i?
interaction V between the system and the bath. The purpose of the master equation
is to find the dynamics of the system by tracing over all degrees of freedom of the bath.
This is not always possible, and we will assume that the interaction V is sufficiently
weak, so that perturbation theory is applicable.
In the interaction representation the evolution of the total density matrix ρ̂tot
becomes,
dρ̃tot
i~ = [Ṽ (t), ρ̃tot ] . (16.21)
dt
where ρ̃tot (t) ≡ U0† ρtot U0 and Ṽ (t) ≡ U0† V U0 and U0 = e−ı(Ĥsys +Ĥres )t/~ . This
evolution is, for the time being, very general, and the solution can be formally written,
Z
1 t
ρ̃tot (t) = ρ̃tot (0) + dt1 [Ṽ (t1 ), ρ̃tot (t1 )] . (16.22)
i~ 0
Iterating once again:
Z Z t Z t1 h i
1 t 1
ρ̃tot (t) = ρ̃tot (0)+ dt1 [Ṽ (t1 ), ρ̃tot (0)]+ 2
dt1 dt2 Ṽ (t1 ), [Ṽ (t2 ), ρ̃tot (t2 ] .
i~ 0 (ı~) 0 0
(16.23)
In the following, we will call several approximations to simplify the calculations, in
particular, the Born approximation, the assumption that the initial state is a product
state and, later-on, the Markov approximation.
This way of terminating an iterative equation by ρtot (0) is generally known as the
Born approximation. Here, we will just go to second order in Ṽ . Tracing over the
bath,
ρ̃sys (t) = Trres ρ̃tot (t) , (16.25)
we extract the density matrix for only the system,
Z
1 t
ρ̃sys (t) = ρ̃sys (0) + dt1 Trres [Ṽ (t1 ), ρ̃tot (0)] (16.26)
i~ 0
Z t Z t1 h i
1
+ 2
dt1 dt2 Trres Ṽ (t1 ), [Ṽ (t2 ), ρ̃tot (0)] .
(ı~) 0 0
Another assumption, which is not essential but often valid, is that Trres [Ṽ (t1 ), ρ̃tot (0)] =
0. If this is the case, then the first-order term will vanish. In second order, we can
write,
is a superoperator acting on the operator density of the system. Taking the temporal
derivative, we have the explicit master equation,
dρ̃sys (t) d
dt = Lρsys (t) = dt (M(t)ρ̃sys (t))
Rt h i . (16.29)
1
= (ı~)2 0 dτ Tr res Ṽ (t), [Ṽ (τ ), ρ̃ sys (t) ⊗ ρ res ]
In other words, the two-point correlation function is significant only, when t ' τ , and
it is valid to extend the upper bound to infinity. This is the Markov approximation.
P −ıωk t
where Γ̂(t) = k gk b̂k e , the bosonic operators â and b̂k act, respectively, on
the system (with the frequency Ω) and the bath (with the frequency ωk ). Here, gk
16.2. OPEN SYSTEMS AND THE MASTER EQUATION 567
characterizes the coupling force between the oscillators of the system and the bath.
Hence,
a† Γ(t)eıΩt + aΓ† (t)e−ıΩt a† Γ(τ )eıΩτ + aΓ† (τ )e−ıΩτ ρ̃(t) ⊗ ρres
Z t
dρ̃sys − a† Γ(t)eıΩt + aΓ† (t)e−ıΩt ρ̃(t) ⊗ ρres a† Γ(τ )eıΩτ + aΓ† (τ )e−ıΩτ
=− dτ Trres .
dt 0
− a† Γ(τ )eıΩτ + aΓ† (τ )e−ıΩτ ρ̃(t) ⊗ ρres a† Γ(t)eıΩt + aΓ† (t)e−iΩt
+ρ̃(t) ⊗ ρres a† Γ(τ )eıΩτ + aΓ† (τ )e−ıΩτ a† Γ(t)eiΩt + aΓ† (t)e−ıΩt
(16.33)
Let’s take a closer look at one of the terms,
Z t
T ≡− dτ Trres a† Γ(t)eıΩt aΓ† (τ )e−ıΩτ ρ̃(t) ⊗ ρres (16.34)
0
Z t
= −a† aρ̃(t) dτ Γ(t)Γ† (τ ) res eıΩt e−ıΩτ .
0
where P V denotes Cauchy part of the principal value. These correspond to a ’Lamb
shift’ and a ’Stark shift’ of the frequency, which are considered to be small in com-
parison to Ω and should be neglected here,
X Z
−1 ∞
T = −a† aρ̃(t) gk2 eβ~ωk − 1 dτ eı(ωk −Ω)(t−τ ) (16.38)
k 0
X −1
= −a† aρ̃(t) gk2 eβ~ωk − 1 πδ(ωk − Ω)
k
X −1 X γ
= −a† aρ̃(t) gk2 eβ~Ω − 1 π = −πN a† aρ̃(t) gk2 δ(ωk − Ω) = N aa† ρ̃(t)π .
2
k k
−1 P
where N ≡ eβ~Ω − 1 . We define γ2 ≡ k gk2 δ(ωk − Ω). The procedure can be
repeated for all terms in the master equation. We then obtain the master equation
for a damped harmonic oscillator,
dρ̃ γ γ
= (N + 1) 2aρ̃a† − a† aρ̃ − ρ̃a† a − N 2a† ρ̃a − aa† ρ̃ − ρ̃aa† . (16.39)
dt 2 2
568 CHAPTER 16. QUANTUM MEASUREMENT
d X
= L†  = ~ı [Ĥ, Â] + γk L̂†k ÂL̂k − 12 L̂†k L̂k  − 12 ÂL̂†k L̂k ,
dt
k
as we will show in Exc. 16.2.6.2. The former stresses the Schrödinger picture and
the latter the Heisenberg picture. In particular, the dissipative Lindbladt terms are
equivalent.
16.2.6 Exercises
16.2.6.1 Ex: Master equation for cavities
Be κ the loss rate for photons by transmission through the mirrors of a cavity. Show
that the master equation gives,
dρ̂ κ κ
= (n̄ + 1) 2âρ̂↠− ↠âρ̂ − ρ̂↠â − n̄ 2↠ρ̂â − â↠ρ̂ − ρ̂â↠,
dt 2 2
P
where the density operator is defined by ρ ≡ m,n Pm,n |mihn| and n̄ is the number
of thermal photons.
Interestingly,Introdução
Zeno’s problem has a counterpart in quantum mechanics. Let us do
the following Gedankenexperiment: A laser beam passes through a dense series of n
A evolução de um sistema instávelπ é notoriamente governada por três tempos distintos: o tempo
polarizers, each curto,
one being rotated by an angle 2n with
onde predomina uma função quadrática,
respect to the preceding one.
o tempo intermediário, onde temos propriamente a
Each polarizer performs a local measurement of beam polarization.
governança do padrão exponencial e o tempo longo, onde a função The result
potência of this
prevalece.
arrangement is that, in the limit of an infinitely dense series, the continuous measure-
ment of the system completely
A equação degoverns its nos
Schrödinger evolution and rotates
leva inevitavelmente the polarization
a termos em tempos curtos,by esse padrão é
9
an angle of π/2 predominado
. por uma função quadrática que foi batizada por Misra e Sudarshan em 1977 por região
“Zeno”,
In every version of em
thealusão
Zeno ao effect,
famoso filósofo Zenão de
the system is Eleia que propôs
inhibited fromo paradoxo
evolving da freely,
flecha.
because of too frequent Zenão propôs vários paradoxos,
measurements no caso state.
of its current específico da flecha,would
Achilles ele diz que se uma
surely be flecha em voo
able to overtake instantaneamente
the turtle, if he ocupa
did sempre o seu espaço,
not always checkeon queher
algoto
parado também
assess ocupa sempre o seu mesmo
the remaining
distance [369] 10espaço,
. então uma flecha em voo em qualquer instante também está em repouso.
hψ0 |ψ(t)i = hψ0 |e−ıĤt/~ |ψ0 i and P (t) = |hψ0 |ψ(t)i|2 . (16.41)
of an atom.
10 The quantum Zeno effect was often used to justify the physical relevance of the state reduction
postulate. It was shown, however, that this postulate is not essential for understanding the quantum
Zeno effect [57]. The effect already follows directly from the Schrödinger equation and therefore
has a purely dynamical nature. This shows that the projection is a purely mathematical construct
without physical reality (see Sec. 16.1.2).
570 CHAPTER 16. QUANTUM MEASUREMENT
whether the system is still in its initial state, but each measurement projects our
system back to its initial state, from where it has to start the transition process from
scratch. Under these conditions, the population of the initial state will evolve like,
P (N ) (τ ) = P (N ) (t/N )N . (16.44)
Fig. 16.9 shows the evolution for five measurements separated by time intervals τ . In
comparison with the evolution, when no measurements are taken (dashed line), the
evolution with measurements inhibits the depletion of the initial state. Extrapolating
the number of measurements to infinity, the probability (16.44) converges to 1,
" 2 #N
t N large −t2 /N τ 2 N →∞
1− −→ e Z −→ 1 . (16.45)
N τZ
Podemos ilustrar esse efeito com um gráfico.
See Excs. 16.3.4.1 and 16.3.4.2.
Onde os estados + e – são descritos pelas matrizes de Pauli e σ1,2,3 são auto estados.
16.3. REPEATED MEASUREMENTS 571
where P̂± is the projection operator. The result of a measurement of the population
is afflicted by an inherent uncertainty expressed by the variance,
In other words, the random projection of the system on the eigenstate basis induces
a noise called quantum projection noise 12 . This noise inhibits the determination
of the probabilities p± in a single measurement. On the other hand, by measuring
populations on a sample of N atoms or by repeating the measurement N times with a
single atom under identical conditions, we can reduce the uncertainty. The probability
of finding an atom r times in the state |+i is [368],
N r
PN,r,+ = p (1 − p+ )N −r . (16.48)
r +
The expectation value and variance of this binomial distribution are [368],
N
X N
X
r̄ = rPN,r,+ = N p+ , (∆r)2 = (r−N p+ )2 PN,r,+ = N p+ (1−p+ ) . (16.49)
r=0 r=0
Thus, the standard deviation decreases with the number of atoms or measurements,
∆r 1
σ= ∼√ . (16.50)
r̄ N
The increase of knowledge on the population of a two-level system by repeated
measurements can be illustrated by a simple simulation exhibited in Fig. 16.10(a).
Experiments are performed by comparing p2 with a random number η. Two outcomes
are possible r = 0, 1. The histograms narrow as the number of experiments increases.
In practice, a measurement is performed using the quantum amplifier method illus-
trated in Fig. 16.6 [360].
but we note that a discussion based on statistical mixtures described by density matrices gives the
same results.
12 Projection noise can be interpreted as shot noise. However, the optical shot noise in photode-
tectors is generated by the repartition of the field energy into discrete photons, the projection noise
is the consequence of the discretization of the electronic excitation levels.
572 CHAPTER 16. QUANTUM MEASUREMENT
(a) 10 (b) 1
N = 100
N = 80
PN,r,+ (Gt)
PN,r,+ (p+ )
N = 60
5 N = 40 0.5
N = 20
N =1
0 0
0 0.5 1 0 0.5 1
p+ Gt (π)
Figure 16.10: (code) (a) Simulation of the increase of knowledge on the population of a
R1
two-level system by repeated measurement, PN,r,+ (p+ )/ 0 PN,r,+ (p+ )dp+ . The population
of the state |+i was set to p+ = 0.2. (b) Determination of the most likely Rabi pulse length
Gt = 2 arcsin(r/N )1/2 .
of the two-level system (which we assume free of spontaneous emission) performs Rabi
Ω2 2 Gt
√
oscillations, ρ++ (t) = G 2 sin 2 , where G = ∆2 + Ω2 . The probability of finding
the system in state |2i, therefore, varies in time, p+ (t) = ρ++ (t), and the binomial
distribution (16.48) becomes,
N
Ω 2N 2N −2r Gt
PN,r,+ (t) = G sin2r Gt
2 cos 2 . (16.51)
r
∆r 1
= . (16.53)
(∂r/∂ω)|(ω−ω0 )T =π/2 TN
Quantum correlations and spin-squeezed states are the same phenomena seen in
different bases. Often the term of quantum entanglement is used in an energy bases,
while spin squeezing is described in a Dicke state bases.
We may also consider real space interferometers. This allows us to use for creating
the Fock states the recently discovered method of transforming BECs into Mott insu-
lator states. Of course it is important to use condensates. First of all, we need Fock
states, i.e. states with strong on-site quantum correlations with well-defined phase re-
lations between every pair of atoms from the same site. Furthermore, Mott insulators
have thus far only been seen with CBEs.
574 CHAPTER 16. QUANTUM MEASUREMENT
system remains in the initial state without loss of probability: [P (T )]N = 1. Interpret
the result.
c. One of the simplest imaginable systems, a laser-driven two-level system executing
Rabi oscillation, is described by the Hamiltonian:
0 Ω
Ĥ = .
Ω 0
Find the expression for τz as a function of the Rabi frequency Ω for the initial state
(1 0)† .
d. Choosing the evolution time t = 0.01τz τz and performing N = 5 measurements
during this time interval, how likely is the system to remain in the initial state?
e. Let us now include a decay channel for the state (0 1)† with Γ = 4γ, such as to
simulate a system with continuous measurement. The system is initially prepared in
the state (1 0)† . If we now observe emission by decay, it means that the system left
the initial state. We now have the effective Hamiltonian:
0 Ω
Ĥ = .
Ω −2ıγ
For this system, the probability amplitude for the initial state is:
1 γ −(γ−∆)t/~ 1 γ −(γ+∆)t/~
hΨ0 |Ψ(t)i = 1+ e + 1− e ,
2 ∆ 2 ∆
p
with ∆ = γ 2 − Ω2 . For a decay rate γ Ω, calculate the probability that the
system remains in the initial state. Interpret the result.
Formulas:
x2
ex = 1 + x + 2 + O(x3 ) , (1 − x)N = 1 − N x + O(x2 )
2 2 3 0 Ω −iĤt/~ cos Ωt −ı sin Ωt
cos (x) = 1 − x + O(x ) , Ĥ = →e = ~ ~ .
Ω 0 −ı sin Ωt
~ cos Ωt
~
assuming that at any instant of time the system stays in a eigenstate |n(R)i,
that, while the first exponential describes the standard dynamical phase, the phase
γn : C → γn (C) is not integrable, i.e. γn cannot be expressed as a function of R(t) and
is not single-valued meaning that from R(T ) = R(0) we cannot infer γn (T ) = γn (0).
Substituting (16.56) in (16.54), we find,
Ĥ(R(t))|ψ(t)i = ı~|ψ̇(t)i = En + ı~ıγ̇n (t) + ı~Ṙ(t) · ∇R |ψ(t)i . (16.57)
For a closed loop, continuity of the gauge field requires ξ(T ) − ξ(0) = 2πm. Hence, up
to an integer multiple of 2π, closed loop Berry phases remain gauge-invariant under
arbitrary gauge transformation and may be related to physical observables.
16.4. GEOMETRIC AND TOPOLOGICAL PHASES 577
we obtain, ZZ
γn (C) = ~ n (r) · dS .
V (16.65)
S
and thus equal to half the enclosed solid angle ΩS . The Berry phase can be
measured in Ramsey experiments, as discussed in Exc. 16.4.3.2.
We will now drop the conditions imposed to the Hamiltonian with regard to its
adiabatic behavior and request that the state must be a eigenstate [144]:
The dynamic phase δ can be zeroed by an appropriate choice of Ĥ(t), but not the
topological phase β. β does not depend on Ĥ(t), but is a geometric property of the
curve, which projects H onto P. In contrast to eıβ , the phase β is only determined
modulo 2πn.
16.4. GEOMETRIC AND TOPOLOGICAL PHASES 579
Figure 16.11: (a) Scheme for measuring the Aharonov-Bohm effect. The electrons propa-
gate as wavepackets whose centers-of-mass are not subject to forces, but whose de Broglie
waves are phase-shifted by the vector potential. (b) Aharonov-Casher effect: the electrodes
Φ do not produce electric fields inside the conductors; even so, one observes constructive
or destructive interference at the output of the interferometer, depending on the applied
potential.
The idea of this effect is schematized in Fig. 16.11. An electron beam is coherently
divided into two arms (e.g. by a double slit) passing both sides of an infinitely extended
and perfectly shielded solenoid. In this way the magnetic field B ~ vanishes in the region
outside the solenoid, but there must exist nonetheless a potential vector A, because
we observe on the screen an interference pattern of the two arms of the electronic
interferometer. When pass a current through the solenoid, we observe a displacement
of the interference pattern.
boundary conditions that the fields E~ and B~ must fulfill, we can determine the fields
as solutions of Maxwell’s system of partial differential equations.
In classical electrodynamics, observed from some inertial frame, the electromag-
netic force Fem acting on a point-like body with charge q, at position r, and with
velocity v, is given by the Lorentz force:
~
Fem (r(t), t) = q E(r(t), ~
t) + qv(t) × B(r(t), t) . (16.72)
Electrodynamic theory affirms the existence of two functions Φ(r, t) and A(r, t), such
that,
~ t) = ∇ × A(r, t)
B(r, and ~ t) = −∇Φ(r, t) − ∂A(r, t) .
E(r, (16.73)
∂t
Thus, we can use the equations (16.73) to rewrite the Maxwell equations.
The so-defined potentials Φ and A are not unique, but any Φ and A leading to the
same fields E~ and B,
~ and thus to the same physics, are equivalent. We will, however,
fix Φ and A adopting an additional condition that must be obeyed, i.e. we will adopt
a particular gauge. For the discussion of the Aharonov-Bohm effect, we will adopt
the Lorentz gauge defined by,
1 ∂Φ(r, t)
∇ · A(r, t) + =0, (16.74)
c2 ∂t
where c is the propagation velocity of light in vacuum.
corresponds to the gauge transform (2.199) and, given the condition (2.198), satisfies
the same Schrödinger equation as Ψ(r, t). We showed this explicitly in Exc. 16.4.3.3.
Thus, the presence of a potential vector in the region R, even in the absence of
fields, causes a phase shift eıg(r) of the wavefunction. An interesting issue studied in
Exc. 16.4.3.4 is, whether this implies that the freedom of choice of the gauge field is
lost.
where ΦB = πa2 B(0, t) is the magnetic field flux B~ through the cross section of
the solenoid.
At a point rsim of the screen, located in the plane of symmetry of the system,
we calculate g(rsim ) from Eq. (16.76) for two different paths: both starting at
the source and ending at the screen, but one going through the left slit, the
other through the right slit:
q rsim
Z Z
qΦB 1 qΦB
g(rsim ) = A(x) · dx = φ̂ · (ρφ̂dφ) = ± .
~ 0 2π~ ρ 2~
The + sign means, that the integration was done in the sense parallel to A, and
thus in the sense of I in the solenoid. The sign - holds for the other integration
path. The phase difference, at point rsim , between these two paths will be:
qΦB
δ= .
~
That is, the phase difference (experimentally observable via a shift of the in-
~ even
terference pattern) is directly proportional to the magnetic field flux B,
though the wavefunction is zero in the region S, where the B~ field is confined.
Let us also imagine another situation: Instead of an electron source producing
582 CHAPTER 16. QUANTUM MEASUREMENT
16.4.3 Exercises
16.4.3.1 Ex: Derivation of the Berry curvature
Derive the expression (16.66) for the Berry curvature.
Figure 16.12: (left) ’Time persistence’ by Salvatore Dali. (center) Recurrent events. (right)
Periodic astronomical cycles.
584 CHAPTER 16. QUANTUM MEASUREMENT
In our current physical understanding of the universe the most elementary events
are collisions between (real or virtual) particles. Obviously, our universe is full of
time. In order to bring any succession of such events into a logical and causal order,
a reference time line is needed. It allows for historical book keeping of sequences of
events, that we may call histories. And in order to facilitate a comparison between
different histories, this reference time line should follow a simple deterministic rule,
e.g. be periodic or exponentially decaying. In fact, both types of processes are cur-
rently used for time measurements. E.g. exponential processes, such as radioactive
decay is commonly used in radioactive dating.
The most common practical approach to the measurement of time, however, is
based on the observation of recurrent phenomena that we think of being periodic,
such as a day on Earth, the dripping of a water pipe, or the oscillation of a pendulum
or of an atomic excitation. Assuming the time intervals separating the recurrent
phenomena as being all the same, we build a ruler for time which we call clock.
But now comes a tricky question: How do we know whether a clock is really
periodic? In fact, we never know for sure whether the time intervals are really all
the same. We only know that some clocks deserve a greater degree of confidence
based on the fact that more care has been taken in their design and construction, or
based on the observation that clocks built in a certain manner tend to deviate less
from each other than clocks built in a different manner. The other approach would
be to compare several totally independent clocks and to give preference to those who
deviate less from each other.
An important criterion for a useful clock concerns its duty cycle. The shorter the
cycling time, the faster we can extract information from the clock, and the higher
is the accuracy we can reach in a given integration time. Ancient time standards
had been link to the periodic motion of celestial bodies, e.g. the revolution time
of the Earth around the sun. Therefore, clocks with smaller duty cycles have been
engineered, such as the clepsydra or the hour glass. But their calibration to periodical
astronomical cycles remained tedious and slow. Historically, the development of ever
precise clocks has been motivated by navigation. Indeed, 1 minute of inaccuracy in
the clock generates an uncertainty of 28 km in global positioning. And this motivation
still prevails nowadays, although, meanwhile, atomic clocks are reaching uncertainties
of below 10−18 and extremely short duty cycles on the order of femto-seconds.
The left part of Fig. 16.13 shows the basic idea of any human-made clock, which
consists in locking an oscillating mechanism, whose time constant can be manipulated,
for instance a pendulum or a laser, by a very precise periodic process, that we call
resonance. This resonance can be an astronomic period or the transition of an atomic
between two states of excitation. Expressing the stability of a clock that can be
obtained in a given integration time as,
Γnat + Γpert 1
σ= , (16.79)
ω S/Nnoise
we see that we better look for oscillators and resonances operating at high frequencies
ω, having very narrow widths Γnat , being subject to very weak perturbations and
line broadenings Γpert , and delivering a good signal-to-noise ratio. The right part of
Fig. 16.13 shows the concept of a cesium beam atomic clock. A microwave oscillator
operating at ω/2π = 9.1 GHz excites cesium atoms passing through a microwave
16.5. FREQUENCY METROLOGY AND SENSING IN QUANTUM MECHANICS585
Figure 16.13: (left) Principle of any clock and possible perturbations. (right) Cesium beam
atomic clock and Ramsey interference fringes.
Figure 16.14: Gravitational redshift is due to the dilation of time near heavy masses.
∆ν ∆λ g∆z
= ' 2 , (16.80)
ν λ c
where ∆z is the change in height. Accounting for the accompanying gravitational time
dilation affecting the atomic clock in the satellite is crucially important for accurate
navigation. For this reason, metrology of time and gravimetry are closely interrelated.
Do the Exc. 16.5.4.1.
computation and quantum communication (see Sec. ??), one of the key technologies
anticipated to drive the quantum revolution 2.0.
approaches investigate the possibilities of overcoming these limits via the creation of
strong interparticle correlations via entanglement, spin-squeezing [745], or superra-
diant lasing [554]. Such correlations have been created with atoms, which therefore
qualify as building blocks of quantum computers or sensing devices capable of over-
ruling the standard quantum limit.
Figure 16.16: (left) Single qu-bit. (center) Larger signals can be observed using ensemble of
qu-bits. (right) Collective manipulations on entangled qu-bits allow precisions beyond the
standard quantum limit.
Sometimes spontaneous emission can be controlled by confining both, the atom and the interrogating
light field, in a cavity.
16.5. FREQUENCY METROLOGY AND SENSING IN QUANTUM MECHANICS589
Typically, the internal Hamiltonian is static and defines the energy eigenstates |0i
and |1i,
Ĥqbit = E0 |0ih0| + E1 |1ih1| . (16.82)
The qubit internal Hamiltonian may contain additional interactions, such as cou-
plings to other qubits or time-dependent stochastic terms due to interactions with an
environment.
Many quantum sensing protocols require to manipulate the qubit either before,
during, or after the sensing process. This is achieved via a control Hamiltonian,
which allows implementing a sequence of appropriate quantum gates, such as the
Hadamard gate or Pauli X and Y gates. These gates are nothing else than what,
in the language of atom interferometry is known as π/2 or π-pulses around different
axes. Advanced sensing schemes employing more than one sensor qubit may further
require conditional gates, especially controlled-NOT gates to generate entanglement,
swap gates to exploit memory qubits, and controlled phase shifts in quantum phase
estimation (see Sec. ??). Finally, the control Hamiltonian can include control fields
for systematically tuning the transition frequency ~ω0 = E1 − E2 .
Finally, the signal Hamiltonian represents the coupling between the sensor qubit
and a signal V (t) to be measured. When the signal is weak (which is assumed here)
Ĥint adds a small perturbation to the internal Hamiltonian. The signal Hamiltonian
can then be separated into two qualitatively different contributions,
Ĥint = Ĥint,k + Ĥint,⊥ , (16.83)
where Ĥint,k is the parallel (commuting) and Ĥint,⊥ the transverse (non-commuting)
components, respectively. The two components can quite generally be captured by,
Ĥint,k = γ2 Vk (t)[|1ih1| − |0ih0|] (16.84)
Ĥint,⊥ = γ
2 [V⊥ (t)|1ih0| + V⊥† (t)|0ih1|] ,
where γ is the coupling or transduction parameter of the qubit to the signal V (t).
Examples of coupling parameters include the Zeeman shift parameter or the Stark
shift parameter of electric dipoles in an electric field. The parallel and transverse
components of a signal have distinctly different effects on the quantum sensor. A
commuting perturbation Ĥint,k leads to shifts of the energy levels and an associated
change of the transition frequency ω0 . A non-commuting perturbation Ĥint,⊥ , by
contrast, can induce transitions between levels, manifesting through an increased
transition rate. Most often, this requires the signal to be time-dependent (resonant
with the transition) in order to have an appreciable effect on the quantum sensor.
An important class of signals are vector signal V(t), in particular, those provided
by electric or magnetic fields. The interaction between a vector signal and a qubit
can be described by the signal Hamiltonian,
ˆ,
Ĥint = γV(t) · ~σ (16.85)
ˆ is the vector of Pauli matrices (2.47). For a vector signal, the two signal
where ~σ
functions Vk (t) and V⊥ (t) are,
Vk (t) = Vz (t) (16.86)
V⊥ (t) = Vx (t) + ıVy (t) ,
590 CHAPTER 16. QUANTUM MEASUREMENT
where the z-direction is defined by the qubit’s quantization axis. The corresponding
signal Hamiltonian,
is just the one of the Rabi model extensively discussed in Sec. 14.2.
Figure 16.18: Difference between clocks and interferometers. (left) Ramsey-type clock.
(right) Ramsey-Bordé interferometer.
frequency given,
mg
νblo = . (16.88)
2~klat
Figure 16.19: Working principle and pictures of the Sr Bloch oscillation interferometer con-
structed at the IFSC, USP.
16.5.4 Exercises
16.5.4.1 Ex: Red-shift of sun light and on Earth
T. Erber et al., Resonance Fluorescence and Quantum Jumps in Single Atoms, Test-
ing the Randomness of Quantum Mechanics [DOI]
D.J. Wineland et al., Squeezed atomic states and projection noise in spectroscopy
[DOI]
16.6.3 on sensing
A.D. Cronin et al., Optics and interferometry with atoms and molecules [DOI]
Nonlinear optics
The discipline of nonlinear optics studies phenomena that occur as a consequence of
modifications of the optical properties of materials by the presence of light. Such
modifications are appreciable only, when the interacting light is sufficiently intense,
i.e. of the order of the interatomic electric field,
e2
Eat = ' 5.14 · 1011 V/m , (17.1)
4πε0 a2B
which explains that nonlinear effects could only be studied properly after the advent
of the laser.
In general, light-matter interaction is expressed through the relationship between
the polarization induced in the medium, P (ω), with the optical light field, E(ω). In
linear optics this relation can be expressed as,
P̃ (ω) = ε0 χ(ω)Ẽ(ω) . (17.2)
But in the perturbative regime of non-linear optics this expression must be generalized
to a series of powers of the electric field,
P̃ (ω) = ε0 [χ(1) (ω)Ẽ(ω) + χ(2) (ω)Ẽ 2 (ω) + χ(3) (ω)Ẽ 3 (ω) + ...] , (17.3)
so that higher order polarization terms, P (n) (ω) = ε0 χ(n) E N (ω), are considered.
Therefore, the phenomena are non-linear in the sense that they depend non-linearly
on the optical field applied to the material. In a more complete treatment, in terms
of the optical properties of materials, P̃ and Ẽ are vector fields and the electric
susceptibility is a tensor. However, to simplify the treatment, we consider the fields
as scalars and χ(n) as constants independent of the frequency ω.
Examples of non-linear optical phenomena are: parametric processes of sum and
difference frequency generation, as shown in Fig. 17.1, optical parametric oscillation,
and the dependence of the refractive index with the optical intensity. Examples of non-
parametric processes are: multi-photonic absorption, stimulated Raman scattering,
and saturated absorption. In the latter example, the absorption coefficient of the
material decreases with increasing light intensity:
α0
α= . (17.4)
1 + I/Is
Saturated absorption is an example where a perturbative approach is not capable of
providing good results, and its most reliable description is given by the approximation
of a two-level quantum system.
595
596 CHAPTER 17. NONLINEAR OPTICS
Figure 17.1: Parametric (a) sum and (b) difference frequency generation. (c) Third harmonic
generation.
n2 ∂ 2 1 ∂2
∇Ẽ − Ẽ = P̃ , (17.5)
c2 ∂t2 ε0 c2 ∂t2
~
B ê3 × ĝ ê2 × ê3
ê3 = , ê2 = , ê1 = , (17.6)
B |ê3 × ĝ| |ê2 × ê3 |
where ĝ is an arbitrarily chosen direction, e.g. gravity. The relative amplitude of the
transitions ∆mJ = 0 is proportional to the projection of the polarization vector on
the magnetic field axis ζπ = (ε̂ · ê3 )2 . To estimate the amplitude of the transitions
∆mJ = ±1, we must project onto the coordinates,
and we obtain ζσ± = (ε̂ · ê± )2 . Resolve the Excs. 17.1.2.1 and 17.1.2.2.
17.1.2 Exercises
17.1.2.1 Ex: Coherent trapping and dark states
Consider a three-level system a, b, and c in Λ-configuration interacting with two
electromagnetic fields with frequencies ω1 and ω2 , as shown in the figure. I.e. the two
states of lower energy b and c are coupled with a more energetic state a. Assume that
Aprisionamento Coerente - Estados Escuros
Figura 1: Conguração Λ com três níveis energéticos interagindo com dois campos resso-
the the transition |bi ↔ |ci
nantes de frequência ω1 eisω2forbidden
. and that the Hamiltonian of the system has
the form Ĥ = Ĥ0 + ĤI , where,
Supondo que as transições |bi ↔ |ci são proibidas e que o Hamiltoniano do sistema possui
a forma H = H0 + HI , onde
Ĥ0 = ~ωa |aiha| + ~ωb |bihb| + ~ωc |cihc|
~ H0 = ~ωa |ai ha| + ~ωb |bi hb|~+ ~ωc |ci hc|
ĤI = − (Ω1 e−ıω1 t |aihb| + Ω∗1 eıω1 t |biha|) − (Ω2 e−ıω2 t |aihc| + Ω∗2 eıω2 t |ciha|) .
e 2 2
~ −iω1 t ∗ iω1 t
HI = − state
a. Assuming that the system’s (Ω1 e is|aidescribed
hb| + Ω1 e |biby ha|)|ψi = K |ai + K |bi + K |ci
2 a b c
find the system of equations that~ describe
−iω t the dynamics
∗ iω t of probability amplitudes Ki
− (Ω 2 e |ai hc|
2
+ Ω 2 e |ci ha|)
2
2
(i = a, b, c).
b. Rewrite the equations for the case where 1 the frequencies of the applied fields are
resonant (that is, ω1 = ωa − ωb e ω2 = ωa − ωc ). Simplify the system by writing in
terms of the variables uk = Ck eiωk t (k = a, b, c). √
c. Assuming the initial condition |ψ(0)i = (|bi+|ci)/ 2, solve the system of equations
for the resonant case and interpret the result.
Simulate the Hanle effect for the Ca 3-level system using optical Bloch equations.
Figure 17.2: Dressed states level scheme for LWI. Quantum interference between two possible
Raman transitions (green) leads to a cancellation of the ground state population. When the
upper level |hi is incoherently pumped, gain can be reached for a weak probe field. Inserting
the atoms in a cavity one can reach lasing [358].
17.2.3 Exercises
17.3. FURTHER READING 601
603
604 CHAPTER 18. ATOMIC MOTION IN FORCE FIELDS
where for a simple two-level atom, |ψ(t)iele = cg (t)|gi + ce (t)|ei, with the atomic
ground state |gi and the excited state |ei. The external states are eigenstates of the
momentum in the case of a free particle, |ψiext = |pi. For particles confined in a po-
tential the external states are the vibrational eigenstates, |ψiext = |ni. The temporal
evolutions of the internal and external degrees of freedom are governed by indepen-
dent Schrödinger equations. For cold atomic clouds the kinetic energy is much smaller
than the excitation energy, which allows the separation of the energy scales. That is,
the internal degrees of freedom are frozen in the ground state. Many phenomena, for
example, Bose-Einstein condensation and the dynamics of condensates are described
in this regime.
Figure 18.1: The internal degrees of freedom of cold atoms are thermally frozen.
Nevertheless, the fact that it is thermally frozen does not prevent the intentional
excitation of the internal degree of freedom by irradiating electromagnetic fields tuned
close to resonances and coupling electronic energy levels. In the case of coupling, the
external and internal degrees of freedom must both be considered.
1 −~2 2 ı~q
Ĥint = (−ı~∇ − qA)2 + qΦ ' ∇ + A · ∇ + qΦ . (18.5)
2m 2m m
From this formula we can, in principle, calculate all electromagnetic forces.
The coupling of external and internal degrees of freedom is mediated by the pho-
tonic recoil transferred to the atom during absorption and emission processes. That
18.1. ELECTROMAGNETIC FORCES 605
is, the feature that the interaction with light simultaneously excites the atom and
exerts a force couples the degrees of freedom. This fact manifests itself in the Hamil-
tonian of the atom interacting with a light field (15.17) by the appearance of terms
joining operators acting on different degrees of freedom,
Ĥ = ~ω↠â + Ĥint + Ĥatm where Ĥint = ~g(r̂)eık·r̂ ↠σ̂ + c.c. , (18.6)
P
where σ̂ ≡ |gihe| and â ≡ n n|nihn + 1| and ~g(r̂) ≡ d12 · E~1 (r̂) is the coupling
constant or one-photon Rabi frequency . The Hamiltonian is that of the Jaynes-
Cummings model, except that in addition to the field operators â and the atom
transition operators σ̂, appears an operator for the position of the atom r̂, whose
quantum features we have not taken very seriously so far. It appears in the Rabi
frequency and also in the term eık·r̂ . Now, we must remember, that
Urecoil = e−ık·r̂ = |p + ~kihp| (18.7)
is the unitary operator of the photonic recoil in the absorption process introduced
in Sec. 2.5.2 and extensively discussed in Sec. 3.6.2. We shall shortly see, that it is
precisely this term in the Hamiltonian that gives rise to all phenomena related to light
forces on atoms.
The presence of the position operator in the Jaynes-Cummings Hamiltonian in-
troduces a new degree of freedom. With no external potential (that is, the sys-
tem is invariant to spatial translations), this degree of freedom is simply the atomic
center-of-mass momentum, such that the new set of quantum numbers is |j, n, pi.
Strictly speaking we have to span the whole Hilbert space by an external product,
Ĥele ⊗ Ĥrad ⊗ Ĥext .
Often
√ a semi-classical description treating the light field a classically is sufficient,
â ' n with n the number of photons. Then the Hamiltonian (18.6) simplifies to,
Ĥint = ~Ω(r̂)eık·r̂ σ̂ + c.c. , (18.8)
√
with the Rabi frequency Ω(r) = ng(r).
Atoms naturally do not exhibit permanent electric dipole moments, when they are
not subject to external electric fields. In contrast, polar molecules (such as heteronu-
clear dimers), which have permanent electric dipole moments can have their motion
be influenced by inhomogeneous electric fields (see Sec. 24.5.3).
magnetic fields. We have already shown in the calculation (9.15) and (9.19) that the
interaction energy (18.5) can be written as,
Here, J = L + S is the total angular momentum resulting from the coupling of the
total angular orbital momentum and the total spin of all electrons. If the atom has
a nuclear spin I other than zero, then F = J + I replaces J in Eq. (18.9), and the
g-factor generalizes to (9.35) 1 ,
F (F + 1) + J(J + 1) − I(I + 1)
gF ' gJ . (18.11)
2F (F + 1)
In Sec. 9.2 we used the formula (18.9) to calculate the Zeeman shift of internal
energy levels. But, according to the formula (18.4), the interaction also generates a
force acting on the center-of-mass of atom,
f = −gF µB mF ∇B . (18.12)
~2 Ω2
E± (z) ' ± 12 [µB gF B(z) − ~ω] ± , (18.15)
4[µB gF B(z) − ~ω]
where the second term can be interpreted as the dynamic Stark shift of the energy
levels.
To illustrate the influence of the radiofrequency, we calculate the potential en-
ergy and the dressed states assuming a linear 1D magnetic field gradient B(z) ≡ zb.
Fig. 18.2(a) illustrates the radiofrequency coupling and Fig. 18.2(b) the dressed states
for two magnetic substates coupled by a radiofrequency. The minimum emerging in
(a) 4 (b)
1
(kHz)
(kHz)
2
0 ω 0 ΩΔ
Δ/2π
Δ/2π
-2
-1
-4
0 0.2 0.4 0 0.2 0.4
z (μm) z (μm)
Figure 18.2: (code) (a) Potential energies for a hyperfine structure F = 12 with a g-fator
of g = − 23 (as e.g. in the ground state 2 S1/2 of 6 Li). A radiofrequency (arrow) couples the
substates mF = ± 12 . Here, b = 200 G/cm and ω = 2π × 5 kHz. (b) Uncoupled dressed states
(dotted line), coupled dressed states (solid line), and dynamic Stark shifts (dash-dotted)
approximated far away from resonance. The Rabi frequency is Ω = 2π × 700 Hz.
the upper curve of Fig. 18.2(a) may serve as a confinement potential. Using an rf-
radiation composed by several frequencies, potential minima can be realized at several
distances z. In Exc. 18.1.4.3 we calculate an example.
In the dressed states basis with the Hamiltonian (18.13), the force is calculated
from,
X
F(r) = hF̂(r)i = −Tr atom,laser ρ̂∇r Ĥ = − hn, j|ρ̂∇r Ĥ|j, ni . (18.16)
n,j
We consider only one dimension and disregard the degrees of freedom of the radiation
field,
If the atoms enter the coupling area adiabatically, the populations of the adiabatic
potentials will only depend on z. This is analogous to the adiabatic transfer via adi-
abatic sweeps or STIRAP pulse sequences. If the atoms are too fast, the populations
also depend on history (i.e. the recent trajectory of the atoms), which can result in
Landau-Zener transitions to other (possibly untrapped) states.
18.1.4 Exercises
18.1.4.1 Ex: The Stern-Gerlach effect
Consider initially motionless 87 Rb atoms trapped in a superposition of two the trap-
pable Zeeman states |F, mF i = |2, +2i and |1, −1i. Suddenly a magnetic gradient of
∂z B = 100 G/cm is applied for 2 ms. Calculate the spatial separation of the atoms
being in either one of the two states after 10 ms of ballistic expansion.
defining the energetic zero in the middle between the states. Now, assume that the
magnetic field grows linearly along the axis z, B(z) = z∂z B, where ∂z B is the gradient.
Also assume that the radiofrequency is tuned in resonance with the difference of the
energies of the Zeeman states at some distance z0 such that, ~ω = µB z0 ∂z B.
a. Calculate the eigenenergies of the coupled system as a function of z.
b. Expands eigenenergies around the position z0 .
c. What would be the oscillation frequency of the trapped atoms inside the adiabatic
potential?
d. Expands the eigenenergies in ~Ω for locations away from resonance.
using light forces to control the motion of neutral atoms, the fundamental bases for
understanding the physics of light forces were not developed before the late 1970s.
Unequivocal experimental demonstrations of cooling and trapping of atoms were not
performed before the mid-1980s. In this section we will discuss some fundamental
aspects of light forces. Practical schemes used to cool and trap neutral atoms will be
presented in Secs. 24.2 and 24.3.
The light force acting on an atom can be of two types: a spontaneous dissipative
force and a conservative dipole force. The spontaneous force arises from the recoil
experienced by an atom when it absorbs or emits a quantum of light. As we saw
in Sec. 1.2.6, when an atom scatters light, the resonant scattering cross section can
λ2
be written as in Eq. (1.73), σ0a = gg21 2π0 , where λ0 is the resonant wavelength. In
the optical region of the electromagnetic spectrum the wavelengths of light are of the
order of several hundred nanometers, and the resonant cross sections for scattering
become very large, (∼ 10−9 cm2 ). Each absorbed photon transfers a quantum of
momentum ~k to the atom in the direction of propagation. Spontaneous emission
following an absorption process occurs in random directions and, hence, averaged
over many absorption-emission cycles, it cancels to zero. Consequently, the total
spontaneous force acts on the atom in the propagation direction of the light, as shown
schematically in the diagram of Fig. 18.3. The saturated photon scattering rate via
spontaneous emission (the reciprocal value of the excited state’s lifetime) sets the
upper limit for the magnitude of the force. This force is called radiation pressure
force.
The dipolar gradient force can be easily understood by considering light as a clas-
sical wave. It is simply the time-averaged force resulting from the interaction of the
transition dipole –induced by the oscillating electric field of the light– with the gra-
dient of the electric field amplitude. The strength of this gradient can be controlled,
e.g. by focusing the light beam. By tuning the optical frequency below or above an
atomic transition, we can control the sign of the force acting on the atom: Tuning
the light below resonance attracts the atom to the center of the light beam, tuning
it above resonance repels it. The dipole force is a stimulated process without en-
ergy exchange between the field and the atom. Photons are absorbed in one light
mode and reappear by stimulated emission in another one. However, conservation
of momentum requires that the change in the propagation direction of the scattered
photons from an initial mode to a final mode leaves the atom with a recoil. Contrary
to spontaneous force, there is, in principle, no upper limit for the magnitude of the
dipole force, since it is a function of the field gradient only and the detuning.
F(0) = − 12 ~∇r Ω(0)(ρ12 e−ı∆t + ρ21 eı∆t ) − 2ı ~kΩ(0)(ρ12 e−ı∆t − ρ21 eı∆t ) . (18.20)
The quantities ρ12 ≡ h1|ρ̂|2i = ρ∗21 are the coherences, which develop in a two-
level system excited by a laser beam. Inserting the stationary solutions of the Bloch
equations (14.99),
Ω2 (2∆ − ıΓ)Ω
ρ22 = and ρ12 = e−ı∆t . (18.21)
4∆2 + 2Ω2 + Γ2 4∆2 + 2Ω2 + Γ2
we obtain
4∆Ω ΓΩ2
F(0) = − 12 ~ ∇ r Ω + ~k . (18.22)
4∆2 + 2Ω2 + Γ2 4∆2 + 2Ω2 + Γ2
2
Γ
With the definition of the cross section, σa (∆) = σa0 4∆2 +2Ω2 +Γ2 ,
2Ω2 Ω2 σa (∆)
F(0) = − 12 ~∆∇r ln 1 + + ~k . (18.23)
4∆2 + Γ2 Γ σa0
The resonant cross section for a ’classical’ transition is σa0 = 3λ2 /2π.
4 Treating the motional and the optical degrees of freedom as classical the density operator only
Apparently, the force comprises two contributions. The dipolar gradient force can
be derived from a potential. It is proportional to the intensity gradient and can
be interpreted as resulting from absorption processes immediately followed by self-
stimulated emission. Near resonance it is dispersive. Far from resonance it can be
approximated by,
−~∆Ω2 |∆|Γ ~Ω2
Fdp = ∇r −→ −∇ r . (18.24)
4∆2 + Γ2 4∆
The radiation pressure force is dissipative. Close to resonance it is absorbing. It is
proportional to the phase gradient and the only force exerted by plane waves. It
can be interpreted as resulting from absorption processes followed by spontaneous
emission. With Ω2 = σa0 ΓI/~ω we get a formula,
I
Frp = ~k σa (∆) = ~kγsct , (18.25)
~ω
which describes the force as a product of the number of photons in the incident beam,
I/~ω, the absorption cross section, σa (∆), and the recoil momentum per photon, ~k.
γsct is the scattering rate. The dipole gradient force (and the associated potential) is
Figure 18.3: (a) An atom with mass m and velocity vA moves to the right and absorbs a
photon propagating to the left with momentum ~kL . Center: An excited atom suffers a
change of momentum pA = mvA − kL . (b) The isotropic reemission of a photon results,
averaged over many absorption-emission cycles, in a momentum change for the atom of
hpA i = mvA − kL .
often used to spatially confine atoms, and the radiation pressure force is often used
to cool them down. Note that we still need to correct Eqs. (18.24) and (18.25) to
take into account the square of the average over the possible spatial orientations of
the transition matrix element d12 /3.
The saturation parameter,
1 2
2Ω
s= , (18.26)
∆ + 41 Γ2
2
612 CHAPTER 18. ATOMIC MOTION IN FORCE FIELDS
allows to write the dipolar gradient force and the radiative pressure force as,
~∆ 1 ~∆ ~kΓ s
Fdp = − ∇s = ∇ ln[1 + s] and Frp = . (18.27)
6 1+s 6 6 1+s
Eq. (18.27) shows that the radiation pressure force ’saturates’ as s increases, and
is therefore limited by the spontaneous emission rate. The saturation parameter
essentially describes the relative importance of terms appearing in the denominator
of the line profile function for the light forces. The spontaneous emission rate is an
intrinsic property of the atom, proportional to the square of the atomic transition
dipole moment, whereas the square of the Rabi frequency is a function of the incident
laser intensity. If s 1, the spontaneous emission is fast compared to any stimulated
process, and the light field is said to be weak. If s 1, the Rabi oscillation is fast
compared to spontaneous emission and the field is considered as strong. √The line
profile factor indicates a ’power broadening’ by saturation of a factor of 2. Note
that the dipolar gradient force and potential, Eqs. (18.27), do not saturate when the
intensity of the light field is increased. Usually Fdp and Udp are used to manipulate
and trap atoms in a laser beam tuned far away from resonance in order to avoid
absorption.
Often, the transition moment can be oriented using circularly polarized light. In
this case, all previous expressions for Fdp , Frp , and Udp should be multiplied by 3.
From now on we will abandon the average over the orientations and only use d212 for
the square of the transition dipole moment. Solve Excs. 18.2.5.1 and 18.2.5.2.
describes, under the influence of the dispersive interaction, a precession around the
polar axis. This was discussed in Exc. 14.5.5.7. The Stark shift causes a rotation
of ~−1 U (r)t. Simultaneously, the atom is subjected to a force, which corresponds to
the gradient of the potential −∇U (r), as illustrated in Fig. 18.4(a). We see that the
phase shifts of the Broglie wave and the Bloch vector are equal. Finally, the light
mode phase is also shifted by the same amount in an effect called refraction. That is,
the internal, external, and optical degrees of freedom are entangled.
This fact has a practical use in atomic interferometers, because it is often easier
to detect an interference of internal excitation states rather than of Broglie waves.
Because of the entanglement, it is sufficient to measure one interference pattern to
know the other one.
By local variations of the potential U (r), e.g. induced by a focused laser beam,
it is possible to manipulate a Broglie wavefront in the same way that, in classical
optics, we manipulate the wavefront of a light beam by lenses or other objects, such
as for instance, the refractive index represented by an atomic cloud near resonance,
as illustrated in Fig. 18.4(b).
614 CHAPTER 18. ATOMIC MOTION IN FORCE FIELDS
Figure 18.4: Diagram (a) shows product states and dressed states for blue detuning. Note
that the population is in the upper level and that the atom is subject to a repulsive weak field
seeking force when it enters the laser beam. Diagram (b) is similar, but for red detuning.
The population is in the lower level and the atom is subject to an attractive high field seeking
force.
The orientation of the force depends on the light frequency as compared to the
resonant frequency. The dipolar force attracts the atom to regions where the light field
is strong, when the frequency is tuned below ω0 , and it attracts the atom to regions
of weak fields, when tuned above ω0 . Integration over the relevant spatial coordinates
results in an effective potential or barrier to the atom. The qualitative behavior of
the dipolar potential and its effect on the motion of atoms is easily visualized in the
dressed states picture. Fig. 18.5 shows what happens when an atom enters a well
defined region of an optical field, for example a focused laser beam.
Outside the atom-dipole coupling zone the expression ~Ω is despicable and the
’dressed states’ are just the atom-field product states. When the atom enters the
field, this expression becomes nonzero and the atom-field states combine to produce
a set of dressed states. The energy levels of the product states ’repel’ each other and
approach the dressed states levels. Assuming that the laser is sufficiently detuned to
maintain the absorption rate negligible, the population remains in the ground state.
We see that blue (red) detuning leads to a repulsive (attractive) potential for atoms
remaining in the grounded state. In addition, since ~Ω is directly proportional to the
root of the laser intensity, an increase in that intensity (optical power per unit area)
obviously amplifies the force on the atom (F ' ∇R Ω).
18.3. PHOTONIC RECOIL ON FREE AND CONFINED ATOMS 615
18.2.5 Exercises
18.2.5.1 Ex: Dipole force for large detunings
Verify that in the limit of large detunings the dipole potential Eq. (18.23) tends to
Ω2
−→ 4∆ .
~ef f an effective
in the xy-plane. αv is the vector part of the atomic polarizability and B
magnetic field [199, 442, 14].
potential and on cooperative effects (bosonic stimulation), but this will be discussed later.
616 CHAPTER 18. ATOMIC MOTION IN FORCE FIELDS
Figure 18.6: (a) Kicking an atom along its dispersion relation. (b) Scheme of the Compton-
scattering of light.
We will calculate in the following the frequency distribution of the light scattered
by an atom as a function of its initial velocity p1 , of the frequency ω1 of the incident
light and of the scattering angle, that is, the angle between the modes k1 and k2 . We
begin by writing the laws of conservation of energy and momentum,
The photonic recoils of the incident and of the scattered light are almost equal,
~k12 ~k 2
ωrec ≡ ' 2 , (18.37)
2m 2m
such we can approximate,
p1
1− ~ω1
mc2 − mc cos ^(k1 , p1 )
ω2 = ω1 p1 , (18.38)
1− ~ω1
mc2 cos ^(k1 , k2 ) − mc cos ^(p1 , k2 )
where we call the angles ϑin = ^(k1 , p1 ), ϑout = ^(k2 , p1 ), and ϑ = ϑin − ϑout =
^(k1 , k2 ). For non-relativistic velocities, the denominator is approximately 1:
with p1 = mv1 . The first term describes the recoil shift and the second term the
Doppler shift.
The second term vanishes for initially at resting atoms, p1 = 0, and Eq. (18.40)
simplifies to,
ω2 − ω1 = 2ωrec (−1 + cos ϑ) . (18.41)
It also vanishes for atoms which have no velocity component in the scattering plane
spanned by the wavevectors k1 and k2 , that is ϑout = 180◦ − ϑin 6 , for which case we
get the maximum recoil shift,
ω2 − ω1 = −4ωrec . (18.42)
2
p
excitation (Ĥele ), the kinetic energy of their center-of-mass (Ĥcm = 2m ), the radiation
field (Ĥrad ) (which may be treated classically under the circumstances discussed here),
and the coupling (Ĥint ) of all three degrees of freedom,
Transition probabilities are readily calculated using Fermi’s Golden rule, because
the density-of-states distributions for the final radiation modes receiving the scat-
tered photons (photonic density-of-states) and the recoil modes receiving the scat-
tered atoms (phononic density-of-states) are white, that is, without resonances, and
isotropic.
In the presence of an imposed trapping potential, an additional term appears
in the Hamiltonian, which has the capacity of dramatically changing the scattering
features,
p2
Ĥcm = + V̂trap (r) . (18.45)
2m
The confining potential may or may not depend on the internal state of the atom.
For ions in a Paul trap it does not depend, but for atoms in magnetic traps it usually
depends,
where ω0 is the frequency of the atomic transition. In the following, however, we will
treat potentials that are independent of the internal atomic state.
confinement sidebands
ωtrp
weak η>1 unresolved Γ <1
ωtrp
strong η < 1 resolved Γ >1
620 CHAPTER 18. ATOMIC MOTION IN FORCE FIELDS
Figure 18.7: Absorption profile in the regimes of (a) weak binding (η > 1) for the cases of
and unresolved sidebands (blue curve) and resolved sidebands (red curve), and (b) strong
binding (Lamb-Dicke regime η < 1) for the same cases as in (a).
The role of photonic recoil in the scattering of light by confined atoms has been
unraveled by Mößbauer, who performed scattering experiments of γ-photons on a 57 Fe
crystal on a narrow transition of this isotope at 14 keV (0.086 nm). The linewidth
of this transition, Γ ≈ (2π) 1 MHz, is much narrower than the recoil-shift, ωrec ≈
(2π) 500 MHz Γ, so that we should expect the scattered light to be considerably
recoil-shifted and Doppler-broadened. In fact, the recoil-shift should be so large, that
scattered photons cannot be reabsorbed by other atoms on this transition being at
rest. This is not what Mößbauer observed in his experiments. He found that scattered
photons can be reused for subsequent scattering, which means that the scattering must
be elastic.
The explanation for this unexpected observation is that, if the nucleus is embedded
in a crystal, the vibrational frequencies, which are even higher than the photonic recoil
frequency, ωtrp ≈ 2π · 3 THz, are unreachable. That is to say, we are in the resolved
sidebands Lamb-Dicke regime, ωtrp ωrec Γ. Here, the phonons corresponding
to the vibrations cannot be excited, so that the recoil momentum must be absorbed
by the whole lattice, whose entire mass is so large, that the photon frequency is not
recoil-shifted by the scattering process. Hence, the first-order Doppler effect is avoided
and the measured width of the transition is just the natural linewidth. Furthermore,
the atom remains in the same vibrational state of their localizing potential. This is
the Mößbauer effect.
γ for 57
Fe optical for 172
Yb+
photon energy ≈ 14 keV ≈ 4 × 1018 Hz ≈ 3 eV ≈ 640 THz
recoil ≈ 2 meV ≈ 500 MHz ≈ 10−10 ≈ 20 peV ≈ 5 kHz ≈ 10−11
linewidth ≈ 5 neV ≈ 1 MHz ≈ 2 × 10−13 ≈ 5 feV ≈ 1 Hz ≈ 2 × 10−15
lattice vibrations ≈ 10 meV ≈ 3 THz ≈ 3 neV ≈ 1 MHz
A similar effect can be observed with trapped atoms driven on very narrow tran-
sitions, as we will discuss in the following.
18.3. PHOTONIC RECOIL ON FREE AND CONFINED ATOMS 621
and insert it together with the Hamiltonian (18.48) into the Schrödinger equation.
Projecting onto the states hn, g| and hn, e|, we easily derive the following equations of
motion,
∞
dcn,g ıΩ X
= −ıωtrp (n + 12 )cn,g − cm,e hn|e−ıkẑ |mi
dt 2 m=0
. (18.51)
∞
dcn,e ıΩ X
= −ıωtrp (n + 1
2 )cn,e − ı(ω0 − ω)cn,e − cm,g hn|eıkẑ |mi
dt 2 m=0
9 Note, that this Hamiltonian cannot be used to describe spontaneous emission. To do so, we need
The fact that only terms proportional to hn|eikẑ |mi contribute can be understood
in terms of the Franck-Condon overlap between the vibrational states to be coupled.
And the fact that the energy of the harmonic oscillator, and thus the effective detuning
∆n ≡ ω − ω0 − ωtrp (n + 21 ), depend on the vibrational state couples the internal and
the external dynamics.
Figure 18.8: (a) Absorption (1) and Rayleigh scattering (2) in a weakly confining trap, where
the momentum must stay conserved. (b) Illustration of the regime ωrec > Γ, where the atom
moves during the absorption process. (c) Absorption and Rayleigh scattering in a strongly
confining trap.
Nevertheless, the off-diagonal coupling elements only contain the degrees of free-
dom of the harmonic oscillator.
X X 2 |α|2n X 2 |α|2n−2
n~ωtrp hn|αihα|ni = n~ωtrp e−|α| = ~ωtrp |α|2 ne−|α|
n n
n! n
(n − 1)!
~2 k 2
= hα|αi = ~ωrec . (18.54)
2m
Thus energy and momentum conservation are automatically satisfied by the way the
kick is implemented.
Let us now consider a single atom (e.g. a trapped ion) confined in an 3-dimensional
anisotropic trap (e.g. strong confinement in one and weak confinement in the other
direction) and address the question, whether the scattering will be anisotropic, as well.
We generalize the problem to threeP dimensions by allowing for quantized vibrational
states in three dimensions, σ̂ † = n |nihn| ⊗ |eihg| with |ni = |nx , ny , nz i.
In first-order perturbation, using the Hamiltonian (18.48), the transition rate for
absorption or emission is given by Fermi’s Golden Rule,
1 d 2π ık·r̂ †
= |hn, e|eıĤt/~ |0, gi|2 ' 2 |hn, e| ~Ω 2 (e σ̂ + e−ık·r̂ σ̂)|0, gi|2 (18.55)
τ dt ~
πΩ2
= |hnx , e|eıkx̂ |0, eihny , e|eıkŷ |0, eihnz , e|eıkẑ |0, ei|2 .
2
Obviously, it is thus sufficient to calculate hnj |eıkẑ |0i, i.e. we can trace over the
internal degrees
√ of freedom. Using our previous results (3.130), we find with αj =
−ıkj atrp,j / 2,
n
1 πΩ2 πΩ2 −|αx |2 −|αy |2 −|αz |2 |αxnx αy y αznz |2
= |hnx |αx ihny |αy ihnz |αz i|2 = e .
τ 2 2 nx !ny !nz !
In Exc. 18.3.4.1 we calculate and illustrate the transition matrix elements h0|eıkẑ |0i
and hn|eıkẑ |0i.
formula,
dσ d 2π X Ĥ (1) |m, eihm, e|Ĥ (1)
= |hn, g, kf |eıĤt/~ |0, g, ki i|2 ' 2 |hn, g, kf | |0, g, ki i|2
dΩ dt ~ m
ω i − ω m
2
2 4 X −ıq·r̂
π~ Ω hn, kf |e |mihm|eıq·r̂ |0, ki i
=
8 m Ei /~ + (0ωx + 0ωy + 0ωz + 23 ) − Em /~ − (mx ωx + my ωy + mz ωz + 23 )
2
π~2 Ω4 X hnx , ny , nz |e−ikf ·r̂ |mx , my , mz ihmx , my , mz |eıki ·r̂ |0, 0, 0i
=
8 m ,m ,m (Ei − Em )/~ − mx ωx − my ωy − mz ωz
x y z
2
π~2 Ω4 X Y hnj |e−ikjf rj |mj ihmj |eikji rj |0i
.
=
8 m ,m ,m j=x,y,z (Ei − Em )/~ − mx ωmx − my ωmy − mz ωmz
x y z
(18.56)
The transition matrix elements can be calculated via (3.135). This expression repre-
sents (in the same time) the dynamic structure factor of the single trapped atom.
Example 80 (Axial incidence): Let us consider the particular case of Rayleigh
scattering from the ground state of light incident in the direction ki ≡ kiz êz .
We can then simplify,
2
dσ π~2 Ω4 X h0|e−ıkxf x̂ |0ih0|e−ıkyf ŷ |0ih0|e−ıkzf ẑ |mz ihmz |eıkzi ẑ |0i
=
dΩ 8 m −mz ωtrp
z
mz 2
π~2 Ω4 −|αf x |2 −|αf y |2 −|αf z |2 −|αiz |2 X
1 αfmzz αiz
= e
8 (E m − Ei )/~ − mz ωtrp,z m z!
mz
2
π~2 Ω4 −(kf ·atrp )2 /2−(ki ·atrp )2 /2 X
1 (− 12 kf z kiz a2trp,z )mz
= e ,
8 (Em − Ei )/~ − mz ωtrp,z mz !
m
z
where atrp ≡ (atrp,x atrp,y atrp,z ). Looking into scattering into transverse di-
rection, kf,z = 0, we get,
π~4 Ω4
dσ 2 2 2 2 2 2
= e−kf x atrp,x /2−kf y atrp,y /2−kiz atrp,z /2 .
dΩ ⊥ 8|Em − Ei |2
Looking into backscattering, kf,z = −ki,z ,
2
2 2
π~2 Ω4 −ki atrp,z )2 X (− 21 kiz atrp,z )mz
dσ 1
= e .
dΩ k 8 (Em − Ei )/~ − mz ωtrp,z mz !
m z
18.3.4 Exercises
18.3.4.1 Ex: Transition elements in anisotropic harmonic traps
a. Calculate and illustrate h0|e−ık·r̂ |0i for an anisotropic trap with cylindrical sym-
metry. P
b. Calculate hn|e−ık·r̂ |0i and nx ,ny ,nz hn|e−ık·r̂ |0i.
18.4. FURTHER READING 625
627
Chapter 19
Cooperativity in light
scattering
In Chp. 14.6.1 we discussed the interaction of light with individual atoms. In the prac-
tice of spectroscopy, however, we often work with ensembles of scatterers. Depending
on their spatial distribution (e.g. disordered, quasi-continuous, periodic), their motion
(hot gas or cold cloud), the possible existence of correlations between them, the pres-
ence of boundary conditions (e.g., free space, cavities, or photon bands), and finally, in
case of degenerate quantum gases, of a possible bosonic stimulation of the scattering
process, we expect new collective effects. In this chapter we will discuss several exam-
ples, in particular, localization effects induced by disorder, super- and subradiance,
Bragg scattering, and the formation of forbidden photonic bands in periodic lattices.
Phenomena leading to self-organization, such as the collective atomic recoil laser, will
be discussed in Chp. 22.6.1.
Figure 19.1: (a) Artist’s view of multiple scattering of a photon through a dilute cloud. (b)
Atomic cloud as a bulk object characterized by a refraction index n(r). (c) Illustration of a
photonic band in an optical lattice.
This chapter starts with introducing as the main theoretical tools the structure
factor and the coupled dipoles model. In Sec. 19.1 we will derive the formalism and
learn, how to employ it to calculate, e.g. the spatial distribution of light scattered by
an atomic cloud and the radiation pressure force on the atoms focusing on situations,
where interatomic interactions induced by the incident light can be neglected. In
Sec. 19.2 we will, disregarding the atomic’s cloud graininess and disorder, introduce
the smooth density approximation and compare it to macroscopic Mie scattering.
In Sec. 19.3 we will focus on cooperative effects inducing collective lineshifts and
broadenings, such as the Lorentz-Lorenz and the Lamb shift. We will also discuss
disorder-induced localization effects, super- and subradiance. In Sec. 19.2 we investi-
629
630 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
gate Bragg scattering, i.e. scattering from periodically ordered atoms, and finally in
Sec. ?? we study scattering from correlated atoms.
that is, the effective scattering cross section is reduced to the effective Rayleigh scatter-
ing cross section by an isolated atom times a geometric term called dynamic structure
factor,
Z
S(q, ω) ≡ 2π1
dteıωt hρ̂(q, t)ρ̂† (q, 0)i , (19.2)
where, Z
ρ̂(q, t) = n̂(r, t)eıq·r d3 r (19.3)
V
In the equations (19.2) and (19.3) we have written the quantities ρ̂(q, t) and n̂(r, t)
as operators, in order to allow for the possibility, that the atomic ensemble is a
quantum gas, i.e. a Bose-Einstein condensate characterized by a single wavefunction
ψ̂(r, t) normalized to the density distribution, ρ̂(r, t) = ψ̂ † (r, t)ψ̂(r, t). We shall return
to this subject in Sec. 28.1.1. In the following, we will restrict ourselves to atoms
localized in space with well-defined velocities and calculate the structure factor (i) for
disordered clouds of atoms excited by the passage of a photon in Sec. 19.1.6 and
(ii) for Bragg scattering by optical lattices in Sec. 19.4. In these cases, we find in the
literature often another definition of the structure factor as the expectation value of
the Fourier transform of the atomic density,
Thus, it describes the amplitude of the electric field of the scattered radiation. We
study the structure factor for various atomic density distributions in Excs. 19.1.8.1
to 19.1.8.3.
Example 81 (Structure factor of a discrete cloud ): In case of discrete
clouds, n(r, t) = j δ (3) (r − rj ), we can disregard the quantum nature of the
P
operators. The relationship (19.3) immediately gives,
X ıq·r (t)
ρ(q, t) = e j (19.6)
j
regime. In the multiple scattering regime other forms of collective phenomena, such as collective
Lamb shifts emerge, as we will see later on.
632 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
Atoms are supposed to be fixed in space. That is, we disregard the Doppler shift
of moving atoms and the photonic recoil. Therefore, it will suffice to consider
the static structure factor.
Atoms are initially uncorrelated and not degenerate. That means that we ne-
glect effects such as bosonic stimulation or Fermi blocking 2 .
Be ω0 the frequency of the incident light, ωa the frequency of the atomic resonance,
and ω the frequency of the scattered light. The Hamiltonian is nothing more than
the generalization of (15.123) to several assumed atoms located at the positions rj ,
N
X
Ĥ = ~gk0 (σ̂j e−ıωa t + σ̂j† eıωa t )(â†k0 eıω0 t−ık0 ·rj + âk0 e−ıω0 t+ık0 ·rj )
j=1
N . (19.9)
XX
+ ~gk (σ̂j e −ıωa t
+ σ̂j† eıωa t )(â†k eıωk t−ık·rj −ıωk t+ık·rj
+ âk e )
k j=1
√
Here, Ω0 = 2gk0 n0 is the Rabi frequency of the interaction between an atom and
the incident light (which is treated as a classical field with n0 photons), σ̂j is the
deexcitation
p operator for the j-th atom, âk is the photon annihilation operator, and
gk = d ω/(~0 Vph ) describes the coupling between the atom and the vacuum modes
the volume of which is Vph . The j-th atom has its lower and upper states denoted by
|gj i and |ej i, respectively. That is, we treat the atoms as simple two-level systems.
We also assume that all atoms are excited by the same unperturbed incident laser
beam, thus neglecting their dephasing along the laser path or induced by near-field
effects (which could arise for large spatial densities).
Within the rotating wave approximation RWA the Hamiltonian simplifies to,
N
X
Ĥ = ~ [gk0 σ̂j â†k0 eı(ω0 −ωa )t−ık0 ·rj + h.c.] (19.10)
j=1
N X
X
+~ [gk σ̂j â†k eı(ωk −ωa )t−ık·rj + h.c.] .
j=1 k
N
X
|Ψ(t)i = α(t)|0ia |n0 ik0 |0ik + βj (t)|jia |n0 − 1ik0 |0ik (19.11)
j=1
X
+ γk (t)|0ia |n0 − 1ik0 |1ik .
k
With this ansatz we imply that at every instant of time, there can be at most only
one excitation in the atomic cloud. The temporal evolution of the amplitudes is
obtained by inserting the Hamiltonian and the ansatz into the Schrödinger equation,
ı~∂t |Ψ(t)i = Ĥ|Ψ(t)i. Once the evolution of the amplitudes is calculated, we can
determine the observables of the system, such as the radiative pressure force or the
amplitudes of the scattered radiation fields or the fields inside the cloud.
Figure 19.2: Scheme of the interaction of a light beam with a sample of atoms.
such that,
[σ̂z , Ĥ] = ~g −eı∆0 t ↠σ̂ + e−ı∆0 t âσ̂ †
[σ̂z , [σ̂z , Ĥ]] = ~g eı∆0 t ↠σ̂ + e−ı∆0 t âσ̂ † = Ĥ
h i
[σ̂z , [...[σ̂z , Ĥ]]...] = ~g (−1)n eı∆0 t ↠σ̂ + e−ı∆0 t âσ̂ † .
3 This ansatz is well adapted to situations where the RWA holds. Otherwise, additional counter-
The time evolution of the amplitudes is obtained by inserting the Hamiltonian Ĥ and
the ansatz |Ψ(t)i into the Schrödinger equation,
∂ ı
|Ψ(t)i = − Ĥ|Ψ(t)i . (19.12)
∂t ~
√
one obtains with Ω0 = 2gk0 n0 ,
N
X
α̇(t) = −ı Ω20 eı∆0 t βj (t)e−ık0 ·rj (19.13)
j=1
X
β̇j (t) = −ı Ω20 α(t)e−ı∆0 t+ık0 ·rj − ıgk γk (t)e−ı∆k t+ık·rj
k
N
X
γ̇k (t) = −ıgk eı∆k t βj (t)e−ık·rj .
j=1
For low excitation rate, we can set α(t) ' 1. Integrating the third equation,
N
X Z t
0
γk (t) = −ıgk e−ık·rj eı∆k t βj (t0 )dt0 , (19.15)
j=1 0
X N
X Z t
0
β̇j (t) = −ı Ω20 α(t)e−ı∆0 t+ık0 ·rj − gk2 eık·(rj −rm ) eı∆k (t −t) βm (t0 )dt0 .
k m=1 0
(19.16)
19.1. THEORETICAL TOOLS AND MODELS 635
where we substituted t00 ≡ t − t0 in theRlast integral. Now, using the Markov approx-
t 0
imation β̃m (t − t00 ) ' β̃m (t), with lim 0 e−ı(ωk −ω0 )t dt0 = πδ(ωk − ω0 ), and with the
t→∞
rate of spontaneous emission,
Vph 2 2
Γ≡ k g , (19.19)
πc 0 k0
the third term becomes for the case m = j,
X Z t 00 X
gk2 e−ı(ωk −ω0 )t β̃j (t − t00 )dt00 ' gk2 β̃j (t)πδ(ωk − ω0 ) (19.20)
k 0 k
Z
Vph Vph 1 Γ
= β̃j (t) gk2 πδ(ωk − ω0 )d3 k = β̃j (t)4πgk20 πk02 = β̃j (t) .
(2π)3 (2π) 3 c 2
ThePthird term Rbecomes for the case m 6= j, evaluating the sum over the wavevectors
Vph
by k → (2π) 3 d3 k,
X N
X Z t
00
ı(k−k0 )·(rj −rm )
gk2 e e−ı(ωk −ω0 )t β̃m (t − t00 )dt00 (19.21)
k m6=j 0
X N
X
' gk2 eı(k−k0 )·(rj −rm ) β̃m (t)πδ(ωk − ω0 )
k m6=j
N Z N
Vph X 2 ı(k−k0 )·(rj −rm ) 3 Γ X
= β̃ m (t) gk e πδ(ω k − ω0 )d k = γjm β̃m (t) ,
(2π)3 2
m6=j m6=j
with
Z ∞ Z π Z 2π
2 Vph π
γjm ≡ gk2 eı(k−k0 )·(rj −rm ) δ(k − k0 )k 2 sin θdkdθdφ . (19.22)
Γ (2π)3 0 0 0 c
Finally,
N Z N
d Ω20 X t Γ X
β̃j (t) = ı∆0 β̃j (t) − β̃m (t0 )dt0 − γjm β̃m (t) , (19.23)
dt 4 m=1 0 2 m=1
or
N
d Ω0 X
α(t) = −ı β̃j (t) (19.24)
dt 2 j=1
N
d Ω0 Γ X
β̃j (t) = ı∆0 β̃j (t) − ı α(t) − γjm β̃m (t) .
dt 2 2 m=1
This means that the problem is reduced to finding the γjm . Continuing the eval-
uation of Eq. (19.22),
Z Z Z
1 2 Vph −ık0 ·(rj −rm ) ∞ 2π 1 ık|rj −rm | cos θ
γjm = gk 2 e e d cos θdφδ(k − k0 )k 2 dk
Γ 4π c 0 0 −1
Z
1 2 Vph −ık0 ·(rj −rm ) ∞ sin k|rj − rm |
= gk e δ(k − k0 )k 2 dk
Γ πc 0 k|rj − r m |
−ık0 ·(rj −rm ) sin k0 |rj − rm |
=e . (19.25)
k0 |rj − rm |
Isolating the self-decaying term and assuming low saturation, α(t) = 1, we get,
˙ Γ ıΩ0 Γ X
β̃j = ı∆0 − β̃j − − γjm β̃m . (19.26)
2 2 2
m6=j
In Exc. 19.1.8.4 we will analyze the validity of the Markov approximation for typical
cold atoms experiments.
The whole expression leads to the exponential kernel, the first two terms are the
cosine part, the third term is the sine part. The rotating wave approximation consists
in neglecting the first term, i.e. it only concerns the cosine part of the kernel.
19.1. THEORETICAL TOOLS AND MODELS 637
5,6
Within the RWA we got ,
ı sin(k0 |rj − rm |)
γjm = e−ık0 ·(rj −rm ) . (19.28)
ık0 |rj − rm |
Without the RWA we would have found [646, 641, 642, 695],
There has been a controversy between Friedberg and Scully about the role of
virtual photons (or collective Lamb-shift) [258, 694, 259, 643]. These terms result
from counterrotating terms in the rotating wave approximation. Scully assumes timed
Dicke states in infinitely large clouds and finds the contributions weak. Friedberg
does a mode expansion of the cloud and finds that different modes decay at different
velocities. This yields time-dependent radiation patterns, which can be temporarily
larger in backward direction.
Normally, the RWA is a good approximation, when Ω ω. Deviations from this
approximations lead e.g. to the Bloch-Siegert shift and important corrections for very
far-detuned (quasi-electrostatic) optical trapping.
√ The above requirement is not well
satisfied for our experiment, since ΩN = N Ω ' 1..10 THz.
Be |b1 b2 ...aj ...bN i the state with all atoms in the ground state except atom j being
in the excited state. Hence, the cloud’s state is simply expressed by the wavefunction,
N
X
Ψatom (t) = βj (t)|b1 b2 ...aj ...bN i . (19.30)
j=1
For large σ the radiation pressure is independent on the choice of the kernel.
In fact we may even set the kernel to 0. For small σ there appears a considerable
deviation. Interestingly, the imaginary part of the kernel gets important for higher
densities, even when the optical density is maintained, e.g. by compressing the cloud
in z-direction. This means that the collective Lamb shift becomes more apparent is
small compressed clouds. But we postpone a more thorough discussion to Sec. 19.3.
The analytic expansion into eigenmodes assumes the RWA. Hence, the numerics
deviate from the analytics for small σ and large N . Is it possible to generalize the
expansion to the exponential kernel (see [695])? The authors also suggest that the
scattered radiation be frequency-shifted due to the imaginary part of the kernel. This
might be an interesting observable for experiments.
they are included in the Hamiltonian when the RWA is NOT applied.
6 Note that Maxwell’s equations contain non-RWA terms.
638 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
the cloud means, that higher Dicke states are populated. Then we may expect a
complicated many-body dynamics, if decay into other states than the timed Dicke
state is possible.
Even though the single atom √ excitation rate may be small in case of large detuning,
the collective Rabi frequency N Ω can be large. The presence of many excitations
in the cloud means, that higher Dicke states are populated. Then we may expect a
complicated many-body dynamics, if decay into other states than the timed Dicke
state is possible.
The model we use (ground state + first excited state in the times Dicke basis) does
not allow more than 1 photon for N atoms. For such a model, neglecting saturation
means ’much’ less than 1 atom in the excited state, i.e. the probability of having the
any atom in the excited state is less than 1. But having N atoms, this means that
each atoms should have a excited state population of much less than 1/N . Including
the saturation in the naive way, means that when this term N Ω20 is not negligible,
than we will have less atoms in the first excited state of the timed Dicke basis, than
if when we would neglect saturation. This is precisely why we call this saturation,
we cannot take more atoms away from the ground state, because the system cannot
absorb more than one photon. But when this term is not longer negligible, then
in a cloud of N atoms, this does not prevent us from taking atoms away from the
ground state (we have N atoms which each can take one photon). Either the term is
negligible (and we could drop it) or we will try to keep its contribution (even at first
order in N Ω20 ), but then we cannot neglect the possibility of having 2 atoms excited.
d X ı
β̄j = Mjm β̄m − Ω̄0 (rj ) . (19.34)
dt m
2
19.1. THEORETICAL TOOLS AND MODELS 639
where the trick with the δjm -symbol helps to remove divergences for equal atom
positions.
In this form the solution is immediately suitable for numerical implementation of
the coupled dipoles model, although in practice the number of atoms is limited to
N < 10000 for ordinary PCs.
1
(μm)
0
y
-1
-8 -7 -6 -5 -4 -3 -2 -1 0
z (μm)
Figure 19.3: Phase delay in the excitation of the atomic dipoles (a) without rescattering
(γjm = 0) and (b) with rescattering.
19.1.4.5 Time-dependence
In order to calculate the time-dependence, we reconsider the equation (19.34) for the
excitation amplitudes. Its solution is formally given by the sum of the general solution
640 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
of the homogeneous equation and a particular (e.g. the asymptotic) solution of the
inhomogeneous equation,
XN
dâk 1
= [âk , Ĥ] = −ıgk eı(ωk −ωa )t σ̂j e−ık·rj . (19.41)
dt ı~ j=1
where the fast oscillating term proportional to eı(ωk +ωa )t has been neglected. Now,
X
Êsct (r, t) = Ek âk (t)eı(k·r−ωk t) , (19.42)
k
p
where Ek = ~ωk /2ε0 Vph . Integrating Eq. (19.41) with âk (0) = 0, inserting it in
Eq. (19.42), and approximating the sum over the modes k by an integral, we obtain,
N Z Z
Vph X t 0 0 ıωa t 0
Êsct (r, t) = −ı 3 dt σ̂j (t − t )e d3 k Ek gk eık·(r−rj )−ıckt . (19.43)
8π j=1 0
Introducing spherical coordinates, d3 k = dkk 2 dφdθ sin θ, and integrating the angular
part Eq. (19.43) becomes,
N Z t
Vph X 1 0
Êsct (r, t) = −ı 2
dt0 σ̂j (t − t0 )eıωa t × (19.44)
4π j=1 |r − rj | 0
Z
0 0 0 0
× dkkEk gk [e−ıck(t −|r −rj |/c) − e−ıck(t +|r −rj |/c) ] .
Assuming the radiation spectrum centered around k ' k0 , we approximate kEk gk '
k0 Ek0 gk0 . Then, extending the lower limit of integration of k to −∞, we obtain for
t < |r − rj |/c [572, 456],
N
dk02 X eık0 |r−rj |
Êsct (r, t) ' − σ̂j (t − |r − rj |/c) . (19.45)
4πε0 j=1 |r − rj |
where we may neglect the radiation retardation in the limit t σR /c. Using (19.19)
and expressing the coupling strength,
s
~ωk
~gk = dEk with Ek = , (19.46)
2ε0 Vph
19.1. THEORETICAL TOOLS AND MODELS 641
we get,
N
ı~Γ X
Êsct (r, t) ' − γoj (r)σ̂j (t) , (19.47)
2d j=1
eık0 |r−rj |
γoj (r) ≡ . (19.48)
ık0 |r − rj |
When applied on the state of Eq. (19.11), neglecting virtual transitions, it yields
Êsct |Ψi = Esct |g1 , ..., gN i, where Esct is the electric field radiated by the excited atoms.
Once the excitation amplitudes βj (∞) are known, the scattered light field and the
total field can easily be calculated via [618],
N
ı~Γ X
Esct (r) = hÊsct (r)i = − γoj (r)βj (∞) and Etot = Elas + Esct , (19.49)
2d j=1
Note that, for the chosen parameters, the result does not depend on the inter-
action terms. I.e. we can as well set the kernel to 0.
Note, that the phase factor eık0 ·r can either be attributed to the atomic dipole
moments or to the field. Here, N = 125 atoms are periodically arranged in a three-
dimensional cubic lattice.
(a) 0 (b)
1
(μm)
(μm)
-2
0
z
-4
y
-1
2
0 2 -2 0 2
-2 -2 0
x (μm)
y (μm) x (μm)
Figure 19.4: (code) (a) Geometry of light scattering from a cubic lattice with lattice constant
d = 100 nm. The light is irradiated from below and is partially reflected. (b) Intensity
distribution of scattered light along the yellow plane.
642 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
This is the expression for the force acting on an atom at the position rj . Knowing
the stationary excitation amplitudes βj (∞) and assuming α(∞) ' 1, the radiation
pressure force can numerically be calculated. Remembering Ω0 = 2gk0 we get for the
absorption force acting on atom j,
To evaluate the emission force acting on atom j, for every atom m of the sum, we
need to average over all possible scattering angles. To do so, we choose a reference
r −r
frame in which the z-component of k is directed along r̂jm = |rjj −rm
m|
, that is,
k = êx,jm k sin θjm cos φjm + êy,jm k sin θjm sin φjm + r̂jm k cos θjm , (19.57)
P Vph R 3
and evaluate the sum over the wavevectors by k → (2π) 3 d k,
N Z sin θjm cos φjm
X ∗ Vph π
Fej =− β̃j (∞)β̃m (∞) dθjm dφjm dk ~k sin θjm sin φjm × (19.58)
(2π)3 c
R 3
m=1 cos θjm
× gk2 [eı(k0 −k)·(rj −rm ) δ(k − k0 )k2 sin θjm − c.c.
N Z π Z 2π sin θjm cos φjm
Γ X
= −~k0 ∗
eık0 ·(rj −rm ) β̃j (∞)β̃m (∞) dθjm dφjm sin θjm sin φjm ×
8π m=1 0 0
cos θjm
× e−ık·|rj −rm | cos θjm sin θjm − c.c. .
V
ph 2 2
remembering Γ = πc k0 gk0 from Eq. (19.19). The integrals over φjm vanishes
whereas the integral over θjm becomes, using,
Z π
α cos α − sin α
dθ sin θ cos θe−ıα cos θ = 2ı = −2ıj1 (α) , (19.59)
0 α2
we find,
N
Γ X ık0 ·(rj −rm ) ∗
Fej = −~k0 e β̃j (∞)β̃m (∞)r̂jm × (19.60)
4 m=1
Z π
× cos θjm sin θjm e−ık|rj −rm | cos θjm dθjm − c.c.
0
N
Γ X ık0 ·(rj −rm ) ∗
= ı~k0 e β̃j (∞)β̃m (∞)r̂jm j1 (k|rj − rm |) − c.c.
2 m=1
N
Γ X ∗
= −ı~k0 fjm β̃j (∞)β̃m (∞) − c.c. ,
2 m=1
where we defined 7 ,
fjm = −j1 (k|rj − rm |)eık0 ·(rj −rm ) r̂jm . (19.61)
In summary we got,
7 For exploitation in MATLAB we may express the spherical Bessel function by a Bessel function
p
of the first kind: jn (x) = π/2xJn+1/2 (x).
644 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
The steady state absorption and the emission part of the radiation pressure force on
the center of mass of the atomic cloud follow from,
1 X 1 X
Fa = hF̂aj i and Fe = hF̂ej i . (19.63)
N j N j
0.5
Fx /h̄k0
Fx /h̄k0
0 0
-0.5
-1 -0.05
0 0.5 1 1.5 0 0.05 0.1
Fz /h̄k0 Fz /h̄k0
Figure 19.5: (code) Force distribution upon scattering of a photon by an atom of an spherical
Gaussian cloud (a) without and (b) with rescattering according to Eqs. (19.63).
As long as the RWA and the Markov approximation are valid and only the lowest
Dicke state is considered (no collective saturation), this simulation is supposed to be
exact and contains all the physics including, e.g. multiple scattering. Fig. 19.5 shows
a comparison of the cases when the off-diagonal components of the kernel (19.37)
are present or not. Apparently, the presence of rescattering dramatically spoils the
radiation pressure force, a phenomenon that we will extensively study in Secs. 19.1.7
and 19.2 [252, 34, 35, 565, 73, 36].
1
β̃j (t) = βj (t)eı(∆0 t−k0 ·rj ) ≡ √ β(t) , (19.64)
N
where β is the macroscopic dipole moment. Note, that the fact that β(t) does not
depend on the atomic position does not imply a continuous density distribution. The
atoms are still sitting at their positions rj ; only are their dipole moments synchronized
to the incident light wave.
The assumption of a timed Dicke state for the atomic cloud is, nevertheless, an
approximation which is not always good [252]. For example, it neglects dispersive
phase shifts of the excitation of the atomic dipole moments by the pump laser beam
19.1. THEORETICAL TOOLS AND MODELS 645
being delayed on its propagation due to its interaction with the atoms. If such phase
shifts (and absorption as well) are radially inhomogeneous, this can lead to deforma-
tion of the pump laser beam’s phase front and thus to lensing. We will discuss this
in Exc. 19.1.8.5 and in Sec. 19.2.
The evolution of the cloud very much depends on the initial conditions, e.g. βj (0) = 1
for the uniformly excited symmetric state βj (0) = eık0 ·rj for timed Dicke states.
In cylindrical coordinates the structure factor can be written,
N
1 X ı(kxj sin θ cos φ+kyj sin θ cos φ+(k cos θ−k0 )zj )
S(k, θ, φ) = e . (19.72)
N j=1
It basically tells the angular distribution of the scattered light. Fig. 19.6 shows nu-
merical calculations and analytical approximations of the structure factor for various
shapes and sizes of the atomic cloud. Obviously, the radiation pattern very much
depends on the size of the spherical cloud. For R < λ it is isotropic, for R > λ
scattering mainly occurs in forward direction. Furthermore, if the cloud is ellipsoidal
the radiation pattern is shifted into forward direction.
S(k, θ)
S(k, θ)
0 0 0
-1 0 1 -1 0 1 -1 0 1
θ/π θ/π θ/π
Figure 19.6: (code) Numerical calculation of S(k, θ) according to (19.68) for 104 atomic
positions rj chosen by a random number generator. For the graphs (a-b) the cloud is homo-
geneous and spherical and has the sizes R = λ (red) and R = 10λ (blue). For the graphs
(c-d) the cloud is Gaussian and spherical and has the same sizes as (a-b). For the graphs
(d-e) the cloud is homogeneous and ellipsoidal with aspect ratio Rz /Rρ = 3 and has the same
radial sizes as in (a-b).
19.1. THEORETICAL TOOLS AND MODELS 647
We are interested in the power emitted into the solid angle Ωk , P (t) ∝ |γk (t)|2 . For
timed Dicke state (19.64) we get immediately from the equations of motion,
N
X
γ̇k (t) = −ıgk eı∆k t βj (t)e−ık·rj (19.73)
j=1
XN
1 √
= −ıgk eı(∆k −∆0 )t √ β(t) eı(k0 −k)·rj = −ıgk eı(∆k −∆0 )t N SN (k) .
N j=1
We see that the time-dependence factorizes from the structure factor, which is the
only component containing an angular dependence. The same holds for symmetric
Dicke states defined by |βj | = |βm |.
10−7 50
Fa,e /h̄k0
10−8 0
β
10−9 -50
0 5 10 0 5 10
Γt Γt
Figure 19.7: (code) Time-dependent of (a) the dipole amplitudes and (b) the forces for
σ = 1, ∆0 = 20Γ, Ω0 = 0.01Γ, and η = 1.
The scalar electric field scattered by an arbitrary distribution of atoms has been
calculated in (19.49). Substituting the stationary timed Dicke state (19.69), we get,
√ N
ı~Γ N Ω0 X eık0 |r−rj |
Esct (r) = − . (19.74)
2d 2∆0 + ıΓN sN j=1 ık0 |r − rj |
Note that the ’timed Dicke’ state starts from the assumption of an infinitely extended
plane wave incident light field. This excludes situations where the incident beam size
is smaller than the cloud’s size.
648 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
using Ω0 = 2gk0 .
The factors sN and fN can be calculated exactly, as will be done in Exc. 19.1.8.6,
N
1 X sin(k0 |rj − rm |)
sN = cos[k0 (zj − zm )] (19.83)
N 2 j,m=1 k0 |rj − rm |
r N
π 1 X J3/2 (k0 |rj − rm |)
fN = k0 (zj − zm ) sin[k0 (zj − zm )] .
2 N 2 j,m=1 (k0 |rj − rm |)3/2
However, this is only practicable for atom numbers small enough for numerical sim-
ulations. For larger atom number we may use the analytic expressions including the
disorder term [74].
Finally, inserting the the expression (19.69) for the dipole moment, we obtain for the
timed Dicke state,
using Ω0 = 2gk0 . Inserting the expressions (19.67) for sN and (19.86) for fN , we may
also write,
Z π Z 2π
F0a + F0e (2∆0 /Γ)2 + 1 1
= N |SN (k, θ, φ)|2 (1 − cos θ) sin θdθdφ .
Fz1 (2∆0 /Γ)2 + N 2 s2N 4π 0 0
(19.90)
Let us assume in the following that the scattering of every single photon can
be treated independently [642]. In particular, the density distribution may change
between two scattering events. The force is something like the first moment of the
structure factor. This makes it so adapted to measure fluctuation-induced deviations
from the structure factor 8 .
S(θ, φ = 0)
S(θ, φ = 0)
0.3 0.333
0.2
0.25 0.332
0 0.2 0.331
0 0.5 1 0 0.5 1 0 0.5 1
θ θ θ
(d) 0.015 (e) 0.3 (f) 1
0.01 0.2
I(θ)
I(θ)
I(θ)
0.5
0.005 0.1
0 0 0
0 0.5 1 0 0.5 1 0 0.5 1
θ θ θ
Figure 19.8: (code) (a-c) Calculation of the structure factor |S(k = k0 , θ, φ = 0)|2 for an
R 2π
isotropic homogeneous density distribution. (d-f) Calculation of the force I(θ) = 0 |S(k =
k0 , θ, φ)|2 (1 − cos θ)dφ. The red curves show numerical calculations based on (19.68), the
green curves show analytical calculations according to (19.98). (a,d) k0 R = 1 and N = 100.
(b,e) k0 R = 5 and N = 100. (c,f) k0 R = 5 and N = 1000.
a collective effect coming from the structure factor of the cloud. We did NOT say that interatomic
interactions are essential...!
19.1. THEORETICAL TOOLS AND MODELS 651
ically,
Z ∞ Z π Z 2π
Vph I
hFunc,z i = σ(∆) ~k0 |SN (k, θ, φ)|2 k 2 sin θdθdφdk . (19.91)
(2π)3 ~ω 0 0 0
For such a cloud the structure factor is SN = N −1/2 . Note that this is unlike N
atoms in the Dicke limit, where SN = 1.
A dense homogeneous cloud with SN (k) ∝ δ 3 (k0 − k) does not scatter light and
experiences no force, hFhom i = 0. This is however not true any more in the limit of
small extended clouds, where fluctuations introduce disorder. This can be shown by
simulating a random atomic distribution rj and integrating the resulting force over all
possible k. For simplicity we assume a very sharp momentum distribution, |k| = |k0 |
V R∞
or 2πph2 0 k 2 dk = 1,
I
hFz i = σ(∆) k0 (sN − fN ) . (19.92)
~ω
Finally to compare with experiment we evaluate the ratio,
hFz i fN
=1− . (19.93)
hFunc,z i sN
We describe the
√ cloud as being made of two fractions: An isotropically scattering
fraction of N0 = N disordered atoms, whose structure factor is Siso (k) = 1, and
a forward scattering homogeneous cloud with structure factor Shom (k) = δk0 ,k . The
surface integration of the total structure factor,
N0 N − N0 N0 N − N0 N0
SN (k) = Siso (k) + Shom (k) = + δk0 ,k ' (19.94)
N N N N N
yields the same N -dependence of the force,
Z π Z 2π √
hFz i 1 1
= | N |2 (1 − cos θ) sin θdθdφ = . (19.95)
hFiso,z i 4πN 2 0 0 N
The interpretation is the following. In the experimentally realized situation, we
are very far in the large cloud limit completely dominated by forward scattering,
which means that if the cloud were homogeneous no radiation pressure force should
be expected at all. However diffuse scattering from disordered atoms (or fluctuations)
disturbs the forward scattering. It is this scattering which gives rise to radiation
pressure.
There is an interesting analogy: Diffuse scattering not only inhibits forward scat-
tering in homogeneous clouds, but also coherent backscattering from ordered struc-
tures. E.g. in optical lattices [662] it disturbs the detection of photonic band gaps.
In an optical lattice the atoms are in the Dicke limit and do not absorb photonic re-
coil. Hence, no displacement due to radiation pressure is expected. However, diffuse
scattering is observed as absorptive features in the spectra [76, 662].
The Bragg scattering is expressed by a periodic structure factor. A widely used
approach to describe the impact of disordered atoms in lattices is to divide the cloud
into a perfectly ordered part with density nfDW and isotropically scattering part
with density n(1 − fDW ) [662]. The factor fDW is known as Debye-Waller factor.
In Exc. 19.1.8.7 we try an alternative traetment of radiation pressure based on a
Monte-Carlo simulation.
652 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
19.1.8 Exercises
19.1.8.1 Ex: Structure factor of a linear array
a. Based on the definition (19.1) compute the structure factor of a linear array of
point-like scatterers.
b. Based on the definition (19.2) compute the structure factor of a linear array of 10
Gaussian density distributions.
19.1.8.5 Ex: Lensing by a dense atomic cloud with the coupled dipoles
model
Simulate the pump laser phase shift and lensing by a small dense cloud by the coupled
dipoles model for red and blue detuning. Discuss the influence of rescattering by re-
moving artificially the off-diagonal terms from the scattering kernel. Discuss whether
lensing is observed within the timed Dicke approximation.
superposition of directions into which the atom might have scattered. This is of course not described
by the simulation.
10 Although the atomic motion is frozen, it is not correct to say that the external degree of freedom
is not involved in the collective dynamics, since the spatial atomic distribution shapes the structure
factor.
19.2. CONTINUOUS DENSITY DISTRIBUTIONS AND MIE SCATTERING 655
that can be considered in numerical simulations to a few 1000. On the other hand,
most cold atom experiments nowadays are performed with 105 to 109 atoms. The
following section are devoted to introducing concepts and approximations allowing us
to understand light scattering from large atomic clouds.
The main step will consist in an approximation called smooth or continuous density
approximation, where the discrete distribution of point-like scatterers is replaced by
an inhomogeneous but smooth continuous density distribution characterized by a
refraction index field nrf r (r). The interaction of this density distribution with light is
then treated in the framework of Maxwell’s equations, i.e. inhomogeneities are treated
as macroscopic boundary conditions to the electromagnetic fields. In the following we
will term this regime as Mie scattering.
At first sight Rayleigh scattering from point-like particles and Mie scattering from
extended objects are quite different phenomena. Rayleigh scattering exhibits reso-
nances due to the internal structure of the particles, e.g. an atom. Mie scattering
shows resonances induced by the boundary conditions the scattering objects impose
to the field. On the other hand, from a microscopic viewpoint, any extended object
(e.g. a dielectric sphere) is nothing but an assembly of microscopic scattering parti-
cles. The question we need to study is then whether a description of the diffraction
from this object as the sum of the radiation patterns scattered from the individual
constituent particles is correct; or in how far the graininess of the cloud’s density dis-
tribution and cooperative effects arising from the interaction between the individual
particles play a role [123].
Interesting phenomena are expected in the transition regime between the limits
of a dense bunch of individual scatterers and macroscopic dielectric objects. One of
them is a strong modification of the radiation pressure force [158], which can conve-
niently be studied with atomic ensembles. The reason is that, in the smooth density
approximation, the cloud can be understood as a macroscopic object characterized
by a refraction index, which can be tuned over huge ranges by changing the cloud’s
density and volume, or by tuning the frequency of the incident light exploiting the
existence of atomic resonances. We will see in the following that it is possible to study
radiation pressure with cold atoms in the Rayleigh-Debye-Gans limit of small phase-
shifts, as well as in the Mie limit of large phase shifts [34]. Despite the absence of
sharp boundaries for the atomic cloud, we predict the occurrence of Mie resonances,
which could be detected experimentally [35].
In light scattering experiments, disorder (or granularity) plays a role when the number
of atoms projected onto a cross section perpendicular to the incident beam is small
enough so that a light mode focused down to the diffraction limit (that is ∼ λ2 ) would
be able to resolve and count the atoms. In other words, the stochastic fluctuations
induced by the random positions of the atoms can be neglected when the total number
of atoms N is larger than the number of modes ∼ σ 2 that fit into the cloud’s cross
section, i.e. when the optical density is b0 = 3N/σ 2 1. Under this hypothesis, the
differential equation (19.26) for β̃j can be simplified by replacing the discrete sum
656 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
In the Exc. 19.2.5.1 we calculate the structure factors for a homogeneous spherical
cloud of radius R and for a Gaussian ellipsoidal cloud with the rms-width σρ,z :
3
Shomog.sphere (k) = q 3 R3 (sin qR − qR cos qR) (19.98)
− 21 k2 σρ2 sin 2
θ− 12 σz2 (k 2
cos θ−k0 )
Sgauss.ellipse (k) = e .
The simulated structure factor (red curve in Figs. 19.10) agrees well with the analytical
expression (green curve). Since small clouds have a larger fluctuations, the fact that
the total force is a sum of intensities rather than amplitudes leads to a finite value at
large scattering angles θ.
(a) 1 (b) 0.1
S(k, θ)
S(k, θ)
0.5 0.05
0 0
-1 0 1 -1 0 1
θ/π θ/π
Figure 19.10: (code) (a) The red curve shows a calculation of the structure factor: |S(k =
k0 , θ, φ = 0)|2 for 200 atoms randomly distributed in a homogeneous spherical cloud of
size R = 10/k0 . The blue curve shows an average over 500 realizations of such a cloud
according to 200 2
P
m=1 |S(k, θ, φ = 0)| . The black curve is an analytical calculation according
to Eq. (19.98) [696]. (b) Same curves as in (a) but with a stretched y-axis.
In the continuous limit defined by (19.96) for the equations of motion (19.32) 11 ,
Z
˙ Γ ı Γ sin(k0 |r − r0 |) 0
β̃(r, t) = ı ∆0 + ı β̃(r, t) − Ω0 − d3 r0 ρ(r0 ) 0
β̃(r0 , t)e−ık0 ·(r−r ) .
2 2 2 k0 |r − r |
(19.99)
Transforming back to βj ≡ β̃eık0 ·r , we obtain the fundamental equation for the dipolar
excitation field,
Z
Γ ıΩ0 ık0 ·r Γ sin(k0 |r − r0 |)
β̇(r, t) = ı ∆0 + ı β(r, t) − e − d3 r0 ρ(r0 ) β(r0 , t) .
2 2 2 k0 |r − r0 |
(19.100)
11 Note that the ’timed Dicke’ assumption (19.64) has not been used here.
19.2. CONTINUOUS DENSITY DISTRIBUTIONS AND MIE SCATTERING 657
4πρ(r)
[∇2 + k02 n2rf r (r)]β(r) = 0 defining n2rf r (r) ≡ 1 − .
k03 (2∆0 /Γ + ı)
(19.101)
N Z
˙ Ω2 X t
β̃j (t) = ı∆0 β̃j (t) − 0 β̃m (t0 )dt0 (19.102)
4 m=1 0
X X N Z t
0
ı(k−k0 )(rj −rm )
− gk2
e e−ı(ωk −ω0 )(t−t ) β̃m (t0 )dt0 .
k m=1 0
and Z ∞
Vph −ıωτ
G(τ ) = e eıckτ k 2 I(k)dk , (19.106)
(2π)3 0
with ω = ck0 and νk = ck. The integral I(k) has been solved by Nicola for an
ellipsoidal Gaussian density distribution. It is quite close to,
r
π 1 F 2 /2 √
I(k) ' I(k0 ) ' e [1 − erf (F/ 2)] , (19.107)
2 kσz
150 1
20
100 0.5
0 I(k)
r/λ
-20 50 0
0 -0.5
0 0 -5 0 5 -0.02 0 0.02
z/λ ×10−6
N/λ (k − k0 )/k0
Figure 19.11: (code) Averaging of the structure factor over the whole k-space.
it is easy to show,
Z t Z t
∂
f (t0 )G(t − t0 )dt0 = f (t)G(0) + f (t0 )Ġ(t − t0 )dt0 (19.109)
∂t 0 0
∂n
and to thusly evaluate the integral until ∂tn G(t − t0 ) gets smooth enough to be ne-
glected.
kσr2 2
G(τ ) = e−(cτ /σz ) /2 , (19.111)
kσr2 + icτ
19.2. CONTINUOUS DENSITY DISTRIBUTIONS AND MIE SCATTERING 659
tm+1 = tm + dt (19.113)
m
X
βm+1 = βm − dt N gk2 βm0 G(tm − tm0 )dt .
m0 =1
0.5
0
β
-0.5
-1
0 10 20 30 40
Ωt
Figure 19.12: (code) Calculation of the time evolution of β for a homogeneous
√ spherical
cloud. The parameters are λ ' 780 nm and c/Ω ' 30 cm with Ω = N gk for typically
gk ' 1 MHz and N ' 106 .
With the solution of the integro-differential equation β(t), we can calculate the
probability that atoms are excited [696] 13 ,
X Z
2
P (t) = |β(t)| = |βj (t)| = d3 r|β(t, r)|2 .
2
(19.114)
j
exchanged between the cavity and the atomic excitation, we get Rabi oscillations. Here, the atomic
cloud is the cavity.
We know that the Mollow triplet in the dressed states picture gives rive to an oscillating decay curve
resembling that of Fig. 19.12. Does this curve also have an interpretation in terms of collective
dressed states?
660 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
tm+1 = tm + dt (19.117)
m
X N
X
βj (tm+1 ) = βj (tm ) − dt2 gk2 βj 0 (tm0 )Gjj 0 (tm − tm0 ) ,
m0 =1 j 0 =1
Vph R∞
with Gjj 0 (tm − tm0 ) = (2π)3 0
k 2 eı(ck−ω)(tm −tm0 ) Ijj 0 (k)dk.
As a first approach we will calculate the radiation pressure for a timed Dicke state
from Eq. (19.89) by explicit analytical integration of the surface-integrated structure
factors s∞ and f∞ in the smooth density limit (19.96) for an ellipsoidal Gaussian
smooth density distribution, as shown in Exc. 19.2.5.2. In the spherical case,
2
/2r̄ 2 N
ρ(r) = ρ0 e−r with ρ0 = , (19.118)
(2π)3/2 r̄3
662 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
0 2
(a) (b)
lg fN
lg(Fz12 /Fz0 )
1
-1
0
,
-2
-1
lg sN
Figure 19.13: (code) (a) Numerical evaluation of sN (blue circles) and fN (blue crosses) [123].
Analytical calculation of sN (black solid line) according to Eq. (19.121). Smooth density
limit of s∞ (red solid line) and f∞ (green crosses). The cloud is assumed spherical, η = 1, and
Gaussian with size σ = 5. (b) Numerical evaluation (blue circles) of the corresponding force
ratio
√ as a function of atom number N for ∆0 = (2π) 2 GHz and small collective saturation,
N Ω0 ∆0 . Analytical calculation in the smooth density limit (green solid line) and
according to Eq. (19.126) (black solid line). The magenta line traces the radiation pressure
force resulting from pump photon absorption only.
Fc 1 + N/Nce
' . (19.126)
F1 1 + (N/Ncr )2
Consequently, the radiation pressure is reduced with respect to its purely absorptive
component. This regime is still ruled by disorder, so that the critical atom number
for the impact of disorder on the absorption process, Nca , has no impact. It does not
even show up in the formula (19.126).
In the regime Nce < N < Ncr , disorder steps back and cooperativity wins, so
that the radiation pressure approaches the smooth density limit. Since in this regime,
the smooth density radiation pressure depends on atom number like ∝ N , as already
shown in Ref. [158], we observe an increase of the radiation pressure beyond the single-
atom value. This is only possible, because the collective enhancement of absorption
(magenta line) rises as fast with N , as the collective enhancement of emission.
For even higher atom numbers, Ncr < N , the radiation pressure dramatically
alters its N -dependency from ∝ N to ∝ N −1 . This change of behavior is not caused
by the interplay of cooperativity and disorder, but can be understood within the
framework of Mie scattering, as discussed in the next section.
(a) (b)
Δ 0 /2 π = 4 GHz
Mie
1 1
lg Fc /F1
lg Fc /F1
RDG
0.5 GHz
0.1 0.1
4 5 6 -1 0 1
lg N lg φ/2π
Figure 19.14: (code) (a) Double-logarithmic plot of the measured (symbols) and calculated
(lines) N dependence of the radiation pressure force. The red symbols and lines correspond
to the detuning ∆0 = (2π) 0.5 GHz and the intensity I0 = 95 mW/cm2 . For the blue
symbols and lines, ∆0 = (2π)4 GHz and I0 = 730 mW/cm2 . The red (bottom) and blue
(top) solid lines show calculations based on the full expression (19.120) without adjustable
parameters. The red (bottom) and blue (top) dotted lines represent just the fraction of
expression (19.120). The dash-dotted lines representing just the parentheses coincide for
both values of ∆0 . (b) Same data as (a), but plotted as a function of the push beam phase
shift ϕ. In this diagram the dotted lines representing the fraction coincide for both values
of ∆0 [62].
19.2. CONTINUOUS DENSITY DISTRIBUTIONS AND MIE SCATTERING 665
Depending on the parameter regime chosen, the measurements exhibited in Fig. 19.14
present data for reduced or enhanced radiation pressure for larger N . As explained
above, the reduction is understood as microscopic Rayleigh scattering at disordered
atoms together with superradiant acceleration of the decay, while the enhancement
is observed when the bulk cloud becomes so small and dense that it turns into an
inhomogeneous dielectric sphere refracting and lensing incident light by macroscopic
Mie scattering.
The question then remains why the radiation pressure, with increasing N , after an
initial rise the drops again. To answer this question we must have a look at the phase
shift induced in the pumjp light by the cloud’s refraction index.. From calculations
done in Secs. 1.2.7 and 24.6.1 we know that (below saturation) the optical density b
and the phase shift ϕ are linked to the refraction index nrf r and the optical scattering
cross section σopt via,
Z Z ∞
ıb ω ∞ 2∆0
+ϕ= [nrf r (r) − 1]dz = ı − σopt (∆0 ) ρ(r)dz (19.127)
2 c −∞ Γ −∞
2π Γ2 4πρ(r)
where σopt (∆0 ) = and nrf r (r) − 1 = − ,
k 2 4∆20 + Γ2 k03 (2∆0 /Γ + ı)
such that,
Z ∞
Γ2 2∆0
b = −σopt (∆0 ) ρ(r)dz = b0 and ϕ= b . (19.128)
−∞ 4∆20+ Γ2 Γ
R∞
Estimating coarsely k −∞
[nrf r (r) − 1]dz ' σ[nrf r (r) − 1], where σ ≡ kr̄, we may
simplify,
ıb
+ ϕ = σ[nrf r (r) − 1] . (19.129)
2
At high atom numbers (i.e. in the smooth density limit), the dependency of the radi-
ation pressure on atom number (19.126) exhibits a maximum at Ncr . For an atomic
cloud with density n and the resonant optical density b0 , the characteristic atom num-
ber can be expressed in terms of the phase shift ϕ experienced by the pump laser beam
on its path across the cloud. Absorption losses for the pump beam (e.g. resonance
fluorescence, whose differential cross section contributes σopt to the total scattering
cross section [456]) is completely negligible at large detunings. Only elastic Rayleigh
scattering occurs, which in the smooth density limit of Mie scattering becomes pure
diffraction (real part of the refraction index). We can than understand the atomic
cloud as a non-absorbing dielectric sphere with a Gaussian index of refraction.
Rewriting Eq. (19.126) in terms of the refraction index,
Fc N 2
' , (19.130)
F1 1 + (2ϕ)2 (2σ)4
we see that, as long as ϕ < 12 , the force increases linearly like ∝ N . The propor-
tionality comes from the perfect (in the sense of not spoiled by disorder) cooperative
enhancement of the scattering rate in the smooth density limit. The enhancement is
only limited by the finite size σ of the cloud, which is accounted for in the second
666 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
14
fraction. This regime, characterized by ,
σ[nrf r (r) − 1] 1 (19.131)
is termed the Rayleigh-Debye-Gans regime.
For larger phase shifts, ϕ > 21 , refraction more and more distorts the wavefront of
the pump beam, which spoils the pump mode depletion and hence reduces radiation
pressure. Consequently, the radiation pressure decreases again like N −1 . This is the
Mie regime of scattering. The maximum is thus a pure diffraction effect, a so-called
Mie resonance. It corresponds to the nrf r , where the Rayleigh-Debye-Gans scattering
approximation looses its validity according to (19.131). In Excs. 19.2.5.3 and 19.2.5.4
we study Mie and Rayleigh-Debye-Gans scattering, and in 19.2.5.5 we discuss the
question is whether recoil is imparted to individual atoms or to the center-of-mass of
the whole cloud.
In the limit of the approximations made, our formula correctly describes the ra-
diation pressure force on extended objects. These objects can either be ensembles of
scatterers like homogeneous, ordered or disordered atomic clouds of arbitrary shapes
and volumes, or macroscopic objects like dielectric spheres. The formula thus rep-
resents a bridge between microscopic Rayleigh scattering and macroscopic Mie scat-
tering. At very low atom numbers, the atomic cloud basically represents a randomly
distributed bunch of scatterers, whose intrinsic disorder spoils cooperativity. The
radiation pressure is then well described by the single-atom value. At large atom
numbers, the atomic cloud forms a smooth density distribution characterized by an
almost perfect a cooperativity, which is only limited by Mie scattering.
shifts induced by the cloud’s refraction index. The timed Dicke states might work
well for homogeneous cylinders, but not for ellipsoidal clouds, which we assume in our
analytical treatments. Therefore, a better approach consists in expanding the cloud
into spherical harmonics. Under the assumption that the cloud is radially symmetric,
ρ(r) = ρ(r), we get [252],
X∞ r
2n + 1
β(r, θ, t) = αn (t)jn (k0 r)Pn (cos θ)e−ık0 r cos θ , (19.134)
n=0
4π
1 1 1
(a) 10 (b) 10 (c) 10
Fa /h̄k0
Fe /h̄k0
Fc /h̄k0
100 100 100
Fe /h̄k0
Fc /h̄k0
100 10−1 100
Figure 19.15: (code) (a) Absorption, (b) emission, and (c) total force for a Gaussian cloud
with the following parameters: σ = 2, ∆0 = 10Γ, Ω0 = 0.001Γ, and η = 1. The magenta
dots show a fully numerical calculation for a randomly generated cloud according to formula
(19.63). The red lines show a calculation according to the complete formula (19.139). The
cyan dots show a calculation assuming timed Dicke states according to (19.82) via numerical
calculation of the structure factor according to (19.83) for the same randomly generated
cloud as for the full simulation. The blue lines are obtained within the timed Dicke state
approximation with structure factors estimated from formula (19.120). (d-f) Same as (a-c)
but with σ = 0.2.
As seen in Fig. 19.15, the results obtained via the complete analytical formula
(19.139) coincide with the fully numerical simulations according to formula (19.63).
But both disagree with analytical and numerical results obtained within the timed
Dicke state approximation.
It is possible to approximate the Bessel function In+1/2 (σ 2 ) and to obtain analyt-
ical solutions. The new formula works for σ 1 and η = 1 [252],
Ω20 σ 2 N 2 Γ2 sN Ω20 4N 2 Γ2 s∞ sN
Fa = ln 1 + 2 = ln 1 + (19.140)
Γ N σ (4∆20 + Γ2 ) 4N Γs∞ 4∆20 + Γ2
Ω2 σ 2 ∆ΓN Ω20 4∆0 ΓN s∞
Fe = −Fa + 0 arctan 2N = arctan 2
∆0 N 2 2 2
σ 4∆0 + Γ + 2σ2 Γ 4N ∆0 s∞ 4∆0 + Γ2 + 2Γ2 N s∞
3N
b0 =
σ2
1 1 1
sN = + s∞ = + .
N N (2σ)2
ˆ ˆ P
The radiated electric field E~(+) (r, t)|ψi where E~(+) (r, t) = k εk âk eık·r−ıωk t ,
ˆ X
E~(+) (r, t)|ψi = εk γk (t)eık·r−ıωk t |0i|1ik (19.141)
k
N
Vph gk0 εk0 k0 −ıω0 t X eık0 |r−rj | − e−ık0 |r−rj |
= −ı √ e βj (t)eık0 ·rj .
4πc N j=1
|r − rj |
We neglect the second term, which describes an incoming wave and go to smooth
densities. We expand into spherical harmonics and obtain,
∞
X Vph gk εk
ˆ λn
E~(+) (r, t)|ψi = e−ıω0 t √0 0 Ω0 (2n+1)ın h(1)
n (k0 r)Pn (cos θ) .
n=0
c N 2∆0 + ıΓ(1 + λ n )
(19.142)
It is an interesting question whether the maxima found in the curves of Fig. 19.15 can
be associated with Mie resonances [90, 667, 65]. To show this we need to apply the
formalism of Mie scattering to atomic clouds with the smooth density approximation,
where it is described by a continuous refraction index, as studied in Ref. [34]. The Mie
formalism had been developed for homogeneous spheres, which can be dielectric or
absorptive. For more general refractive index distribution the formalism gets quickly
cumbersome 15 . Also Mie resonances are generally though of being conditioned to
the existence of sharp boundaries.
The results of Ref. [34] surprisingly show that Mie resonances are expected for
parabolic distributions of atoms, as is the case for example for Bose-Einstein con-
densates in the Thomas-Fermi limit. However, the spectra of Mie resonances exhibit
much less structure. This is understood by the fact that ’whispering galery’ Mie reso-
nances may live on the surface of a sphere whose refraction index drops quadratically
to zero, while cavity type Mie resonances may not.
19.2.5 Exercises
19.2.5.1 Ex: Structure factor of a homogeneous spherical and of a Gaus-
sian ellipsoidal cloud
only that the dipole moment amplitudes and the incident light field are now considered
as vectors. The expression for the matrix Mjm is the same as in (19.36). However,
the vectorial kernel must be calculated in a different way. Using the abbreviations,
ı 1
rjm = rj −rm , rjm = |rj −rm |+δjm , cjm = − , (19.144)
k0 rjm (k0 rjm )2
the new vectorial kernel can be written,
(3)
2
3γjm (1 + cjm )rjm δmn − (1 + 3cjm )rjm r|jm
γjm = 2 . (19.145)
2 rjm
The steady-state scattered light field and the total field are now,
N
ıΓ ~j − (1 + 3coj (r))[(r − rj ) · β
3γoj (r) (1 + coj (r))k02 |r − rj |2 β ~j ](r − rj )
E~sct (r) = −
X
2
.
2 j=1
2 |r − rj|
(19.146)
A movie can be assisted at (watch movie).
19.4. SCATTERING FROM PERIODIC STRUCTURES AND PHOTONIC BANDS671
Only the second term makes speckles. However, the atomic motion smoothes out
the speckle pattern in all directions. Only in the presence of correlations in the
atomic positions they do not. E.g. in crystals or in the backscattering directions.
Therefore, the constructive interference depends strongly on the backscattering angle
ϑ. Fast atomic motion internal excitations can dynamically break the time-reversal
symmetry of the scattering path, esp. when the laser is close to resonance and the
phase delay per scattering process is long. Coherent backscattering is a weak form of
Anderson localization (also strong localization) of light.
19.3.1 Exercises
19.3.1.1 Ex: The green flash
Discuss whether the ’green flash’ at sunset could be due to superradiant extinction.
It can last seconds because refraction sweeps the ray through the spectrum (see the
movie Le rayon vert by Éric Rohmer).
19.3.1.2 Ex: Faraday effect with vectorial kernel in the coupled dipoles
model
Try to simulate the pump laser phase shift and lensing by a small dense cloud by the
coupled dipoles model for red and blue detuning. Discuss the influence of rescattering.
2. They modify the local density of states in a way to suppress spontaneous emis-
sion.
3. In optical lattices, unlike other systems, the scattering is very weak except when
the light is tuned close to atomic resonances. Consequently, the expected for-
bidden bands are very narrow. This bears the advantage that we can adjust
the optical density and tune the photon energy and the Bragg angle (i.e. the
quasi-momentum) independently, which facilitates the mapping of the disper-
sion relation.
where s(r) is the density distribution in an elementary cell 16 . We define the reciprocal
lattice by,
b×c c×a a×b
a∗ ≡ , b∗ ≡ , c∗ ≡ , V ≡ a(b × c) . (19.149)
V V V
The Fourier transform of the density is,
X
ρ(u) = ρhkl δ(u − la − mb − nc) ? s(r) , (19.150)
l,m,n
We consider a wave being scattered from the direction k0 towards the direction
ks ≡ k0 + ∆k. The structure factor ρ describes the amplitude of the radiation field
scattered by the lattice. If the density distribution is the same at all sites of the
lattice, X
n(r) = n1 (r) ? δ (3) (r − R) , (19.153)
R
X ım∆k·R Z
ρ(∆k) = e n1 (r)eı∆k·r d3 r (19.155)
m V
ıNs ∆kR
1−e
Z
2
/2r̄ 2 −y 2 /2r̄ 2 −z 2 /2r̄ 2 ı∆k·r 3
= × n0 e−x e e e d r
1 − eı∆kR V
2 2 2 2
≈ Ns δ(∆k − 2π/R)n0 (2π)3/2 r̄3 e−6∆k r̄
= N δ(∆k − 2π/R)e−6∆kx r̄ .
16 This is similar to the square of the Bloch function defined in Sec. 5.1
674 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
that is, for perfect periodicity, introducing the density distribution of a unit cell and
a 1D lattice.
The Debye-Waller factor describes the diffusion of the density over the sites of the
lattice due to the thermal motion,
Z X
ρhkl = n1 (r)eı∆k·r d3 r eı(jêz λdip /2+u)∆k (19.158)
V j
X
= fi eı(jêz λdip /2+u)∆k = ρhkl eıu∆k ,
j
with 2
eıu∆k ≈ e−|G| u2 /6
. (19.159)
2 2
Example 88 (Optical lattice): The exponential distribution e−6∆kx r̄ is called
Debye-Waller factor and describes the smearing out of the population over the
lattice due to the thermal motion of the atoms. The δ-function sets the wavevec-
tor of emitted light. That is, the power of light is only emitted in particular
directions given by the Bragg condition. Here, the solid angle ∆Ωs does not
depend on the thermal distribution (as long as the atoms are within the Lamb-
Dicke regime), but on the lattice size, which determines the goodness of the
approximation of the Airy function (the sum in the above equation) by a Dirac
δ-function. The √ width of the Airy function for a lattice of size w0 is approxi-
mately ∆k = 2 3/w0 . With this the solid angle is,
12
∆Ωs = . (19.160)
k2 w02
We can also estimate the solid angle from the diameter d of the Gaussian beam
at a distance x away from a scattering medium of size w0 ,
s 2
λx
d = w0 1 + (19.161)
πw02
πd2 λ2 4π
∆Ωs = = = 2 2 .
x2 πw02 k w0
The power scattered into this solid angle is,
1. The structure factor treatment assumes low optical density, which is not neces-
sarily guaranteed when the laser is tuned close to a resonance [661].
2. As the structure factor is independent of the laser detuning, it will not reveal
any spectral structure, such as a band-gap or dip due to diffuse scattering. Also,
absorption is not incorporated into the model.
With the lattice constant d = 12 λdip the interference is constructive when the
difference of the paths of two beams reflected by different layers is a multiple of the
wavelength,
1 1
2 λdip cos βi + 2 λdip cos βs = λ . (19.165)
676 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
the incident electric fields with the transmitted and reflected ones,
or in S-matrix notation,
Ez+ Ez−
= Sβ , (19.172)
E0− E0+
where
ıβ 1 + ıβ r t
Sβ = = , (19.173)
1 + ıβ ıβ t r
t = 1 + ıβ and r = ıβ , (19.174)
which satisfy |t|2 − |r|2 = 1. Recasting the problem into the T -matrix notation,
+ +
Ez E0
= T β , (19.175)
Ez− E0−
where
1 1 + 2ıβ ıβ 1 t2 − r2 r
Tβ = = . (19.176)
1 + ıβ −ıβ 1 t −r 1
We note that the employed model is classical, since we describe the gas by a sequence
We still need to multiply with the transfer matrix describing the propagation of
the incident beam through the layer,
eık·d 0
Td = , (19.178)
0 e−ık·d
such that the transfer matrix for passing the beam through an atomic layer is,
T = Tβ Td . (19.179)
For Ns layers we must obviously concatenate the matrices MNs . The reflection
coefficient is now simply,
(T Ns )12
rNs = , (19.180)
(T Ns )22
and can be evaluated numerically or analytically [198]. If the incident beam hits the
cloud under an angle, k · d = kd cos θ = kz d, and furthermore, as shown in [200]
Eq. (A10), the theory must be generalized by replacing,
β
β→ (cos−1 θ + cos θ) . (19.181)
2
eikz d ıβe−ıkz d
T = Tβ Td ' , (19.182)
−ıβeikz d e−ıkz d
with k · d = kd cos θ = kz d. Near the Bragg angle we have cos θ ' cos θbrg = λsp /λdip
and near resonance we have 2π/k = λ ' λsp , so that, with d = λdip /2, we obtain
kz d ' π and,
Ns Ns
Ns eıπ ıβe−ıπ −1 −ıβ Ns 1 Ns ıβ
T ' = = (−1) ,
−ıβeıπ e−ıπ ıβ −1 −Ns ıβ 1
(19.183)
which we will verify in Exc. 19.4.4.3. The total reflectivity being rNs ' Ns ıβ, we get
the reflection,
Ns2 βsp
2 2
Γsp
|rNs |2 ' . (19.184)
4∆ + Γ2sp
2
The intensity reflection profile, therefore, is a Lorentzian also in the limit of thick
clouds. The additional condition for the occurrence of prohibited photonic bands is
a large number of atomic layers, Ns 1. We note that the model is an extension of
the Kronig-Penney model introduced in Sec. 5.3.
680 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
superradiance.
19.4. SCATTERING FROM PERIODIC STRUCTURES AND PHOTONIC BANDS681
1. At the limit of thin lattices, we do not expect multiple scattering. The absorptive
(imaginary) part of polarizability dominates, β ≈ Im β. Thus, the reflection
coefficient is almost real, reıφ ≈ |r|, the phase shifts are negligible, the profile
of the reflection spectrum is symmetric. In this scheme, the interference of the
radiation patterns of individual atoms is destructive in all directions, except
under the Bragg angle. There are losses caused by scattering in non-paraxial
modes. They are also due to elastic Rayleigh scattering but, because of the
finite Debye-Waller factor, the radiation becomes diffuse and incoherent.
2. At the limit of thick lattices, we have multiple scattering. The (real) dispersive
part of the polarizability is β ≈ Re β. Thus, absorption is suppressed, we
observe large phase shifts and the reflection spectrum profile is asymmetric.
Multiple beam interference gives rise to global scattering.
Physically, the set parameters consistent of the quasi-momentum and the energy
of the Bloch wave, (Θ, ∆), is more relevant because it allows analyzing the dispersion
relation 22 . We observe the existence of energies ∆, where the real part of the quasi-
momentum vanishes (modulo π). The 3D representation in Fig. 19.18(e) illustrates
the occurrence of an avoided crossing due to the band gap at the edge of the Brillouin
zone.
22 We note, that Θ via the relation (19.186) of the angle of incidence θ, but only weakly of the
energy ∆.
682 CHAPTER 19. COOPERATIVITY IN LIGHT SCATTERING
Now we separate the entire structure into two parts T = T (b) T (a) , such that the field
between the two parts is,
+
E1 1
= T (a)
E0+ . (19.196)
E1− −T21 /T22
The local density of states (LDOS) in a photonic band can be evaluated from
[714, 278], X
N (ω) = 2ω δ(ω 2 − ωk2 ) . (19.198)
k
dkef f (ω)
N (ω) = . (19.200)
dω
19.4. SCATTERING FROM PERIODIC STRUCTURES AND PHOTONIC BANDS683
100
local intensity
10−2
Figure 19.19: (code) The intensity decreases exponentially over a infinite lattice. For a finite
lattice (here Ns = 3000), the intensity approaches a constant value at the end of the lattice.
The graph shows in blue the intensity along the lattice, in red the exponential decay due
to absorption in a homogeneous cloud (Lambert-Beer law), in green the hyperbolic decay
following Ohm’s law and in cyan the transmission assuming a lattice of N layers.
19.4.2.6 Impurities
The formula (19.197) allows the calculation of the impact of localized defects in nu-
merical simulations. For example, it is instructive to look at the intensity profile along
the structure in the presence of a localized lattice defect. In Fig. 19.20(c) we observe
an intensity peak located at the 100-th atomic layer, exactly where the defect was
introduced. This peak corresponds to a localized evanescent wave. Photons can not
propagate freely through the lattice: they prefer to tunnel between lattice defects the
rather than propagate by radiation.
The curves (a-b) of Fig. 19.20 illustrate how an empty photonic band fills up with
localized states with well-defined energies when noise is added to the periodicity of
the lattice. This situation is similar to that in semi-conductors doped by donor and
acceptor states.
1. Firstly, being intrinsically 1D, the model does not apply to 2D or 3D lattices.
local intensity
5000 5000
Δ/Γ
Δ/Γ
0 0 0.2
-5000 -5000
0
-0.03-0.02-0.01 -0.01 0 0.01 0 200 400
Re Θ Im Θ N
Figure 19.20: (code) (a,b) Dispersion relation as in Fig. 19.18, but in the presence of ran-
domly distributed defects. (c) Intensity profile along the lattice with a defect located in the
100esima atomic layer.
In contrast, the coupled dipoles model offers several advantages. An optical lat-
tice is, after all, nothing more than a periodically ordered diluted sample of atoms.
Hence, we can apply the coupled dipoles model introduced in Sec. 19.1.2 [619, 618].
The advantages of this model are its applicability to 3D systems and finite and disor-
dered lattices. It also allows the inclusion of all kinds of inhomogeneities such as, for
example, the spatial intensity distribution of a focused incident laser, or the deviation
of the laser beam penetrating an atomic cloud due to refraction. On the other hand,
the model only lends itself to heavy numerical simulations, limiting it to some 10000
atoms.
The Bloch model is another model to describe 3D periodic systems [22]. It was in-
troduced in Chp. 5.4 for 1D optical lattices. Its disadvantages are that it supposes
infinite lattices and the absence of defects.
The Bloch model is commonly used to describe the scattering of electron waves in
a solid, where the band gap originates from the Coulomb interaction of the electron
with the atoms of the solid crystal. In contrast, in photonic crystals, the modes with
high (low) frequency ω concentrate their energy in spatial regions with low (high)
dielectric index . Close to geometric (Mie) resonances this causes a repulsion in
the density distribution of photonic states, and the opening of a gap separating high
frequency bands (air bands) from low frequency bands.
In optical lattices the photons interact with the atomic resonances. Photons and
electrons are distinguished by their different dispersion relations and by the fact that
electronic waves are scalar and photonic waves are vectorial. The electrons of a crystal
are bound to an energy surface and follow the lines of the dispersion relation without
leaving the metal. In contrast, photons are usually injected into the structure, which
gives an additional degree of freedom. Therefore, any point in the phase diagram
can be reached and the dispersion relation only informs, where at which point the
transmission is stronger.
19.4. SCATTERING FROM PERIODIC STRUCTURES AND PHOTONIC BANDS685
Figure 19.21: Attempt of an artistic illustration of the isoenergetic surfaces for electrons in
a metal (left), photons in a 3D optical lattice (center), and photons in a 1D optical lattice
(right). The interaction of atoms with the lattice is much weaker than the interaction of
electrons with the metal, except close to resonances. Therefore, photonic bands are much
narrower than electronic bands.
19.4.4 Exercises
19.4.4.1 Ex: Factor structure of a 1D lattice
Calculate the structure factor of the density distribution (19.163) and the square
|Sk0 (k)|2 .
R. Röhlsberger et al., The collective Lamb shift in nuclear c-ray superradiance [DOI]
C.E. Máximo et al., Spatial and temporal localization of light in two dimensions
[DOI]
W.R. Garrett et al., Large Multiple Collective Line Shifts Observed in Three-Photon
Excitations of Xe [DOI]
M. Weidemüller et al., Local and global properties of light-bound atomic lattices in-
vestigated by Bragg diffraction [DOI]
D. Petrosyan, Tunable photonic band gaps with coherently driven atoms in optical
lattices [DOI]
Jin-Hui Wu et al., Controlling the photonic band structure of optically driven cold
atoms [DOI]
K. Ishizaki et al., Realization of three-dimensional guiding of photons in photonic
crystals [DOI]
D. Petrosyan, Tunable photonic band gaps with coherently driven atoms in optical
lattices [DOI]
Figure 20.1: Scheme of (a) a linear cavity and (b) a ring cavity with optical modes α±
pumped by incident light fields η± and decaying into the void with rate κ.
691
692 CHAPTER 20. COUPLING OF ATOMS AND OPTICAL CAVITIES
with [âω , â†ω0 ] = δω,ω0 . The part of the Hamiltonian Ĥlaser:cavity describes pumping
of the cavity with an external light field matched to the cavity mode. The standard
procedure consists in setting up the quantum Liouville equation for the total density
operator ρ̂total = ρ̂cavity ⊗ ρ̂bath , assuming a quantum Wiener process, and tracing
the over the bath’s degrees of freedom [269, 152]. From this procedure, as shown in
Sec. 16.2.4, we derive a master equation for the reduced density operator ρ̂.
The inevitable coupling of the cavity to the environment, described by Ĥcavity:bath ,
leads to irreversible losses. These losses can be described as spontaneous decay to the
continuous vacuum heat bath. The irreversibility of the process is readily understood
in terms of the phase space offered by the vacuum being much larger. In fact, whenever
discrete states are coupled to a continuum, the equilibrium is very much on the side of
the continuum, that is, discrete states decay. Since an optical cavity can be considered
20.1. LIGHT FIELDS IN CAVITIES WITHOUT ATOMS 693
1
n̄ = , (20.2)
1− e−~ω/kB T )
Lbath ρ̂ = κ2 (n̄ + 1)(2âρ̂↠− ↠âρ̂ − ρ̂↠â) + κ2 n̄(2↠ρ̂â − â↠ρ̂ − ρ̂â↠) (20.3)
† † † 2 † 2
+ β(2n̄ + 1)(2â âρ̂â â − (â â) ρ̂ − ρ̂(â â) ) .
We have already shown this in Exc. 16.2.6.1. The constants are the cavity decay rate
κ and the pump rate β. Note that at room temperature in the microwave regime,
n̄ may be as large as a few hundred photons, whereas in the optical regime we may
neglect the thermal excitation,
or, using the electric field normalized to the amplitude of the field generated by a
single photon, E~cav+
= E~1 α,
h i
E~cav (z, t) = Re E~1 eı(kz−ωt) α(t) + E~1 eı(−kz−ωt) α(t) (20.7)
The result (20.7) shows that, letting η = 0, we see that the cavity field decays with
the time constant κ from the stationary situation. κ also corresponds to the HWHM
2
of the field intensity, |α(|∆c | = κ)| = 12 |α(0)|2 . Note, that the intensity decays as
√
2κ, and the HWHM of the field amplitude is α(|∆c | = 3κ) = 21 |α(0)|.
Re α
0
-1
-2
0 1 2 3 4 5
κt
Figure 20.2: (code) Transient oscillations in a cavity pumped out of resonance.
Frequently, we are interested in the light reflected from a cavity. The reflective
response of the cavity to an incident pump beam Ein is,
E~ref
+ ~+ ~+
l = rin Ein + tin Ecav . (20.8)
Also we may want to consider a temporal variation of the input field, for example of
its detuning, ∆c (t). In such cases, the equation (20.5) can not be integrated easily,
and we need to resort to numerical methods. The simplest (and least convergent)
method in a Newtonian iteration like,
h i
E~ref
+
l (t + dt) = ~+ (t) + dt tin (−κ − ı∆c (t))E~cav
Eref l
+
(t) + η ~+ (t) .
Ein (20.9)
We will study in Exc. 20.1.5.1 how a sudden change of the pump laser detuning may
lead to ringing with a fixed frequency ∆c . In Exc. 20.1.5.2 we will show, how it is
possible to empty an optical cavity in times shorter than 1/κ.
The two counterpropagating field modes will, provided they have the same polar-
ization, interfere and form a standing light wave. In contrast to linear cavities, the
phase of this standing wave is represents a degree of freedom, as it depends on the
phases of the two field modes α± , which in turn can be controlled by the incident
20.1. LIGHT FIELDS IN CAVITIES WITHOUT ATOMS 695
laser fields η± . To better understand the behavior of the phase as a function of the
pump beams, we analyze the interference of the two modes, which is described by,
η ∗ α+ (0) ∗
η+ η ∗ η+ η ∗
η+ α− (0)
∗
α+ α− = ∗
α+ (0)α− (0) − − − + 2 − 2 e−2κt + 2 − 2 + (20.12)
κ − ı∆c κ + ı∆c κ + ∆c κ + ∆c
∗
η+ η ∗ η+ η ∗
η− η+
+ α+ (0) − 2 − 2 e(−κ−ı∆c )t + α− ∗
(0) − 2 − 2 e(−κ+ı∆c )t .
κ − ı∆c κ + ∆c κ + ı∆c κ + ∆c
∗ η2
α+ α− = (1 − eıφ )2 e−2κt + e2ıφ + 2eıφ (1 − eıφ )e−κt cos ∆c t . (20.13)
κ2 2
+ ∆c
∗ η2 2
α+ α− = 2
(1 − e−κt )eıφ + e−κt . (20.14)
κ
This results shows that the cavity phase adjusts itself to the pump field in exponential
time κ−1 :
∗
Im α+ α− (1 − e−κt )2 sin 2φ + 2(1 − e−κt )e−κt sin φ
tan θ = ∗ = . (20.15)
Re α+ α− (1 − e−κt )2 cos 2φ + 2(1 − e−κt )e−κt cos φ + e−2κt
For small phase slips φ π, this reduces to,
For small amplitude oscillations, we can expand the pump term into a Fourier series
of Bessel functions,
Z T
α± (t) = e(−κ−ı∆c )t α± (0) + η e(κ+ı∆c )τ ± φ0 (κ+ı∆c )τ +ıΩτ
2
e ∓ φ0 (κ+ı∆c )τ −ıΩτ
2
e dτ
0
At low frequencies, Ω κ, the phase of the cavity field goes as, θ(φ0 ) ' 2φ0 sin Ωt,
κ
and for high frequencies, Ω κ, as, θ(φ0 ) ' −2φ0 Ω cos Ωt. Thus, we observe a
low-pass behavior of the phase excursions of the cavity field with a cut-off frequency,
κ−1 :
2φ0
θm ' . (20.24)
1 + Ω/κ
πδf sr
κ = τκ−1 = . (20.26)
F
20.1. LIGHT FIELDS IN CAVITIES WITHOUT ATOMS 697
The intensity decay rate of the cavity, measured by ’cavity ring-down’ is κint = 2κ.
Note, that κint is also the FWHM width of the intensity transmission spectrum 1 ,
such that the finesse
δf sr
F = (20.27)
κint /2π
is simply the ratio between the free spectral range and the FWHM of the cavity
intensity transmission curve, both measured in Hertz.
For a cavity with a given geometry filled with a Gaussian mode of light with power
P , the intensity is determined by Gaussian optics 2 ,
s 2
2P 2 2
−2ρ /w (z) λz
I(r) = e and w(z) = w0 1 + . (20.28)
πw2 (z) πw02
R
Defining the mode volume via I(0)Vm ≡ I(r)dV and evaluating the spatial integral
over the Gaussian mode along the cavity, we obtain,
π
Vm = Lw02 . (20.29)
4
Defining the amplitude of the electric field generated by a single photon via,
where n is the number of photons in the cavity, we calculate for the energy stored in
the cavity,
Z Z Z
I(r) 1 2P 2 2
ε0 |E~1 (r)|2 dV = ε0 ρdρdφdz = 2
e−2ρ /w (z) ρdρdφdz (20.31)
nε0 c nc πw (z)
1 P 1 n~ωδf sr ~ω
= L= L= .
nc 2 nc 2 2
Hence, s s r
I(0) 1 2P 1 Ln~ωδf sr
|E~1 (0)| = = = .
nε0 c nε0 c πw02 nε0 c 2Vm
Inserting the expression (20.25) for the free spectral range of a cavity we get,
r
~ω
|E~1 (0)| = . (20.32)
2ε0 Vm
1 See script on Electrodynamics (2020), Sec. 7.3.5 and Exc. 7.3.6.17.
2 See script on Electrodynamics (2020), Sec. 7.4.1 and Exc. 7.4.3.1.
698 CHAPTER 20. COUPLING OF ATOMS AND OPTICAL CAVITIES
which will be derived in Excs. 20.1.5.3 and 20.1.5.4, and the anisotropy by,
where Ωcav is the solid angle covered by the cavity mode. The formula,
expresses that the density of states is nothing more than the structure factor of the
cavity.
8πb2
Ωcav,con ' . (20.44)
L2
Hence,
1 w0 h −2(x2 +y2 )/w2
√ √ i
F 1
~ω umn (r)
= F e Hm ( 2x/w)2 Hn ( 2y/w)2 (20.48)
~ω w
1 w0 h 2 2 √ i h 2 2 √ i
= F e−2x /w Hm ( 2x/w)2 F e−2y /w Hn ( 2y/w)2
~ω w
1 w0 −2kx2 w2 √ 2 2 √ m,n=0 1 w0 −2(k2 +k2 )w2
= e Hm ( 2kx w)2 e−2ky w Hn ( 2ky w)2 −→ e x y .
~ω w ~ω w
Finally,
Z
umn (r) k
Rmn (êk ) = F d kkx d ky (20.49)
~ω
Z √ Z √
1 w0 2 2 2 2 k
= e−2kx w Hm ( 2kx w)2 d kx
e−2ky w Hn ( 2ky w)2 d ky .
~ω w k
1 a2 k2 w02 1 πa2
X 1 π X π
F umn (r) = 2 2 = = .
mn
~ω 2k w 2 /2L2
2 2
2k w 13.72 2L 2 13.72 4L2
13.72 m,n<a2 k2 w0
20.2. INTERACTION OF ATOMS WITH CAVITIES 701
20.1.5 Exercises
20.1.5.1 Ex: Quick ullage of an optical cavity
Consider a linear cavity resonantly pumped by a laser beam until a stationary state
is reached. Suddenly, the phase of the incident light is changed by 180◦ . Based on
equation (20.6), analyze the evolution of the light field inside the cavity.
The quantization of the electromagnetic field has been presented in Sec. 15.1.1 for
the case of plane wave radiation modes. In the following, we want to generalize the
treatment to arbitrary field modes characterized by mode functions ukλ (r) labeled by
a wavevector and a polarization. These are classical vector functions satisfying the
vector Helmholtz equation and the transversality condition,
with kλ = ωkλ /c [503]. These classical functions are chosen to form an orthonormal
set,
Z
u∗kλ (r) · uk0 λ0 (r)d3 r = δkk0 δλ,λ0 . (20.54)
V
~λ
ufkλree (r) = √ eık·r , (20.55)
V
where ~kλ is a polarization vector such that k·~kλ = 0 and V is the photon quantization
volume.
Let us now rewrite the quantized transverse vector potential of the radiation field
(15.7) as,
Xq †
∗
Â(r, t) = 2ωkλ ε0 [ukλ (r)âkλ (t) + ukλ (r)âkλ (t)] .
~
(20.56)
kλ
The quantum properties of the electric and magnetic field operators are determined by
the bosonic annihilation and creation operators, âkλ (t) and â†kλ (t), respectively, with
usual commutation relations: [âkλ (t), âk0 λ0 (t)] = 0 and [âkλ (t), â†k0 λ0 (t)] = δkk0 δλλ0 .
For weak-coupling between the atoms and the field, one has the contributions to
20.2. INTERACTION OF ATOMS WITH CAVITIES 703
the Hamiltonian,
Ĥatom = ~ω0 σ̂ z
X
Ĥf ield = ~ωk [â†kλ (t)âkλ (t) + 12 ]
, (20.57)
k
X
Ĥatom:f ield = −ı~ (σ̂ + + σ̂ − )[gkλ (r)âkλ − gkλ
∗
(r)â†kλ ]
k,λ
and, r
ωk
gkλ (r) ≡ d · ukλ (r) (20.58)
2ε0 ~
is a complex function associated with the coupling strength between the atom and
the field.
By solving the Heisenberg equations of motion for the atomic and field operators
within the Born and Markov approximations (or using the procedure leading to the
expression (15.135)), one obtains the spontaneous emission rate on a transition |ei →
|gi of frequency ω0 : X
Γ(r) = 2π |gkλ (r)|2 δ(ωk − ω0 ) . (20.59)
k,λ
which is the same result obtained by the Weisskopf-Wigner theory [503]. To evaluate
the sum we need go to a continuous k-space via,
X X V Z
−→ lim d3 k . (20.60)
V →∞ 8π 3
k,λ λ
Plugging in the density of states for free space (1.53),we finally get,
X Z |d · ~kλ |2 E~2
k
Γ(r) = 2πδ(ωk − ω0 )ρ(ωk , k)dΩk dωk , (20.62)
~2
λ
where the density of states can now be arbitrarily shaped by the presence of boundary
conditions, such as optical cavities or dielectric or metallic surfaces.
Example 92 (Two atoms): Let us now consider two excited atoms labeled
by q = 1, 2, located at r1 and r2 , and described by two of their eigenstates,
{|gq i, |eq i}. As usual, |gq i is the eigenstate with lowest energy (Egq = − 21 ~ωq )
704 CHAPTER 20. COUPLING OF ATOMS AND OPTICAL CAVITIES
and longer lifetime, i.e. the ground state, whereas |eq i is the eigenstate with
highest energy (Eeq = + 21 ~ωq ). In presence of an external electromagnetic field,
the Hamiltonian of the two atom-field system in the electric dipole approxima-
tion is,
(q) (q)
X
Ĥ = [Ĥatom + Ĥatom:f ield ] + Ĥf ield ,
q=1,2
(q)
where Ĥatom ≡ ~ωq ŝzq
is the atomic Hamiltonian, with ŝzq ≡ 12 (|eq iheq |−|gq ihgq |)
(q)
being the energy operator of the q-th atom, and Ĥatom:f ield ≡ −d̂q · E(r ~ˆ q ) is
the atom-field interaction Hamiltonian. The electric dipole moment operator
satisfies heq |d̂q |eq i = 0 = hgq |d̂q |gq i and has non-vanishing off-diagonal ele-
ments, i.e. the eigenstates have no permanent dipole moment. We define the
dipole-moment matrix element as dq ≡ heq |d̂q |gq i. The description of the field
is identical to the case of one atom (20.55).
For weak-coupling between the atoms and the field, the interaction Hamiltonian
becomes,
X X +
Ĥatom:f ield = −ı~ (ŝq + ŝ− ∗ †
q )[gkλ (rq )âkλ − gkλ (rq )âkλ ] ,
k,λ q=1,2
where ŝ+ −
q ≡ |eq ihgq | and ŝq ≡ |gq iheq | are the electric dipole raising and low-
ering operators, respectively, satisfying the well-known commutation and anti-
− ± + −
commutation relations: [ŝ+ z z ±
q , ŝq 0 ] = 2ŝq δqq 0 , [ŝq , ŝq 0 ] = ±ŝq δqq 0 , and [ŝq , ŝq 0 ]+ =
± 2
0 with (ŝq ) = 0.
By solving the Heisenberg equations of motion for the atomic and field opera-
tors within the Born and Markov approximations, one obtains the spontaneous
emission rate on a transition |eq i −→ |gq i of frequency ωq :
X πωq X
Γq = 2π |gkλ (rq )|2 δ(ωk − ωq ) = |dq · ukλ (rq )|2 δ(ωk − ωq ) ,
ε0 ~
k,λ k,λ
natural width due to spontaneous decay and the line shift are calculated by integration
over the coupling force between the atom and every available field mode [see (20.61)],
ZZ
|deg · ~k |2 |E~k |2
Γ = 2πδ(ω0 − ωk )ρ(ωk , k)dΩk dωk
~2
, (20.64)
X Z Z |dei · ~k |2 |E~k |2 1
∆ω = ρ(ωk , k)dΩk dωk
i
~ 2 ωei − ωk
p
with the field amplitude per photon, |E~k | = ~ωk /2ε0 V , derived in (20.32), and the
index i running over all internal atomic states. These formulas are simply applications
of second-order perturbation theory (6.16) and of Fermi’s golden rule (6.112) for
transition probabilities both weighed with dynamic structure factor of the cavity.
We use the number of modes per unit of frequency range and per unit of solid
angle in free space derived in (20.39) (ρ/V is the density of states),
V ωk2
ρf ree (ωk ) = . (20.65)
(2π)3 c3
Setting θk as the angle between the atomic dipole moment and the cavity axis (which
is not in place, yet, as we are still in free space), d · ˆk = d cos αk = d cos(90◦ − θk ) =
d sin θk ,
Z
d2 sin2 θk ~ω0 V ω02
Γf ree = 2π sin θdθdφ (20.66)
~2 2ε0 V (2π)3 c3
Z π
d2 k 3 d2 k 3
= sin3 θdθ = .
4πε0 ~ 0 3πε0 ~
Since the integration covers the whole solid angle of space 4π and the free space is
isotropic, there is no preferred orientation, and we can perform the integration setting
θ ≡ θk . The result coincides with earlier calculations (13.43).
Figure 20.3: Illustration of cooperativity in a cavity. An excited atom can decay by emitting
light into a cavity mode or into free space. Note that the cavity is not pumped directly by
an incident laser beam.
Now, we put the cavity in place, but tune it off resonance, such that no photons
can be emitted into the solid angle covered by the cavity. That is, in the presence of
the cavity, the solid angle of free space into which photons can be emitted is restricted.
We repeat the calculation (20.66), but now we are not free to choose the axis of the
706 CHAPTER 20. COUPLING OF ATOMS AND OPTICAL CAVITIES
coordinate system arbitrarily. Instead, we we assume that θk (θ, φ) ' const over the
small solid angle covered by the cavity, θ, φ ∈ Ωcav ,
Z
d2 sin2 θk ~ω0 V ω02
Γf ree∈cav
/ = 2π sin θdθdφ (20.67)
êk ∈cav
/ ~ 2 2ε0 V (2π)3 c3
Z
d2 k 3 sin2 θk 3 2
' Γf ree − sin θdθdφ = Γf ree 1 − Ωcav sin θk .
8π 2 ε0 ~ êk ∈cav 8π
Finally, we study the situation, when the cavity is in place but now tuned on
resonance. Then the density of states in the photon emission directions k ∈ cavity is
modulated by the Airy function (20.40),
ρcav (ωk , k) = ρf ree (ωk )L(ωk ) . (20.68)
That is, photons can be emitted into the cavity mode with increased probability.
Once again we repeat the calculation (20.66), but now inserting the density of states
of the cavity,
Z
d2 sin2 θk ~ω0 V ω02
Γcav = 2πL(ω 0 ) sin θdθdφ (20.69)
êk ∈cav ~2 2ε0 V (2π)3 c3
Z
d2 k 3 sin2 θk 3
' L(ω 0 ) sin θdθdφ = Γf ree L(ω0 ) Ωcav sin2 θk ,
8π 2 ε0 ~ êk ∈cav 8π
where, in the last step, we substituted the free space decay rate (20.66). An analogous
calculation for the cooperative Lamb shift ∆ωcav is left to the Exc. 20.2.6.1:
Γcav = 3
Γf ree 8π Ωcav sin2 θk L(ω0 )
0 . (20.70)
∆ωcav = 3
Γf ree 32π Ωcav sin2 θk LL(ω
(ω0 )
0)
2δf sr
The quantities Γcav and Γf ree∈cav are scattering rates into complementary solid
angles. Now, the total spontaneous emission rate, which determines the lifetime of
the radiating excited state and the spectral width of the transition is simply the sum
of the partial scattering rates,
Γ = Γf ree∈cav
/ + Γcav (20.71)
3
2 3 2
= Γf ree 1 − 8π Ωcav
sin θk + Γf ree L(ω0 ) 8π Ωcav sin θk
3
2
= Γf ree 1 + (L(ω0 ) − 1) 8π Ωcav sin θk .
Assuming for simplicity, θk = π/2, we find the extremes of the spontaneous emission
rate, Γenh and Γinh , when the cavity is on or off resonance. With a resonant high
finesse cavity the emission is obtained via kL → 0,
q
2
Γenh = Γf ree 1 + 1 + 2Fπ − 1 3 3
8π cav ' Γf ree 1 + 4π 2 F Ωcav . (20.72)
Ω
For a small solid angle, we can expect a big increase of the scattering into the cavity,
but without noticeable inhibition of the total decay, Γinh ' Γf ree .
If the atoms are saturated by an incident laser, they scatter light into the cavity
at a rate (apart from a factor 12 ),
3
Γcav = Γf ree L(ω0 ) 8π Ωcav . (20.74)
Without cavity the emission into the same solid angle is obtained via F → 0,
3
Γf ree∈cav = Γf ree 8π Ωcav . (20.75)
The scattering will fill the cavity with photons, until the leakage equalizes the
pumping. When the balance is reached, Γcav will also be the rate at which photons
are emitted by the cavity mode. We calculate an example in Exc. 20.2.6.2.
3F 3F 8πb2
Υ' Ω cav = , (20.76)
4π 2 4π 2 L2
where b is the clear aperture of the cavity mirrors. That is,
F 6b2
Υ= . (20.77)
π L2
that is,
F 6
Υ= , (20.79)
π k 2 w02
8π F 6 F δf sr
Ωcav = , Υ= , = , (20.80)
k 2 w02 π k 2 w02 π κ
which we will call the atom-field coupling strength, we may rewrite the prefactor
3
Γf ree 8π Ωcav for the case of Gaussian cavity modes,
3 g2
Γf ree Ωcav = . (20.82)
8π 2δf sr
π
Inserting this into the formulae (20.70), where we set θk = 2,
g2 κ
Γcav ' ∆2a +κ2
. (20.84)
−2g 2 ∆a
∆ωcav ' ∆2a +κ2
Obviously, the atom-field coupling strength g plays a central role in the modification
of the emission spectra of atoms interaction with cavities. Therefore, we will have to
study this quantity more deeply in the next sections.
(a) 10 ×10
4
(b) (c) 0.5 (d) 1
2
Δωcav /Γfree
Γcav /Γfree
intensity
ρcav /ρfree
5 1 0 0.5
×105
0 0 -0.5 0
-5 0 5 -5 0 5 -5 0 5 -5 0 5
Δa /κ Δa /κ Δa /κ Δa /κ
Figure 20.4: (code) Spontaneous emission on the strontium atom cooling transition (λ =
689 nm, Γ = (2π) 7.6 kHz) in a L = 3.6 cm long cavity with F = 110000 and w0 = 50 µm.
(a) Cooperativity L(ω) from Eq. (20.40) as a function of the detuning of the cavity.
(b) Linewidth Γcav from (20.70)(ii) and (c) frequency shift of the strontium transition from
Eq. (20.70)(ii). (d) Spontaneous emission into the cavity (blue) and into the open space
(red).
atoms, the light power scattered into the cavity is Pj = 20 fW. Estimation
20.2. INTERACTION OF ATOMS WITH CAVITIES 709
the cavity transmission by T ' π/F , we expect that a number of T P~ω cav
=
107 s-1 photons can be detected by a photodetector recording the photon number
leaking out of the cavity.
where E~1 (0) the electric field produced by a single photon inside the cavity mode
volume Vm calculated in (20.29) and Γ is the spontaneous decay rate. See also (watch
talk).
σ0 6 2g 2
β0 = 2
= 2 2 = . (20.88)
πw k w δf sr Γ
The frequency shift accumulated during a round trip in the cavity, δf sr β0 , becomes
noticeable, when it exceeds the linewidth of the cavity κ. From this condition, we
obtain the optical depth for a single passage through the atomic sample multiplied
by the finesse of the cavity, which is precisely the cooperativity parameter,
δf sr β0 F 6 2g 2
Υ≡ = = . (20.89)
κ π k2 w2 κΓ
The sensitivity to the atom number can be measured in terms of a critical atom
number Ncrt , which the system can resolve. It goes to infinity as the resonant single
atom reflection coefficient disappears, β0 → 0, and converges to zero when β0 → 1,
4π k2 w2 Γδf sr 2κΓ 4
Ncrt = = 4π = 2π = 2 = . (20.90)
F β0 6F F g2 g Υ
The total reflection coefficient is increased by the number of atoms N and the
number of photon round trips F in the cavity, which gives out of resonance,
6 Γ g 2 kV L N U0
N F β∆ ' N F = NF = N U0 F = , (20.91)
k2 w2 2∆a ∆a ωπw2 c κ
where we introduced the light-shift produced by a single off-resonant photon and the
Rayleigh scattering rate produced by a single photon,
g2 g2 Γ
U0 ' and γ0 ' . (20.92)
∆a ∆2a
p
Using the dipole moment d = 3πε0 ~Γ/k 3 and defining the electric field am-
plitude per photon E1 by the photon number n, the power P , and the intensity
I = ε0 cnph E12 = 2P/πw02 , we calculate for the single-photon Rabi frequency (or atom-
field coupling strength),
s s s
dE1 1 3πε0 ~Γ I 1 3πε0 ~Γ 1 2P 6 P
g= = = = Γ .
~ ~2 k 3 ε0 cnph ~2 k 3 ε0 cnph πw02 k 2 w02 nph ~ω
(20.93)
The single-photon light shift (or atom-atom coupling strength) is,
g2 6 Γ P P
U0 = = 2 2 = β∆ , (20.94)
|∆a | k w0 |∆a | nph ~ω nph ~ω
20.2. INTERACTION OF ATOMS WITH CAVITIES 711
g2 Γ 6 Γ2 P
γ0 = = . (20.95)
∆2a k 2 w02 ∆2a nph ~ω
In a cavity the flux δf sr = P/nph ~ω is just the photon round trip rate (or free
spectral range). 1/δf sr is the time a photon interacts with an atom. In free space,
this does not make sense to me, because I don’t know how define the mode volume. I
can however consider the interaction of a photon with an extended sample of length
L. In this case, a photon interacts with the sample for the time L/c, so that,
g2 6 Γ c c
U0 = = 2 2 = β∆ , (20.96)
|∆a | k w0 |∆a | L L
and
g2 Γ 6 Γ2 c
γ0 = 2
= 2 2 2 . (20.97)
∆a k w0 ∆a L
This means, we can use the CARL equations derived for a ring cavity by just substi-
tuting κ = δf sr = c/L, where L is now not the cavity but the sample length.
Lambert-Beer’s law relates the amplitude of the field transmitted to the incident (see
also Eq. (24.138)),
Z ∞
E~t ı ∆a
= exp ıσ(∆a ) − n(r)dz = e−b/2 eıϕ . (20.98)
E~0 2 Γ −∞
I.e. the field is shifted by an amount ϕ and absorbed by an amount b. The expo-
nent (called optical density) can be rewritten, if we assume a number of N atoms
homogeneously distributed over a cylinder with length L and radius r̄ = w0 ,
Z ∞ Z L
ı ∆a Γ2 1 ı∆a N
OD = ıσ(∆a ) − n(r)dz = σ0 2 − − 2L
dz (20.99)
2 Γ −∞ ∆ a 2 Γ 0 πr̄
Γ2 1 ı∆a N 6N Γ2 1 ı∆a
= σ0 2 − − = − −
∆a 2 Γ πw02 k 2 w02 ∆2a 2 Γ
2
6N 1Γ ıΓ L
= 2 2 − − = N (−γ0 − ıU0 ) .
k w0 2 ∆2a ∆a c
Hence, the parameters N γ0 and N U0 are nothing else that the optical density per
photon round trip time δf sr = c/L,
−N (γ0 + ıU0 )
OD = . (20.100)
δf sr
712 CHAPTER 20. COUPLING OF ATOMS AND OPTICAL CAVITIES
g2
s= . (20.101)
Γ2
Therefore, the number of photons needed to saturate an atomic transition is,
Γ2 Γ k2 w2 1
nsat = 2
= = . (20.102)
2g δf sr 6 2s
We see, that there is a symmetry between Υ and s, that is, between Ncrt and nsat .
The regime N Υ > 1 denotes the collective behavior of N atoms in the same way as
nsat > 1 indicates saturation. While Υ depends only on the phase matching between
the atomic antenna and the cavity, s also depends on the cavity mode volume and
the natural decay rate.
The equations of motion (20.10) describe the light fields classically. Hence, it is not
surprising that they can be derived with entirely classical arguments. Here, we will
show this for a ring cavity in the presence of an atom sitting at a fixed position
z on the cavity’s optical axis. We understand the atom as a beamsplitter located
within the mode volume of the resonator and partially reflecting and transmitting
incident light with reflection and transmission coefficients (rβ , tβ ) = (ıβ, 1 + ıβ) such
that rβ2 + t2β = 1. Transmitted (forward scattered) photons may be phase-shifted,
while reflected photons are backscattered into the counterpropagating mode of the
resonator.
Similarly, we treat the incoupling mirror as a beamsplitter with coefficients (rin , tin ).
The incident field Ein produces, in the cavity, field amplitudes of α± for the co- and
counterpropagating waves. As in (20.7), we normalize the amplitudes by the numbers
of photons n± ≡ |α± |2 . After a round-trip time τ = δf−1 sr through the mode volume
we have in the position of the incoupling mirror the field,
L is the total length of the ring cavity. Obviously, we have kL = ω/δf sr . In the
vicinity of a resonance we have, ∆c δf sr , and the quantity ω/δf sr is almost integer,
ω ' 2πN δf sr − ∆c , such that we can expand the exponential, eıkL = 1 − ı∆c /δf sr .
Thus, we obtain,
2 −2ıkz
τ α̇+ = − 1 − rin (1 + ıβ)eıkL α+ + ıβrin e α− + tin α+ in
(20.106)
τ α̇− = − 1 − rin (1 + ıβ)eıkL α− + ıβe2ıkz α+ + tin (1 + ıβ)α−
in
.
We now connect the transmission of the coupling mirror tin with the decay constant
κ assuming that the light can only leave the cavity through this mirror. We define,
T
κ≡ (20.107)
τ
714 CHAPTER 20. COUPLING OF ATOMS AND OPTICAL CAVITIES
√ p
as
p the part of the light lost during one round trip. Thus, tin = T ' π/F =
κ/δf sr . Besides that, tin is very small, such that,
t2in = 1 − rin
2
' 2(1 − rin ) . (20.108)
It gives the cavity losses for the two modes during one round trip. We assume here that
losses can only occur via the coupling mirror. However, all losses can be included in a
single appropriate κ. There are usually other losses due to scattering on the surface of
the mirrors or absorption by the atoms. Finally, we obtain for weak atomic reflection
and in resonance, that is, for β κ and ∆c = 0 the system of equations,
p
α̇+ = −κα+ + ıδf sr β(1 − t2in )e−2ıkz α− + κδf sr α+ in
(20.110)
2ıkz
p in
α̇− = −κα− + ıδf sr βe α+ + (1 + ıβ) κδf sr α−
~Ω(r)2
φ(r) = , (20.114)
4∆a
The Rabi frequency Ω(r)2 = 4g 2 |α+ eıkZ + α− e−ıkZ |2 is normalized to the frequency
of Rabi generated by a photon g. Using the frequency shift (light-shift) by photon
g2
(20.92), U0 = ∆ a
, we can also write,
In the above derivation we assumed, for simplicity, the pump laser on resonance
with the cavity, ∆c = 0. Relaxing this condition, an analogous derivation yields 3 ,
mode. This approximation is only good far from resonance, ∆a Γ. We will see in Sec. 22.1.3, how
to generalize the equations of motion for near resonance cases.
716 CHAPTER 20. COUPLING OF ATOMS AND OPTICAL CAVITIES
~˙ = W α
ıα ~ + ı~η , (20.125)
where α~ and ~η regroup the amplitudes α± and η± . This equation takes the form of a
Schrödinger equation, where,
U0 − ∆c − ıκ U0 e−2ıkz
W ≡ (20.126)
U0 e2ıkz U0 − ∆c − ıκ
This normal mode splitting 4 of the cavity results from the coupling of the two
cavity modes â†+ â− . Obviously, the energies and widths of the eigenvalues do not
depend neither on the pump intensities η± nor the z-position of the atom. On the
other hand, the spectral behavior of α± , and hence the observable quantities, such as
the transmission
κα± 2
T± ≡ (20.128)
η±
depend on these parameters. In the Exc. 20.2.6.8 we will calculate the spectrum by
the condition 0 = ∂∆c α± (∞).
In the following, we will discuss two relevant cases: unidirectional pumping and
symmetrical pumping.
χ ıU0 e2ıkz
α+ (∞) = η+ and α− (∞) = η+ . (20.129)
χ2 + U02 χ2 + U02
These formulas show that, for weak coupling, U0 κ, the counterpropagating mode
receives little light. On the other side, for strong coupling (or very high finesse) and
∆c = 0, the intensity is equally distributed, |α+ |2 = |α− |2 = η2κ
+ 5
. We calculate the
splitting of normal modes in Exc. 20.2.6.9.
The counterpropagating modes form, by interference, a standing light wave giving
rise to a dipole potential in the form of a one-dimensional optical lattice. Defining
the phase θ through,
and,
1
2ε0 cE12
ı(z, t) = 1
E12
E + (z, t)E − (z, t) (20.132)
2 2
= |α+ (t)| + |α− (t)| + 2|α+ (t)||α− (t)| cos(2kz + θ) .
That is, the phase indicates the positions of the potential maxima. Inserting the
stationary solution for unilateral pumping (20.129), we derive the expression,
∗
Im α+ (∞)α− (∞) Im χıe−2ıkz κ cos 2kz + (U0 − ∆c ) sin 2kz
tan θ = ∗ = = .
Re α+ (∞)α− (∞) Re χıe−2ıkz κ sin 2kz − (U0 − ∆c ) cos 2kz
(20.133)
Or in other words, the phase of light determines the equilibrium position of the atom
(or vice versa). Two cases are interesting: (i) For ∆c = 0 and U0 κ the condition
(20.133) turns into tan θ = − tan 2kx. In this case, the phase of the backscattered
field adjusts in such a way, that the atom stays at the valleys of the anti-nodes.
(ii) For ∆c = U0 (or alternatively, when κ U0 , ∆c ) the condition (20.133) turns
into tan θ = − tan(2kx+ π2 ), such that the atom is at half height of the potential slope,
exactly at the position, where it is able to backscatter the maximum of photons from
the pumped mode α+ to the mode α− .
η+ Us
α+ (∞) ' and |α− (∞)| = |α+ (∞)| .
κ κ
Looking at short times, we find that the dynamics of the unpumped mode is
delayed, since,
giving,
t2 −2ıkrs
α+ (0) ' η+ t and α− (0) ' −ıUs η+
e .
2
The complete solution of the equations (20.128) with unidirectional pumping,
η− = 0, will be derived in Exc. 20.2.6.10.
φ=0 uni
-2 -2 -2 -2
0 0 0 0
2 -5 2 -5 2 -5 2 -5
5 0 5 0 5 0 5 0
φ = π/4
-2 -2
0 0
2 -5 2 -5
5 0 5 0
φ = 2π/4
-2 -2
0 0
2 -5 2 -5
5 0 5 0
φ = 3π/4
|α+ |
-2 -2
U0 02 0
2
5 0 -5 5 0 -5
Δc
Figure 20.6: (code) (columns 1 and 2) Steady state field values according to (20.122) for
bidirectional pumping with a single scatterer located at φ. (columns 3 and 4) Same for
unidirectional pumping. Here the location of the scatterer does not matter.
The potential is calculated in the same way as in (20.132). The coherences are,
∓ıkz ±ıkz
∗
α± (∞)α± (∞) = η 2 (κ−ı∆c ∓2U 0e sin kz)(κ+ı∆c ∓2U0 e
(κ2 +2U0 ∆c −∆2 )2 +4κ2 (U0 −∆c )2
sin kz)
c
2 ∓ıkz 2
+(∆c ∓2ıU0 e
∗
α± (∞)α∓ (∞) = η 2 (κ2κ+2U sin kz)
0 ∆c −∆ ) +4κ (U0 −∆c )
2 2 2 2 , (20.135)
c
α∓ α∗
e±2ıkz = ±
|α− α∗
+|
such that the relative phase of the counter-propagating waves adjusts itself to,
∗
Im α+ (∞)α− (∞) κ2 sin 2φ + 4U02 sin2 (kz + φ) sin 2kz
tan θ = ∗ = 2 . (20.136)
Re α+ (∞)α− (∞) κ cos 2φ − 4U02 sin2 (kz + φ) cos 2kz
The quantity η 2 /κ2 denotes the number of intracavity photons. According to the
formula (20.136), for a weak atom-field coupling, U0 κ, the phase adjusts itself to
the external pumps, θ → 2φ, while for strong coupling, it adjusts to the position of
the atom, θ → 2kz.
φ = π/2,
κ + ıU0 ± U0 e∓2ıkz
α± (∞) = ±ıη (20.137)
κ2 + 2ıκU0
−4U02 cos2 kz sin 2kz −8U02
θ(∞) = arctan 2 2
' arctan 2 kz ,
2
κ + 4U0 cos kz cos 2kz κ + 4U02
expanding the last formula around kz = 0. Fig. 20.7 shows how, with the
increase in the coupling force between the field and the atoms, U0 , the phase
tends to lock to the atomic position. However, the phase imposed by the external
pump competes for this privilege. Curiously, this is independent of laser power,
but depends only on the ratio between U0 and κ.
0.5
(π)
0
θ
-0.5
-0.5 0 0.5
kz (π)
√
Figure 20.7: (code) ’Pulling’ of the phase by the atoms for U0 /κ = 0, 0.2, 0.5, 1 and 2.
The solid curves show the exact phase, the ’dash-dotted’ curves the linear approximation
(20.141).
making the transition to the Lamb-Dicke regime by z → 0. The phase of the standing
wave is,
∗ 2 2
Im α+ (t)α− (t) z→0 α0+ − α0−
tan θ = ∗ (t) −→ sin 2U0 t . (20.141)
Re α+ (t)α− 2α0+ α0−
We see that, in resonance and without pumping, the field adjusts its phase to the atom
and also decays with the rate κ, while the atom redistributes the photons between
modes with the (Rabi-)frequency 2U0 . The formula (20.141) does not show any
damping of the phase dynamics in the Lamb-Dicke regime. Thus, in the absence of
pumping, the cavity dissipation reduces only the field amplitudes, but does not damp
the adjustment of the phase to the atomic position.
~ (∞) = −ıW −1 ~η .
α (20.142)
This particular solution of the inhomogeneous equation, added to the general solu-
tion of the homogeneous equation, gives the general solution of the inhomogeneous
Schrödinger equation,
~ (t) = e−ıW t α
α ~ (0) + (1 − e−ıW t )~
α(∞) . (20.143)
where,
−κ − ıU0 −ıU0 J0 0 −ıU0 J1
L= and X= . (20.147)
−ıU0 J0 −κ − ıU0 ıU0 J1 0
P (n) (n)
We insert the ansatz α± = n α± eınωt , where α̇± = 0, into the equations and
project onto the basis eınωt :
X (n) X (n) ınωt
ınωx α± eınωt = L − X(eıωt − e−ıωt ) α± e + η± (20.148)
n n
(n) (n+1) (n−1)
=⇒ (L − ınω)α± + X(α± − α± ) = −η± δn0 .
(n+1) (n) (n−1) (n)
We define operators S↑↓
n by α± = S↑n α± for n ≥ 0 and α± = S↓n α± for n ≤ 0
and obtain,
(n)
L−ınω + X(S↑n − S↓n ) α± = −η± δn0 . (20.149)
For n = 0, we get,
h i−1
(0)
α± (∞) = α± = − L + X(S↑0 − S↓0 ) η± . (20.150)
By recursive substitution of the lower into the upper equation, the stationary solution
can now be written by means of continued fractions,
1
S↑↓
0 ≡ 1 X. (20.152)
L − ıω ± X L−2ıω±X 1 X
···
α± (∞) in Eq. (20.150) gives us the stationary solution of the differential equation,
time-averaged over an oscillation period.
If we are deep in the Lamb-Dicke regime, we need only consider the first order of
the continued fractions. We set S↑↓
1 = 1 and obtain the simplified equations,
−1
S↑0 = [L − ıω + X] X and S↓0 = [L − ıω − X]−1 X (20.153)
h i−1
(0)
α± = − L + X S↑0 − S↓0 η±
(1) (0) (−1) (0)
α± = S↑0 α± and α± = S↓0 α± .
Explicitly,
U0 kz0 −U0 (1 ± kz0 ) ı(κ + ıU0 + ıω)
S↑↓
0 =
(κ + ıU0 + ıω)2 + U02 (1 − k 2 x20 ) −ı(κ + ıU0 + ıω) −U0 (−1 ± kz0 )
−2U02 k 2 x20
S↑0 − S↓0 = I2
(κ + ıU0 + ıω)2 + U02 (1 − k 2 x20 )
(0) (1) (−1) −ıωt
α± = α± + α± eıωt + α± e . (20.154)
722 CHAPTER 20. COUPLING OF ATOMS AND OPTICAL CAVITIES
Finally,
α± = −[1 + S↑0 eıωt + S↓0 e−ıωt ][L + X(S↑0 − S↓0 )]−1 η± . (20.155)
The Fig. 20.8 shows how the phase of the ring cavity behaves in time for various
atom-field coupling constants.
Figure 20.8: (code) (a) Photon number and (b) phase shift of the cavity in response to
a modulated atomic position. The parameters are kz0 = 1, κ = (2π) 20 kHz, ω = 2.5κ,
and η± = 10κ. (c) Spectrum recorded behind a cavity mirror with the transmittivity thr .
The spectrum is obtained as the Fourier transform of the g (1) (τ ) correlation function of the
transmitted field Eout = thr eıka Ecav , where eıka is a fixed phase factor depending on the
mirror position a (see Sec. 15.3.2).
α̇± = (−κ − ıU0 )α± − ıU0 e∓2ıkz0 sin ωt α∓ + ηe±ıφ0 sin ωφ t . (20.156)
20.2.6 Exercises
20.2.6.1 Ex: Cooperative Lamb shift in a cavity
Calculate the cooperative Lamb shift in a cavity from the second formula (20.64) and
plot the result as a function of the cavity detuning.
20.2. INTERACTION OF ATOMS WITH CAVITIES 723
20.2.6.5 Ex: Filling and drainage of a ring cavity with one fixed atom
Calculate by simulation of the classical CARL equation, how a laser-pumped ring
cavity fills and looses photons in the presence of a single immobile atom.
α ~ 1 e−λ1 t + α
~ =α ~ 2 e−λ2 t ,
Derive from the equations (20.121) the equations of motion for intermodal coherences
∗ ∗ ∗ ∗
α+ α+ , α− α− , α+ α− , and α− α+ .
Xiaodong Zeng et al., Enhancement of the vacuum Rabi oscillation via surface plasma
modes in single-negative metamaterials [DOI]
20.3.1 on cavities
B. Yurke et al., Quantum network theory [DOI]
B.M. Peden et al., Nondestructive cavity QED probe of Bloch oscillations in a gas
of ultracold atoms [DOI]
D. Meiser et al., Prospects for a Millihertz-Linewidth Laser [DOI]
M.A. Norcia et al., Strong coupling on a forbidden transition in strontium and non-
destructive atom counting [DOI]
M.A. Norcia et al., Cold-Strontium Laser in the Superradiant Crossover Regime
[DOI]
M.A. Norcia et al., Cavity-mediated collective spin-exchange interactions in a stron-
tium superradiant laser [DOI]
A. Bychek et al., Superradiant lasing in inhomogeneously broadened ensembles with
spatially varying coupling [DOI]
Chapter 21
Correlated atoms,
entanglement and quantum
gates
In Chps. 13.4 to 18.4 we concentrated on understanding the interaction of single atoms
with light fields. In particular, we introduced the Rabi and the Jaynes-Cummings
model. In Chp. 19.5.4 we introduced the coupled dipoles model for the interaction
of many atoms with a single photon. In the following chapters, we will extend these
models to several and many atoms and many photons 1 .
model RWA symmetry atom # photon # section
Rabi no P 1 n 14.2
Jaynes-Cummings yes U (1) 1 n 15.2
Dicke no P >1 n 21.1
Tavis-Cummings yes U (1) >1 n 21.2.2, 21.2.3
It does not come as a surprise that totally new phenomena arise from the collective
interaction of several atoms with a single light mode. For instance, the atomic cloud
can evolve toward a spin-squeezed or an entangled state, or it can emit light in a
super- and subradiant way. The interplay between collective processes and processes
favoring an individualization of the atom-light interaction is subtle, and the different
models used to understand the processes only grasp partial aspects. The difficulty
arises from the complexity of the task of describing the dynamics of N evolving in
a Hilbert space of dimension 2N . Approximations used to reduce the complexity of
the Hilbert space come at the price of eventually loosing some interesting features.
On the other hand, they may also help to crystallize fundamental symmetries, which
allow us to deepen our intuition on the collective behavior of the many-body system.
Famous models used in the description of collective scattering are (among oth-
ers) the Dicke model and the Tavis-Cummings model (see table above). The Dicke
model (presented in Sec. 21.1) assumes a total indistinguishability of the atoms, the
Tavis-Cummings (discussed in Sec. 21.4 and used for the purpose of quantum com-
1 Other models, such as the Ising model, the Heisenberg model, or the Bose-Hubbard model are
727
728 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
putation) makes use of the rotating-wave approximation. Both models are based on
the assumption of non-interacting atoms, neither by ground state collisions nor via
radiation exchange, which certainly is a good assumption in the case of dilute atomic
samples. On the other hand, we saw in the discussion of the coupled dipoles model
in Chp. 19.5.4 that already the presence of a single photon in an atomic cloud leads
noticeable interatomic interaction effects. In Sec. 21.3 we will extend the Dicke model
to interacting atoms.
Finally, new phenomena arise from the presence of optical cavities shaping the
spatial and spectral distribution of electromagnetic vacuum modes (see Chp. 20.3.2)
and from the consideration of photonic recoil (see Chp. 22.6.1 and Chp. 23.6).
Sec. 21.1.1).
3 Note, that here and in the following we set ~ ≡ 0 for simplicity.
21.1. THE DICKE MODEL IN THE MEAN-FIELD APPROXIMATION 729
and related to the Pauli spin matrices via ŝj = 12 σ̂j . The Hamiltonian of the Dicke
model is,
N
X N
X
Ĥ = ωc ↠â + ωa ŝzj + 2g(â + ↠) ŝxj . (21.3)
j=1 j=1
√
Frequently, the coupling strength is normalized to the atom number, g ≡ λ/ N . The
coupling can be written as the sum of two terms: a co-rotating term that conserves
the number of excitations and is proportional to âσ̂ + + ↠σ̂ − and a counter-rotating
term proportional to âσ̂ − + ↠σ̂ + .
The above Hamiltonian assumes that all the spins are identical, i.e. they have the
same transition frequency, they do not interact with each other, and they equally
couple to the radiation field (e.g. a cavity mode). For the simple system of only
two not mutually interacting spins, s1 and s2 , simultaneously coupling to the same
radiation field, the Dicke model has been introduced in Sec. 4.4.2. There, we have
shown that the spin operators can be added, Ŝ = ŝ1 + ŝ2 , and the total system be
represented in a coupled basis, where [ŝ1 · ŝ2 , Ŝ2 ] = 0 = [ŝ1 · ŝ2 , Ŝz2 ]. This concept can
be generalized to an arbitrary number of spins, that is, under the above assumption,
one can define macroscopic collective spin operators,
N
X
Ŝα ≡ ŝα
j with [Ŝx , Ŝy ] = ıŜz , (21.4)
j=1
and α = x, y, z. Using these operators, one can rewrite the above Hamiltonian as
We see that the Dicke states are not made to unambiguously label degenerate
states. States |S, M i with S < |M | are largely degenerate. The degeneracy of a Dicke
state with S ≤ N2 , that is, the number of states |+iN+ |−iN− composing a single Dicke
state labeled by |S, M i is [270],
(2S + 1)N !
#= . (21.7)
( N2 + S + 1)!( N2 − S)!
Transitions between energetically degenerate states |S, M i and |S, M 0 i with M = M 0
but S 6= S 0 are prohibited.
Figure 21.1: (a) Illustration of the Bloch sphere. (b) Bloch spheres of the various su-
per/subradiant states with N = 6 atoms.
yielding,
cos ϕ sin ϑ
hϑ, ϕ|Ŝ|ϑ, ϕi = 12 sin ϕ sin ϑ . (21.11)
cos ϑ
We will also denote the probability of finding the system in state |±i by,
Ŝz = Ŝ+ Ŝ− − Ŝ− Ŝ+ with Ŝz |S, M i = M |S, M i (21.13)
2 (N I
1 1
N̂± ≡ ± Ŝz ) with N̂± |S, M i = 2 (N ± M )|S, M i .
From this we conclude that S = N/2 is half the number of atoms and M the inversion.
We consider an ensemble of N = N+ + N− two-level atoms excited by a definite
number of M photons, such that,
N = N+ + N− and 2M = N+ − N− . (21.14)
Hence, N± = N2 ± M is the number of atoms in each of the two states. The N atoms
can occupy 2N different collective states. However, when the atoms are identical and
couple uniformly to the same light mode, all states where the same number of atoms
is excited are energetically degenerate, and the total energy available to the system
is,
E = M ωc . (21.15)
The degeneracy of each many-body state with a given inversion M is given by the
binomial coefficient,
N
X
N N N N
#= N = = such that = 2N . (21.16)
2 +M N+ N− N+
N+ =0
732 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
Each block has two degenerate non-shifted eigenvalues and two non-degenerate shifted
eigenvalues,
p
E = nωc and E = nωc ± $n with $n ≡ (ωc + ωa )2 − 4g 2 n . (21.21)
5 In Sec. 4.4.3 we used for the coupling of two spins the notation |(s , s )s, mi. For coupling N
1 2
spins, we should write in analogy,
N
z }| {
|( 12 , .., 12 ), S, M i ≡ |N, S, M i .
The description of the dynamics can be simplified by reducing the order of the matrix
by calculating the average of the lines 2 and 3,
√
√ c − ωa
(n + 1)ω g 2n √0
ĤD = g 2n nω g 2n , (21.22)
√c
0 g 2n (n − 1)ωc + ωa
with c12 n = c21 n . The new Hamiltonian (21.22) has exactly the same eigenvalues as
the complete one (21.22), λ = nωc , nωc ± $n , but without degeneracies. If we assume
2
√ ωc = ωa , with high intensity, n = |α| for
furthermore that the excitation is resonant,
all n, defining the Rabi frequency Ω ≡ 2g n our matrix becomes,
Ω
|α|ω √
2
0
√Ω |α|ω Ω
ĤD ' 2 √
2 , (21.24)
Ω
0 √
2
|α|ω
Hence, the spin evolves on the surface of a Bloch sphere with radius,
khϑ, ϕ|N Ŝ|ϑ, ϕiN k = S while still khϑ, ϕ|N Ŝ2 |ϑ, ϕiN k = S(S + 1) . (21.31)
hN+ i = N
2 + 21 hŜz i = N cos2 ϑ
2 = N p+ = N (1 − p− ) = N − hN− i . (21.32)
6 An illustration of the generalized Bloch sphere is attempted in Fig. 21.18.
21.1. THE DICKE MODEL IN THE MEAN-FIELD APPROXIMATION 735
eıF (Ŝz ) Ŝ+ e−ıF (Ŝz ) = Ŝ+ eı[F (Ŝz +I)−F (Ŝz )] , (21.33)
where F is an arbitrary function. For F (Ŝz ) ≡ φŜz the unitary transform eıF (Ŝz )
denotes a rotation about the z-axis, which we will study in the example below. For
F (Ŝz ) ≡ ζ Ŝz2 it generates squeezing along the z-axis, which we will study in the next
section. Furthermore, we define the rotation matrices about the Cartesian axis,
1 0 0 cos γ 0 − sin γ
Rx (γ) ≡ 0 cos γ − sin γ , Ry (γ) ≡ 0 1 0 ,
0 sin γ cos γ sin γ 0 cos γ
cos γ − sin γ 0
Rz (γ) ≡ sin γ cos γ 0 , (21.34)
0 0 1
for which it is possible to show (with α = x, y, z),
We also find,
cos(ϕ + φ) sin ϑ
hϑ, ϕ|N Rz (φ)Ŝ|ϑ, ϕiN = N N N
sin(ϕ + φ) sin ϑ = hϑ, ϕ + φ| Ŝ|ϑ, ϕ + φi .
2
cos ϑ
To vary the polar angle ϑ of a coherent spin state |ϑ, ϕi, we first rotate the
coordinate system about the z-axis until ϕ = 0, then rotate about the y-axis by
the desired angle θ, and finally rotate back about the z-axis to reach the initial
azimuth ϕ,
cos ϕ sin(θ + ϑ)
hϑ, ϕ|N Rz (ϕ)Ry−1 (θ)Rz−1 (ϕ)Ŝ|ϑ, ϕiN = N2 sin ϕ sin(θ + ϑ) = hϑ+θ, ϕ|N Ŝ|ϑ+θ, ϕiN .
cos(θ + ϑ)
736 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
The rotation about the x-axis is derived in Exc. 21.1.6.5, and in Exc. 21.1.6.6
we write down the explicit rotation matrix for two atoms.
Rotations such as the ones described by Rα (γ) are generated by the Dicke Hamil-
tonian (21.8), since the solution of the Schrödinger equation is,
That is, the Dicke Hamiltonian generates rotations Rx (Re Ωx t), Ry (Im Ωy t), and
Rz (∆t). This confirms that rotations do only transform coherent states into each
other. Nevertheless, there are other unitary operations that transform coherent states
into states that cannot be represented by coherent states. One example for this is
squeezing.
Since there are no quantum correlations between the particles, the uncertainty of
coherent spin states is additive (see Exc. 4.3.4.5),
N
X
(∆Ŝα )2 = (∆ŝα 2
k) . (21.40)
k=0
For a coherent spin state we can calculate explicitly [see Exc. 21.1.6.3(b)],
∆Ŝx2 1 − sin2 ϑ cos2 ϕ
hϑ, ϕ|N ∆Ŝy2 |ϑ, ϕiN = N4 1 − sin2 ϑ sin2 ϕ . (21.41)
2 2
∆Ŝz sin ϑ
Figure 21.2: Illustration of the uncertainty of the spin components of a coherent spin state.
Note, that spin squeezing along the z-axis could be obtained by quantum non-
demolition measurement of the inversion, that is, by measuring Ŝz without influencing
the populations of the ground and excited state.
Figure 21.3: Illustration of the uncertainty in (a) a coherent state and (b) a spin-squeezed
state.
Specifying the rule (21.33) for the particular case F (Ŝz ) ≡ ζ Ŝz2 , we get,
2 2
eıζ Ŝz Ŝ+ e−ıζ Ŝz = Ŝ+ e2ıζ(Ŝz +1/2) , (21.43)
and hence,
1
(Ŝ+ e2ıζ(Ŝz +1/2) + e−2ıζ(Ŝz +1/2) Ŝ− )
2 2 21
Qz (ζ)Ŝ = eıζ Ŝz Ŝe−ıζ Ŝz = 2ı (Ŝ+ e2ıζ(Ŝz +1/2) − e−2ıζ(Ŝz +1/2) Ŝ− ) . (21.44)
Ŝz
Let us now apply the squeezing operator to the state | π2 , 0i. In Exc. 21.1.6.7 we
show that,
1
2 2
h π2 , 0|N eıζ Ŝz Ŝe−ıζ Ŝz | π2 , 0iN = 0 N2 cosN −1 ζ (21.45)
0
Ŝx2 N +1 1
2 2
h π2 , 0|N eıζ Ŝz Ŝy2 e−ıζ Ŝz | π2 , 0iN = N + 1 N8 + −1 N (N8−1) cosN −2 2ζ .
Ŝz2 2 0
Nevertheless, the prescription does generate quantum correlations in Ŝx and Ŝy ,
which can be transformed to squeezing by subsequently rotating the collective spin
about the x-axis [410]. As shown in Exc. 21.1.6.8, a rotation by an angle ν does not
modify the x-component,
2 2
h π2 , 0|N eıν Ŝx eıζ Ŝz ∆Ŝx2 e−ıζ Ŝz e−ıν Ŝx | π2 , 0iN (21.46)
N (N +1) N (N −1) 2
= 8 + 8 cosN −2 2ζ − N
4 cos2N −2 ζ ,
21.1. THE DICKE MODEL IN THE MEAN-FIELD APPROXIMATION 739
(a) 1 (b) 3
0.8
2
ΔŜj /S
0.6
Ŝj /S
Ŝx /S
0.4 Ŝx2 1/2 /S
ΔŜx /S
1
ΔŜy /S ΔŜx /S
0.2 ΔŜy /S
ΔŜz /S
ΔŜx ΔŜy /S 2 ΔŜz /S
0 0
0 0.5 1 0 0.5 1
ζ ν/π
Figure 21.4: (code) (a) Uncertainties calculated in (21.45) as a function of the squeezing
parameter. (b) Uncertainties after application of squeezing operator as a function of the
rotation angle ν about the x-axis.
In a cloud this latter idea is not realizable, but perhaps inside an optical cavity.
spin operator Ŝ defined by its commutation behavior [Ŝx , Ŝy ] = ıŜz and characterized
by its eigenvectors |S, M i,
Ŝ2 |S, M i = S(S + 1)|S, M i , Ŝz |S, M i = M |S, M i , (21.50)
with the projection quantum number M = −S, −S + 1, . . . , S − 1, S. We consider
a single particle of spin S and take the state |S, M = +Si as a vacuum for a set of
boson operators, and each subsequent state with lower projection quantum number
as a boson excitation of the previous one,
1
|S, S − ni 7→ √ (b̂† )n |0i . (21.51)
n!
Each additional boson then corresponds to a decrease of M in the spin projection.
Thus, the spin raising and lowering operators Ŝ± = Ŝx ± ıŜy satisfying so that
[Ŝ+ , Ŝ− ] = 2Ŝz correspond to bosonic annihilation and creation operators, respec-
tively. The precise relations between the operators must be chosen to ensure the cor-
rect commutation relations for the spin operators. The resulting Holstein-Primakoff
transformation can be written as,
√ q † √ q
†
Ŝ+ = 2S 1 − b̂2Sb̂ b̂ , Ŝ− = 2S b̂† 1 − b̂2Sb̂ , Ŝz = (S − b̂† b̂) . (21.52)
The transformation is particularly useful in the case where S is large, when the
square roots can be expanded as Taylor series, to give an expansion in decreasing
powers of S [693, 75]. We will verify the commutation relations in Exc. 21.1.6.14.
The Holstein-Primakoff approximation allows us to rewrite the Hamiltonian (21.5)
as,
ĤHP = −∆c ↠â − ∆a Ŝz + 2g(â + ↠)Ŝx (21.53)
q q
√
b̂† b̂ b̂† b̂
= −∆c ↠â − ∆a (S − b̂† b̂) + 2g(â + ↠) 2S 1− 2S b̂ + b̂† 1− 2S
In contrast ().
21.1. THE DICKE MODEL IN THE MEAN-FIELD APPROXIMATION 741
21.1.6 Exercises
21.1.6.1 Ex: Dicke Hamiltonian for 3 atoms
Generalize the Dicke Hamiltonian (21.22) to 3 atoms.
Prove the relationship eıF (Ŝz ) Ŝ+ e−ıF (Ŝz ) = Ŝ+ eı[F (Ŝz +I)−F (Ŝz )] .
Calculate explicitly for the case of two atoms the rotation matrices eıγ Ŝα for α =
x, y, z. Check the relationship eıγ Ŝα Ŝe−ıγ Ŝα = Rα (γ)Ŝ by explicit calculation.
with Ŝϑ,ϕ defined in Exc. 21.1.6.2 in view of its squeezing features of an initial coherent
spin state |0, ϕi [410].
(∆Ŝz )2 1
≥
hŜx i2 + hŜy i2 N
N (N +2)
hŜx2 i + hŜy2 i + hŜz2 i ≤ 4
h∆Ŝx2 i + h∆Ŝy2 i + h∆Ŝz2 i ≥ N
2
hŜk2 i + hŜm
2
i− N
2 ≤ (N − 1)h∆Ŝn2 i
N (N −2)
(N − 1)[h∆Ŝk2 i + h∆Ŝm
2
i] ≥ hŜk2 i + 4 ,
for (kmn) = (123). Verify that, according to these criteria, coherent Dicke states are
not entangled.
21.2. SUPER- AND SUBRADIANCE IN OPEN SYSTEMS 743
Verify that the relations (21.52) satisfy [Ŝ+ , Ŝ− ] ' 2Ŝz and [Ŝx , Ŝy ] ' 2ı Ŝz .
[Ĥ, Ŝ2 ] = 0 with Ŝ2 = 21 (Ŝ+ Ŝ− + Ŝ− Ŝ+ ) + Ŝz2 (21.54)
but [Ĥ, Ŝz ] = ı(Im Ω Ŝx − Re Ω Ŝy ) 6= 0 .
Hence, under the effect of the Dicke Hamiltonian an initial state |N, S, M i can only
change its magnetic quantum number |N, S, M i −→ |N, S, M 0 i, and the manifolds
with a given S form closed sub-spaces (see Exc. 21.3.3.2). In other words, once we
start in a superradiant state |N, S, M i = |N, N2 , + N2 i, spin conservation excludes
subradiant states, which allowed us to restrict to the superradiant Dicke subspace.
Transitions between Dicke subspaces are only possible via physical processes that act
on individual atoms, e.g. decay or phase fluctuation processes, as we will see later
[273].
Figure 21.6: (a) Illustration of the Dicke states for N = 6. Hamiltonian interactions are in
depicted in red. Superradiant decay occurs through a cascade from state M = S to state
M = −S. Spontaneous emission and phase noise leads to transitions along the blue arrows.
The lowest states in each S subspace are dark and can only decay through a (subradiant)
dark states cascade. (b) Scheme of the Dicke model.
frame we have,
P ˙ P
ρ̂˙ = ı[ρ̂, ĤgD ] + j Lγj ,L̂j ρ̂ or  = −ı[Â, ĤgD ] + j L†γ Â
j ,L̂j
ĤgD = −∆c ↠â − ∆a Ŝz + 2g(âŜ+ + Ŝ− ↠) + 2g 0 (âŜ− + Ŝ+ ↠) + η(â + ↠)
.
and Lγ,L̂ ρ̂ ≡ γ(2L̂ρ̂L̂† − L̂† L̂ρ̂ − ρ̂L̂† L̂)
Γ and κ describe collective decay, while γ1 and β stand for spontaneous emission
and phase fluctuation of individual atoms. In Exc. 21.3.3.3 we derive the Heisenberg
21.2. SUPER- AND SUBRADIANCE IN OPEN SYSTEMS 745
In Exc. 21.3.3.4 we verify that these equations of motion do not change the spin Ŝ.
The Eqs. (21.56) can be rewritten in terms of observables as,
2(g + g 0 )(â + ↠) −Ŝx + Ŝx Ŝz + Ŝz Ŝx
Ŝ˙ = 2ı(g − g 0 )(â − ↠) × Ŝ + Γ −Ŝy + Ŝy Ŝz + Ŝz Ŝy . (21.57)
−∆a 2 2
−2Ŝz − 2Ŝx − 2Ŝy
Only the terms Lγ1 and Lβ can change Ŝ2 . The Ŝ2 and Ŝz eigenvalues determine
the coupling strength of the many-atom (Dicke) state to the cavity mode and the
coherent, external drive. This coupling determines the rate of cavity photon gen-
eration as well as the pumping strength. The magnitude of the coupling strength
distinguishes between superradiance and subradiance. For superradiant states the
coupling strength scales superlinear in N , while for subradiant states the scaling is
sublinear in N , and some subradiant states are dark. Dark means that the collec-
tive coupling to the cavity and the coherent, external drive of these states vanishes,
meaning these states cannot decay via collective interactions e.g. by creating a cavity
photon. However these states still decay into other states via the decay and dephasing
processes Lγ1 and Lβ acting individually on the emitters, see Fig. 21.6. Hence, spon-
taneous decay is an individualization process [273]. Generally, the spin preserving
contributions in the master equation (21.59) generate quantum correlations leading
to collective behavior (both super- and subradiance are collective effects) and the
non-preserving terms destroy correlations leading to individualization (all properties
scale exactly linear in N ). However only the spin non-preserving contributions intro-
duce coupling between superradiant and subradiant states, thus in order to prepare
subradiant states an interplay of collectivity and individualization is necessary. Based
on these considerations, we may distinguish between collective versus individual be-
havior and superradiant versus subradiant behavior. The latter are special cases of
collective behavior. This twofold distinction seems crucial when investigating super-
and subradiance in the presence of dephasing and individual decay.
That is, the Dicke Hamiltonian preserves the excitation number, except for the
counter-rotating terms, which can only change the excitation number by ±2.
The Dicke model without RWA, g 0 = g, has one global symmetry 8 ,
P : (â, σ̂ ± ) → (−â, −σ̂ ± ) . (21.60)
Because P squares to unity, it has two eigenvalues, 1 and −1. This symmetry is
associated with a conserved quantity: the parity of the total number of excitations,
P = (−1)Nex . This parity conservation is a consequence of the preserved excitation
number. A state of the Dicke model is said to be normal when this symmetry is
preserved, and superradiant when this symmetry is spontaneously broken.
Z(α) is the partition function, ĥj the single-atom Hamiltonian, and β ≡ 1/kB T .
Carried out in Exc. 21.3.3.5, the calculation results in,
F (α) = ωc α2 − N
ln(2 cosh βE)
β (21.63)
q
ωa 2
where ± E ≡ hĥj i = ± 2 + (2gα)2
are the single-atom energy eigenvalues. The minimum of the free energy as a function
of the field amplitude, F 0 (α) = 0, yields a critical coupling strength gc ,
√ q
gc N = 12 ωc ωa coth βω2 a , (21.64)
Below gc the free energy minimizes for α = 0, and beyond gc it minimizes for α > 0,
as seen in Fig. 21.7.
8 For a Dicke picture of CARL see Exc. 22.5.7.1.
21.2. SUPER- AND SUBRADIANCE IN OPEN SYSTEMS 747
200
F (g, α) (MHz)
100
-100 0
500
0 1
α -500 2 g/gc
Figure 21.7: (code) Free energy as a function of coupling strength and photon number.
Beyond the critical coupling strength gc the minimum of the free energy splits opening the
way for two possible equilibrium states of the mean-field phase.
interaction between ferromagnetic atoms and the spontaneous ordering of excitations below the
critical temperature.
748 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
Figure 21.8: (a) Principle scheme of standard lasing. Here, the coherence is stored in
the cavity field. The gain profile is much larger than the cavity width (good-cavity limit,
κ Γg ), as shown in (c). The laser frequency follows any (technical) cavity fluctuation:
ωlas = ωcav + ωg Γκg . (b) Superradiant lasing. Here, the coherence is stored in the gain. We
are in the bad-cavity limit, κ Γg , as shown in (d). The laser frequency is robust to cavity
Γ
fluctuations: ωlas = ωg + ωcav κg .
the field oscillator is so strong that the system collapses in a ferroelectric-like phase
transition. In Exc. 21.3.3.6 we will derive the Heisenberg equations for the open Dicke
model in the Holstein-Primakoff approximation.
Example 101 (Finding instabilities via retarded Green’s functions): Let
us consider a set of linear Heisenberg equations,
v̇i (t) = Mij vj (t) and v̇i† (t) = Mij
† †
vj (t) .
The retarded Green’s function is defined by,
Gij (t) ≡ −ıh[vi (t), vj† (0)]iθ(t) .
Its time derivative is,
Ġij (t) ≡ −ıh[vi (t), vj† (0)]iδ(t) − ıhv̇i (t)vj† (t)iθ(t) + ıhvj† (t)v̇i (t)iθ(t) .
Defining the equal-time correlation function by,
Sij ≡ h[vi (0), vj† (0)]i ,
we get,
Ġij (t) = −ıSij δ(t) − ıMik hvk (t)vj† (t)iθ(t) + ıMik hvj† (t)vk (t)iθ(t)
= −ıSij δ(t) − ıMik h[vk (t), vj† (t)]iθ(t) = −ıSij δ(t) + Mik Gkj (t) .
R∞
With the Fourier transform f (ω) = −∞ dteıωt f (t) we finally get,
G(ω) = (M + ıωI)−1 ıS .
For example, we may consider the system of two coupled oscillators studied in
Sec. 15.6.7. In this case,
â
â†
v≡ such that S = diag (1, −1, 1, −1) .
b̂
b̂†
21.2. SUPER- AND SUBRADIANCE IN OPEN SYSTEMS 749
Thus,
ω − ωa + ıκa 0 − Ω2 − Ω2
0 −ω − ωa − ıκa − Ω2 − Ω2
GR (ω)−1 −1
= S (ωI−ıM ) = .
− Ω2 − Ω2 ω − ωb + ıκb 0
− Ω2 − Ω2 0 −ω − ωb − ıκb
0 = det GR (ω)−1 = ωa2 ωb2 + ωa2 κ2b − ωa Ω2 ωb + κ2a ωb2 + κ2a κ2b ,
yielding, s
(ωa2 + κ2a )(ωb2 + κ2b )
Ω= .
ωa ωb
−2ıg 2
gâ = Ŝ− = −2ı(κc + ıUc )Ŝ− . (21.65)
κ − ı∆c
where we additionally assumed η ≡ 0 and introduced the abbreviations,
g 2 ∆c g2 κ
Uc ≡ and κc ≡ .
∆2c + κ2 ∆2c + κ2
Inserting this into the above Heisenberg-Liouville equations of the Dicke model (21.57)
without RWA, g 0 = g, we obtain neglecting Γ,
−16κc Ŝy + 16Uc Ŝx
˙Ŝ = × Ŝ . (21.66)
0
−∆a
750 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
0
Sz
-2
2
2
0 0
-2 -2
Sy Sx
21.2.4 Exercises
11 This is somewhat analogous to the photon number-dependent phase shift observed in dispersive
21.2.4.5 Ex: Heisenberg equation for the open Dicke model in the
Holstein-Primakoff approximation
Derive the Heisenberg equations for the open Dicke model in the Holstein-Primakoff
approximation.
when half of the atoms are deexcited. The correlated atoms can be in a Dicke state
(then the total dipole moment is always zero) or in a product state (then the net
dipole moment is non-zero at half-deexcitation). In the second case, we also talk
about superfluorescence. In this case, an excited initially incoherent sample develops
correlations due to the emission process. One can also imagine the case that the
emission patterns pairwise cancel, and the decay is thus inhibited. This is called sub-
radiance. Superradiance has been used in the microwave domain as a spectroscopic
method in the observation of photon echoes.
The mean-field Dicke model totally neglects interactions between the atoms due to
the exchange of real of virtual photons, i.e. neither resonant dipole-dipole interactions
nor van der Waals interactions are considered [24, 72, 216, 235, 596, 604, 650]. That
is, interaction terms such as,
N
X
ĤIsing = − Vij σ̂j+ σ̂i− (21.69)
i,j6=i
are absent from the Hamiltonian. Spin-spin interactions are studied in the so-called
Ising model, which is interesting in the context of (anti-)ferromagnetism [263, 264,
274, 409, 772]. The negligence of interaction is of course the price to pay for the
simplicity of the model and its manageability for large atom numbers. On the other
hand, in Sec. 19.1.2 we have studied dipole-dipole interactions in the linear optics
regime allowing for at most a single photon to interact with the cloud.
The starting point is the collective many-atoms Hamiltonian (19.9) of the scalar
coupled dipoles model. After tracing over the vacuum modes, one obtains the master
21.3. INTERACTING ATOMS 753
12
equation ,
with rji = |rj − ri |. In Exc. 21.3.3.1 we derive the expression for the line shifts ∆ij
and the decay rates Γij from the linear optics scalar coupled dipoles model.
Assuming ∆ji = 0 = Γji and Ω(rj ) = Ω we recover the mean-field Dicke model,
where interaction terms are completely neglected,
X X
Ĥ = −∆a σ̂j+ σ̂j− + 21 Ω (σ̂j+ + h.c.) . (21.71)
j j
Example 104 (Two atoms with dipole-dipole interactions): For the case
of only two atoms located at rj , using the basis defined in (4.73), we find the
Hamiltonian,
1 ∗
0 2
Ω (r2 ) 21 Ω∗ (r1 ) 0
1 Ω(r ) 1 1 ∗
2 −∆a ∆
2 21
Ω (r1 )
Ĥ = 21 2
, (21.73)
1 1 ∗
2 Ω(r1 ) 2
∆12 −∆ a 2
Ω (r 2 )
1 1
0 2
Ω(r1 ) 2
Ω(r2 ) −2∆a
with Ω(r) = Ω0 eık·rj . For two atoms the master equations (21.70) can easily be
solved numerically by setting 13 ,
as usual and,
hi, j|ρ̂|m, ni = ρij,mn , (21.75)
where the indices i, m = 1, 2 refer to the first atom and the indices j, n = 1, 2
to the second. The populations of the Dicke states |11i, |12i, |21i, and |22i
12 In return, the equations of motion (19.26) for the excitation amplitude should follow as the
Heisenberg equation with the above Hamiltonian.
13 Remember, that the formal solution of coherent part of the master equation can be written as
are then given by ρij,ij , and the populations of the (anti-)symmetric states
|ψi(s,a) = √12 (|1, 2i ± |2, 1i) are calculated via,
The temporal evolution of the populations in one and two atom systems, initially
driven by a laser field which is then suddenly switched off, is shown in Fig. 21.10.
1 1 1
(a) (b) (c)
ρ
sub
super
0 0 0
0 5 10 0 5 10 0 5 10
Γt Γt Γt
Figure 21.10: (code) (a) Response of a single two-level atom driven by a laser light with
Ω = 5Γ, ∆ = −Γ. The curves show (blue) the ground state and (red) the excited state
populations. The light is switched off at Γt = 0.5. (b) Response of two atoms j = 1, 2 located
at kzj = ±5 driven by the same laser light as in (a). (cyan and magenta) populations of
the states |egi and |gei. The two black lines show the populations of the (anti-)symmetric
states √12 (|egi ± |gei). (c) Same as (c) but with kzj = ±0.5.
where 21 (σ̂iz + 1) = |eii he| is the probability of finding the i-th atom in an excited
state and C6 interatomic van der Waals interaction coefficient of the transition.
21.3. INTERACTING ATOMS 755
The master equation can be numerically solved using the procedure outlined in
example 104. The result of such a simulation is shown in Fig. 21.11. Comparing
(a) 1 (b) 1
0.5 0.5
ρ
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Ωt Ωt
Figure 21.11: (code) Populations in a system of two two-level atoms interaction via van der
Waals forces. Initially (Ωt < π) only one atom is driven, after that only the other. We
assume Ω Γ and C6 = 4 · 107 . The interatomic distance is (a) kr12 = 0.5, respectively,
(1)
(a) kr12 = 5. (blue) ρ11,11 , (cyan) ρ12,12 , (magenta) ρ21,21 , (red) ρ22,22 , (black dotted) ρ22 ,
(2)
and (black) ρ22 .
the evolutions calculated in Fig. 21.11 for large and small interatomic distances,
we see that the excitation of the first atom impedes the excitation of the second
one when the interaction is strong.
Note, that an interesting situation occurs when the coupling is completely uniform
(e.g. mediated by a cavity), κij ' κ,
as pointed out in Eq. (21.49). Such Hamiltonians may be interesting for the generation
of spin-squeezing.
21.3.3 Exercises
21.3.3.1 Ex: Super- and subradiant linewidth and decay rates from the
coupled dipoles model
Calculate super- and subradiance linewidth and lineshifts for two atoms interacting
via dipole-dipole interaction using the linear optics scalar coupled dipoles model cul-
minating in Eq. (19.26) using the exponential kernel (19.29).
756 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
Figure 21.12: (a) Generation of entangled photon via parametric down-conversion. (b)
Energy conservation upon parametric down-conversion. (c) Illustration of the EPR paradox.
On the other hand, quantum mechanics postulates that the reduction of the
wavepacket describing the particle at detector a decides spontaneously on the re-
758 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
sult of the measurement at detector b. This decision has no physical cause. If the
particles move with the speed of light, this decision is not restricted by local causality
[332]. The state, also called Bell state, can be given in the form,
|ψi = | +a +b i + | −a −b i . (21.80)
In Exc. 21.4.3.1 we check that this state cannot be expressed as a product state. Let
us have a closer look at the EPR state (21.80) and imagine that particle ’a’ (the one
that will later be analyzed by detector a) is embarked with Alice on a starship, while
particle ’b’ flies with Bob in opposite direction.
We introduce a basis choosing z as the quantization axis, that is, the basis vectors
1 0
|+iz = , |−iz = (21.81)
0 1
are eigenvectors of the Pauli matrix σ̂x . Quantum mechanics now tells us that, before
any measurement, the wavefunction is in a superposition of states |+i and |−i. That
is, in the z-basis, using the tensor notation, we may write the entangled state (21.80)
as,
|ψi = √1 (|+iz ⊗ |+iz + |−iz ⊗ |−iz ) . (21.83)
2
Hence, if Bob measures along the z-axis, his measurement will be predefined by Alice’s
measurement. On the other hand, if tries to measure along the x-axis, he will find a
21.4. QUANTUM CORRELATIONS AND ENTANGLEMENT 759
random result with a probability of 12 for each outcome 14 . That is, the direction of
measurement in Alice’s experiment fixes the quantization axis for Bob’s experiment.
Two hypotheses have been put forward: The first one is that Alice’s experiment
sends information to Bob’s experiment. Alice’s particle told Bob’s particle which
spin state he should be in. But this hypothesis violates the locality requirement of
special relativity claiming that information cannot travel faster than light. Einstein
called it ’spooky action at a distance’. The second hypothesis is to assume that
quantum mechanics does not violate locality, but that the particles carry with them
’local hidden variables’ (like a proper DNA) whose values would be set right from the
moment of the separation of the particles and which would determine the outcomes
of any future spin measurement. Then, quantum mechanics would not be a complete
theory, because it has nothing to say about these hidden variables.
Let us first do the quantum calculation. Using the rules for Pauli matrices and
the anti-symmetry of the entangled state,
σ̂x |±iz = |∓iz , σ̂y |±iz = ±ı|∓iz , σ̂z |±iz = ±|±iz , (21.88)
as well. But according to quantum mechanics, Alice should be unable to know the spin value in x-
direction, since she already knows the spin the z-direction, which is not allowed by the commutation
rules,
[σ̂za , σ̂xa ] = ı~σ̂ya , [σ̂zb , σ̂xb ] = ı~σ̂yb .
15 The anti-symmetric state is chosen here for symmetry reasons facilitating the quantum calcula-
ˆ · a ⊗ I|ψi = ±1
A(a, λ) ≡ hψ|~σ and ˆ · b|ψi = ±1 , (21.90)
B(b, λ) ≡ hψ|I ⊗ ~σ
with the crucial assumption that A does not depend on b and BR not a. Let us assume a
probability distribution ρ(λ) for the hidden variable satisfying ρ(λ)dλ = 1. Then the
joint probability distribution P (a, b) should coincide with the quantum mechanical
correlation function,
Z
P (a, b) = ˆ · a ⊗ ~σ
A(a, λ)B(b, λ)ρ(λ)dλ = hψ|~σ ˆ · b|ψi ≥ −1 . (21.91)
For example, let the hidden variable now be unit vector ~λ with uniform probability
distribution over all directions, such that,
Then,
Z
P (a, b) = − sign(cos ϑaλ )sign(cos ϑbλ )ρ(ϑλ )dϑλ (21.93)
Z ϑb
=− sign(cos ϑλ )dϑλ = −1 + π2 ∠(a, b) .
ϑa
This probability distribution clearly deviates from the quantum prediction, in partic-
ular at angles ∠(a, b) = π2 , as seen in Fig. 21.13(a).
1
ˆ · b
(a) (b)
|P (a, b) − P (a, c)| − 1 + P (a, c)|
2
ˆ · a ⊗ σ
0.5
0
0
P (a, b) , σ
-0.5 -2
1
1
0 0
-1
0 1 2 -1 -1 (a, b) (π)
(a, c) (π)
(a, b) (π)
Figure 21.13: (code) (a) Joint probability distribution P (a, b) in hidden variables theory
(red) and correlation function hψ|~σ ·a⊗~σ ·b|ψi in quantum mechanics (blue). (b) Verification
of Bell’s inequality.
Let us finally derive Bell’s inequality. Because of (21.90) for the anti-symmetric
state we have the relation B(b, λ) = −A(b, λ) and A(a, λ) = A(a, λ)−1 allowing us
21.4. QUANTUM CORRELATIONS AND ENTANGLEMENT 761
to derive,
Z
P (a, b) − P (a, c) = − [A(a, λ)A(b, λ) − A(a, λ)A(c, λ)] ρ(λ)dλ (21.94)
Z
= A(a, λ)A(b, λ) [A(b, λ)A(c, λ) − 1] ρ(λ)dλ ,
and
Z
|P (a, b) − P (a, c)| ≤ |A(a, λ)A(b, λ) [A(b, λ)A(c, λ) − 1] |ρ(λ)dλ (21.95)
Z
≤ [1 − A(b, λ)A(c, λ)]ρ(λ)dλ
Z Z
= ρ(λ)dλ + A(b, λ)B(c, λ)ρ(λ)dλ = 1 + P (b, c) .
This inequality derived from the hidden variables assumption is violated by quantum
mechanics, since,
ˆ · a ⊗ σ̃
|h~σ ˆ · a ⊗ σ̃
ˆ · bi − h~σ ˆ · ci| = | − a · b + a · c| (21.96)
ˆ · b ⊗ ~σ
≤ 1 − b · c = 1 + h~σ ˆ · ci
is not satisfied for arbitrary choices of a, b, and c. To see this we plot in Fig. 21.13(b),
finding that this quantity becomes negative for some choices of the projection vectors.
Example 106 (EPR paradox using GHZ states): Here, we show a modified
version of the EPR experiment proposed by Greenberger, Horne and Zeilinger
(GHZ) based on the GHZ state [292]. They imagined a source creating three
correlated non-interacting spin 12 particles labeled a, b, and c flying toward
three Stern-Gerlach type magnetic detectors, which measure the spins in x or
in y-direction. The correct state is either quantum,
|ψi = | +a +b +c iz − | −a −b −c iz , (21.98)
or
|ψi = |+iz ⊗ |+iz ⊗ |+iz − |−iz ⊗ |−iz ⊗ |−iz , (21.99)
in tensor notation, or it contains hidden variables, that is, instructions telling
the detectors which measurement result to exhibit upon arrival of a particle,
!
±ax ±bx ±cx
|ψi = . (21.100)
±ay ±by ±cy
σ̂x ⊗ σ̂x ⊗ σ̂x |ψi ≡ σ̂xa σ̂xb σ̂xc |ψi = −|ψi , (21.101)
but also,
σ̂xa σ̂yb σ̂yc |ψi = σ̂ya σ̂xb σ̂yc |ψi = σ̂ya σ̂yb σ̂xc |ψi = |ψi . (21.102)
762 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
Figure 21.14: Scheme of the EPR experiment proposed by Greenberger, Horne and Zeilinger.
Which one of the two pictures is correct, the hidden variables assumption or
quantum mechanics, can actually be tested by Wheeler’s delayed choice experiment.
It consists of using pairs of correlated photons emitted in different directions, where
they are expected by photodetectors. Polarizers located in front of the detectors fix
the quantization axis to êx or êy , but the choice of the axis is made only after the
photons were created by the source, in order to avoid possible backactions of the
polarizers on the source 17 . Bell’s inequality condition [59] predicts a statistics for the
results of repeated measurements where the orientation of the polarizers is randomly
16 Note that the coupling of different degrees of freedom is not an entanglement, but it is the
laser beams are not really conclusive, because the attenuation of a coherent state does NOT result
in a single photon states (non-classical Fock state). Even if on average only less than one photon is
in the interferometer, this may still be a Glauber state. Nonetheless, real single photon experiments
involving the deexcitation of individual atoms by emission of two photons, produced the same result.
21.4. QUANTUM CORRELATIONS AND ENTANGLEMENT 763
varied. The experiment run by Aspect [28, 27] showed that the assumption of hidden
variables violates local realism and thus confirmed quantum mechanics as being a
complete theory.
The set is canonical and the entropy maximal. On the other hand, if the states are
entangled, the entropy is always less than S (N ) = N bits. For the state of maximally
entangled spins, we have,
ρ(N )
corr =
1
2N
(| + +...+ih+ + ... + | + | − −...−ih− − ... − |) , (N )
Scorr = 1 bit . (21.108)
Entropy is a measure for the degree of entanglement of a system. For a Dicke state
|J, M i the entropy is,
2J
SJM = − log2 . (21.109)
J +M
For beam splitters, the information entropy is calculated as follows: A Fock state
|N i = |+iN divided by a beam splitter is described by,
N
X
−N/2 N
2 (|+i + |−i) N
= 1
2N/2
|+iN −n |−in (21.110)
n=0
n
XN
(N ) N N
ρsplit = 1
2N
|+iN −n |−in h+|N −m h−|m
n,m=0
n m
N
X
(N ) N N
ρsplit = 1
2N
|+iN −n |−in h+|N −n h−|n .
n=0
n n
764 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
The division of a beam is an incoherent process in the sense that it increases the
entropy. The process is irreversible. The divided beams can not be recombined by
a coherent process. For example, an interferometer always has two output ports.
However, the phase is preserved.
The quantum information content is defined by the deviation from maximum en-
tropy,
Q = Smx − Sactual . (21.112)
If the system is in a mixed state, the entropy measures deviations from a pure state
behavior [52].
21.4.3 Exercises
21.4.3.1 Ex: Bell states
Show that the Bell states cannot be written as products of two states.
Show that, when the components of polarization are measured along the axes that
have an angle of 45◦ with respect to the original axes, corresponding to the states,
|h0 i = √1 (|hi
2
+ |vi) and |v 0 i = √1 (|hi
2
− |vi) ,
In order to correlate particles, they must be able to exchange information, that is,
they must interact in some way. One method uses ions trapped in a linear Paul trap,
where they form a straight chain, individually driven by laser beams and coupled to
each other via their vibrational degrees of freedom mediated through Coulomb repul-
sion [141]. We present this idea in Sec. 21.5.2. Such scenarios have been implemented
to create the first quantum computers. Another way consists in trapping neutral
atoms in a dual optical lattice and let them undergo controlled collisions [375], as will
be discussed in Sec. 21.5.3.
Alternative approaches to realizing quantum gates use dipole-dipole interactions
[109], conditional Raman adiabatic passages by laser-induced excitation of interatomic
dipole-dipole interactions [460], or interactions between permanent dipoles of atoms
in Rydberg states [376] 18
X X
Ĥ = ω↠â + ωai (σ̂i+ σ̂i− − 21 ) + 1 +
2 gi (t)(âσ̂i + â+ σ̂i− ) . (21.114)
i=1,2 i=1,2
The suffix denotes the individual atoms. Note that the coupling constant gi (t) can
be considered as time dependent, which may be useful for modeling radiation pulse
envelopes. In matrix representation the Hamiltonian acting on the subspace of n
photons is,
c22 n−1
X c21 n
|ψi =
c12 n |ni where cij ≡1 hj|2 hi|hn|ψi (21.115)
n
c11 n+1
for example, hn| ⊗ h2| ⊗ h2| = hn|(1 0 0 0), hn| ⊗ h2| ⊗ h1| = hn|(0 1 0 0), and so on.
The atomic transition operators are generalized to,
1
X X 1
σ̂1+ = |niσ̂ + ⊗ Ihn| = |ni
0
hn|
(21.116)
n n
0
1
X X 0
σ̂2+ = |niI ⊗ σ̂ + hn| = |ni
hn| , etc. .
n n
1
0
We now assume that the light field only interacts with the upper level, as in the case
of the dual resonance configuration studied in Exc. 15.2.5.4. This can be taken into
account, neglecting those terms in the matrix (21.122) rotating with a positive Rabi
frequency. Letting, moreover, ∆1 = ∆2 and Ω1 = Ω2 , we obtain,
0
−nΛ
Ĥn(1) =
.
(21.124)
−nΛ
−2(n + 1)Λ
For Λ = π,
1
π −1
e−ıĤn t −→
.
(21.126)
−1
1
21.5.1.4 Bloch vector and the Q-function in the JC model with 2 atoms
Despite the more complex structure of the Hilbert space, the Bloch vectors of the
individual atoms and the Q-function can be evaluated in analogy to Sec. 15.2.1. The
Bloch vector is,
√
√ 2 Re ρ12(A1)
~A1 = √12 2 Im ρ12(A1)
ρ where ρij(A1) = Tr |ii1 hj|ψihψ| . (21.127)
ρ22(A1) − ρ11(A1)
770 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
In particular we have,
X
ρ22(A1) = 2 hj|1 hi|hn|1i1 h2|ψihψ|ni|ii1 |ji2 (21.128)
i,j,n
X
= 2 hj|1 h2|hn|ψihψ|ni|1i1 |ji2
j,n
X
= hn| c2,1,m |mihm0 + 1|c∗1,1,m0 +1 + c2,2,m |m0 − 1ihm0 |c∗1,2,m0 i |ni
n,m,m0
X
= (c∗1,1,n c2,1,n + c∗1,2,n c2,2,n ) .
n
Summarizing,
P ∗
ρ12(Ai6=Aj) = (c c + c∗i,j,n c2,2,n )
Pn 1,1,n j,i,n
ρ11(Ai6=Aj) = (|c | + |ci,j,n |2 )
2 . (21.129)
Pn 1,1,n 2 2
ρ22(Ai6=Aj) = n (|cj,i,n | + |c2,2,n | )
X
1 1
Q(α) = π
hα|ρ̂f ield |αi = π
hα| 2 hj|1 hi|ψihψ|ii1 |ji2 |αi (21.130)
i,j
X
= 1
π
hα| (c∗2,2,m−1 c2,2,n−1 |n − 1ihm − 1| + c∗1,2,m c1,2,n |nihm|
m,n
The investigations of the last section can be applied to describe the transfer of quan-
tum coherence from one atom to another. The size of our system is now increased
by the additional degrees of freedom, provided by internal states of the second atom.
A suitable basis was defined in equation (2.110). Now, we imagine the following
Gedankenexperiment: Starting from the initial condition that two adjacent atoms
are in their respective electronic ground states, we assume a microwave π/2-pulse to
create simultaneously, but independently on both atoms, a superposition of the HFS
levels. Then, a non-resonant optical π-pulse interacts with the upper HFS level of
the first atom |1iA1 , and afterward the second atom |2iA2 . Instead of reducing the
atomic states and preparing a Schrödinger cat state in the optical field, (as we did
in Sec. 15.2), we now project the field state onto the coherent state |βi and leave
the atoms in a correlated state. The total procedure can be resumed by tracing the
21.5. CREATING QUANTUM CORRELATIONS 771
Finally, the correlation of the spin orientation of the atoms must be probed with
additional laser light fields, which are selectively irradiated onto the atoms and res-
onantly tuned to a rapidly decaying optical transition. The states correlated in this
way show a relatively low order entanglement, the von Neumann information entropy
only being S 0 = −hlog2 ρi = N − 1.
Q(α)
0 0.2
0 0.1
pn
-1 0
1 5
1 0
0 0 -1 0
-5 0 5
Im ρ12 -1 -1 0 50 100 150 0 10 20 -5
Re ρ12 pn n
t n
Figure 21.15: (code) Evolution of the state during a Jaynes-Cummings interaction with 2
atoms: (a) Bloch vector for the two atoms, (b) time evolution of the coherence ρ12 showing
the phenomenon of collapse and revival, (c) distribution of photons, and (d) Q(α)-function.
21.5.1.6 The XOR gate, treating the field phase as degree of freedom
1 1 1
0
−1 ı
0
−1 ı
−1 1 −1
= . (21.134)
0 1 0
−ı −1 0
−ı −1 0
0 1 0
1 1 ı 1
−ı 1 ı 0
ı ı 1 0
−1 ı 1 −1
= = I ⊗ Uπ/2 ⊗ I . (21.135)
1 1 ı 0
−ı 1 −ı
ı
−ı ı 1 −ı
1 ı 1 0
By reducing the Glauber state |αi we obtain, a posteriori, the unitary XOR gate,
1 1 ı 0
−ı 1 ı −1
= = Uπ/2 ⊗ I . (21.136)
ı ı 1 −ı
−1 ı 1 0
|g1 i|g2 i|0i |g1 i|g2 i|0i |g1 i|g2 i|0i |g1 i|g2 i|0i
|g1 i|e2 i|0i step 1 |g1 i|e2 i|0i step 2 |g1 i|e2 i|0i step 3 |g1 i|e2 i|0i
y y y .
|e1 i|g2 i|0i −ı|g1 i|g2 i|1i +ı|g1 i|g2 i|1i |e1 i|g2 i|0i
|e1 i|e2 i|0i −ı|g1 i|e2 i|1i −ı|g1 i|e2 i|1i −|e1 i|e2 i|0i
(21.137)
21.5. CREATING QUANTUM CORRELATIONS 773
Figure 21.16: Scheme for an XOR-gate in an ion trap or a ring-cavity (see text for expla-
nation). The four possible initial collective states respond differently to a predefined laser
pulse sequence. Possible transitions are represented by solid lines, impossible transitions by
dashed lines. |ai is an auxiliary excited level.
Using the tensor notation |ψi ⊗ |ψi ≡ |ψ1 i|ψ2 i and introducing a matrix notation by
defining a basis as,
1 0 0 0
0 1 0 0
|gi ⊗ |gi =
0 , |gi ⊗ |ei = 0 , |ei ⊗ |gi = 1 , |ei ⊗ |ei = 0 , (21.138)
0 0 0 1
we can set up the truth table for this gate, also known as Cirac-Zoller gate, as,
We now define a new basis for the second atom only via |a2 i = √1 (|g2 i + |e2 i)
2
and |b2 i = √12 (|g2 i − |e2 i). The basis transform corresponds to,
|gi ⊗ |ai |gi ⊗ |gi 10 0 1
|gi ⊗ |bi |gi ⊗ |ei 10 0−1
√1 I ⊗ H I ⊗ H = √12
|gi ⊗ |ai = 2
where
|gi ⊗ |gi 0 1 1 0
|gi ⊗ |bi |gi ⊗ |ei 01 −1 0
(21.140)
is the Hadamard gate applied to the second qubit. Finally, we obtain the controlled
NOT gate,
1 0 0 0
0 1 0 0
UXOR = (I ⊗ H)UCZ (I ⊗ H)−1 = 0 0 0 1 .
(21.141)
0 0 1 0
774 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
In particular, if |ψ1 i is initially in a superposition state, the states |ψ1 i and |ψ2 i end
up entangled,
UXOR (|gi + |ei) ⊗ I = √1 (|gi
2
⊗ |gi + |ei ⊗ |ei) . (21.142)
↓↓ ↓↓
↑↓ − ↓↑
−→ . (21.144)
↓↑ ↑↓
↑↑ ↑↑
This gate is equivalent to the XOR port for the qubits defined by ↓ and ↓ ± ↑, since,
↓ (↓ ± ↑) ↓ (↓ ∓ ↑)
−→ . (21.145)
↑ (↓ ± ↑) ↑ (↓ ± ↑)
21.5.4 Exercises
21.5.4.1 Ex: Generating a Bell state
Show that the operation (H1 ⊗ I)C12 , where H1 is the Hadamard gate acting on the
first qu-bit and C12 the controlled NOT acting on the second qubit, applied to the
2-qubit ground state generates entanglement.
A presentation on the subject is available at (watch talk). See also the website IBM
Circuit Composer.
These two gates are particular cases of the most fundamental single qubit quantum
gate, which can be written as,
cos ϑ2 −eıλ sin ϑ2
U3 (ϑ, ϕ, λ) = , (21.148)
eıϕ sin ϑ2 eıλ+ıϕ cos ϑ2
which is a unitary operation since det U3 (ϑ, ϕ, λ) = eı(λ+ϕ) . Special cases are the
Hadamard gate,
π 1 1 1
H ≡ U3 ( 2 , 0, π) = 2
√ , (21.149)
1 −1
α
|εi = α|0i + ıβ|1i = . (21.153)
ıβ
|q0 i
0
|0i , for example |01i = . (21.154)
1
|1i
The only difference is, that now the possible values are not restricted to the binaries
0 or 1, but can be anything between 0 and 1, provide the total wavefunction stays
normalized.
Figure 21.17: (a) Common symbols for fundamental classical gates and (b) for quantum
gates. (b) Example of a quantum circuit generating Bell type entanglement.
21.6. QUANTUM GATES 777
|q0 q1 i
0
|0 0i 1
|0 1i , for example |01i =
0 . (21.155)
|1 0i
0
|1 1i
Single qubit gates can be embedded in multi qubit registers. For example, the negation
of ’NOT’ gate in H4 can be applied either to the first or the second qubit,
ˆ X̂ ⊗ I = |q¯0 , q1 ihq0 , q1 |
X0 =
. (21.156)
ˆ I ⊗ X̂ = |q0 , q¯1 ihq0 , q1 |
X1 =
The most interesting two-qubit quantum gate is the ’controlled NOT’ gate or
antivalence, which we will now discuss in detail. This gate is originally defined on
H2 ⊗ H2 . The quantum operation is implemented by first going to the Hilbert space
H1 ⊗ H4 , applying the unitary transform,
I2
N= , (21.157)
X
c
X 01 ˆ I⊕X
= = |q0 , q0 ⊕ q1 ihq0 , q1 |
c
. (21.159)
X 10 ˆ S(I ⊕ X)S = |q1 ⊕ q0 , q1 ihq0 , q1 |
=
where the ’SWAP’ operator S has been defined in (2.116). When the qubits addressed
by a gate are not identified from their position in the tensor product, an index at the
gate symbol indicates which qubits are involved. For example, X0 inverts the first
c c
qubit, and X 0→1 = X 01 controls the state of the second qubit by the first one.
778 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
It is possible to show that all quantum logic gates can be reduced to a universal
so-called Toffoli gate [50],
U
H8 −→ H8
U
|q0 , q1 , q2 i −→ |q0 , q1 , (q0 ∧ q1 ) ⊕ q2 i (21.165)
U
I8 −→ U ,
or in short-hand notation,
cc I6
X 012 = I ⊕ I ⊕ X = |q0 , q1 , (q0 ∧ q1 ) ⊕ q2 ihq0 , q1 , q2 | = , (21.166)
σ̂x
0 0 0 1
c I 0
controlled X X = CX I ⊕ σ̂x = 2
0 σ̂x
c I2 0
controlled Z Z = CZ I ⊕ σ̂z =
0 σ̂z
cc I
Toffoli X = CCX I ⊕ I ⊕ σ̂x = 6
σ̂x
PN −1
Similarly, the quantum Fourier transform acts on a quantum state |xi = i=0 xi |ii
PN −1
and maps it to a quantum state i=0 yi |ii according to the same formula (21.168).
In case that |xi is a basis state, the quantum Fourier Transform can also be expressed
as the map,
N −1
1 X xk
|xi 7→ √ ωn |ki . (21.169)
N k=0
Equivalently, the quantum Fourier transform can be viewed as a unitary matrix (or
a quantum gate, similar to a Boolean logic gate for classical computers) acting on
quantum state vectors, where the unitary matrix is given by,
1 1 1 1 ··· 1
1 ω ω2 ω3 ··· ω N −1
2 4 6 2(N −1)
1 1 ω ω ω · · · ω
FN = √ 1 ω 3
ω 6
ω 9
· · · ω 3(N −1) (21.170)
N
.. .. .. .. ..
. . . . .
1 ω N −1 ω 2(N −1) ω 3(N −1) · · · ω (N −1)(N −1)
Figure 21.19: Quantum circuit for QFT with n qubits (without rearranging the order of
output states).
For instance, [0.x1 ] = x21 and [0.x1 x2 ] = x21 + x222 . With this notation, the action of
the quantum Fourier transform can be expressed in a compact manner:
1
QFT(|x1 x2 . . . xn i) = √ |0i + e2πı [0.xn ] |1i ⊗ |0i + e2πı [0.xn−1 xn ]
|1i ⊗ · · ·
N
⊗ |0i + e2πı [0.x1 x2 ...xn ] |1i , (21.175)
where we have used [0.x1 x2 ...xm ] = [x1 x2 ...xn ]/2m . This can be seen by rewriting
the formula for the Fourier transform in the binary expansion:
2n −1
1 O
n
1 X xk n−j
QFT(|xi) = √ ωn |ki = ... = √ |0i + ωnx2 |1i . (21.176)
N k=0 N j=1
then a(j) ∈ N0 , because 2n−j−r ≥ 0, for n−j−r ≥ 0, and b(j) = 0.xn−j+1 xn−j+2 . . . xn ,
thus the (2) becomes:
1 O
n
n−j
QFT(|x1 x2 . . . xn i) = √ |0i + ωnx2 |1i (21.179)
N j=1
1 O
n
=√ |0i + e2πı[0.xn−j+1 xn−j+2 ...xn ] |1i
N j=1
1
=√ |0i + e2πı[0.xn ] |1i ⊗ |0i + e2πı[0.xn−1 xn ] |1i ⊗ · · · ⊗ |0i + e2πı[0.x1 x2 ...xn ] |1i .
N
To obtain this state from the circuit depicted above, a swap operations of the
qubits must be performed to reverse their order. After the reversal, the n-th output
qubit will be in a superposition state of |0i and e2πı [0.x1 ...xn ] |1i, and similarly the
other qubits before that (take a second look at the sketch of the circuit above).
In other words, the discrete Fourier transform, an operation on n qubits, can be
factored into the tensor product of n single-qubit operations, suggesting it is easily
represented as a quantum circuit (up to an order reversal of the output). In fact, each
of those single-qubit operations can be implemented efficiently using a Hadamard
gate and controlled phase gates. The first term requires one Hadamard gate and
(n − 1) controlled phase gates, the next one requires a Hadamard gate and (n − 2)
controlled phase gate, and each following term requires one fewer controlled phase
gate. Summing up the number of gates, excluding the ones needed for the output
reversal, gives n + (n − 1) + · · · + 1 = n(n + 1)/2 = O(n2 ) gates, which is quadratic
polynomial in the number of qubits.
21.6.5 Exercises
21.6.5.1 Ex: Three-bit SWAP gate
a. Convince yourself that the SWAP gates defined by (21.161) have the explicit matrix
form,
k = [0, ..., 7]
k = [0, 1, 4, 5, 2, 3, 6, 7]
01
Sij (k, kij ) ≡ 1 where ,
k12 = [0, 2, 1, 3, 4, 6, 5, 7]
k02 = [0, 2, 4, 6, 1, 3, 5, 7]
the index k indicating the column and kij the row where the matrix has an entry.
b. Verify,
−1
S −1 = S , Sij = Sij .
784 CHAPTER 21. CORRELATED ATOMS, ENTANGLEMENT AND QUANTUM GATES
Figure 21.21:
21.7.3 on entanglement
A. Einstein et al., Can quantum-mechanical description of physical reality be consid-
ered complete [DOI]
J.S. Bell et al., On the Einstein-Podolsky-Rosen paradox [DOI]
A. Aspect, Experimental Test of Bell’s Inequality Using Time-Varying Analyzers
[DOI]
S.J. van Enk, Single-particle entanglement [DOI]
789
790 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
with [â, ↠] = 1. With this we obtain, for a single atom coupled to the mode of
the cavity pumped by a laser, within the dipolar approximation and the RWA 1 , the
following relevant contributions to the Hamiltonian (~ = 1),
p̂2
Ĥatom = −∆a σ̂ + σ̂ − + 2m (22.2)
†
Ĥcav = −∆c â â
Ĥatom:cav = g↠σ̂ − e−ıkẑ + h.c.
Ĥlaser:cav = ıη↠+ h.c. ,
displacement operator. ∆a is the detuning between the light and the atomic resonance,
∆c between the light and cavity resonance and g is the light-atom coupling force, also
called the single-photon Rabi frequency. Neglecting the kinetic energy term p̂2 /2m,
the photonic recoil e−ıkvẑ , and the pumping η we recover the Jaynes-Cummings model.
Do the Excs. 22.1.7.1 and 22.1.7.2.
For a ring cavity, we must distinguish the counterpropagating modes â± ,
Ê + (z, t) = Ê1 â+ (t)eıkz + Ê1 â− (t)e−ıkz = (Ê − (z, t))† . (22.3)
p̂2
Ĥatom = Ĥelectron + Ĥmotion = −∆a σ̂ + σ̂ − +
2m
Ĥcav = −∆c â†+ â+ − ∆c â†− â−
. (22.4)
Ĥatom:cav = gâ†+ σ̂ − e−ıkẑ + h.c. + gâ†− σ̂ − eıkẑ + h.c.
Ĥlaser:cav = −ıη+ (â+ − â†+ ) − ıη− (â− − â†− )
We identify the degrees of freedom of the system through the quantum observables
appearing in the Hamiltonian: the counterpropagating modes of light with the am-
plitudes (â± ), the internal degrees of freedom (σ̂ z , σ̂ ± ), and the spatial coordinates of
the atom (ẑ, p̂).
In contrast to linear cavities, ring cavities have the following particularities: 1. The
phase of the standing wave is free to move; 2. the counterpropagating modes of the
cavity have independent photon budgets, each backscattering event conserves mo-
mentum; 3. the backscattering acts on the phase of the standing wave. Atoms can
be trapped by the dipole force within the cavity mode volume. The dipole force
corresponds to a backscattering of photons between modes.
Figure 22.2: Scheme of an atom interacting with a ring cavity showing the relevant degrees
of freedom (x̂, p̂, â± ) and the possible decay processes (κ, Γ) for the derivation of the field
equations.
2 Throughout this chapter we will sometimes emphasize the motional degree of freedom in green
in the description of the coupled dynamics. Here, gatom:vacuum denotes the vacuum
Rabi frequency, that is, the coupling strength between the atomic dipole and the
electromagnetic vacuum into which spontaneously emitted photons may escape. On
the other hand, gcav:bath denotes the coupling strength between the cavity mode and
the spectrum of electromagnetic modes into which cavity photons can escape and
which we will from now on call heat bath to facilitate its semantic distinction from the
spontaneous vacuum. The total density operator, the Hamiltonian and the equation
of motion are, consequently,
ρ̄ = ρ̂atom ⊗ ρ̂electron ⊗ ρ̂cav ⊗ ρ̂vacuum ⊗ ρ̂bath (22.6)
H̄ = Ĥmotion + Ĥelectron + Ĥcav + Ĥatom:cav + Ĥlaser:cav + Ĥatom:vacuum + Ĥcav:bath
ρ̄˙ = −ı[H̄, ρ̄] .
With the inclusion of the vacuum field modes âω and âk the number of degrees of
freedom to be treated literally explodes, and we have to find a way to eliminate them
from the equation of motion. We do this by partially tracing the density operator
over the vacuum field modes using the Weisskopf-Wigner theory,
ρ̃ ≡ Trvacuum Trbath ρ̄ = ρ̂atom ⊗ ρ̂electron ⊗ ρ̂cav . (22.7)
This allows us to reduce the Hamiltonian Ĥ = H̄ − Ĥatom:vacuum − Ĥcav:bath . The
price to pay is, that the equation of motion for this receives dissipative terms: The
Liouville equation turns into a master equation with the following form,
where Ĥ is the Hamiltonian from (22.4). We see that each degree of freedom has its
own loss mechanism. κ for the finite transmission of the resonator mirrors [152, 269] 3 ,
Γ for the spontaneous emission, and γf rc , when we exert a frictional force on the
atoms.
3 Here, we only consider cavity decay into a zero temperature reservoir without quantum phase
fluctuations.
22.1. CAVITY INTERACTING WITH A SINGLE ATOM 793
and similarly for â− . For the Pauli deexcitation matrix, using the following commu-
tation rules, [σ̂ − , σ̂ + ] = σ̂z , σ̂ ± σ̂z = ±σ̂ ± , and [σ̂ − , σ̂ − ] = 0, we calculate,
For the Pauli inversion matrix, using the following commutation rules, [σ̂ + σ̂ + , σ̂z ] = 0
and [σ̂ ± , σ̂z ] = ±2σ̂ ± , we calculate,
Finally, we need to derive the equations governing the motion of atoms. For the
position we obtain,
p̂2
ẑ˙ = ı[Ĥ, ẑ] = ı[ 2m 1
, ẑ] = m p̂ . (22.12)
794 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
δẑ n
[ẑ n , p̂] = ı = ınẑ n−1 (22.13)
δẑ
X∞ X∞
(ık)n n (ık)n−1 n−1
[eıkẑ , p̂] = [ẑ , p̂] = −k ẑ = −keıkẑ ,
n=0
n! n=0
(n − 1)!
The quantum Langevin equations [152] describing the dynamics of the internal and
external degrees of freedom of the atom and the field are coupled,
[Ĥ, ~k(n̂+ − n̂− ) − p̂] = ~kη+ (â†+ + â+ ) − ~kη− (â†− + â− ) , (22.17)
in the bad cavity limit, the fields â± drop out of the dynamics.
22.1. CAVITY INTERACTING WITH A SINGLE ATOM 795
Therefore, the internal state has no separate dynamics of its own, and we can adia-
batically eliminate the internal degrees of freedom. Thus, we can neglect correlations
between degrees of freedom, hâ± σ̂ ± i = hâ± ihσ̂ ± i etc. [558, 265]. The adiabatic elim-
ination of the excited state comes down to treating the atom as a classical antenna.
We obtain the stationary solutions for t → ∞ in the same way as for the optical
Bloch equations [see Eq. (14.99)], assuming in Eq. (22.15),
σ̂˙ − = 0 = σ̂˙ z (22.18)
where hσ̂ − i = ρ21 . Introducing the position-dependent Rabi frequency of the atom in
the standing wave 5 ,
Ω̂(z) = 2g(eıkẑ â+ + e−ıkẑ â− ) , (22.19)
we write,
0 = (−γ + ı∆a )σ̂ − + 12 ıΩ̂σ̂z (22.20)
† − +
0 = −2γ − 2γ σ̂z + ıΩ̂ σ̂ − ıΩ̂σ̂ .
These equations are solved by,
−2(γ 2 + ∆2a )
σ̂z (∞) = (22.21)
2(γ 2 + ∆2a ) + Ω̂† Ω̂
−ı(γ + ı∆a )Ω̂ −ı(γ + ı∆a )
and σ̂ − (∞) = ' Ω̂ .
2(γ 2 + ∆2a ) + Ω̂† Ω̂ 2∆2a
Inserting the approximated expression for σ̂ − (∞) into the equation of motion (22.15)
for the light modes,
â˙ ± = (−κ + ı∆c )â± − ıgσ̂ − (∞)e∓ıkẑ + η± (22.22)
2 2
2 2
g γ ıg g γ ıg
= −κ + ı∆c − 2 − â± − + e∓2ıkẑ â∓ + η± ,
∆a ∆a ∆2a ∆a
and for the atomic momentum,
p̂˙ = ıg~kσ̂ − (∞)(â†+ e−ıkẑ − â†− eıkẑ ) − ıg~kσ̂ + (∞)(â+ eıkẑ − â− e−ıkẑ ) (22.23)
2 2
2~kγg 2ı~kg 2ıkẑ
= (â†+ â+ − â†− â− ) − (e â+ â†− − e−2ıkẑ â†+ â− ) .
∆2a ∆a
Defining the light-shift U0 caused by only one photon and the scattering rate γ0 by,
g2 Γg 2
U0 ≡ and γ0 ≡ , (22.24)
∆a ∆2a
with Γ = 2γ, we finally get a set of equations, where the internal degree of freedom
of the atom has been eliminated,
as in (15.28). Note also, that [Ω̂, Ω̂† ] = 8g 2 6= 0, but this is negligible when the fields are large enough
to be considered as classical. On the other hand the quadratic terms Ω† Ω are negligible when they
are small compared to γ 2 or ∆2a .
796 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
p̂2 X
Ĥ = + (U0 − ∆c )↱ â±
2m ± X , (22.26)
+ U0 (e−2ıkẑ â†+ â− + e2ıkẑ â+ â†− ) − ı η± (â± − ↱ )
±
as shown in Exc. 22.1.7.3. Note that dissipative terms are (naturally) absent from this
Hamiltonian, so that it shall not be used for the description of radiation pressure. On
the other hand, all coherent terms of the equations of motion can be derived from this
Hamiltonian (22.26), and we can verify momentum conservation (22.17) and deduce
transformation properties. Solve the Exc. 22.1.7.4.
These equations, which we will call CARL equations for reasons that we will be-
come clear in Sec. 22.2, describe the coupled dynamics of atoms being accelerated by
the kick eıkẑ imparted by the photonic recoil received upon scattering a photon from
one mode into the counterpropagating one. In the same time, the backscattering an-
nihilates a photon â± in one mode and creates a photon â†∓ in the counterpropagating
mode. From now on we will exclusively use the CARL equations.
Finally, let us summarize, how the operators act on states and observables of the
coupled system, noting that the same transformation rules as for free and trapped
atoms (3.124) also hold for the optical lattice,
scale, adiabatically following the evolution of the other degrees of freedom. Hence,
we set,
â˙ ± = 0 , (22.29)
and obtain from the first Heisenberg equation (22.15),
ıgσ̂ − e∓ıkẑ − η±
â± (∞) = . (22.30)
−κ + ı∆c
Defining,
g 2 ∆c g 2 κc
Uc ≡ and κc ≡ , (22.31)
κ2 + ∆2c κ2 + ∆2c
we can write the position-dependent Rabi frequency (22.19) as,
η+ eıkẑ + η− e−ıkẑ
Ω(z) = 2g(eıkẑ â+ + e−ıkẑ â− ) = 2(κc + ıUc ) − 2ıσ̂ − (22.32)
g
η± =0
−→ −4ı(κc + ıUc )σ̂ − ,
in the limit of no pumping η± = 0. Inserting this into the other Heisenberg equations
(22.15), we immediately get,
η± =0
σ̂˙ − = (−γ + ı∆a )σ̂ − + 2ı Ω(z)σ̂z −→ (−γ + ı∆a )σ̂ − + 2(γc + ıUc )σ̂ − σ̂z , (22.33)
and
η± =0
σ̂˙ z = −2γ σ̂z + ıΩ† (z)σ̂ − − ıσ̂ + Ω(z) −→ −2γ σ̂z − 4κc Î , (22.34)
as well as, ẑ˙ = p̂/m and
η± =0
p̂˙ = ıg~kσ̂ − (e−ıkẑ â†+ − eıkẑ â†− ) − ıg~kσ̂ + (eıkẑ â+ − e−ıkẑ â− ) −→ 0 . (22.35)
Alternatively, the coherent part of these equations can be derived by insert â± (∞)
directly in the Hamiltonian (22.4). This will be done in Exc. 22.1.7.6. The equations
(22.36) tell us that, in the absence of cavity decay and spontaneous emission, the
atomic population will not undergo nutation, but the dipole moment will rotate with
a velocity which depends on the inversion. The dynamics becomes interesting in the
presence of several atoms, as studied in 21.2.3.
section aims at providing a general recipe for choosing the right set of equations
depending on two basic criteria for the nature of the degrees of freedom involved in
the dynamics: (i) Do judge it necessary to treat the degree of freedom as quantum or
may a classical description be sufficient; and (ii) Is the degree of freedom subject to
dissipation (e.g. spontaneous emission of the electronic excitation, cavity decay of the
field mode, or collisions messing up the center-of-mass motion), or not. The procedure,
which will be applied throughout the remaining part of this chapter, leads to very
different descriptions of the system depending on the specific parameter regime.
coupling the atomic motion, the light fields and the electronic excitation, al-
though the electronic excitation |ii is often eliminated adiabatically. The dis-
sipative degrees of freedom related to vacuum modes leading to spontaneous
emission and cavity decay, as well as collisions between moving atoms are traced
away, but may be considered in the master equation for the density operator
and in the Heisenberg equations, where the corresponding decay rates are added
phenomenologically.
This first item has already been solved in the previous sections and led us to
the Hamiltonian (22.4) and the corresponding Heisenberg equations (22.15),
or the adiabatically approximated Hamiltonian (22.26) and the corresponding
Heisenberg equations (22.25).
2. Now, we must decide which degrees of freedom B can be treated as classical.
Typically, those are highly excited degrees of freedom (e.g. fast velocities of
many photons in a mode). The corresponding operators can be substituted by
their expectation value. Purely classical energy terms in the Hamiltonian can
be ignored and removed. For the degrees of freedom  we want to treat as
quantum, we chose an appropriate common basis, which can be discrete {|mi}
where m is a complete set of quantum numbers. It can also be continuous {|ri}
or a combination of both {|r, mi}. Now, we need to expand all operators on the
chosen basis.
3. For the quantized degrees of freedom must now decide, whether they all evolve
coherently or whether they are subject to dissipation. In the first case, perform
steps A4 to A7, in the second case, perform steps B4 to B7.
A4. In the case of coherent dynamics, we expand the state of the system |ψ(t)i on
the whole basis. For example, expanded on a partially continuous basis, the
6 Throughout this chapter we will denote momentum states by the Greek letters |µi or νi. Photonic
states will be labeled by the Latin letters |ni or |mi. Electronic excitation states will be labeled by
the Latin letters |ii or |ji. Finally, expansion coefficients cν,i,n+ ,n− of the state or of the density
matrix elements will sometimes be emphasized in red.
22.1. CAVITY INTERACTING WITH A SINGLE ATOM 799
where hr, α+ , α− , i|ψ(t)i = cn+ ,n− ,i (r, t) are the expansion coefficients depend-
ing on photon numbers n± in the counterpropagating light modes, the electronic
excitation state i, and the atomic position r treated in terms of a continuous
wavefunction in space 7 . When we want to treat the atomic motion as being
quantized in discrete momentum states labeled by some integer number µ, we
adopt the notation,
X
|ψ(t)i = cν,n+ ,n− ,i (t)|νi ⊗ |n+ i ⊗ |n− i ⊗ |ii , (22.39)
ν,n+ ,n− ,i
where hν, α+ , α− , i|ψ(t)i = cν,n+ ,n− ,i (t) are the new expansion coefficients.
A5. Next, we write down the Schrödinger equation for the state |ψ(t)i, insert the
expansion on the basis, and we derive a linear set of equation of motion for the
expansion coefficients ċ{m} (r). This set governs the dynamics of the quantum
degrees of freedom.
A6. The dynamics of the classical degrees of freedom B is obtained by taking the
expectation values of the Heisenberg equations. Here, we need to take care
that the quantum degrees of freedom appearing in the Heisenberg equations are
expressed by their expansions.
A7. The coupled set of equations for the expectation values of projectors of the
system into a particular state, that is, ċ{m} (respectively, ċ{m} (r)), and of ob-
servables B really represents all we need to describe the system dynamics and
to simulate it numerically. Hack everything into your PC, pronto!
B4. In the case of dissipative dynamics, derive the master equation obeying commu-
tation rules for all quantized degrees of freedom including the Lindbladt terms
and expand every quantized degrees of freedom on the common basis, we do the
same expansions, but,
ρ̂(t) = |ψ(t)ihψ(t)| . (22.40)
B5. Derive the linear set of equation of motion for the matrix elements ρ̇{m},{n} .
B6. Take the expectation values Tr ρ̂B̂ of the Heisenberg equations for all degrees of
freedom to be handled classical as expanding the quantized degrees of freedom
on their basis.
B7. The coupled set of equations for ρ̇{m},{n} and Ḃ is sufficient to describe the
dynamics of the system, hack everything into your PC, pronto!
7 Note that, in the absence of other quantum numbers, we rather use to write c(r) = ψ(r).
800 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
Let us apply the procedure to the simplest case that the electronic excitation has
been adiabatically eliminated and all remaining degrees of freedom can be treated
classical. Then we do not require a Hamiltonian, nor the Schrödinger equation. We
just take the expectation value of the Heisenberg equations (22.25) for all degrees of
freedom which is easy to do, because there is no quantum state to be expanded: we
just can replace the operators by c-numbers,
These totally classical equations, called CARL equations, will be studied in the subse-
quent sections, while a thorough discussion of partially quantized equations of motion
is postponed to Sec. 22.4.
p̂2 X
Ĥ = + U0 ↱ â± + U0 (e−2ıkẑ â†+ â− + e2ıkẑ â+ â†− ) , (22.42)
2m ±
and read,
˙
b̂± = −ıU0 e∓2ıkẑ b̂∓ (22.44)
¨
b̂± = −U02 b̂± − 2k żU0 e∓2ıkẑ b̂∓ ,
we see that, if the motion weren’t a dynamic variable, ẑ˙ = 0, the field amplitudes
would just perform harmonic oscillations.
On the other hand, with the substitution ĉ ≡ â+ eıkẑ + â− e−ıkẑ 8 , the complete
Hamiltonian rephrased as,
p̂2
Ĥ = + U0 ĉ† ĉ . (22.45)
2m
Here, we see that, if the motion described by (ẑ, p̂) weren’t a dynamic variable, our
system would simply be a harmonic oscillator vibrating with the frequency U0 .
8 Provided we are allowed to commute the operators ẑ and â± .
22.1. CAVITY INTERACTING WITH A SINGLE ATOM 801
For atoms much hotter than the recoil limit and macroscopic light intensities we may
replace the quantum operators by complex numbers z ≡ ẑ, α± ≡ â± and ρ21 = σ̂ − .
The classical equations of motion (22.41) for the coupled system of a single atom
confined at the position r = z of the dipolar potential of a ring cavity can be cast
into the form,
α̇+ −κ − γ0 + ı(∆c − U0 ) −(γ0 + ıU0 )e−2ıkz α+ η
= + +
α̇− −(γ0 + ıU0 )e2ıkz −κ − γ0 + ı(∆c − U0 ) α− η− .
∗ ∗ ∗
ṗ = 2~kγ0 (α+ α+ − α− α− ) + 2ı~kU0 (α+ α− e−2ıkz − ∗ 2ıkz
α+ α− e )
(22.46)
∗
Recalling that α± α± is the number of photons in the respective mode, we can interpret
this equation as a rate equation: The number of photons in a mode α+ changes by
photon losses at a rate κ from resonator, or by gain due to backscattering from the
counterpropagating mode, or by pumping with an external incident light field at rate
η+ .
The equations (22.46) completely describe our coupled atom-cavity system. They
are totally classical and work for both, atoms and macroscopic particles.
d 2
â = ı[Ĥ, â2+ ] (22.47)
dt +
= −2ı(U0 − ∆c )â2+ + 2U0 e2ıkẑ â+ â†− − e−2ıkẑ â+ â− − 2ıη+ â+ .
This is the lowest order cumulant expansion. The expectation values form a system
of 10 linear first order differential equations [124] from which we can calculate the
steady-state of the system.
The correlation functions and spectra are obtained directly from the CARL equa-
tion for the cavity fields (22.25) using the Wiener-Khinchin theorem (15.101) and
the quantum regression theorem (15.155). With the substitutions, t → 0, B̂ → 1,
 → dÂk /dτ , and ξi (τ )Âi (0) → ζi Âi (τ ), the quantum regression theorem can be
written in the form,
d X d X
hÂk (τ )i = ζi hÂi (τ )i =⇒ hÂk (τ )Ĉ(0)i = ζi hÂi (τ )Ĉ(0)i .
dτ i
dτ i
(22.48)
802 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
Applied, for example, to the correlation function (22.47), we obtain from the CARL
equation (22.25), with γ0 = 0 and ∆c = 0,
d
hâ+ (τ )â+ (0)i = (−κ − ıU0 )hâ+ (τ )â+ (0)i − ıU0 e−2ıkẑ hâ− (0)â+ (0)i + η+ hâ+ (0)i .
dτ
(22.49)
Repeating this calculation for the other first-order correlations functions we obtain a
system of 10 linear first order differential equations, which can be solved via Laplace
transform.
22.1.7 Exercises
22.1.7.1 Ex: Origin of quantum correlations
Derive the Hamiltonian (22.2) from the JCM Hamiltonian (15.28) transforming it
into the co-rotating frame.
N
X
α̇± = (−κ − ıN U0 )α± − ıU0 e∓2ıkzj α∓ + η±
j=1 . (22.50)
∗ 2ıkzj ∗ −2ıkzj
ṗj = −2ı~kU0 (α+ α− e − α− α+ e )
804 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
called bunching parameter, vanishes. That is, the impact of the scatterers on the light
modes cancels out, as we will see in Exc. 22.2.5.1 for the case of two atoms. If on
the other hand, atoms accumulate in the antinodes of the standing wave, it increases
the contrast of it can spread more efficiently collectively by Bragg scattering. The
particularity of the CARL is that during the temporal evolution the bunching process
can amplify itself leading to an exponential growth of the counterpropagating mode,
accompanied by an increasingly pronounced self-bunching.
In general, the equations can not be solved analytically, especially when the pump
varies over time. A first approach consists in iterating them numerically,
In Exc. 22.2.5.2 we will extend the equations (22.50) to the presence of two atomic
species, and in Exc. 22.2.5.3 we will use them to describe the response of the light
fields to an inertially moving atom supposing that is does not feel the CARL force.
happens, because the axial motion influences the number of intracavity photons of the radiation field
which, in turn, determines the depth of the dipole potential.
22.2. CARL: THE COLLECTIVE ATOMIC RECOIL LASER 805
apply the boundary condition (22.54) and eliminate the pump mode α+ from the
dynamics of the system. That is, the following equations are usually sufficient to
describe the CARL:
The frequency offset of the cavity resonances caused by the atom, U0 , can exceed the
width of the cavity κ. From equation (20.124) we know,
χχ∗
|α+ (∞)|2 = 2
η+ .
(χ2 + U02 )(χ∗2 + U02 )
The maxima of |α+ (∞)|2 as a function of ∆c give the shifted resonances of the modes.
Example 107 (Locking on transverse modes): We already mentioned that
as the CARL accelerates, the frequency of the light which is backscattered to the
probe shifts to the red until it escapes from the resonant mode. What happens
if we provide another resonant mode that can receive photons? We will show in
the following calculation, that CARL simply picks up the closest mode to dump
the photons. The starting point is generalized CARL equations to accommodate
a second reverse mode labeled β− ,
×109
600 2
(kHz)
|α+ + α− |2
400
1
200
pcm
0 0
0 10 20 30 0 20 40 60
κt κt
-2
p/h̄k
100
rcm
-4
0
0 10 20 30 0 2 4
kr
Figure 22.5: In the momentum picture the CARL acceleration process occurs as Raman-
anti-Stokes processes along the free-particle dispersion relation.
When only one atom is in the cavity or when the atoms are perfectly bunched
together, it is possible to derive analytical solutions. Particularly interesting is the
following situation: We pump the cavity from one side. The pump is supposed to
be dominant and locked to a resonance, such that we can neglect the feedback of the
system on the pump, that is, we can assume, α+ = η/κ. Using the abbreviations
χ ≡ κ + ıU0 − ı∆c and the photon recoil shift [265],
~k 2
ωrec ≡ , (22.58)
2m
the equations (22.26) then become,
we assume that the atom and the standing wave have the same velocity, that is, they
move in phase. We obtain as solution,
−ıUN α+ 2κ
β= , k v̇ = 8ωrec U02 α+
2
. (22.61)
κ + 2ıkv κ2 + 4k 2 v 2
If κ 2kv, then the differential equation is approximately solved by,
(kv)3 = 3εκU02 α+
2
t. (22.62)
This means that the CARL frequency, that is, the frequency difference between the
emitted probe wave and the incident light, increases temporarily. The frequency
corresponds to the double Doppler shift. As the frequency of the probe light gradually
shifts away from the cavity resonance, the probe light finally stops being amplified,
and the amplitude of the probe field decreases: CARL is only a transient phenomenon.
In fact, the behavior described by the equation (22.61) was observed in experiments
[429].
Example 108 (Universal scaling ): Our formula describing CARL,
ıU0 η+ X 2ıkzj
α̇− = − e − χα− (22.63)
χ j
∗
α− −2ıkzj α− 2ıkzj
kv̇j = −4ωrec ıU0 η+ e − e − γf ric kvj ,
χ∗ χ
where χ = κ + ıN U0 − ı∆c , can be rewritten in terms of a universal ’scaling’ if
we define,
γf ric
τ = 4ωrec ρt and γ̄ = (22.64)
4ωrec ρ
2kvj
θj = 2kzj and Pj =
4ωrec ρ
ı|χ| χ
A= √ α− and χ̄ = .
χ∗ ρN 4ωrec ρ
We obtain,
1 X ıθj
Ȧ = e − χ̄A (22.65)
N j
θ̇ = Pj
Ṗj = −2(A∗ e−ıθj + Aeıθj ) − γ̄Pj .
×109 (c)
(a) (b) 2 -0.2
1
|α+ + α− |2
Udip /η2 U0
-0.205
kvcm /κ
1
0.5
-0.21
0
0 -0.215
0 0.2 0 0.2 0 2 4
κt κt kz/π
Ff rc = −βf rc p . (22.69)
We observe that the frequency shift of the backscattered light stabilizes at a lower
value and that the oscillation amplitude of the interference signal is stronger than
without molasses [722].
(b) 2 ×10
9
(a) (c) -0.18
1
|α+ + α− |2
Udip /η2 U0
-0.2
kvcm /κ
1
0.5
-0.22
0
0 -0.24
0 0.5 0 0.5 0 2 4
κt κt kz/π
Figure 22.8: (code) Influence of an external friction force (switched on at κt = 0.3) propor-
tional to βf rc = 3.8 · 106 s-1 . Other parameters are the same as in Fig. 22.6.
Let us note that, while provide a nice intuition, the description of the impact of
an optical molasses force is incomplete, because it disregards heating effects caused
by random photon scattering processes. The temperature of the atomic cloud is given
810 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
It is also interesting to study the impact of a weak injected probe light on the CARL
dynamics. This light may have the same frequency as the pump light, η− /|η− | =
η+ /|η+ |. This is for example the case, when light is backscattered from imperfection
of the cavity mirror surfaces. Or may have a different frequency (provided the probe
frequency is resonant to a cavity mode),
η− = eı∆νt . (22.70)
The injected light fields η+ and η− create a standing light wave potential, in which
the atoms move as being subject to an array of hurdles. The CARL force is too weak,
the atoms will get stuck. We study this in Exc. 22.2.5.4. In Exc. 22.2.5.5 we study
CARL dynamics for the case that the atoms are subject to an additional harmonic
potential.
∗ ∗
Frp = 2~kγ0 (α+ α− − α− α+ ). (22.71)
But this force is weak when the laser is tuned far enough from the atomic resonance.
Fig. 22.9 shows a simulation of the impact of radiation pressure for the same
system as in Fig. 22.6. For clarity the radiation pressure force has been exaggerated
by a factor of 100.
1
kvcm /κ
0.5
0
0 0.1 0.2 0.3
κt
1
UN = 0.1
|α− |2 /|α+ |2
0.5 UN = 5
UN = 10
0
0 10 20 30
κt
Figure 22.10: (code) Dynamics simulated with equations (22.46). At time κt = 0 the pump
rate η− is reduced by 1%. After a certain time that depends on N U0 , the standing wave
breaks down ejecting atoms from the mode volume. Here, we simulate this via an artificial
linear trap loss process setting in at time κt = 15. As a result of the diminishing N U0 the
standing wave recovers.
The physical explanation for the instability is reminiscent to the CARL behavior
described in previous sections. Because atom-field coupling is strong, a small imbal-
ance of the injected beam intensities is sufficient to displace the atoms to a location,
where the light which they scatter into the reverse direction interferes destructively
with the injected light. A simple argument explains, why the standing wave is at-
tracted towards a position where it gets unstable. We treat the imbalance η+ > η−
as if it was due to a scatterer fixed in space inside the mode volume sitting on an
edge of the standing wave (corresponding to a π/4 phase shift), such that it only
scatters light from α+ into α− , but not the other way round. Whether the phase of
the standing wave adjusts itself to that of pump field or such that the scatterer sits at
the bottom of the potential well depends on the ratio of the coupling to the external
field (given by κ) and to the scatterer (given by U0 ). If U0 is stronger, the field is
pulled towards the scatterer dragging along the atoms which are free to move. The
stationary situation is therefore a displacement of the standing wave and the atoms
by λ/4, which is just the position where the cavity field α− and the injected field
812 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
(in)
α− are out of phase. The injected light is not transmitted through the cavity any
more.
Note that the instability occurs in a plane wave situation, there is no need to
consider the transversal motion. Furthermore, it is a single-atom effect, since we
assume perfect bunching. We may therefore consider a single atom strongly coupled to
the cavity and use the set of equations (22.46). In the undepleted pump approximation
we assume α̇+ = 0, and if the atom adiabatically follows the dynamics of the potential
valley, ẍ = 0, so that,
s
α+ α−∗
0 = −χα+ − ıU0 ∗ α− + η+ , (22.72)
α+ α−
s
∗α
α+ −
α̇− = −χα− − ıU0 ∗ α+ + η− .
α+ α−
Neglecting the impact of the atom on the amplitudes of the fields, we may derive the
steady-state solution,
χ + ıN U0 (1 ∓ 2ıkZcm )
α± = η . (22.77)
χ2 + N 2 U02
As shown in (22.28) the normalized field intensity can be written as,
1
2ε0 cE12
I(z, t) = |α+ |2 + |α− |2 + 2|α+ ||α− | cos(2kz + 2θ) , (22.78)
4N U0 (∆c − 2N U0 )
' 2 kZcm .
κ + (∆c − 2N U0 )2
An individual atom in this optical potential feels the dipolar force,
which describes the force on a single atom. Note that the harmonic force, ω02 kZj ,
on the atom is continuously fed by the center-of-mass force, ω02 µj kZcm . The above
equation can be rewritten as,
µ µ
k Z̈1 0 −ω02 (1 − Nj ) 0 −ω02 Nj . k Ż1
k Ż 1 0 0 0 .
1 kZ1
2 µj 2 µj
k Z̈2 = 0 −ω0 N 0 −ω0 (1 − N ) . k Ż2 . (22.84)
k Ż2 0 0 1 0 . kZ2
. . . . . . .
814 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
p
The eigenvalues of the matrix are e = ıω0 , ıω0 1 − µj . Therefore, we expect, in
addition to the secular frequency ω0 , a second oscillation frequency,
p
ωcm = ω0 1 − µj . (22.85)
We find splitting in the strong coupling regime, g > Γ, but we consider the weak
coupling regime, g < Γ, to implement the phononic coupling.
Example 110 (Two-coupled atoms): Here we rewrite the field equations for
two atoms in center-of-mass coordinates,
z1 + z2
Z= and z = z2 − z1 ,
2
finding,
α̇± = (−κ + ı∆c − 2ıU0 )α± − 2ıU0 α∓ e∓2ıkZ cos kz + η .
22.2.5 Exercises
22.2.5.1 Ex: Anticorrelated atoms
Study the CARL equations (22.50) for the case of two fixed atoms sitting at positions
kz1 = kz2 − π.
22.2. CARL: THE COLLECTIVE ATOMIC RECOIL LASER 815
×107 ×108
(a)
Etot /h (MHz) 8 (b) 3 (c) 2.026
6 2
|α+ |2 − |α− |2
|α+ |2 + |α− |2
2.025
4 1
2.024
2 0
0 -1 2.023
1 2 3 4 1 2 3 4 1 2 3 4
ωφ /ω0 ωφ /ω0 ωφ /ω0
Figure 22.11: (code) Response of the cavity field to a phase modulation of the pump light.
(a) Kinetic energy, (b) photon imbalance, and (c) total photon number. The simulation
assumes N = 106 strontium atoms initially at T = 1 µK driven on the λ = 689 nm line and
κ/2π = 3.4 MHz, ∆c = 2U0 , η = 104 κ, N U0 /κ = −.02, φ0 = 0.01. The calculated secular
√
frequencies are ω0 = 0.35κ and ω0 1 − µN = 0.32κ.
22.3.1.1 RIR-spectroscopy
Two laser beams 1 and 2 having each two different frequencies ω−∆ω/2 and ω+∆ω/2
are irradiated from two different directions enclosing a small angle θ into the atomic
trap,
E(r, t) = E1 + E2 (22.88)
(0)
1
1
=E cos k1 · r − (ω − 2 ∆ω)t + cos(k2 · r − ω + 2 ∆ω)t
(0)
≈ 2E cos (K · r − ωt) cos 21 q · r − 12 ∆ωt ,
Regarding the momentum transfer, the 4WM process can be interpreted as Bragg
scattering af the atoms at one of the two counterpropagating standing waves. We
develop the theory if Bragg scattering in Sec. 28.2.
It is already clear that we should expect a ringing with exactly the time-dependent
frequency ∆c (t). Fig. 22.13 shows
This simple mathematical model only assumes a resonance with a given width
and a frequency-scanned oscillator. The physical nature of the resonance and the
level splitting are not specified and the formalism should be applicable to a variety of
situations. A ringing is, in fact, observed for RIR scans, if the scanning speed is too
22.3. PHENOMENA RELATED TO CARL 819
0.08
0.06
0.04
ρ12
0.02
0
-0.02
-100 0 100
t (μs)
Figure 22.13: (code) Time-evolution of the atomic coherence as the driving laser frequency
is swept across resonance. The parameters are Γ = 2π × 8 kHz, Ω = 0.2Γ, ∆ = −50Γ..50Γ
and t = −150..150 µs.
fast, in particular for atoms trapped in optical lattices. Ringing can also be generated
in classical harmonic oscillators and laser-driven two-level systems as we will see in
Exc. 28.5.4.1.
(analogous to parametric Raman cooling). Thus, the FEL converts the kinetic energy
of the electron beam into laser radiation. The inversion in a FEL can be interpreted
as a relative displacement of the probability distributions for absorption and emission
of photons in momentum space,
Wabs (∆) = sinc 2 12 (∆ + ε/2) (22.93)
Wabs (∆) = sinc2 12 (∆ − ε/2) ,
where ε = ~k 2 /2me . The gain is a convolution of the difference of the above distri-
bution with the momentum distribution of the electrons [640].
D 0, W 0
Figure 22.15: (code) Flight-of-time measurement of the atomic momentum distribution after
matter wave superradiance.
The phenomenon was explained as follows. Let us imagine a first photon scattered
by an atom into the direction of the long axis of the condensate. This atom will be
accelerated by the photonic recoil in a direction of 135◦ with respect to the direction
of the photon, and it will interfere with the rest of the condensate thus generating
a standing matter wave of oriented in such a way, that the following photons are
scattered into the same direction as the first one via Bragg scattering. This reinforces
the contrast of the matter wave, etc.. We obtain an exponential gain of photons in
the mode defined by the first scattered photon, as well as of the mode receiving the
scattered matter wave. As the path of the gain is longer along the long axis of the
condensate, this mode is favored. That is, the condensate can be considered as a
cavity embracing the solid angle Ωsol .
The Rayleigh scattering rate for a single atom is,
I 3Ωsol
R1 = sin2 θσ(∆a ) , (22.94)
~ω 8π
where θis the angle between the polarization of the incident laser (intensity I) and the
Γ2
direction into which the light is scattered. The cross-section is σ(∆a ) = σ0 4∆2 +2Ω 2 +Γ2 ,
2
λ
where σ0 = 3 2π . Now, for the set of atoms, the superradiant scattering rate is not
only amplified by the number of condensed atoms, N , but also by the number of
atoms, Nr , already being in the mode receiving scattered atoms,
Nr + 1
Rsr = R1 N . (22.95)
2
822 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
22.3.5 Exercises
22.3.5.1 Ex: Good and bad cavity regime
What are the characteristics distinguishing the good from the bad cavity limit.
In this exercise we study ringing in (a) an excited classical harmonic oscillator and
(b) in a laser-driven two-level system.
classical adiab.elim.
|ψi = |zimotion ⊗ |α+ i ⊗ |α− i ⊗ |iielectron → |zi . (22.96)
Since the motion, being the only quantum degree of freedom, is not subject to dissi-
pation, we may use the Schrödinger equation in addition to Heisenberg equations for
the light fields. Then, following the procedure outlined in Sec. 22.1.5, with z being
the only quantum number, we find that the system is completely described by the
dynamics of the expansion coefficients,
d
ı~ hz|ψ(t)i = hz|Ĥ|ψ(t)i (22.97)
dt
α̇± = hz| ~ı [Ĥ, â± ] − κâ± |zi .
22.4. QUANTIZATION OF THE ATOMIC MOTION IN CAVITIES 823
Plugging in the Hamiltonian (22.26), we obtain the Schrödinger equation for the
particle’s motional wavefunction 11 ,
d p2
ı~ hz|ψ(t)i = hz|ψ(t)i + U0 hz|â†+ â− e−2ıkẑ + â+ â†− e2ıkẑ |ψ(t)i , (22.98)
dt 2m
and the Heisenberg equations for the light fields,
α̇± = hz|(−κ + ı∆c − ıU0 )â± + ıU0 e∓2ıkẑ â∓ + η± |zi (22.99)
Z
= (−κ + ı∆c − ıU0 )α± +ıU0 α∓ hz| dz|ψ(t)ihψ(t)|e∓2ıkẑ |zi + η±
Z
= (−κ + ı∆c − ıU0 )α± + ıU0 α∓ dz|ψ(z, t)|2 e∓2ıkz + η± .
In summary,
ı~ d2
†
ψ̇(z) = 2
ψ(z) − ıU0 α+ α− e−2ıkz + α+ α−† e2ıkz ψ(z)
2m dz
Z . (22.100)
α̇± = (−κ + ı∆c − ıU0 )α± + ıU0 α∓ dz|ψ(z)|2 e∓2ıkz + η±
These are the new CARL equations that should be used in cases when the particle is
slower than the recoil velocity, in which its motion must be described as a propagating
matter wave. Note, that under this form, the equations can easily be generalized to
apply to macroscopic wavefunctions such as a Bose-Einstein condensate.
The expectation value of the particle’s position is then calculated via,
We can verify that the expectation value of the particle’s position satisfies the classical
equation of motion,
¨
mz̈ = hψ(t)|mẑ|ψ(t)i = 2ı~kU0 â†+ â− hψ(t)|e−2ıkẑ |ψ(t)i − â+ â†− hψ(t)|e2ıkẑ |ψ(t)i
= 2ı~kU0 â†+ â− e−2ıkhψ(t)|ẑ|ψ(t)i − â+ â†− e2ıkhψ(t)|ẑ|ψ(t)i (22.102)
= 2ı~kU0 â†+ â− e−2ıkz − â+ â†− e2ıkz .
be removed form the Schrödinger equation by a simple unitary transformation. Note that this can
be done even though these terms depend on the field amplitudes α± .
824 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
it lies in the monochromaticity of the driving laser fields and the one-dimensional
geometry of the system, which allow us to write the recoil operator (2.190) or (3.127)
as,
Z
e2ıkẑ = |p + 2~kihp|dp . (22.103)
yields,
d
ı~ hp|ψi = U0 â†+ â− hp|e−2ıkẑ |ψi + U0 â+ â†− hp|e2ıkẑ |ψi (22.106)
dt
d
= ı~ c(p) = U0 [â†+ â− c(p + 2~k) + â+ â†− c(p − 2~k)] .
dt
That is, if the initial momentum distribution is narrow, ∆p 2~k, and if CARL-
induced recoil due to backscattering of photons between counterpropagating modes
is the only force acting on the atoms, the momentum of the atoms can only adopt
discrete values in units of 2~k, as if the atomic velocity were quantized. That is, the
quantization of the field is, in some way, imprinted on the distribution of the atomic
moment, so that we may as well use a discrete notation,
X
|ψi = cν |νi , (22.107)
ν
12
such that ,
d
ı~ cν = U0 â†+ â− cν+1 + U0 â+ â†− cν−1 . (22.108)
dt
Example 111 (Analogy to the Bose-Hubbard Hamiltonian): Interestingly
the CARL Hamiltonian with quantized motion (22.104) has, in momentum
space, a similar shape as the 1D Bose-Hubbard Hamiltonian in position space,
X X
Ĥ = ~ωrec ν 2 + U0 |ν − 1ihν|â†+ â− + |ν − 1ihν|â+ â†−
ν ν
X X X
= ~ωrec ν 2 + U0 â†+ â− †ν−1 Âν + U0 â+ â†− †ν+1 Âν .
ν ν ν
12 Note that, instead of expanding the state |ψi, we could expand the motional wavefunction into
plane waves,
X Z X
hz|ψ̃i = ψ̃(z) = √1 cn (t)e2ınkz normalized as dz|ψ̃(z)|2 = 1 = |cn (t)|2 .
2π
n n
Insertion of this expansion into the quantized CARL equations (22.100) yields the same results.
22.4. QUANTIZATION OF THE ATOMIC MOTION IN CAVITIES 825
Applying the recipe detailed at the beginning of Sec. 22.4, we write down the same
Schrödinger equation as in (22.98) 13 , but now projecting on hν| rather than on hz|
and inserting the expansion (22.107) and,
X X
e2ıkẑ = |ν + 1ihν| and p̂ = ν2~k|νihν| . (22.110)
ν ν
We get,
d X X (ν 0 2~k)2
hν|ı~ cν 00 |ν 00 i = hν| |ν 0 ihν 0 | (22.111)
dt 00 0
2m
ν ν
!
X X
+ U0 0
|ν − 1ihν 0
|â†+ â− + 0
|ν + 1ihν 0
|â+ â†− .
ν0 ν0
hψ|â˙ ± |ψi = hψ|(−κ + ı∆c − ıU0 )â± − ıU0 e∓ı2kẑ â∓ + η± |ψi , (22.112)
we get,
X X
hν 0 |c∗ν â˙ ± cν 00 |ν 00 i = hν 0 |c∗ν 0 [(−κ + ı∆c − ıU0 )â± − ıU0 |ν ∓ 1ihν|â∓ + η± ] cν 00 |ν 00 i .
ν 0 ,ν 00 ν,ν 0 ,ν 00
(22.113)
13 Here again, as done in Sec. 22.4.1, we ignore terms of the Hamiltonian (22.26) which do not
depend on z or p.
826 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
Finally,
∗ ∗
ċν = −4ıωrec ν 2 cν − ıU0 α+ α− cν+1 + α+ α− cν−1
X . (22.114)
α̇± = (−κ + ı∆c − ıU0 )α± − ıU0 α∓ c∗ν∓1 cν + η±
ν
In Exc. 22.4.6.1 and 22.4.6.2 we study the equations (22.114) in the presence of a
constant external force.
that is, skipping the quantum number counting the photons, we will treat the light
fields classically, â± = α± ,
X
p̂ = ν~k|ν~kihν~k| ⊗ I
ν
X
e −ıkẑ
= |ν~k − ~kihν~k| ⊗ I
ν
, (22.117)
σ̂ + = I ⊗ |2ih1|
X
ρ̂ = |ψihψ| = c∗µ,i cν,j |µihν| ⊗ |iihj|
µ,ν,i,j
22.4. QUANTIZATION OF THE ATOMIC MOTION IN CAVITIES 827
where in the last equation we defined cν,j = hp, j|ψ(t)i = hν~k, j|ψ(t)i = hν, j|ψ(t)i,
such that,
ρµ,ν;i,j ≡ hµ, i|ρ̂|ν, ji = c∗µ,i cν,j . (22.118)
Adopting the short notation |νi ≡ |ν~ki we can write the state of the system,
X
|ψ(t)i = cν,1 |ν~k, 1i + cν,2 |ν~k, 2i (22.119)
ν
X
hψ(t)|p̂|ψ(t)i = ~k ν(|cν,1 |2 + |cν,2 |2 )
ν
X
±ıkẑ
hψ(t)|e |ψ(t)i = (c∗ν±1,1 cν,1 + c∗ν±1,2 cν,2 )
ν
X
hψ(t)|â|ψ(t)i = α (|cν,1 |2 + |cν,2 |2 ) = α
ν
X
hψ(t)|σ̂ − |ψ(t)i = c∗ν,1 cν,2
ν
X
hψ(t)|σ̂z |ψ(t)i = (|cν,2 |2 − |cν,1 |2 ) .
ν
To describe the dynamics of the system we could use the Schrödinger equation,
but the Hamiltonian (22.4) does not contain spontaneous emission nor cavity decay.
So, let us employ the equations of motion (22.15), which were derived from a master
equation. The equation of motion for the field yields,
α̇± = hψ(t)|â˙ ± |ψ(t)i = hψ(t)|(−κ + ı∆c )â± − ıgσ̂ − e∓ıkẑ + η± |ψ(t)i (22.120)
X
= (−κ + ı∆c )α± + η± − ıg c∗ν∓1,1 cν,2
ν
X
= (−κ + ı∆c )α± + η± − ıg ρν∓1,ν;1,2 .
ν
˙ 1 ν~k
ẋ = hψ(t)|x̂|ψ(t)i = hψ(t)|p̂|ψ(t)i = , (22.121)
m m
828 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
and
˙
ṗ = hψ(t)|p̂|ψ(t)i (22.122)
= hψ(t)|ıg~kσ̂ − â†+ e−ıkẑ − â†− eıkẑ − ıg~kσ̂ + â+ eıkẑ − â− e−ıkẑ |ψ(t)i
X
∗ ∗ ∗ ∗
= ıg~k α+ cν−1,1 cν,2 − α− cν+1,1 cν,2 − α+ cν+1,1 c∗ν,2 + α− cν−1,1 c∗ν,2
ν
X
∗ ∗
= ıg~k(α+ ρν−1,ν;1,2 − α− ρν+1,ν;1,2 − α+ ρν,ν+1;2,1 − α− ρν,ν−1;2,1 ) .
ν
Analogously to the treatment in Sec. 14.2.1, we will now derive the master equation
without adiabatic elimination of the excited state (in this case called Maxwell-Bloch
equations from the Liouville equation [first line of Eq. (22.8)] using the Hamiltonian
(22.4). The coherent part is,
The incoherent part comprises the spontaneous decay [second line of Eq. (22.8)],
hµ, i|Lcavity−vac,± ρ̂|ν, ji = −κhµ, i|↱ â± ρ̂ − 2â± ρ̂↱ + ρ̂↱ â± |ν, ji = 0 . (22.125)
22.4. QUANTIZATION OF THE ATOMIC MOTION IN CAVITIES 829
Finally, using the definition of the recoil shift ωrec = ~k 2 /2m we get,
(22.126)
We note ρ̂µ,ν;2,1 = ρ̂∗ν,µ;1,2 .
The equations (22.126) form a set a equations to describe the quantized CARL
without adiabatic elimination. And as shown in the derivation of the CARL equations
(22.25), they contain radiation pressure. In Exc. 22.4.6.4 we study the quantized
CARL Maxwell-Bloch equations without adiabatic elimination of the excited state
for a three-level system.
Our starting point is the quantum version of the CARL equations (22.25), where we
neglect spontaneous emission, γ0 = 0. Setting α+ = η+ /κ and η− = 0 we get,
where the index j runs over all atoms. For this case, the total momentum is a constant
of motion for N atoms is,
N
X
[Ĥ, 2~kâ†− â− + p̂j ] = 0 . (22.128)
j=1
where the Hamiltonian is obtained from (22.26) by eliminating the mode α+ and
summing over all atoms,
" #
X p̂2j η+ −2ıkẑj 2ıkẑj †
Ĥ = + U0 e â− + e â− + (U0 − ∆c )â†− â− . (22.129)
j
2m κ
and calculate the expected value of the equations (22.127) regarding the atomic mo-
tion,
dâ− X
= (−κ − ı∆c )â− − ıU0 ηκ+ hψ(zj )|e−2ıkẑj |ψ(zj )i (22.131)
dt j
X
= (−κ + ı∆c − ıU0 )â− − ıU0 ηκ+ c∗j,µ cj,ν j hµ|e−2ıkzj |νij
j,µ,ν
X
= (−κ + ı∆c − ıU0 )â− − ıU0 ηκ+ c∗j,ν cj,ν+1 .
j,ν
We do not use the second equation (22.127), but instead we use the Schrödinger
d|ψ(z i
equation ı~ dt j = Ĥ|ψ(zj )i, which yields,
X dcj,ν X 1 X
ı~ |νij = p2j cj,ν |νij + ~∆c â†− â− cj (ν)|νij (22.132)
ν
dt ν
2m ν
X †
+ ~U0 η (â− e−2ıkzj + â− e2ıkzj )cj,ν |νij .
ν
ċj,µ = −4ıωrec µ2 cj,µ − ı∆c â†− â− cj,µ − ıU0 η[â†− cj,µ+1 + a− cj,µ−1 ]
X , (22.133)
â˙ − = −(κ + ı∆c )â + N U0 η c∗j,ν cj,ν+1
ν
where we took the expectation value of Eq. (22.131). The equations (22.133) can be
used for numerical simulations [598].
22.4. QUANTIZATION OF THE ATOMIC MOTION IN CAVITIES 831
Figure 22.17: Probe light and ’bunching’ when the temperature is raised.
The equations (22.133) allow to calculate the derivatives of the density matrix given
by,
X
%µ,ν ≡ e−ı(µ−ν)∆c t N1 c∗j (µ)cj (ν) , (22.134)
j
yielding,
d%µ,ν X
= e−ı(µ−ν)∆a t N −1 ċ∗j,µ cj,ν + c∗j,µ ċj,ν − ∆c c∗j,µ cj,ν ı(µ − ν) (22.135)
dt j
X
= e−ı(µ−ν)∆a t N −1 [ıωr (µ2 − ν 2 ) − ∆c ı(µ − ν)]c∗j,µ cj,ν +
j
+ ıU0 η ac∗j,µ+1 cj,ν − ac∗j,µ cj,ν−1 + ↠c∗j,µ−1 cj,ν − ↠c∗j,µ cj,ν+1 .
d%µ,ν
= ı(µ − ν)[ωr (µ + ν) − ∆c ]%µ,ν
dt
+ ıU0 η ã(%µ+1,ν − %µ,ν−1 ) + ㆠ(%µ−1,ν − %µ,ν+1 ) . (22.136)
dã X
= −κã − ıN U0 η %ν,ν+1
dt ν
These are the CARL equations for the density matrix. In 22.4.6.3 we show an alterna-
tive derivation obtained by directly inserting the adiabatically simplified Hamiltonian
(22.26) into the Liouville equation (22.8).
P ∗
We note, that ν %ν,ν+1
P is the ’bunching’ and that %µ,ν = %ν,µ . The average
moment is given by hpi = ν ν%ν,ν . In the Figs. 22.18 and 22.19 we show simulations
in the semi-classical regimes ρ 1 in the ’bad-cavity’ limit, κ > 1, and the ’good-
cavity’ limit κ 1.
Here are some movies illustrating the quantum CARL dynamics, simulated using
the momentum state expansion (watch movie) and (watch movie). The following
dynamics were calculated via direct integration of the Schrödinger equation with-
out momentum state expansion: Bloch oscillations (watch movie), CARL dynamics
(watch movie), and joint Bloch and CARL dynamics (watch movie). See also (watch
talk).
832 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
0.02 1 0 0.6
(a) (b) (c) (d)
|α− |2 /|η|2
0.4
p
0.01 0.5 -1
pn
b
0.2
0 0 -2 0
0 50 -3 -2 -1 0 1 2 3 0 50 0 50
t (μs) n t (μs) t (μs)
Δ / ω r = 36
Figure 1
0.3 22.18: (code) Simulation of the CARL equation (22.133) in the superradiant, semi-
classical, 1
’bad-cavity’ regime for κc = 4, ρ = 4, ∆c = 0.
0.2
Im λ
0.5
φ
0.1 0.5
10 1 0 0.6
(a) (b) (c) (d)
0 (e) 0
(f)
|α− |2 /|η|2
-0.5 2 0 2 0.4
-0.1 0 0
p
5 -2
50 0.5 -2
pn
b
0 m -2 0 2
n
t (μs) -1 0.2 δr
0 0 -1.5 0
0 50 100 -3 -2 -1 0 1 2 3 0 50 100 0 50 100
t (μs) n t (μs) t (μs)
Δ / ω r = 2.8571
Figure
10 22.19: (code) Simulation of the CARL equation (22.133) in the 0.15
superradiant, semi-
classical, ’good-cavity’ regime for κ c = 0.02, ρ = 1 ∆c = 0.
4,
5
0.1
0.5 Im λ
φ
0 (e)
0.05
22.4.5
-5 Approximation for a bimodal
0 momentum
(f)
distribution
2
0 2
We
-10 observe that in the quantum regime the momentum -2 -2 distribution of-2 the 0matter
0 0
0 50 100 m 2
wave is tbimodal, n
that is, only two momentum states are simultaneously populated.
(μs) δr
This justifies a simplification of the equations (22.136), assuming that at a given time,
atoms must either be in a specific state |µij or in a superposition of this state is an
adjacent state |µ − 1ij . Hence,
d%µ,µ d%µ−1,µ−1
= −ıU0 η(ã%µ,µ−1 − ㆠ%µ−1,µ ) = − (22.137)
dt dt
d%µ−1,µ
= −ı[(2µ − 1)ωr − ∆c ]%µ−1,µ + ıU0 ηã(%µ,µ − %µ−1,µ−1 )
dt
dã
= −ıN Uµ η%µ−1,µ − κc ã .
dt
Introducing the coherence Sµ = %µ−1,µ and the inversion Wµ = %µ,µ − %µ−1,µ−1 and
postulating the normalization 1 = %µ,µ + %µ−1,µ−1 ,
dWµ
= −2ıU0 η(ãSµ∗ − ㆠSµ ) (22.138)
dt
dSµ
= −ı[(2µ − 1)ωr − ∆c ]Sµ + ıU0 ηñaWµ
dt
dã
= −ıN U0 ηSµ − κc ã .
dt
22.4. QUANTIZATION OF THE ATOMIC MOTION IN CAVITIES 833
We assume that the atoms are initially ’bunched’ in a specific state |µij . Only adjacent
momentum states are coupled,
d%µ+1,µ
= ı[ωr (2µ + 1) − ∆c ]%µ+1,µ + ıU0 ηㆠ(%µ,µ − %µ+1,µ+1 ) (22.139)
dt
d%µ−1,µ
= −ı[ωr (2µ − 1) − ∆c ]%µ−1,µ + ıU0 ηã(%µ,µ − %µ−1,µ−1 )
dt
dã
= −ıN U0 η(%µ,µ+1 + %µ−1,µ ) − κc ã .
dt
Conjugate the upper equation and build the sum and difference, Bµ ≡ %µ,µ+1 +%µ−1,µ
and Dµ ≡ %µ,µ+1 − %µ−1,µ ,
dBµ
= ı(∆c − 2µωr )Bµ − ıωµ Dµ + ıU0 ηã(%µ+1,µ+1 − %µ−1,µ−1 ) (22.140)
dt
dDµ
= ı(∆c − 2µωr )Dµ − ıωr Bµ − 2ıU0 ηã + ıU0 ηã(%µ+1,µ+1 − 2%µ,µ + %µ−1,µ−1 )
dt
dã
= −ıN U0 ηBµ − κc ã .
dt
dBµ
= −ıδµ Bµ − ıωr Dµ (22.141)
dt
dDµ
= −ıωr Bµ − ıδr Dµ − 2ıU0 ηã
dt
dã
= −ıN U0 ηBµ − κc ã .
dt
1 0.6
(a) κ = 0.2ω r κ = 21.4ω r
(b)
ρ = 0.1...6.8 ρ = 0.1...6.8
G/ωr ρ 0.4
G/ωr ρ
0.5
0.2
0 0
-100 -50 0 50 -100 -50 0 50
Δc /ωr Δc /ωr
×106
1 10
(c) κ = 0.03 (d) κ = 0.03
0.1 0.1
0.3 0.3
Δω/ωr
G/ωr ρ
1 1
0.5 3 5 3
0 0
0 2 4 6 0 2 4 6
ρ ρ
The gain bandwidth ∆ωG being much larger than the recoil frequency, the Bragg con-
dition for scattering between different momentum states is approximately fulfilled for
a large number of initial momenta. I.e. although the momentum transfer is quantized,
∆c = nωr , the atoms can be accelerated to high velocities. From Eq. (22.140)(a) we
see |D̄µ /B̄µ | = ρ|λ̄| 1, i.e. %µ,µ+1 ' %µ−1,µ .
In the superradiant regime, κ > 1, of the semiclassical
√ limit the characteristic
equation reduces to λ2 = −ı/κ, i.e. λ = ±(1 − ı)/ 2κ.14 Hence, the gain is,
p
G = 12 ωr ρ 2/κ (22.148)
∆ωG ' κc = ωr ρκ ωr .
1/3
In fact, the relative gain bandwidth is on the order of ∆ωG /ωr ' ρ ∝ nN U02 .
Since recoil can be neglected we can have absorption or emission. The gain results
from the difference bewteen the average rates of both.
Now we turn to the quantump limit, ρ < 1. In the good-cavity regime, κ ' 0,
λ = ρ−1 + 21 (δm − ρ−1 ) − 12 (δm − ρ−1 )2 − 2ρ. Hence, the gain is,
p p
G = 12 ωr ρIm (δm − ρ−1 )2 − 2ρ ' 12 ωr ρ 2ρ (22.149)
3/2
∆ωG = ωr ρ < ωr .
p
In fact, the relative gain bandwidth is on the order of ∆ωG /ωr ' ρ3/2 ∝ nN U02 .
Here recoil plays a role so that we have emission without absorption. Gain re-
sults exclusively from emission. From Eq. ((22.140)a) we see |D̄m /B̄m | = ρ|λ̄| ' 1,
i.e. %m,m+1 %m−1,m .
In the superradiant regime, κ > 1, of the quantum limit, λ = ρ−1 + ρ2 [(δm − ρ−1 ) +
ıκ]−1 . Hence,
ρκ ωr ρ2
G = 12 ωr ρ ' 1
2 (22.150)
(δm − ρ−1 )2 + κ2 κ
∆ωG = κc > ωr ρ .
solution.
836 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
1 (a) 1 (b)
0.5 0.5
G
0 0
-0.5 -0.5
semi-classic semi-classic
quantum quantum
-1 -1
0 1 2 3 0 1 2 3
Figure 22.21: (code) Analytical approximations of the characteristic equation for the various
regimes in the good-cavity limit κ = 0.03 (a) and the superradiant limit κ = 2 (b).
22.4.6 Exercises
22.4.6.1 Ex: Quantized CARL equations in the presence of a constant
external force
a. Generalize the quantized CARL equations (22.100), respectively, (22.114) for the
presence of a constant external force.
b. Now, we consider a unidirectionally pumped ring cavity, with the pump laser locked
to a cavity mode, in the presence of an external periodic potential. Show that this
system is equivalent to CARL in a ring cavity pumped from both sides.
Derive the directly from the Liouville equation (22.8) using the Hamiltonian (22.26)
in adiabatic elimination.
Derive the Maxwell-Bloch equations for a three-level system coupled to a ring cavity
without adiabatic elimination, but with quantized motion.
d −ı
ċn+ ,n− ,i = hr, α+ , α− , i|ψ(t)i = hr, α+ , α− , i| Ĥ|ψ(t)i (22.151)
dt ~
α̇± = hψ(t)|ȧ± |ψ(t)i = hψ(t)| ~ı [Ĥ, â± ] − κâ± |ψ(t)i .
Example 112 (CQED by reducing the ring cavity mode volume): Tech-
nically a ring cavity design with (w, L, F ) = (30 µm, 4 cm, 200000) is feasible.
This is enough to get below the critical atom number, but this is not sufficient
to get into the CQED regime. Reducing Γ seems unavoidable.
The isotope 88 Sr posses a narrow transition which can be used for optical cooling.
The following table compares the various systems, i.e. a CQED example taken
from Rempe, the macroscopic ring cavity with rubidium used in our Tübingen
CARL experiments and a cavity tuned close to the strontium intercombination
line.
838 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
Figure 22.22: (a) Bragg scattering at a 1D optical lattice. (b) Same as in (a), but now the
optical lattice is generated by the mode of an optical ring cavity.
22.5. QUANTIZED LIGHT INTERACTING WITH ATOMS MOVING IN CAVITIES839
Note that the master equation using the Lindbladt operator traces over the reser-
voir. Hence, the master equation only treats the cavity modes, but does not al-
low predictions on the quantum behavior of outcoupled fields. In order to describe
e.g. quantum correlations in output field, one needs an input-output theory [152, 269].
In the case of classical motion quantum light fields,
In Exc. 22.5.7.1 we show how to cast the Hamiltonian into a matrix form using an
appropriate basis already used in the discussion of the beam splitter in the photon
representation in Sec. 15.6.1.
In Exc. 22.5.7.2(a) we derive the equations of motion for the components cn+ ,n− of
the state vector from the Schrödinger equation cavity decay. In Exc. 22.5.7.3(a) we de-
rive the equations of motion for the components ρm+ ,n+ ;m− n− = hm+ , m− |ρ̂|n+ , n− i
of the density operator. Simulations performed based on these equations of motion
are shown in Fig. 22.23).
The simulations reveal a number of interesting facts:
The field amplitudes |α± |2 execute oscillations due to CARL coupling, but with
preserved Poissonian shape of the photon number distributions.
When simulations are done with initial Fock states, they eventually relax to a
Glauber state.
The mean photon numbers and the atomic coordinates evolve in a continuous
way.
0.4 7 4
(a) (b) (c)
0.3 6
2
p/h̄k
|α|2
0.2 5
pn
0
0.1 4
0 3 -2
0 10 20 0 50 0 50
photon# time time
Figure 22.23: (code) Time-evolution of CARL with one classical atom and two quantized
field modes. (a) Photon number distributions of two modes after some evolution time t. (b)
Time-evolution of the mean photon number; at time t = 10, the atom is suddenly displaced.
(c) Time-evolution of the (classical) atomic trajectory.
15
of type (15.216) ,
This means, that we expect (for a fixed location z of the atom) an oscillatory behavior
∗
of the field amplitudes |α± |2 and also of the phase (encoded in α+ α− ), but no modi-
fication of the coherent photon statistics. On the contrary, as shown in Sec. 15.6.1, a
beam splitter tends to transform sub-Poissonian states into Poissonian ones. In fact,
CARL is nothing else than a movable beam splitter.
Non-linearity may come into play, if photonic recoil is included, i.e. if the atomic
motion itself becomes a degree of freedom. This is the next step of our problem, that
needs to be solved.
Ĥint ' U0 (α∗+ e−2ıkz â− + α+ e2ıkz â†− ) ≡ ı~β ∗ â− − ı~βâ†− ,
ıU0
since defining β ≡ ~
α+ e2ıkz , we find,
†
β ∗ tâ− −βtâ−
B(βt) = e−ıĤint t/~ = e .
22.5. QUANTIZED LIGHT INTERACTING WITH ATOMS MOVING IN CAVITIES841
photonic recoil. The question is now, how to conceal this fact with the observation of
discrete momentum sidemodes [660]. We have seen earlier, that the quantization of
the motion is transferred from the quantized photon fields via the operator (22.103)
to the atom: An initially resting atom can only adopt motional states with momenta
equal to a multiple of 2~k.
We will now take the quantization of the atomic momentum for granted and study
quantized light modes coupled by recoiling atoms, we extend our Hilbert space like,
adiab.elim.
|ψi = |νimotion ⊗ |iielectron ⊗ |ni+ ⊗ |ni− → |ν, n+ , n− i , (22.159)
16
Expanding the motion and fields ,
X
|ψi = cν,n+ ,n− |ν, n+ , n− i , (22.160)
ν,n+ ,n−
The equations of motions derived in Exc. 22.5.7.2 from the Schrödinger equation are
now generalized to,
increases a lot, that is like dim ν · dim n+ · dim n− , where ν, n± are the numbers of states considered.
842 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
However, the mere inclusion of a third degree of freedom coupled to the other two
degrees in the same linear fashion will modify the behavior of the system, which will
continue to oscillate as it did before, now just involving the atomic motion in this
dynamics. This behavior can only change, when we additionally consider the force
equation,
p̂˙ = 2~kU0 (â†+ â− e−2ıkz − â+ â†− e2ıkz ) , (22.165)
which will lead to feedback. Inserting the expansion (22.160),
p p
ċν,n+ ,n− = 2~kU0 cν,n+ ,n− n+ (n− + 1) − cν,n+ ,n− n+ (n− + 1) . (22.166)
Note that, in contrast to CARL equations, the fully quantized equations (22.162)
are linear. That is, it is the ’classization’ 17 of the degrees of freedom which intro-
duces the non-linearities, which are typical for CARL, e.g. the feedback introduced by
classical light fields. The process of ’classization’ corresponds to tracing over degrees
of freedom, which we want to treat classically and thus remove from the Hamiltonian.
and similarly for α− . The probability distribution for momentum states is given by,
X
Pν (t) = |cν,n+ ,n− (t)|2 , (22.169)
n+ ,n−
hp̂2 i X (ν~k)2
Ekin (t) = = |cν,n+ ,n− (t)|2 (22.171)
2m ν,n+ ,n−
2m
X
= ν 2 ~ωrec |cν,n+ ,n− (t)|2 ,
ν,n+ ,n−
Epot (t) = ~U0 he−2ıkẑ â†+ â− + e2ıkẑ â+ â†− i (22.172)
X p
= ~U0 c∗ν+1,n+ +1,n− −1 cν,n+ ,n− (n+ + 1)n−
ν,n+ ,n−
p
+c∗ν−1,n+ −1,n− +1 cν,n+ ,n− n+ (n− + 1) .
22.5.4.2 Master equation for CQED with atomic recoil in the adiabatic
approximation
Using the Hamiltonian (22.26) or (22.98), the expansion of the recoil operator (22.103),
the expansion of the photon field operators (22.154), and the matrix representation
of the density operator (14.16), that is,
X
ρ̂ = |µ, m+ , m− iρµ,ν;m+ ,n+ ;m− ,n− hν, n+ , n− | , (22.176)
µ,ν;m+ ,n+ ;m− ,n−
2
p̂
hµ, m+ , m− |[ 2m , ρ̂]|ν, n+ , n− i = ωrec (µ2 − ν 2 )ρµ,ν;m+ ,n+ ;m− ,n− (22.178)
hµ, m+ , m− |[(U0 − ∆c )↱ â± , ρ̂]|ν, n+ , n− i = (U0 − ∆c )n± ρµ,ν;m+ ,n+ ;m− ,n−
hµ, m+ , m− |[−ıη+ (â+ − â†+ ), ρ̂]|ν, n+ , n− i
√ p
= −ıη+ m+ ρµ,ν;m+ −1,n+ ;m− ,n− − m+ + 1ρµ,ν;m+ +1,n+ ;m− ,n−
√ p
+ n+ ρµ,ν;m+ ,n+ −1;m− ,n− − n+ + 1ρµ,ν;m+ ,n+ +1;m− ,n−
0.4 10 15
(a) (b) (c)
pn+ , pn− 10
p̂/h̄k
|α|2
0.2 5
5
0 0 0
0 5 10 15 0 50 0 50
#photon n+ , n− time time
1 20
(d) (e)
0 0
0 1 2 3 4 5 6 0 50
#momentum N time
Figure 22.24: (code) Time-evolution of CARL with one classical atom and two quantized
field modes. (a) Photon number distributions of two modes after some evolution time t. (b)
Time-evolution of the mean photon number; at time t = 10, the atom is suddenly displaced.
(c) Time-evolution of the (classical) atomic trajectory.
22.5.5.1 Kicking
Let us first describe the kick as an incoherent disruptive one time event transforming
operators, the Hamiltonian and the system’s state like,
and in particular,
The Schrödinger equation tells us the system’s evolution after the kick,
d p2
ı~ hp|ψi = hp|ψi + U0 hp|e−2ıkẑ |ψiâ†+ â− + hp|e2ıkẑ |ψiâ+ â†− . (22.182)
dt 2m
We remind that this equations contains CARL feedback via the simultaneous presence
of ẑ and p̂. If we want to disregard the CARL force (assuming, for example, that the
motion is totally imposed by an external force, as done in Sec. 20.2.5 in order to focus
on the behavior of the light fields), we must not project the Schrödinger equation on
hp| but treat the motional degree of freedom as classical.
22.5.5.2 Forcing
Let us now describe the kick as a force F (t) = mgθ(t) being switched on at a given
time, but being constant afterward,
p̂2
Ĥ = + U0 e−2ıkẑ â†+ â− + e2ıkẑ â+ â†− + mgẑ . (22.183)
2m
846 CHAPTER 22. ATOMIC MOTION IN OPTICAL CAVITIES
We note that with (2.153) the Hamiltonian transformed into the accelerated frame
reads,
†
d ımgẑt/~
Ĥ −→ e−ımgẑt/~ Ĥeımgẑt/~ + ı~ e eımgẑt/~ = Ĥ(ẑ, p̂ − mgt) − mgẑ
dt
(p̂ − mgt)2
= + U0 e−2ıkẑ â†+ â− + e2ıkẑ â+ â†− . (22.184)
2m
The Schrödinger equation for the transformed wavefunction |ψi = |ψ̃ieımgẑt/~ , which
tells us the system’s evolution during the force,
d (p − mgt)2
ı~ hp|ψ̃i = hp|ψ̃i + U0 hp|e−2ıkẑ |ψ̃iâ†+ â− + hp|e2ıkẑ |ψ̃iâ+ â†− ,
dt 2m
(22.185)
has a similar shape to Eq. (22.182).
22.5.5.3 Vibrating
Another option might be to additionally confine the atom in a harmonic potential
[762],
p̂2 m
Ĥ = + U0 e−2ıkẑ â†+ â− + e2ıkẑ â+ â†− + ω 2 ẑ 2 (22.186)
2m h 2 i
= ~ω(†  + 21 ) + U0 D† (α)â†+ â− + D(α)â+ â†− ,
†
−α∗ Â 2ıka
where D(α) ≡ eα with α ≡ √ ho .
2
22.5.7 Exercises
22.5.7.1 Ex: Analogy between CARL and two-atom Dicke states
a. Write the Hamiltonian (22.156) in matrix form using the basis,
ψk = {|0, 0i, |0, 1i, |1, 0i, |0, 2i, |1, 1i, |2, 0i, ...} .
b. Now restrict to the finite number of states ψk = {|0, 0i, |0, 1i, |1, 0i, |1, 1i} and
discuss the analogy between CARL and two-atom Dicke states.
S. Barbay et al., Pump-probe spectroscopy of the sodium D line and the question of
recoil-induced gratings in hot vapors [DOI]
M. Vengalattore et al., Optical bistability at low light level due to collective atomic
recoil [DOI]
V. Vuletic et al., Three-dimensional cavity Doppler cooling and cavity sideband cool-
ing by coherent scattering [DOI]
R. Culver et al., Collective strong coupling of cold potassium atoms in a ring cavity
[DOI]
S. Bux et al., Cavity-controlled matter wave superradiance at the recoil limit [DOI]
22.6.1 on CARL
R. Bonifacio et al., The self-consistent pendulum picture of the free electron laser
revised [DOI]
Statistical and
thermodynamic approaches to
CARL
The CARL phenomenon introduced in the previous chapter raises a variety of ques-
tions, such as: How does it compare to an ordinary laser? Is there a phase transition?
Of what kind would be this transition (in the Ehrenfest or Landau classification
scheme)? What are the coherence properties (measured by correlation functions)?
How do these properties depend on the random motion (temperature) of the atoms?
These issues will be addressed in this chapter and in the later chapter within the
models of Langevin, Fokker-Planck, Vlasov, and Kuramoto.
851
852 CHAPTER 23. STATISTICAL AND THERMODYNAMIC APPROACHES TO CARL
xn , pn
ap a
Optical molasses obviously are subject to a cooling limit coming from the ran-
dom scattering of photons. As a consequence, atoms follow a random walk in the
momentum space, which leads to the diffusion and heating of atoms and impedes the
bunching of atoms. It also turns out that a minimal grouping is required to initial-
ize CARL. Therefore, there is a threshold behavior as a function of the equilibrium
temperature of the molasses,
P
α̇− = −κα− − ıU0 α+ m e2ıkzm
. (23.3)
k v̇n = 4ıωrec U0 α+ (α− e−2ıkzn − α−
∗ 2ıkzn
e ) − γf rc kvn + ξn (t)
The equation corresponds to a Langevin equation, where the stochastic term ξn (t) de-
scribes white noise. We can simulate this equation by a Runge-Kutta method, where
the atoms are continually exposed to random momentum changes. For N atoms we
need to solve 2N + 2 Langevin equations to describe the dynamics of all degrees of
freedom. The Langevin equations are associated to so-called Fokker-Planck equations
[722, 600, 378, 721]. These describe the temporal evolution of the atomic density
along the optical x-axis. With these equations we replace the 2N trajectories of indi-
vidual particles by a one-dimensional field P (x, t). The Vlasov equation represents a
different approach: Here we assume that the equilibrium between cooling and heating
is achieved by a continuous thermalization process described by a single rate γth 1 .
friction force to the atoms [326, 266]. This implies the existence of diffusion and a finite equilibrium
temperature even at if the atoms are initially at T = 0.
23.1. THE LANGEVIN MODEL 853
where s = I/Is is the saturation parameter. This formula estimates the maximum
friction force, when the lasers are tuned close to an atomic resonance 2 .
A more fundamental problem is the interdependence of the molasses friction and
the CARL. In fact, because the dipole potential influences the detuning of the molasses
beams by light-shifting the D2 line, the γth coefficient depends on ∆a and η+ (the
reverse field |α− | may be neglected). The threshold equations must then be solved
in a self consistent way. It might however be possible to determine γth only slightly
(thresh)
above threshold, where the modification is small, γth ≈ γth .
to cooling phenomena called ’polarization gradient cooling’, which can cause much higher friction
coefficients.
We also note that, when the molasses is applied to atoms confined to a potential, the atomic levels
can be displaced (e.g., by light-shift or the Zeeman effect). This causes an inhomogeneous effective
detuning of the laser beams generating the molasses.
854 CHAPTER 23. STATISTICAL AND THERMODYNAMIC APPROACHES TO CARL
as follows,
Now, we apply this method to the CARL subject to an optical molasses. Making the
adiabatic approximation θ̈n = 0, the starting point is,
X
α̇− = −κα− − ıU0 α+ eıθm ≡ B(α− , θn )
m
.
8iωrec U0 α+ ξn (t) ξn (t)
θ̇n = (α− e−ıθn − α−
∗ ıθn
e )+ ≡ F (α− , θn ) + γf rc
γf rc γf rc
(23.13)
In order to apply the Runge-Kutta method (23.11), we identify the variables and
functions,
α− (t) B(α− , θn ) 0
x(t) ≡ , f (x) ≡ , g(t) ≡ , (23.14)
θn (t) F (α− , θn ) ξn (t)/γf rc
such that,
α̃− α− (0) B(α− (0), θn (0)) 0
x̃ = = + dt + √ (23.15)
θ̃n θn (0) F (α− (0), θn (0)) ζn 2DT dt
and
α− (dt)
x(dt) = (23.16)
θn (dt)
α− (0) dt B(α− (0), θn (0)) B(α̃− , θ̃n ) 0
= + + + √ .
θn (0) 2 F (α− (0), θn (0)) F (α̃− , θ̃n ) ζn 2DT dt
The Langevin equation can be used to simulate the temporal evolution of CARL.
Fig. 23.2 shows a simulation of the frequency and amplitude of the CARL based on
prescription (23.11).
iθm
3.6
|α+ + α− |2
me
|α− |2
5 3.4 0.5
3.2
b=
3
0 0
0 2 4 0 2 4 0 2 4
t (μs) t (μs) t (μs)
×109
φ̇ (MHz)
3 (e) N = 100000
(d)
-3.2 κ/2π = 550 kHz
2 ωrec /2π = 4.5 kHz
-3.3
p/h̄k
η = 200 ns−1
1 -3.4 U0 ≡ g12 /Δa = (2π) -0.5 Hz
,
-3.5 N U0 /κ = -0.091
γf rc = 1100000 s−1
pcm
0
0 2 4 0 5 10 Tmol = 1000 μK
t (μs) kr
Figure 23.2: (code) Time evolution with phase transition of the viscous CARL in the adi-
abatic approximation. Shown are (a) the number of photons in the probe mode, (b) the
beat signal, (c) the bunching, (d) the phase of the standing wave and the position of the
center-of-mass of the cloud, and (e) the dipole potential calculated from (22.21) with the
atomic distribution.
such that,
0
α̃− α− (0) B (α− (0), θn (0))
x̃ = θ̃n ≡ θn (0) + dt Vn (0) +
q 0
Ṽn Vn (0) F (α− (0), θn (0), Vn (0)) ζn 2γf2rc DT dt
(23.19)
856 CHAPTER 23. STATISTICAL AND THERMODYNAMIC APPROACHES TO CARL
and
α− (dt) α− (0) B (α− (0), θn (0))
dt
x(dt) = θn (dt) = θn (0) + Vn (0) (23.20)
2
Vn (dt) Vn (0) F (α− (0), θn (0), Vn (0))
0
B(α̃− , θ̃n )
+ Ṽn + q 0
.
F (α̃− , θ̃n , Ṽn ) ζn 2γf2rc DT dt
iθm
3.6
|α+ + α− |2
10
me
|α− |2
3.4 0.5
5 3.2
b=
3
0 0
0 2 4 0 2 4 0 2 4
t (μs) t (μs) t (μs)
×106
φ̇ (MHz)
3 5 N = 100000
(d)
κ/2π = 550 kHz
2 ωrec /2π = 4.5 kHz
2kr
0 η = 200 ns−1
1 U0 ≡ g12 /Δa = (2π) -0.5 Hz
,
N U0 /κ = -0.091
γf rc = 1100000 s−1
pcm
0 -5
0 2 4 0 2 4 6 Tmol = 1000 μK
t (μs) p/h̄k
Figure 23.3: (code) Time evolution of the complete dynamics of the viscous CARL with
phase transition. Same parameters as in Fig. 23.2. Shown are (a) the number of photons in
the probe mode, (b) the beat signal, (c) the bunching, (d) the phase of the standing wave
and the position of the center-of-mass of the cloud, and (e) the dipole potential calculated
from (22.21) with the atomic distribution.
23.1.3 Exercises
23.1.3.1 Ex: Viscous CARL
Langevin simulations
23.2. THE FOKKER-PLANCK AND THE VLASOV MODEL 857
dQ
+ Q∇ · v = D∇2 Q , (23.21)
dt
d ∂
represents a generalization of the continuity equation, since with dt ≡ ∂t + v · ∇, we
obtain,
∂Q
= −∇ · (vQ) + D∇2 Q . (23.22)
∂t
If ρ is a local density, it the continuity equation reads,
dρ
+ ρ∇ · v = 0 . (23.23)
dt
Knowing,
d ∂
≡ +v·∇ , (23.24)
dt ∂t
we obtain,
∂ρ
= −∇ · (ρv) . (23.25)
∂t
The Fokker-Planck equation is just a generalization to include a diffusion process,
dρ ∂2ρ
+ ρ∇v = Dx 2 , (23.26)
dt ∂x
or
∂ρ ∂2ρ
= −∇ · (ρv) + Dx 2 . (23.27)
∂t ∂x
α̇− = −κα− − iN U0 α+ b
hk 2 v 2 i
Dθ = . (23.29)
γf rc
We will derive the results (23.32) and (23.33) in Exc. 23.2.5.1. Also,
d Im α−
kv = arctan , (23.34)
dt Re α−
and,
Z 2π X Z 2π
hθ̇i ≡ Q̇(θ, t)dθ = Q̇ν (t) eiνθ dθ (23.35)
0 ν 0
X 8ωrec U0 α+ 16ωrec U0 α+
∗
= ν α− Qν+1 α− Qν−1 δν0 = Re (α− Q∗1 ) .
ν
γf rc γf rc
for N U0 > κ, the CARL becomes unstable again: It will unlock from the self-determined frequency
and start to oscillate strongly. However, this effect is not described by the equations used for the
simulation, which suppose an adiabatic elimination of the inertia.
23.2. THE FOKKER-PLANCK AND THE VLASOV MODEL 859
iθm
|α+ + α− |2
3.6
me
|α− |2
1 3.4 0.5
3.2
b=
3
0 0
0 2 4 0 2 4 0 2 4
t (μs) t (μs) t (μs)
p (MHz)
3 (e) 1 N = 100000
(d)
κ/2π = 550 kHz
2 ωrec /2π = 4.8 kHz
0.5 η = 200 ns−1
P
γf rc = 1200000 s−1
0 0
kv
0 2 4 0 5 10 Tmol = 1000 μK
t (μs) kr
Figure 23.4: (code) Temporal evolution of the complete dynamics with phase transition
of the viscous CARL. The same parameters as in Fig. 23.2. Shown are (a) the number
of photons in the probe mode, (b) the beat signal, (c) the bunching, (d) the phase of the
standing wave and the position of the center-of-mass of the cloud, and (e) the atomic density
distribution.
(a) 15 ×10
6
(b) 300 (c) 1 (d) 300
p (kHz)
10 200 200
kHz
|α− |2
0.5
b
kv
5 100 100
0 0 0 0
0 5 0 5 0 5 0 5
P (W) P (W) P (W) P (W)
Figure 23.5: (code) Bistability of CARL near the threshold when the power is ramped at
different paces.
tionary distribution Q0 as time goes on. Following Bonifacio et al. [97] and Javaloyes
[379, 380, 381], we replace the friction term by an additional (Vlasov) equation:
θ̇ = ωrec V
. (23.38)
V̇ = iU0 α+ (α− e−iθ − α−
∗ iθ
e )
0 = ∂t Q + θ̇∂θ Q + %̇∂% Q + γf rc (Q − Q0 )
where iλ ≡ λg + iλω into the field equation (23.38)a, the first order expansion of the
Vlasov equation (23.41) (only retaining co-rotating terms) and into the expression for
23.2. THE FOKKER-PLANCK AND THE VLASOV MODEL 861
bunching (23.39). This ansatz accounts for the fact that, in steady state, we expect a
fixed CARL frequency ν and a spatially modulated density distribution of the atoms.
The set of equations becomes,
κ + N U02 η̃+
2
2πRe Γ(σ, γ, iλω ) = 0 (23.46)
ν+ N U02 η̃+
2
2πIm Γ(σ, γ, iλω ) = 0 .
Now the condition λω Re Γ(σ, γth , iλω ) = κ Im Γ(σ, γth , iλω ) leads to,
Z ∞ 2
d% %e−σ% (κλω − 2κε% + λω γth )
2 =0. (23.47)
−∞ (λω − 2ε%)2 + γth
The values λω where this integral is zero are inserted into one of the equations (23.46).
Finally,
2 −1
η̃+ = 2
, (23.48)
N Ũ0 2πRe κΓ(σ, γth , iλω )
is the expression for the pump power threshold as a function of temperature σ, friction
γth , cavity damping κ and coupling constant U0 . The intracavity CARL power is,
2
P+ = ~ωδ η̃+ . (23.49)
λω is the associated CARL frequency, i.e. the frequency difference between probe and
pump.
862 CHAPTER 23. STATISTICAL AND THERMODYNAMIC APPROACHES TO CARL
7
x 10
2 1 300
250
0.8
1.5
200
0.6
kv (kHz)
| α- | 2 1 150
b
0.4
100
0.5
0.2
50
0 0 0
0 5 10 0 5 10 0 5 10
P (W) P (W) P (W)
Figure 23.6: (code) Intracavity threshold power as a function of temperature and friction
coefficient. The coupling strength is set to Ũ0 = −10−7 , the atom number is N = 106 .
The final equations show that the threshold power drops as N U02 increases. To
see the dependencies of the threshold power on T and γth , we have to evaluation of
the integral (23.47). Fig. 23.6 shows a numerical evaluation of Eqs. (23.47),(23.48)
and (23.49) for finite T and γth .
Apparently, the threshold pump power drops with vanishing friction and with
low temperatures. For typical experimental situations, T ≈ 100 µK and γth ≈ 10κ,
we expect threshold powers on the order of about P+ = 1 W, corresponding to
(out)
P+ = 4 µW leaking out of the cavity.
23.2.5 Exercises
23.2.5.1 Ex: Fokker-Planck equations
Derive the equations (23.32) and (23.33) from (23.30) and (23.28) applying the ex-
pansion (23.31).
the perturbation. There are no delay times effects and no spatial constraints due to
a finite size of the individual oscillators.
The viscous CARL system is representative for of the vast class of Kuramoto
systems [691] introduced in Sec. ??. This can be seen most appropriately by rewriting
the CARL equations in terms of phase and amplitude of individual atoms. Another
approach is via the Fokker-Planck equation, for which there already exists a Kuramoto
equivalence.
Why is it interesting to investigate yet another coupled oscillator system? Our
CARL is fully classical, although we deal with microscopic particle, such as atoms and
photons. However, our system bears the possibility of being transferred to quantum
situations, and thus to study the coupling of large ensembles of quantum oscillators.
Furthermore, the coupling mechanism is well understood and controlable by experi-
ment. Because the coupling goes through the standing wave fraction, it depends on
the atom number and the coupling strength independently. Furthermore, the tunable
friction force (temperature) corresponds to a variable width of the distribution of the
oscillator frequencies, which we can manipulate in-situ and on-line.
Another realization of the Kuramoto model would be atoms in a ring cavity stand-
ing wave. Cold trapped atoms have the same oscillation frequencies. And thus do
not need to synchronize. For hot atoms, however, the oscillation period depends on
their kinetic energy.
The equation for the phases of the atoms is similar to the Kuramoto model [381],
KX
θ̇n = sin(θn − θm ) − ξn (t) . (23.50)
N m
where the quantities with numeric suffixes are real, such that,
|α+ + α− |2 = α12 + α22 + 2α1 α2 cos φ . (23.52)
By this ansatz we assume that a pumping laser is locked to the mode α+ of the ring
cavity, such that the phases of the fields α+ and η+ are equal and, without loss of
generality, zero. On the other hand, the phase φ of the probe mode is a dynamic
variable. Inserting the ansatz into the equations (??) we obtain,
X
α̇1 = −(κ + iN U0 − i∆c )α1 − iU0 e−iθn +iφ α2 + η1 , (23.53)
n
X
α̇2 + iφ̇α2 = −(κ + iN U0 − i∆c )α2 − iU0 eiθn −iφ α1 + η2 ei∆−iφ ,
n
ξn κK
θ̇n ≈ + |b| + K|b| sin(ψ − θn ) . (23.66)
γf r ωca
Oscillators with ωn ≤ K|b| are locked. For a reasonable ’bunching’ this is satisfied in
the ’good cavity’ limit.
16ωrec N U02 α+
2
1 X
θ̈n = −γf r θ̇n + ξn + 2 2
[κ cos(θm − θn ) + ω sin(θm − θn )] (23.67)
ω +κ N m
κ
= −γf r θ̇n + ξn + K|b| cos(ψ − θn ) + K|b| sin(ψ − θn ) .
ω
especially, Z 2π
1
ρ(θ, t)dθ = 1 =⇒ ρ0 = 2π , (23.72)
0
En ≡ eıθn , (23.74)
ıξn 8ωrec U0 α+ ∗
Ėn = En − (α− − α− En2 ) . (23.75)
γf r γf r
23.3.4 Exercises
23.3.4.1 Ex: Kuramoto simulations
Implement the Kuramoto model with pendulum clocks.
Note that numerical simulations revealed that in certain regimes the CARL cor-
responds to a first-order phase transition: At high temperatures the probe to pump
power diagram shows bistability [379] and [599]DOI .
The correspondence between CARL and a common laser is illustrated by the table
below. A number of questions arise from the analogy:
The CARL being a laser without inversion, what is the equilibrium parameter? Is it
the temperature of the atomic cloud? How to calculate the density of states?
ferromagnet laser
control parameter external magnetic field H pump intensity S
equilibrium parameter temperature T population inversion σ
magnetization electric field
( (
order parameter 0 T > Tc 0 σ > σc
hM i = q
Tc −T
hEi = q
σc −σ
c T
T < Tc c σ
σ < σc
P -representation
probability density P (M ) ∝ e−F (M )/kB T
P (x, y) ∝ e−G(x,y)/Kσ
thermodyn. free energy F = F (T, H) G(x, y)
)
heat capacity C(T ) = ∂E(T
∂T
?
dBn
= ın(aBn−1 + a∗ Bn+1 ) − n2 DBn (23.78)
dτ
da
= B1 − κa ,
dτ
23.4. THERMODYNAMICS OF THE CARL PROCESS 869
d2 a da
+ (κ + D) + (κD − i)a = 0 (23.80)
dτ 2 dτ
d2 B1 dB1
+ (κ + D) + (κD − i)B1 = 0
dτ 2 dτ
The determinant is,
−D − λ i
det = (D + λ)(κ + λ) − i = 0 (23.81)
1 −κ − λ
dA dA |eλτ A|2
eλτ 2λ + eλτ (κ + D) =− λeλτ A + κeλτ A (23.86)
dτ dτ 2D
da λ+κ
= λa − |a|2 a (23.87)
dτ 2D(2λ + κ + D)
= λa − C|a|2 a
or,
d|a|2
= 2|a|2 Re λ − 2|a|4 Re ReC (23.88)
dτ
in steady-state,
0 = λa − C|a|2 a . (23.89)
0.8
0.6
intensity
0.4
0.2
0
0 5 10
time
A similar treatment with a(τ ) = A(τ )eλτ and B1 (τ ) = β(τ )eλτ and d2 β/dτ 2 = 0 and
|A|2 dβ/dτ = 0 and βd|A|2 /dτ = 0 results in,
dB1 2λ + λ∗ + κ
= λB1 − |a|2 B1 (23.90)
dτ 2D(2λ + κ + D)
= λB1 − C̃|a|2 B1 ,
i.e. the instability comes from the field and it drives the bunching.
23.4. THERMODYNAMICS OF THE CARL PROCESS 871
Apparently, noise mainly affects the field’s phase and not its amplitude. Note that the
first-order coherence is just the normalized autocorrelation g (1) (τ ) ≡ Rα (τ )/Rα (0),
2 2
σ%2 /2
g (1) (τ ) = eıωca τ e−ωca τ (23.98)
iωca τ −Dθ |τ |/2
=e e .
The power spectral density is a Lorentzian,
Z ∞
Sα (ω) = |α0 |2 Rα (τ )e−ıωτ dτ (23.99)
−∞
|α0 |2 Dθ2
= .
(ω − ωca )2 + Dθ2
The CARL emission bandwidth is thus 21 Dθ . The above results show that the CARL
laser bandwidth increases linearly with temperature and reduces with the friction
force. In our experiment we have γf r = 5κ and σ = 10κ, so that Dθ = σ 2 /γf r = 20κ.
The CARL bandwidth is extremely large as compared to the CARL frequency ωca =
5κ. The reason is that we only considered a single atom. The impact of several atoms
will partially compensate and reduce the linewidth.
874 CHAPTER 23. STATISTICAL AND THERMODYNAMIC APPROACHES TO CARL
The light mode α− appears to be a superposition of coherent waves having the same
frequencies, αm (t) = α0 eiωca t+iϕm (t) , but interrupted by random phase jumps. We
may thus try an analogous argumentation as for pressure broadening,
Z X X
∗
hα− (t)α− (t + τ )i = |α0 |2 e−ıωca t−ıϕn (t) eıωca t+ıωca τ +ıϕn (t+τ ) dt
n m
XZ
= |α0 |2 eıωca τ eıϕm (t+τ )−ıϕn (t) dt = N hαn∗ (t)αm (t + τ )iδnm .
n,m
Cross-terms vanish, the autocorrelation function is just the sum of the single-atom
components. However, while we may view the noise impact of the atoms as coming
from a single atom, the rate of the phase jumps is much higher, than for a single
atom. For one atom the probability density for encountering a coherent interval of
length τ is given by p1 (τ )dτ = 21 Dθ e−Dθ |τ |/2 dτ [456]. Thus,
Z ∞
∗ 2 iωca τ −Dθ |τ |/2
hα1 (t)α1 (t + τ )i = |α0 | e e 2 ıωca τ
= |α0 | e p1 (τ 0 )dτ 0 . (23.101)
τ
Fig. 23.8 shows the dependencies of the correlation functions on N and T . For
N = 106 atoms the CARL bandwidth becomes βN = 0.01κ, which is already well
below any mechanical noise. The linewidth reduction with increasing atom numbers
is quantitatively supported by numerical simulations (see below).
23.5. COHERENT PROPERTIES OF CARL 875
1 1 2 2
(a) (c) (b) (d)
1.5 1.5
(1)
(2)
0.5 0.5 1 1
g
g
0.5 0.5
0 0 0 0
-2 0 2 -2 0 2 -2 0 2 -2 0 2
κτ κτ
Figure 23.8: (code) First and second-order correlation functions. The solid and dotted lines
in (a) and (b) are calculated for N = 1 and N = 100, resp.. The solid and dotted lines in
(c) and (d) are calculated for T = 100 µK and T = 300 µK, resp..
X Xk
ık k/2
Rα (τ ) = |α0 |2 eıωca τ hϕn (t + τ )2 ij/2 hϕm (t)2 ik/2−j/2
n,m,k 2k/2 (k/2)! j=0 j/2
(23.110)
X k/2
1 1 1
= |α0 |2 eıωca τ − hϕn (t + τ )2 i − hϕm (t)2 i
n,m,k (k/2)! 2 2
X X
− 21 hϕn (t+τ )2 i − 21 hϕm (t)2 i 1 2 1 2
2 ıωca τ
= |α0 | e e e = |α0 |2 N 2 eıωca τ e− 2 hϕn (t+τ ) i e− 2 hϕn (t) i .
n m
so that,
∗
Rα (τ ) ≡ hα− (t)α− (t + τ )i = he−κt β ∗ (t)e−κt−κτ β(t + τ )i (23.112)
−κτ −2κt ∗
=e heβ (t)β(t + τ )i
Z Z t Z t+τ
00 0 0 00
2 −κτ
= U02 α+ e eıθ(t )−ıθ(t )+κt +κt −2κt dt00 dt0 dt
Z Z t Z t+τ
00 0 00 0 0 00
2 −κτ
= U02 α+ e eıωca t −ıωca t +ıϕ(t )−ıϕ(t )+κt +κt −2κt dt00 dt0 dt .
Rt 0
1
Using the representation of half the δ-distribution e(ıωca +κ)t dt0 = δ1/2 = limκ→0 κ+ıωca .
itself, but the random momentum kicks imparted by molasses-cooled atoms. The
CARL linewidth is β = κ2 /δ|α− |2 = 10−14 κ.
The time-lap between two scattering events for a single atom be distributed ac-
cording to p1 (τ )dτ = γe−γτ dτ . What is the waiting time distribution for two atoms
[629]? Schenzle related the waiting time distribution to the autocorrelation function
g (2) (τ ).
F (ω)
Pbeat
g (1)
0 0.5
5
0
-5 0
0 50 100 -10 0 10 0 2 4
Δω (κ)
(c) 3 ×10
6
(d) 10
2 (f)
2
|α− |2
g (2)
1 5
1
0 0 0
0 50 100 -10 0 10 0 2 4
κt κτ Δω (κ)
Figure 23.9: Simulated time-dependence of the (a) beat and the (b) intensity signals. Calcu-
lated (c) first and second-order correlation functions. The temperature was set to T = 100 µK
and the atom number to N = 40. (e-f) Spectra of the CARL laser for (e) 4 and (f) 40 atoms.
The total coupling constant N U0 has been held constant, only the number of atoms whose
trajectories are simulated is varied.
(ph)
Phody ∝ b† b − c† c (23.114)
√ p p √
= |α+ η + α− 1 − η|2 − |α+ 1 − η − α− η|2
p
= (1 − 2η)(|α− |2 − |α+ |2 ) + 4 η − η 2 α+ Re α− .
(ph)
For a 50% beamsplitter the offsets disappear, Phody ∝= 2α+ Re α− , which means that
the homodyne signal is insensitive to intensity noise in the individual ports. Similarly
we obtain for the quadrature component,
(qu) √ p p √
Phody ∝ |α+ η + ıα− 1 − η|2 − |α+ 1 − η − ıα− η|2 (23.115)
= −2α+ Im α− .
By mounting a piezo on one of the mirrors in the homodyne loop, we can influence
which quadrature component to map. Is there a way how to get Im α− and Re α−
simultaneously?
All these quantities can be evaluated from numerical simulations of the Langevin
equations.
23.5. COHERENT PROPERTIES OF CARL 879
P− ∝ |α− |2 . (23.119)
So it may be used directly for the intensity correlation function (23.117) analogous to
Hanbury-Brown-Twiss.
The beat signal is obtained in a Young type experiment,
because we may assume α+ real. Obviously, the beat signal oscillates between the
limits ±2α+ |α− |. Currently, we only use one port of the beamsplitter. Thus we
miss information on the other quadrature phase. The question arises now, how to
calculate the spectrum if only the real part of the field Re α is known. The interesting
quantity is |g (1) (τ )|, because it contains the information on the photon statistics.
This function is smooth (it does not oscillate) and should in our case describe an
exponential decay (23.98). Fortunately, from numerical calculation it seems that
|g (1) (τ )| is just the convolution of hRe α(t) Re α(t + τ )i/hRe2 α(t)i, so that we may
recover the informations. For the spectrum, which is calculated from the complex
quantity g (1) (τ ) the question is more delicate. It comes down to asking if,
Z
F (ω) ∝ Re hRe α(t)Re α(t + τ )i eiωτ dτ (23.121)
The spectrum Sα− (ω) = |Fa− (ω)|2 has a bandwidth of β = 5.6/t0 . For example
an oscillation observed for a period 100 times longer than the cavity decay time,
t0 = 100/κ, the bandwidth will be β = 0.056κ. Simulations are based on the Langevin
equation.
880 CHAPTER 23. STATISTICAL AND THERMODYNAMIC APPROACHES TO CARL
0.5
0.01
(W)
0.5 0.5
F (ω)
0
Pbeat
P−
0
-0.5
0 0
-1 -0.01
0 0.005 0 0 0.005 -0.01 0 0.01
t (s) τ (s) t (s) τ (s)
Figure 23.11: (code) Measured trajectories (a) and (b) and correlation functions (c) and (d)
of the CARL laser. The coupling constant was N U0 = −0.1.
Figure 23.12: Spectrum of the CARL calculated from the first-order correlation function in
Fig. 23.11(c). The coupling constant was N U0 = −0.1κ.
23.5.3 Exercises
23.5.3.1 Ex: Autocorrelation functions
Atom Optics
883
Chapter 24
scale, the corresponding degree freedom is frozen and does not influence the kinetics.
885
886 CHAPTER 24. MANIPULATION OF ATOMIC GASES
The incomparable success of atomic optics has been rewarded with 20 nobel prices
in the last 25 years (Dehmelt, Paul, Ramsey, Cohen-Tannoudji, Chu, Phillips, Cornell,
Wieman, Ketterle, Hänsch, Glauber, Hall, Wineland, Haroche, Ahskin) plus several
indirectly related noble prices (De Gennes, Leggett, Thouless, Haldane, Kosterlitz).
For review articles on BEC see [551, 171, 157, 281, 609, 672], [403] or check in internet
sites http://amo.phy.gasou.edu/bec.html and http://jila.edu/bec.html.
In this chapter we review the basic techniques of Atomic Optics, emphasizing the
cooling, trapping and measurement of cold atomic gases. See also (watch talk) and
(watch talk).
Hamiltonian,
~2 d2
Ĥ = − . (24.1)
2m dx2
Therefore, the general solution of the stationary Schrödinger equation,
is,
r
ıkx −ıkx 2mE
ψ(x) = Ae + Be with k= . (24.3)
~2
Note, that the wavefunctions eıkx are not quadratically integrable. On the other
hand, they do not represent real physical systems. In practice, we need to consider
wavepackets or specify a finite volume for the particle.
Note also that the eigenvalue spectrum of is continuous. To warrant the interpre-
tation Rof the wavefunction as a probability density we will require quadratic integra-
bility, |ψ|2 d3 r = 1. This means that the wavefunction can not be infinite in a finite
volume, but it can be infinite in an infinitely small volume.
The description of the atomic motion by a wave equation emphasizes the fact
that microscopic particles have wave properties with each atom corresponding to a
velocity-dependent de Broglie wave,
h
λdB = , (24.4)
p
~
kB TD = Γ. (24.5)
2
We can also compare the kinetic energy with the energy transferred to an atom by
the absorption of a single photon. The photonic recoil energy is given by,
~2 k 2
kB Trec = . (24.6)
2m
888 CHAPTER 24. MANIPULATION OF ATOMIC GASES
Atomic clouds with temperatures around T ' TD = 10..1000 µK are called cold.
Clouds with temperatures around and below T . Trec = 0.1..10 µK are called ultra-
cold.
In most atomic optical experiments we do not work with individual atoms (or
ions), but with relatively dilute ensembles of atoms, called clouds. In general, clouds
can not be described by a single wavefunction. Either we describe every atom by a
separate and independent wavefunction (which only works when the atoms do not
interact), or we describe the cloud by probability distributions (such as the ’density
matrix’). Let us now consider a thermal cloud. The Maxwell-Boltzmann distribution
of velocities is,
r 3
m 2
g(v) = e−mv /2kB T . (24.7)
2πkB T
R R∞
This distribution is normalized, g(v)d3 v = 0 4πv 2 g(v)dv = 1. Average velocity is
now Z r
kB T
v̄ = vg(v)dv = . (24.8)
m
×10−7
2
1.5
g(v)
1 T = 10 K
0.5
100 K
0
-500 0 500
v
It represents an average over the de Broglie wavelengths of all atoms of the sample.
When a dense gas is sufficiently dense, so that this quantity exceeds the average
distance between atoms,
ρ ≡ nλ3therm > 1 , (24.10)
where n is the atomic density, we enter a new regime, where the Maxwell-Boltzmann
law ceases to be valid. Since λtherm ∝ T −1/2 , this regime corresponds to low tempera-
tures. The quantity ρ is called phase space density. A phase space density approaching
1 means an increased probability of finding more than one atom per elementary phase
space cell. We then enter the regime of quantum degeneracy, where the Boltzmann
24.1. THE ATOMIC MOTION 889
or U (x) = m 2 2 2 2 2 2 2
2 ωr ρ , where ρ = x + y + λ z with λ = ωz /ωr . We define ω̄ =
(ωr2 ωz )1/3 = λ1/3 ωr . The single-particle levels of this Hamiltonian are, nx ny nz =
~ωx nx + ~ωy ny + ~ωz nz , where the nj are integer numbers.
(a) (b)
100 100
(MHz)
50 50
U
0 2 0 2
0 0
2 0 -2 -2 1 0 -1 -2
x (mm) x (mm)
y (mm) y (mm)
(c) 40 (d) 2
(108 )
(MHz)
30
20
g()h̄ω 1
U
10
0
-1 0 1 0 50 100
x (mm) /kB (μK)
Figure 24.3: (code) (a) The figure shows two dimensions of a Ioffe-Pritchard type magnetic
trapping potential (characterized by being approximately linear at large distances from the
center and harmonic near the center). (b) Harmonic approximation (most experimentally
feasible potentials are approximately harmonic near the center). (c) One-dimensional cut
through the potential of (a,b). (d) Density of states for a harmonic (dotted line) and a
Ioffe-Pritchard type potential (solid line).
We now introduce the density of states η() for an arbitrary potential via,
Z Z Z Z p
1 (2m)3/2
η()d ≡ d3 rd3 k = d3 r d − U (r) . (24.19)
(2π)3 (2π)2 ~3
For the cylindrical harmonic trap defined in (24.18), we find with a little help from
Dr. Bronstein [111],
Z r
(2m)3/2 3 m
η() = d r − ωr2 ρ2 (24.20)
(2π) ~
2 3 2
Z 1 Z √1−x̃2 Z √1−x̃2 −ỹ2 p
1 82
= dx̃ √ dỹ √ dz̃ 1 − x̃2 − ỹ 2 − z̃ 2 .
(2π)2 (~ω̄)3 −1 − 1−x̃2 − 1−x̃2 −ỹ 2
2
η() = (harmonic potential) . (24.21)
2(~ω̄)3
24.2. OPTICAL COOLING 891
Another example is the box potential. In this case we can simply obtain,
Z
(2m)3/2 √ (2m)3/2 √
η() = d3 r = V (box potential) . (24.22)
(2π)2 ~3 V (2π)2 ~3
24.1.4 Exercises
24.1.4.1 Ex: Fundamental temperature limits
Calculate the Doppler limit, the recoil limit, and the threshold to quantum degeneracy
for an atomic cloud of density n = 1014 cm−3 for the sodium D2 transition (λ =
590 nm, Γ/2π = 10 MHz) and the rubidium D2 transition (λ = 780 nm, Γ/2π =
6 MHz).
where the term kvz is the Doppler shift. The force F− acting on the atom will be in
the direction opposite to the motion. In general,
~kΓ s
F± = ± . (24.26)
2 (2(∆ ∓ kvz )/Γ)2 + 1 + s
Supposing now, that we have two light fields propagating in directions ±z, the total
force will be F = F+ + F− . If kvz is small compared to Γ and ∆, we find through a
Taylor expansion,
kvz (2∆/Γ)
Fz ' 4~ks . (24.27)
[1 + s + (2∆/Γ)2 ]2
This expression shows that, if the detuning ∆ is negative (that is, on the red side of
the resonance), then the cooling force will oppose the motion and be proportional to
the atomic velocity. Fig. 24.4 shows this decelerating dissipative force as a function
of vz at a detuning ∆ = −Γ with I = Isat /2. The one-dimensional motion of the
atom is thus behaving like being subject to a friction force which is proportional to
the atomic velocity,
−4k 2 (2∆/Γ)
Fz ' αd vz with αd = s . (24.28)
1 + s + (2∆/Γ)2
0.4
0.2
(h̄kΓ)
0
F
-0.2
-0.4
-20 -10 0 10 20
vz (m/s)
Figure 24.4: (code) Doppler force due to one-dimensional radiative pressure as a function of
atomic velocity along the z-axis for red detuning ∆ = −Γ at a light intensity of I = 2Isat .
The solid line shows the exact expression for the restoring force [Eq (24.26)]. The broken line
shows the approximate linear expression of the velocity dependence according to Eq. (24.27).
section we compare the expressions for these quantities in the sub-Doppler regime to
those found by the conventional one-dimensional Doppler cooling model for optical
molasses.
Figure 24.5: The upper line shows, how the polarization changes as a function of position (in
units of a wavelength) for the ’lin-perp-lin’ standing wave configuration. The figure below
shows a simplified picture of the Sisyphus cooling mechanism for an atom with two levels,
Jg ↔ Je .
state Zeeman levels. To illustrate the cooling mechanism, we assume the simplest
case, a transition Jg = 12 −→ Je = 32 . As shown in Fig. 24.5 an atom moving through
the region z ' λ/8, where the polarization is σ− , will see its population pumped
to Jg = − 12 . In addition, the Clebsch-Gordan coefficients that control the dipolar
coupling of the Je = 32 require that the Jg = − 21 couples to σ− with a force three
times larger than the Jg = + 12 does. The difference of the coupling forces leads to
the light-shift between the two fundamental states shown in Fig. 24.5. As the atom
continues to move toward +z, the relative coupling forces are reversed near 3λ/8,
where the polarization is essentially σ+ . Thus, the relative energy levels of the two
hyperfine fundamental states oscillate ’out of phase’ when the atom moves through
the standing wave.
The fundamental idea is that the optical pumping rate, which always redistributes
population to the lower hyperfine level, delays the light-shift of the atom moving
through the field. The result is a ’Sisyphus effect’, where the atom spends most
of its time in sub-levels climbing a potential hill and thus converting kinetic energy
into potential energy. This accumulated potential energy is subsequently dissipated
by spontaneous emission to the electromagnetic modes of the vacuum. Simultane-
ously, the spontaneous emission transfers the population back to the lower one of
two ground state levels. The lower diagram of Fig. 24.5 illustrates the phase delay
of optical pumping. For this cooling mechanism to work, the optical pumping time,
which is controlled by the intensity of the light, must be sufficiently slow to give the
atom enough time to climb a noticeable part of the light-shift potential. This time
essentially depends on the speed of the atom. As the atom is moving slowly, having
previously been cooled by the Doppler mechanism, the light field must be weak in or-
der to decrease the optical pumping rate. Interestingly, this physical picture combines
the conservative dipole optical force, whose spatial integral gives rise to the mounts
and valleys of the potential on which the atom moves, and the irreversible dissipation
of energy by spontaneous emission, which is necessary for any type of cooling.
We can obtain simple expressions for the friction coefficient and the capture ve-
locity after some definitions. As in the Doppler cooling model we define the friction
coefficient αlpl as the proportionality constant between the force F and the atomic
velocity v,
F = −αlpl v . (24.30)
We assume that the light field is tuned to the red of the transition Jg - Je ,
∆ = ω − ω0 , (24.31)
λ/4
τp = (24.32)
v
or
Γ0 ' kv , (24.33)
24.2. OPTICAL COOLING 895
where Γ0 = 1/τp and λ/4 ' 1/k. Now, the energy W dissipated during a cycle of esca-
lation and spontaneous emission is essentially the average energy difference between
the light-shifted ground states, ∆ls ≡ ∆+ + ∆− , that is W ' −~∆ls . Therefore, the
rate for energy dissipation is,
dW
= Γ0 ~∆ls . (24.34)
dt
At the same time, every temporal energy change of a system can always be expressed
as dW
dt = F · v. Therefore, in this one-dimensional model, considering Eq. (24.31), we
can write,
dW
= −αlpl v 2 = −Γ0 ~∆ls , (24.35)
dt
such that with (24.33),
Γ ∆2
0
' (24.38)
Γ 4Ω2 .
Therefore, the sub-Doppler friction coefficient can also be written,
~k 2 ∆
αlpl = − (24.39)
4Γ
Eq. (24.39) makes two remarkable predictions: Firstly, in the ’lin-perp-lin’ configura-
tion the sub-Doppler friction coefficient can be a large number in comparison to αd .
Note that from Eq. (24.28), with I . Isat and ∆ Γ,
3
2 Γ
αd ' ~k , (24.40)
∆
and 4
αlpl ∆
' . (24.41)
αd Γ
Secondly, αlpl is independent of the intensity of the applied field. This last result is
different from the friction coefficient, which is proportional to the field intensity up
to until saturation (see Eq. (24.28)). However, although αlpl seems impressive, the
range of atomic velocities where it can operate is constrained by the condition,
Γ0 ' kv . (24.42)
896 CHAPTER 24. MANIPULATION OF ATOMIC GASES
The ratio of the capture velocities for sub-Doppler versus Doppler cooling is therefore
only,
vlpl 4∆ls
' . (24.43)
vd ∆
Fig. 24.6 graphically illustrates the comparison between the Doppler and the ’lin-
perp-lin’ sub-Doppler cooling mechanisms. The dramatic difference of the capture
ranges is evident. Note also that the slopes of the curves give the friction coefficients
and that, within the capture range, the slope is much steeper for the sub-Doppler
mechanism.
0.4
0.2
(h̄kΓ)
0
F
-0.2
-0.4
-20 -10 0 10 20
vz (m/s)
Figure 24.6: (code) Comparison of slopes, amplitudes, and capture ranges for Doppler and
Sisyphus cooling.
24.2.2.2 σ + -σ − molasses
The second mechanism operates with two counterpropagating light beams, which are
circularly polarized in opposite directions. When the two counterpropagating beams
have the same amplitude, the resulting polarization is always linear and orthogonal
to the propagation axis, but the tip of the polarization vector traces a helix with a
periodicity of λ [see Fig. 24.7(b)]. The physics of this sub-Doppler mechanism does
not involve hill-climbing nor spontaneous emission, but an imbalance of the photon
scattering rates by the two counterpropagating light waves as the atom moves along
the z-axis. This imbalance leads to a velocity-dependent force counteracting the
atomic displacement. The essential factor leading to the different scattering rates
is the creation of a population orientation along the z-axis between the sub-levels
of the atomic ground state. The more populated sub-levels scatter more photons.
Now, considering the energy level diagram (see Fig. 24.5) and the symmetry of the
Clebsch-Gordan coefficients, it is evident that transitions Jg = 12 ↔ Je = 23 coupled
by linearly polarized light can not produce an orientation of the population in the
ground state. In fact, the simplest system exhibiting this effect is Jg = 1 ↔ Je = 2.
A measure for this orientation is the magnitude of the matrix element hJz i between
the sublevels Jgz = ±1. For an atom at rest at the position z = 0 interacting with the
light polarized along the y-direction, the light-shifts ∆0 and ∆± of the three sub-levels
of the ground state would be,
∆+1 = ∆−1 = 43 ∆0 , (24.44)
24.2. OPTICAL COOLING 897
and the stationary populations would be 4/17, 4/17 e 9/17, respectively. Obviously,
linearly polarized light does not produce a stationary orientation, hJz is = 0. But when
the atom begins to move along the z-axis with velocity v, it sees a linear polarization
precessing about the axis of propagation at an angle ϕ = kz = −kvt. This precession
gives rise to a new term in the Hamiltonian, V = kvJz . Furthermore, when we
transform to a rotating coordinate system, the eigenfunctions of the Hamiltonian of
the atom moving in this new ’inertial’ system become linear combinations of the basis
functions of the resting atom. The expectation value of the stationary orientation
operator Jz , is now zero in the inertial system [174],
40 ~kv
hJz i = = ~(Π+ − Π− ) . (24.45)
17 ∆0
Figure 24.7: Polarization as a function of position (in units of a wavelength) for the σ + -σ −
standing wave configuration.
Note that, as the expectation value of the orientation is nonzero only when the
atom moves. In Eq. (24.45) we denote the populations of the sub-levels |± > as
Π± , and we interpret nonzero matrix elements as a direct measure of the population
difference between the ground state levels |± >. Note that, since ∆0 is negative (red
tuning), the Eq. (24.45) tells us, that the population Π− is larger than Π+ . Now, when
an atom traveling in +z direction is exposed to two light waves with polarizations
σ∓ propagating in the ∓z directions, the preponderance of population in the state
|−i will result in a higher scattering rate from the wave propagating in −z direction.
Therefore, the atom will be subject to a total force opposite to its movement and
proportional to its velocity. The differential scattering rate is,
40 kv 0
Γ . (24.46)
17 ∆0
40 2 Γ0
αcp = − ~k , (24.48)
17 ∆0
which is a positive quantity, since ∆0 is negative for red detuning. Comparing αcp
with αlpl we see, that αcp must be much smaller, because we always assumed that
the light shifts ∆ are much larger than the linewidths Γ0 . However, the heating rate
due to fluctuations of the recoil is also much lower. Thus, the minimum temperatures
that can be reached with the two sub-Doppler mechanisms are comparable.
Although the Doppler cooling mechanism also depends on an imbalance of scat-
tering from counterpropagating light waves, in this case the imbalance comes from
the fact that the Doppler shift experienced by the moving atoms leads to different
probabilities for photon scattering. For the sub-Doppler mechanism the scattering
probabilities from the two light waves are the same, but the ground state populations
are not. The state with the largest population suffers the highest scattering rate.
Jn denotes the Bessel function of nth order. In essence, the system is governed
by three time-constants: The natural decay width of the cooling transition Γ is a
measure for the innerratomic time scale, since it determines the average duration of
absorption-emissions cycles. The secular frequencies ζr,z determine the time scales
for changes in the external degrees of freedom, i.e. for changes of the ion’s location
and velocity. The Doppler-shift kv0 of the resonance frequency in the return point of
the ion motion finally, is a measure for the kinetic energy of the ion.
The relative importance of these three characteristic frequencies reveal the state
of the ion in the trap. The modulation index kv0 /ζr,z decides on the hight and the
number of sidebands in the excitation spectrum. The better the ion has been cooled,
24.2. OPTICAL COOLING 899
the smaller the modulation index and the smaller the hight and number of sidebands.
The kinetic energy of the ion is,
1 1
Ekin = mv 2 = mζr,z
2
x20 . (24.50)
2 0 2
The modulation index kv0 /ζr,z = kx0 = 2πx0 /λ is also called Lamb-Dicke parameter.
By cooling the der Lamb-Dicke parameter is so much reduced and the ion is so well
localized that its motional sidebands are smaller than the wavelength of the exciting
light. It then is in the so-called Lamb-Dicke regime x0 λ [?] and has so small
motional sidebands that they do not contribute to the line shape and do not influence
the line width. Therefore the linear Doppler effect vanishes.
The quantity ζr,z /Γ defines the resolution of the sidebands. If the resolution is
poor, we talk about weak confinement, else about strong confinement. Therefore the
same ion can be weakly confined with respect to an allowed transition and strongly
confined with respect to a forbidden transition. The cooling processes in the two cases
of strong and weak trapping must be described differently. At weak confinement the
oscillation frequency ζr,z is so slow that many absorption-emission cycles with the time
constant Γ−1 can occur during one oscillation period. Cooling process and cooling
limit are approximately the same as for free particles and are described by Doppler
cooling.
In the case of strong confinement for the description of the cooling process we must
consider the quantization of the motional energy in the harmonic potential. The two
levels coupled by the narrow transition split into vibrational sublevels |nr,z i, which
are populated in thermal equilibrium according to the Bose-Einstein distribution and
have the kinetic energies Ekin ,
1
nr,z = and Ekin = ~ζr,z (nr,z + 21 ) . (24.51)
e~ζr,z /kB T −1
To perform the so-called optical cooling sideband cooling [746] the laser is tuned to
the first lower sideband. The laser light is then scattered in a Raman-Anti-Stokes
process at the excited electronic state with a vibrational quantumnumber lower by 1
|e, nr,z − 1i. The subsequent spontaneous decay occurs most probably to the same
vibrational substate of the ground state |g, nr,z −1i. The net effect of such a scattering
process therefore is a transition to the next lower vibrational quantum number. The
zero point energy of the ion in the trapping potential cannot be underscored by
cooling, Ekin > 21 ~ζr,z (for the Yb+ ion it is Ekin > 2 neV). However, the uncertainty
of the kinetic energy, and the temperature T given by (24.51) have no lower limit [204].
At every absorption process, free particles carry away the momentum of the pho-
tons ~k. The recoil of a free Yb+ ion corresponds to the frequency shift ε/2π =
5.3 kHz. On a narrow transition, it yields a resonance at the frequency ε. For trapped
ions, this is not the case, because the momentum is absorbed by the whole trap (see
analogy to the Mößbauer effect).
900 CHAPTER 24. MANIPULATION OF ATOMIC GASES
kB T X πn Z p2 X πn Z (n+1)~Q0 /2 p2
= dp = 2 dp (24.53)
2 n
~Q0 (n+1)th Bloch band 2m n
~Q0 n~Q0 /2 2m
2 X∞ 2
Q0 3n2 + 3n + 1 Q0 1 + 4fB + fB2
= 2Erec πn = Erec . (24.54)
k n=0
24 k 12(1 − fB )2
Furthermore,
n̄ fB
fB = ⇐⇒ n̄ = .
1 + n̄ 1 − fB
24.2.5 Exercises
24.2.5.1 Ex: Optical molasses
Optical molasses are created (in one dimension) by two beams counterpropaganting
lasers tuned to red of an atomic transition. Each of the laser beams exerts on the
atoms the radiative pressure force F± = ~k Γ2 [2(∆±kv)/Γ]
s
2
+1+s
. ∆ is the detuning of
the laser, ν is the velocity of an atom.
a. Show that for small velocities (|kv| Γ and ∆ ≤ Γ) the optical molasses can be
understood as a friction force and calculate the friction coefficient.
b. Heating processes caused by spontaneous emission limit the minimum temperature
that can be reached in optical molasses. Calculate the laser tuning, where the tem-
perature reaches its minimum value and specify the cooling limit.
Help: Suppose a one-dimensional molasses and assume, that the spontaneous emis-
sion only happens along this dimension. The heating rate follows from the scat-
2
d hp i 2 2
tering rate R through dE = dt k
= ~2m 2R, the cooling rate follows from
dE
dt heat 2m
dt cool = F v.
24.3. OPTICAL AND MAGNETO-OPTICAL TRAPS 901
µB gF mF dB
∆B = z ≡ ∂ωzeem z , (24.55)
~ dz
see Fig. 24.8. We also apply counterpropagating laser beams along the directions
±z with circular polarizations of opposite signs and tuned to the red of the atomic
transition. It is clear from Fig. 24.8 that an atom moving in ±z direction will scatter
σ∓ type photons at a faster rate than σ± type photons, because the Zeeman effect
902 CHAPTER 24. MANIPULATION OF ATOMIC GASES
Figure 24.8: Left: Diagram of the energy level shift in an MOT, when an atom moves out of
the center of the trap. A restoring force is observed around the indicated resonance positions.
Right: Scheme of a typical MOT set up showing the six laser beams and the current-carrying
coils in anti-Helmholtz configuration producing the quadrupolar magnetic field.
will pull the ∆mJ = ∓1 transition closer to the laser frequency. The expression for
the radiation pressure force extends Eq. (24.26) to include the Doppler effect kvz and
the Zeeman effect,
~k 2Ω2
F±z = − Γ . (24.56)
2 4(∆ ± kvz ± z∂ωzeem )2 + 2Ω2 + Γ2
The atom will, therefore, feel a restoring force which pushes it back to the origin.
If the laser beams are red-detuned by an amount ∆ = −Γ, the Doppler shift of the
atomic motion introduces velocity-depending term to the restoring force, such that
for small displacements and velocities the total restoring force can be expressed by
the sum of a term which is linear in the velocity and a term which is linear in the
displacement,
FM OT = F1z + F2z = −αż − κz , (24.57)
as we will study in Exc. 24.3.3.1. From Eq. (24.57) we can derive the equation of
motion of a damped harmonic oscillator with mass m,
2α κ
z̈ + ż + z = 0 . (24.58)
m m
The damping constant α and the spring constant κ can be written compactly in terms
of atomic parameters and the field as,
Figure 24.9: Picture of a strontium MOT operated at 461 nm. The atomic cloud, which
consists of about 106 atoms at 5 mK temperature is visible as a diffuse spot located inside a
three-mirror ring cavity.
quadrupolar field in the z-direction is twice the gradient in the radial directions x and
y, such that κz = 2κx = 2κy .The damping term, which proportional to the velocity,
implies that the kinetic energy E is dissipated from the atom (or a cloud of atoms)
as,
E/E0 = e−2αt/m , (24.64)
where m is the atomic mass and E0 the kinetic energy at the beginning of the cooling
process. Therefore, the dissipative force term cools the atomic cloud and, at the same
time, combines with the position-dependent term to confine it. The time constant for
the damping,
m
τ= (24.65)
2α
is typically dozens of microseconds. It is important to keep in mind that a MOT is
anisotropic, since the restoring force is proportional to the anisotropic field gradients.
Because of its dissipative non-conservative nature, it is is more accurate to characterize
a MOT by the maximum capture rate, rather than by a ’potential depth’.
In early experiments MOTs were loaded from a decelerated atomic beam. Later it
was shown, that the low-velocity tail of the Maxwell-Boltzmann distribution provides
a sufficient amount of atoms that can be captured by a MOT, so that it can be
loaded directly from an atomic vapor at room temperature. Now many groups in the
world use these assemblies for applications ranging from precision spectroscopy to the
optical control of reactive collisions; the MOT has become the working horse of atom
optics.
2 E.g. a photon at the center of the sun will take thousands of years to get out.
24.3. OPTICAL AND MAGNETO-OPTICAL TRAPS 905
%)P
+ ¢¡ £¤ ¥¦
§©¨«ª¬®ª %
+ ¢¡ £¤ ¥& 3
maximum) or repulsive.
6.4.5 Magnetic traps
Compared to MOTs, optical
Pure magnetic traps
traps have also (far off-resonance
been used optical
to study cold collisions, trap,
and they are FORT) are tuned
far away from resonances, where
critical for the study of the
dilute population in excited
gas-phase Bose-Einstein condensatesstates
(BECs) inis insignificant and
which collisions figure importantly. We anticipate therefore that magnetic traps
spontaneous forces are absent. Note from Eq. (18.22), that spontaneous forces fall off
with the square of the detuning while the potential derived from the dipolar force only
906 CHAPTER 24. MANIPULATION OF ATOMIC GASES
decreases linearly with the detuning. The off-resonant optical density is negligible,
so that radiation trapping is not an issue. The most simple FORT consists of a
single focussed, linearly polarized gaussian laser beam tuned far to the red of an
atomic resonance. For large detunings and strong field gradients the Eqs. (18.24) and
Eqs. (18.25) become [300],
2
~Ω(r)2 3πc2 Γ I(r) 3πc2 Γ
U (r) ' = I(r) and ~γsct (r) ' σa (∆) = I(r) ,
4∆ 2ω03 ∆ ω 2ω03 ∆
p (24.66)
using the Rabi frequency ~Ω = d12 E, the dipole moment d12 = 3πε0 ~Γ/k 3 , and the
intensity I = ε20 c|E|2 . This shows that the potential becomes directly proportional
to the light intensity and inversely proportional to the detuning. Therefore, at large
detuning but very high intensity, the depth of the FORT can be maintained, although
the atoms do not absorb photons. Important advantages of FORTs as compared to
MOTs are: (1) high densities (∼ 1012 cm-3 ) and (2) a well-defined polarization axis
along which the atoms can be aligned or oriented (polarization of the spins).
Since lasers beams can easily be manipulated in position, intensity, and frequency,
they can realize a large wide variety of possible geometries. For example, with a
focused laser beam, one may influence the local density of a condensate and stir
it around by moving the position of the laser beam. Strongly focused laser beams
are often used for transporting or manipulating microscopic objects in arrangements
called optical tweezers. And with standing light waves, it is possible to form periodic
optical lattices in one, two or three dimensions (see Sec. 27.4.2).
which is U0 < 0 for red-detuned light, we can approach the potential near its center,
that is, near the optical axis, r 12 w0 , and within the range of the Rayleigh length,
z πw02 /λ, by a harmonic potential 4 ,
2x2 + 2y 2
(−2x2 −2y 2 )/w02 −z 2 /zR
2 z2
U (r) ' U0 e e ' U0 1 − − 2 (24.71)
w02 zR
m 2 2 m 2 2 U0 r2 z2
≡ U0 + ωr r + ωz z ≡ kB T + 2+ 2 .
2 2 kB T 2r̄ 2z̄
Example 113 (Dipole trap for rubidium): The formulas (24.66) hold for a
two-level system. In case of the D1- and D2-lines of rubidium, we must consider
all contributions weighted by the respective detunings,
~Γ 3~πc2 Γ I0
1 gD2 /gD1 I0
U0 ≡ σ0 + ' ,
4 ∆D1 ∆D2 ~ω 2ω 2 ∆ ~ω
πc2 Γ2
1 gD2 /gD1 I0
γsct ' + .
2ω 2 ∆2D1 ∆2D2 ~ω
3 See script on Electrodynamics (2020).
4 The diameter of a Gaussian beam can be characterized in several ways,
r̄1/e2 -radius √ r̄1/2-radius
r̄1/√e-radius = √ = 2 r̄1/e2 -radius = ,
2 2 ln 2
and r̄-rms ≡ r̄1/√e-diam and r̄-hwhm ≡ r̄1/2-diam and r̄-diam = 2r̄-radius .
908 CHAPTER 24. MANIPULATION OF ATOMIC GASES
The spontaneous emission rate decays faster with detuning than the potential
depth. Thus, heating can be avoided by working at large detunings and provid-
ing higher laser intensities. Defining the recoil temperature by,
~2 k2
Trec = ,
kB m
the heating rate is [300],
1 ~2 k2
Ṫ = Trec γsct = γsct .
3 3mkB
(a) (b)
1 1
0.5 0.5
0 0
1 1
0 0.5 0 0.5
0 0
-1 -0.5 -1 -0.5
r/w0 r/w0
z/zR z/zR
Figure 24.11: (code) (left) Dipole potential created by a Gaussian beam. (right) Dipole
potential created by a stationary light wave.
where we neglect terms containing higher powers of the coordinates than squared.
This leads to the identities,
q q r √
2 U0 U0 2U0 √ P+ P−
ωr = w0 m , ωz = k 2m , ωlat = k √
m ( P+ ++ P− )2
q √ q .
w0 kB T 2 kB T
r̄ = 2 U0 , z̄ = k U0
(24.76)
Solve the Excs. 24.3.3.5 and 24.3.3.6.
24.3.3 Exercises
24.3.3.1 Ex: Linearization of the MOT
Derive the friction coefficient and the spring constant for a MOT.
~2 p 2
Epq = U0 + ~ω(q + 21 ) + .
2mR02
Figure 24.13: (a) Quadrupolar magnetic trap generated by a pair of current-carrying wires
in anti-Helmholtz configuration. (b) Ground state energy levels of 23 Na, 2 S1/2 , F = 1 as a
function of axial distance from the trap center. (c) Illustration of Majorana spin-flips: The
red atom passes through the hole, while the green one moves adiabatically avoiding the hole.
Close to the trap center an expansion of the magnetic field generated by anti-
Helmholtz coils yields,
x
B~ = y ∂r B , (24.77)
−2z
where the field gradient ∂r B along radial direction r2 ≡ x2 + y 2 in the trap center
depends on the applied current and the geometry of the coils. However, the 1:2 aspect
ratio is generic for all quadrupolar potentials, as we will see in Exc. 24.4.6.2. We easily
verify that,
x
~=0 ~ =√ r ∂ B y .
∇·B but ∇|B| (24.78)
r2 + 4z 2
4z
Thus, the quadrupolar magnetic potential is linear in the spatial coordinates,
p
U (r) = −|~ ~ = µB gJ mJ ∂r B r2 + 4z 2 ,
µ||B| (24.79)
where 2∂r B = ∂z B.
To calculate the rms-radius r̄ of a cloud of temperature T confined to this poten-
tial, we set,
kB T ≡ U (r̄, 0) = µB r̄∂r B , (24.80)
and obtain the density distribution,
√
n(r) = n0 e−U (r)/kB T = n0 e− r 2 +4z 2 /r̄
. (24.81)
Normalization requires,
Z Z ∞Z ∞ √
e− r +4z /r̄ 2πr drdz
2 2
N= n(r) d3 r = n0 (24.82)
R3 −∞ 0
Z ∞Z ∞ Z ∞ 2|z|
2|z|
= n0 2πr̄2 ξe−ξ dξdz = n0 2πr̄2 e− r̄ 1 + r̄ dz
−∞ 2|z|/r̄ −∞
Z ∞
3 −ζ
= n0 2πr̄ e (1 + ζ) dζ = n0 4πr̄3 .
0
24.4. MAGNETIC TRAPS 913
Therefore, the effective volume is, Vef f = 4πr̄3 . In application example is discussed
in Exc. 24.4.6.3.
Considering for simplicity only radial motion, v = vêr , then by equating (24.85) and
(24.87),
µB v
rsf ∂r B = ωLarmor (rsf ) = , (24.89)
~ rsf
914 CHAPTER 24. MANIPULATION OF ATOMIC GASES
Let us now estimate the spin relaxation rate from the flow of atoms through the
volume,
3
1 rsf v̄
=N , (24.91)
τsf Vef f rsf
3
where rsf /Vef f is simply the fraction of the cloud’s volume overlapping with the
spin-flip volume. Then,
1 N 2 N ~v̄
= r v̄ = 3 v̄ (24.92)
τsf 4πr̄3 sf kB T µB ∂r B
4π µB ∂r B
N~ kB T N ~(µB ∂r B)2
= 3
(µB ∂r B)2 = .
4π(kB T ) m 4πm(kB T )2
That is, the problem gets worse when the cloud is cooled to low temperatures.
Figure 24.14: (a) Magnetic trap in Ioffe-Pritchard configuration. (b) Time-Orbiting Poten-
tial (TOP) trap. (c) Death-circle in a TOP trap.
The Earth’s gravitational field deforms the trapping potential and, in the case of a
harmonic potential, causes a gravitational sag without changing the secular frequen-
cies of the potential. Assuming the potential to be given by,
m 2 2 m 2 2 m m m
U= ωr r + ωz z − mgz = ωr2 r2 + ωz2 (z − g/ωz2 )2 − g 2 /ωz2 , (24.97)
2 2 2 2 2
the atoms sag to a height of g/ωz2 . In time-dependent traps, gravity causes a more
complex behavior [315]. Important works have been done by [137, 562, 238, 304, 2,
177, 438, 441, 197]. We study the impact of gravitation in Exc. 24.4.6.6.
Figure 24.15: Creation of a repulsive hole by light tuned to the blue of an atomic transition.
916 CHAPTER 24. MANIPULATION OF ATOMIC GASES
Figure 24.16: The basic idea of evaporation consists in removing hot particles from the
sample.
Figure 24.17: Principle of (a) rethermalization due to elastic collisions and (b) evaporation.
(~ω)3
Z=N , (24.106)
(kB T )3
and
Z
~ωrf
6 + 6β~ωrf + 3(β~ωrf )2 + (β~ωrf )3
Ẽ = εn(ε)η(ε)dε = E 1 − . (24.109)
0 6eβ~ωrf
As the truncation removes the hottest atoms from the cloud, we loose atom number
and energy. Assuming the existence of some rethermalization mechanism, we may
now use the new values for N and T to calculate the new equilibrium Boltzmann
distribution starting all over from Eq. (24.104),
Ẽ
N ←− Ñ and T ←− (24.110)
3N kB
Repeating this over and over the temperature will gradually reduce. The cooling
process can be speed up by readjusting the truncation frequency to the actual tem-
perature. This is called forced evaporation (see Fig. 24.18).
Figure 24.18: (code) (a) Forced evaporation by truncating the Boltzmann distribution over
and over again. (b) Evolution of the temperature and (c) of the phase space density with
number of remaining atoms.
24.4. MAGNETIC TRAPS 919
24.4.4.2 Rethermalization
As already mentioned, rethermalization occurs due to elastic collisions. It needs more
or less three collisions per atom to rethermalize a cloud [513, 756], so that the collision
rate determines the speed of the evaporation process. A large collision rate is desirable
to keep the evaporation process faster than trap loss processes. Evaporation ramps
between several seconds and a minute are typical.
The maximum rate of elastic collisions between trapped atoms (in the trap center)
is, √
γcoll = n0 σel v̄ 2 ∝ ρ3 N 2/3 , (24.111)
where n0 is the peak density,
σel = 8πa2s , (24.112)
is the cross-section
√ for elastic collisions and, v̄ being the average thermal velocity of
the cloud, 2v̄ is the average relative velocity between two of its atoms [401]. This
formula gives the average collision rate at the center of the cloud, where the density
is highest. To calculate the total collision rate, we need to integrate over the entire
volume of the cloud,
Z R
σel v̄n2 (r)d3 r
γ̄coll = N1 γcoll (r)n(r)d3 r = R . (24.113)
n(r)d3 r
√
For harmonic potential we find an average rate reduced by 2 2, for linear potentials
by 8. We verify this in Exc. 24.4.6.7. Finally, the rate for collision events is two times
smaller, as it involves two atoms at a time.
Obviously, the evaporation process slows down when the cloud cools more, unless
the edge of the potential is lowered, such that the hotter atoms of the colder cloud can
be evaporated. By continually lowering the edge of the potential, while the atomic
cloud keeps on rethermalizing (this procedure is called forced evaporation) very low
temperatures in the nano-Kelvin regime can be achieved, and the phase space density
can be increased by many orders of magnitude (between a MOT and a BEC there are
6 orders of magnitude) up to the threshold of Bose-Einstein condensation. Of course,
this is only possible by sacrificing many hot atoms. Even with a well optimized
evaporation ramp (i.e., a controlled lowering of the potential edge), usually only some
0.1% of the atoms reach the condensation phase after about 500 collisions per atom.
Two aspects should be mentioned regarding the optimization of the evaporation
ramp. The first aspect is, that elastic collisions with atoms from the residual back-
ground vapor of the vacuum chamber limit the lifetime of the trap. Therefore, the
evaporation must be sufficiently fast, which requires either a high rate of elastic col-
lisions or a good vacuum. A compromise must be found between a slow but efficient
evaporative cooling and a minimization of the losses, which come into play when the
evaporation takes too long. The second aspect is, that the dimensionality of the evap-
oration surface determines the effectiveness of the cooling. In the first demonstration
of evaporation, H↑ atoms of a hot cloud were ejected over a saddle point. The saddle
was located a small region away from the trap center, and only atoms with sufficient
kinetic energy along a certain direction, Ez > Uedge , could leave the trap. In such
cases, evaporation is called one-dimensional. Even though ergodic redistribution due
to anharmonicities of the potential will drive, sooner or later, all the atoms to this
920 CHAPTER 24. MANIPULATION OF ATOMIC GASES
region, this effect becomes less pronounced when the cloud cools down, because the
atoms accumulate at the bottom of the approximately harmonic (and therefore sep-
arable) potential. This fact has inhibited efficient evaporation of H↑ below 120 µK
[257].
A second evaporative technique has been demonstrated in traps called time-
orbiting potential (TOP) [560]. It is a feature of TOP traps to display a spatial
region called a ’death-circle’, where passing atoms are ejected from the trap. This
fatal circle can act as a 2-dimensional evaporation surface, provided the radius of the
circle is large enough [320]. However, under the influence of gravity the dimensionality
is further reduced to 1D [401].
The most successful evaporation technique implemented so far is based on a ra-
diative coupling of confined and free states. We discuss this technique in the following
sections. Publications on evaporative cooling are [458, 499, 560, 578, 66, 493, 513, 237,
240, 333, 535, 345, 478, 179, 402, 755, 321, 473, 218, 347]. See ([157], Sec. 3.1.4) for
an overview.
The population of the quantized levels should not change under adiabatic decom-
pression, e~ωi /kB Ti = e~ωf /kB Tf , and the phase space density remains unchanged,
ni λdB,i = nf λdB,f . If this is true, then the temperature and density change as,
3/2
ωf Tf nf
= = . (24.115)
ωi Ti ni
Solve Exc. 24.4.6.8.
(a) 40 (b) 40
mF = 2
20 20
Δ (MHz)
Δ (MHz)
h̄ω
0 0
-20 -20
mF = −2
-40 -40
-1 0 1 -1 0 1
z (mm) z (mm)
Figure 24.20: (code) (Left) Potentials due to the Zeeman structure of an atom in the ground
state with F = 1. (Right) Adiabatic potentials resulting from the coupling of Zeeman
levels via radiofrequency radiation being resonant with the difference of Zeeman levels at the
position 0.7.
2π|V12 |2
P = 1 − e−ξ with ξ= . (24.118)
~gµB ∂z Bv
The Landau-Zener theory is strictly valid only for a two-level system, which we use
here only for a qualitative discussion of two following limiting cases.
For a weak rf-field, ξ 1, P is much smaller than 1, i.e., the atoms remain pre-
dominantly on the diabatic surface shown in Fig. 24.20(a). The probability for a spin
flip transition is, P ≈ t, which describes the diabatic limit of rf-induced evapora-
tive cooling: The atomic energy levels are almost unperturbed, the atoms often spill
across the resonance surface, and only after 1/P oscillations, they spin-flip from the
hyperfine state |F, F i to the |F, F − 1i.
The adiabatic limit is clearly the ideal situation for evaporative cooling. However,
the evaporation process in a trap (with oscillation time Tosc ) saturates at a lower
rf-power. The condition for saturation is P ≈ Tosc /τel , where τel is the average time
between two collisions. This means that an energetic atom is evaporated before it
collides again.
Only the component of the magnetic field of the rf-radiation which is perpendic-
ular to the magnetic trapping field induces spin-flips. In certain geometries of the
confinement potential, for example the quadrupole trap, the magnetic field covers the
entire solid angle. Consequently, there are two points where the trapping field and
the rf-field are parallel and the elements of the transition matrix consequently zero.
Within an area around these points, the coupling is diabatic. In practice however, the
rf-transition can be sufficiently saturated that this area is small and does not strongly
affect the evaporation efficiency.
Note also that gravitation deforms the equipotential surfaces of the confinement
potential, which can reduce the evaporation efficiency [401]. Solve Exc. 24.4.6.10.
24.4. MAGNETIC TRAPS 923
Figure 24.21: Effective potential due to a rapid modulation of the trap’s location.
Figure 24.22: Creation of a repulsive hole by light tuned to the blue of an atomic transition.
4m1 m2
ξ= . (24.119)
(m1 + m2 )2
Around 3/ξ collisions per atom on average are required for complete thermalization
of a gas. For example, for the Rb-Li mixture, we have 3/ξ = 12.4. The collision rate
is,
Z
Γcoll = σ12 v̄ n1 (r)n2 (r)d3 r , (24.120)
924 CHAPTER 24. MANIPULATION OF ATOMIC GASES
Analytic solutions can be derived for harmonic traps. This will be studied in
Excs. 24.4.6.11 and Exc. 24.4.6.12.
24.4.6 Exercises
24.4.6.1 Ex: Lack of trapping potentials for strong field seekers
Show that it is not possible to create magnetic trapping potentials for atoms in low-
field seeking Zeeman states.
Φdc denotes the amplitude of the dc part of the voltage, Φac the amplitude of the ac
part. The potential Φ(r, z, t) exerts, in the temporal average, a central force on the
ion, if the radiofrequency field satisfies specific conditions.
condition for the existence of minima in a potential, ∂i ∂j φ < 0, does not obey the Laplace equation.
24.5. OTHER TRAPS 927
potential around the position r0 of the potential minimum, which depends on the
geometry of the electrodes and the applied voltages,
In the last step, we assume that the potential has an almost cylindrical shape. For
a given geometry, the curvatures bz,r can be extracted from numerical simulations.
From the continuity equation, we find bz = −2br . The polarity of the electrodes is
modulated with frequency Ω,
These equations predict stable orbits, provided that the parameters a and q are within
the so-called stability diagram shown in Fig. 24.25.
Φ(r, z) (arb.units)
-2
10
10
0 0
-10 -10
r (mm)
z (mm)
According to these equations, the ion goes through oscillatory motions that are
defined by the trap parameters ai and qi . For the motion of the ion to be finite, its
oscillation amplitude may not exceed the boundaries defined by the electrodes. This
condition imposes an allowed regime for the trap parameters called stability diagram
[488].
928 CHAPTER 24. MANIPULATION OF ATOMIC GASES
0.2
-0.2
ar
-0.4
-0.6
0 0.5 1
az
In the limit |ai |, qi 1 the ion travels only a short distance s r0 during one
modulation period Ωa . Then the ion undergoes a slow periodic motion called macro-
motion within the trapping potential with the secular frequency ζi . This motion is
modulated by a rapid oscillation called micromotion, which is excited by the modula-
tion field Ωa . Without dc voltage applied between the ring and the endcaps, ai = 0,
the motion of the ion is described by the following simple equation:
ri (t) = ri0 1 − 12 qi cos Ωa t cos ζi t , ζi = √18 qi Ωa , i = r, z . (24.131)
The orbit of the ion is confined to the inner region of the trap, if its kinetic energy is
less than mζr2 r02 + M ζz2 z02 . Since the trap is, at any instant of time, focusing in some
directions and defocusing others, it is not a conservative potential. The oscillatory
motion (disregarding micromotion) of the ion, however suggests a model, where the
trap is described by a pseudo-potential [246, ?] whose depth is,
qz
Dz = 8 eVac = 2Dr if ai = 0 . (24.132)
(μm)
0 0
-2
-5
y
-4
-5 0 5 0.5 1 1.5 2
x (μm)
1
0
Other geometries, deviating from the perfect quadrupole, are possible for the
electrodes. These traps are also well described by equation (24.124), -1 as long as the
v
ion is near the trap center. For example, higher order multipolar traps have been used
0.5 [725].
for trapping ion clouds [731], as well as Paul-Straubl traps [636] and storage rings 1 1.5 2
Particularly important for the storage of arrays of cooled ions with applications in
40
eF (eV/cm)
20
0
-20
-40
24.5. OTHER TRAPS 929
quantum computing is the linear Paul trap [576, 249, 575], where immobilized ions
are aligned on a linear chain. The advantage of the linear trap, as compared to other
traps designed for many ions, is the easier optical access to individual ions by focused
laser beams and the possibility of canceling the micromotion.
24.5.2 Micromotion
The motion of an ion in a Paul trap is a superposition of two vibrations with the
respective oscillation frequencies Ωa (modulation frequency) and ζr,z (secular fre-
quencies for radial and axial direction vibrations). For an ion in thermal equilibrium
(i.e., without active cooling), the mean kinetic energies of the micro- and macromotion
are equal [80].
The macromotion can be reduced by cooling, in contrast to the micromotion,
which is constantly excited by the modulation of the applied electric field [140]. On
the other hand, the amplitude of the micromotion decreases with the distance of the
ion from the trap center and, in the minimum of the pseudo potential, disappears
completely. Therefore, to suppress the micromotion, it is imperative to cool the
macromotion and push the ion to the trap center, if necessary, using additional static
electric fields. Since the frequency of the micromotion is much higher than that of the
macromotion, the dynamic sidebands can be resolved on a sufficiently narrow optical
transition. When the modulation frequency Ωa is very high, the secular frequencies
of the macromotion are also high, so that even large optical transitions are able to
resolve the macromotional sidebands. This is called the strong coupling regime.
Because of Coulomb repulsion, only a single atom can be at the center of a Paul
trap, such that it is difficult to zero micromovement. One solution is to use a linear
trap, where the center is smeared out over a straight line. Solve Exc. 24.5.4.1.
Example 114 (Numerical calculation of the electric field created by a
charged surface): To calculate the trapping potential for a charged particle
held on top of a planar microtrap structure, we proceed as follows. The energy
of a charge in an electric field is H = −eΦ. The electrostatic potential is given
by Coulomb’s law,
ρ(r0 ) r − r0 E(r0 ) 0
Z Z Z
1 X 0 1 X 0 1 X
Φ(r) = 0
dV − Φn 0 3
df + df ,
4π0 n Vn |r − r | 4π n Sn |r − r | 4π n Sn |r − r0 |
where φn is the voltage applied to the n-th boundary. In practice, electric fields
are generated by electrodes set to specific voltages. Using the Dirichlet boundary
conditions, we only retain the second term. Furthermore, to account for the
planar geometry of the chip electrodes, we only consider surface boundaries in
the y 0 = 0 plane,
ydx0 dz 0
Z
1 X
Φ(r) = − Φn 3 .
4π n
p
Sn (x − x0 )2 + y 2 + (z − z 0 )2
This implies that the field lines cross the chip surface orthogonally, which in
reality is only true if the chip electrodes cover the whole area. Therefore, we
only consider small gaps between the electrodes. We digitize the integral by
dividing every electrode Φn into a number of identical surface elements ∆fm ,
1 X y∆fm
Φ(r) = − Φn p 3 .
4π n,m (x − xm ) + y 2 + (z − zm )2
2
930 CHAPTER 24. MANIPULATION OF ATOMIC GASES
f1
10 mm
f2
f3
f4
frf
Figure 24.27: Possible design for a microchip ion trap. Φn are static potentials except for
Φ0 , which is alternates sign with radio frequency.
0 -0.1 -0.1
(eV)
(eV)
50 40
20
0 0 20
-20 10 -20 -100
20 -50 20 0
30 0 100 0
x (μm) z (μm) y (μm)
y (μm) x (μm) z (μm)
35 -0.18
30
-0.2
(μm)
25
20 -0.22
y
15
-0.24
-20 0 20 -50 0 50
x (μm) x (μm)
Figure 24.28: (code) Two-dimensional cuts through the electric potential generated by the
microchip shown in Fig. 24.27 for Φrf = 100 V and Φj = 0.
(r − r0 ) × j 0
Z
~ µ0
B(r) = dV .
4π V |r − r0 |3
ds0 × (r − r0 )
Z
~ µ0 I
B(r) =
4π C |r − r0 |3
p
~ µ0 I X ds2y,n (z − zn )2 + ds2z,n (x − xn )2 + ds2x,n (y − yn )2
|B(r)| = 3 .
4π n
p
(x − xn )2 + (y − yn )2 + (z − zn )2
rf
Figure 24.29: Resonance circuit for electronic ion detection. The trap is operated by a
radiofrequency, while a DC voltage is scanned across the stability diagram. Simultaneously,
an oscillating field is tuned near one of the trap’s secular frequencies. When the stability
point is such, that the secular frequency coincides with the frequency of the oscillating field,
the motion of the ions is parametrically excited and the resonance circuit is damped. This
damping is detected by a narrow-band amplifier.
24.5.3 QUEST
Homonuclear atoms and dimers do not have a permanent electrical dipole moment,
but they may have a permanent magnetic dipole moment. Therefore, homonuclear
dimers must be confined by magnetic field gradients, or else an electric dipole moment
must be induced by an oscillating electromagnetic field. In the optical regime, this
was demonstrated with the quasi-electrostatic trap (QUEST).
In contrast, heteronuclear dimers are polar molecules with a permanent electric
dipole moment, which can be quite large if the molecules are deeply bound. According
to Earnshaw’s theorem, there is no static magnetic field maximum in free space. Thus,
no ’high-field seeking’ state can be trapped. In principle, QUEST-type dipolar optical
traps can also be used for heteronuclear dimers. The problem is, however, that in
contrast to homonuclear molecules, transitions between the vibrational ground state
levels are possible. Thus, the light generating the QUEST also induces transitions
leading to a redistribution of the population over all vibrational states.
932 CHAPTER 24. MANIPULATION OF ATOMIC GASES
I(r)
Udip (r) = −αstat . (24.133)
2ε0 c
Loosely bound homonuclear molecules are subject to the sum of the restoring forces
exerted by magnetic traps on the individual atoms, µm = 2µa and dm = 2da . This is
also true for heteronuclear molecules as long as the trapping potential is much weaker
than the binding energy.
Thus, two identical dipoles with 1 Debye = 10−27 /2.998 Cm= 10−19 /c Cm2 /s=
39.36 eaB parallel oriented at a distance r = 1 µm have the energy,
1 p2
Ĥint = ≈ h × 1.5 MHz ≈ kB × 73 µK .
4πε0 r3
For example, LiRb has an electrical dipole moment of between -2 and -4.2 Debye
depending on the vibrational state of the molecule.
24.5.4 Exercises
24.5.4.1 Ex: Coulomb repulsion in linear Paul trap
Coulombian repulsion prevents that two ions confined in a linear Paul trap be si-
multaneously in the ground state. Determine the spatial extent of the ground state
and the depth of the potential in the pseudo-potential approximation. What is the
equilibrium distance of the ions?
p χ 4πn(r)
η= 1+χ'1+ with χ=− . (24.135)
2 k 3 (2∆/Γ + ı)
where n(r) is the density distribution of the cloud. The imaginary part of the suscep-
tibility is related to the absorption coefficient α and the real part to the dispersion
coefficient δ,
α 2δ
Im χ = and Re χ = . (24.136)
ω/c ω/c
Now, the absorption and dispersion coefficients can be related to the optical cross-
section σ(∆) defined in (1.102) [456], where ∆ is the detuning of light frequency from
an atomic resonance, whose linewidth is Γ. This result is called the optical theorem,
∆
α = nσ(∆) and δ = nσ(∆) , (24.137)
Γ
Finally, we obtain the Lambert-Beer law,
Z ∞
ı ∆
E = E0 eıωL/c exp ıσ(∆) − n(r)dz ≡ E0 eıωL/c e−b/2 eıϕ . (24.138)
2 Γ −∞
The absorption b describes the loss of intensity for the laser beam due to scattering
by the (disordered) atoms. It is strong near resonance, but diminished quadratically
with the detuning ∆. The scattering is necessarily accompanied by radiation pressure
accelerating and heating the atoms. The dispersion ϕ describes the refraction of the
laser beam by the atomic density distribution (which for this purpose can be consid-
ered as continuous) [158, 252]. It disappears in resonance and diminishes slowly with
increased detuning (∝ ∆). It is connected to the dipole force and, thus, does not heat
the atomic cloud. The coefficient ϕ describes the phase shift of the electromagnetic
wave transmitted through the atomic cloud.
934 CHAPTER 24. MANIPULATION OF ATOMIC GASES
Figure 24.31: Absorption images after a time-of-flight allow to identify the presence of a Bose-
condensate through its characteristic momentum distribution. Shown are images (a,b) above,
(c,d) slightly below, and (e,f) well below the critical temperature for a Bose-Einstein phase
transition (figures [320]).
Fig. 24.31 shows examples of absorption images of an atomic cloud taken at differ-
ent stages of the evaporation process. Fig. 24.31(a,b) was taken at a temperature of
320 nK; the cloud is large and isotropic and therefore purely thermal. At 250 nK [see
Fig. 24.31(c,d)] an elliptically shaped part appears in the center of the thermal cloud.
And at 180 nK [see Fig. 24.31(e,f)] the thermal cloud almost completely disappeared
for the benefit of the condensate. A quantitative evaluation of the condensed fraction
is given in Sec. 25.1.4. Solve the Excs. 24.6.6.2 and 24.6.6.1.
the first one will give different results. However, there is a non-destructive imaging
technique called dispersive imaging or phase contrast imaging. In this technique, the
laser light is tuned sufficiently far from resonance, |∆| Γ, for spontaneous emission
and heating induced by random photonic recoil to be negligible [19]. This permits to
take a series of consecutive images and create a movie of the temporal evolution of
the cloud. Another advantage of this technique is the low off-resonant optical den-
sity, which allows to take pictures of very dense clouds in situ, that is, while they are
confined in a trap.
The physical quantity which is measured by this method is the local phase shift of
the wavefront of the probe laser. Wavefront distortions are difficult to measure. To
transform the phase profile into an intensity profile, a method known from classical
optics called Schlieren method is used. It is based on the interference of the probe
beam with its distorted wavefront and a reference plane wave. In practice, there
are several possibilities. For dark-ground imaging, the part of the incident beam
not having interacted with the atoms is blocked behind the interaction zone (see
Fig. 24.32)
2
I¯dg = 12 |E − E0 |2 = I0 e−b/2+ıϕ − 1 (24.141)
b→0 ∆2
−→ I0 ϕ2 = I0 b .
Γ2
The intensity signal I¯dg is quadratic in optical density b.
For phase contrast imaging, the part of the beam not having interacted with the
atoms receives a phase shift of λ/4 with respect to the part of the beam having
interacted with the atoms:
2
I¯pc = 21 |E − E0 + E0 e±ıπ/2 |2 = I0 e−b/2+ıϕ − 1 + e±ıπ/2 (24.142)
b→0 ∆
−→ I0 (±1 + ϕ)2 ' I0 1 ± b .
Γ
The intensity I¯pc is linear in b and, consequently, more sensitive to weak signals.
Finally, a third technique, called polarization contrast imaging, detects the local bire-
fringence of the atomic cloud [106, 611].
The imaging techniques shown so far only allow to visualize the instantaneous
density distribution of the atomic cloud n(r). If we are interested in other quantities,
we have to conceive the experiment in such a way, that the desired information leaves
its signatures in the density distribution. For example, to measure the excitation
24.6. ANALYSING TECHNIQUES 937
frequencies of a condensate, which can perturb its shape and observe the subsequent
time evolution of n(r, t) via dispersive imaging [385, 494, 20, 403].
I(y, z) R
= e−σ n(r,z)dx = e−σf (y,z) . (24.143)
I0 (y, z)
24.6.6 Exercises
24.6.6.1 Ex: Lensing by cold clouds
The interaction of light with two-level atoms generates a susceptibility which gives
rise to a refraction index,
s
4πn(r)
η(r) = 1 − 3 ,
k (2∆/Γ + i)
where n(r) is the cloud’s density distribution and Γ/2π = 30.5 MHz for strontium.
a. Calculate the phase-shift suffered by a light beam crossing an ultracold atomic cloud
(N = 105 , T = 1 µK) confined in an isotropic harmonic trap (ωtrp = (2π) 100 Hz) as
a function of detuning.
b. Estimate the focal distance of the cloud for ∆ = −Γ.
938 CHAPTER 24. MANIPULATION OF ATOMIC GASES
Calculate the inverse Abel transform using Bessel of an arbitrary function in 2D.
S.C. Bell et al., A slow atom source using a collimated effusive oven and a single-layer
variable pitch coil Zeeman slower [DOI]
R.A. Cline et al., Spin relaxation of optically trapped atoms by light scattering [DOI]
P.A. Vicharelli et al., Iterative method for computing the inverse Abel transform
[DOI]
24.7.1 on gravimetry
S. Abend et al., Atom-chip fountain gravimeter [DOI]
G. Ferrari et al., Long-Lived Bloch Oscillations with Bosonic Sr Atoms and Appli-
cation to Gravity Measurement [DOI]
K. Gietka et al., A supersolid-based gravimeter in a ring cavity [DOI]
T.H. Loftus et al., Narrow line cooling finite photon recoil dynamics [DOI]
Y.B. Ovchinnikov et al., A permanent Zeeman slower for Sr atomic clock [DOI]
J.E. Sansonetti et al., Wavelengths, transition probabilities, and energy levels for the
spectrum of neutral strontium [DOI]
S. Snigirev et al., Fast and dense magneto-optical traps for strontium [DOI]
F. Sorrentino et al., Laser cooling and trapping of atomic strontium for ultracold
atoms physics, high-precision spectroscopy and quantum sensors [DOI]
Z.W. Barber et al., Direct excitation of the forbidden clock transition in neutral 174 Yb
atoms confined to an optical lattice [DOI]
A. Traverso et al., Inelastic and elastic collision rates for triplet states of ultracold
strontium [DOI]
G. Verma et al., A compact atomic beam based system for Doppler-free laser spec-
troscopy of Strontium atoms [DOI]
Xinye Xu et al., Cooling and trapping of atomic strontium [DOI]
T. Zelevinsky et al., Optical clock and ultracold collisions with trapped strontium
atoms [DOI]
S. Stellmer et al., Creation of Ultracold Sr2 Molecules in the Electronic Ground State
[DOI]
S. Stellmer et al., Laser Cooling to Quantum Degeneracy [DOI]
942 CHAPTER 24. MANIPULATION OF ATOMIC GASES
Chapter 25
Thermodynamics of ideal
quantum gases
The journey of the quest for Bose-Einstein condensation (BEC) begins with its pre-
diction by Bose and Einstein in 1926. The first hint, that the condensation was more
than just a theoretical fantasy came from London [455], who linked the newly dis-
covered phenomenon of superfluidity in 4 He to BEC. However, the interpretation of
the λ-point in terms of BEC was not obvious, because strong interactions between
particles concealed the role of quantum statistics, and the thermodynamic potentials
exhibited divergences at the critical temperature instead of discontinuities, as ex-
pected for an ideal gas BEC. These uncertainties triggered an intense search for other
systems. In 1954, Schafroth pointed out that electron pairs can be seen as composite
bosons and may form Bose-Einstein condensates at low temperatures [626]. In 1957,
Bardeen, Cooper and Schrieffer developed the microscopic theory of superconductiv-
ity [49], after other researchers, including Blutt, Schaffrot, Fröhlich and Bogolubov,
had suggested a relationship of this phenomenon to Bose condensation of electron
pairs (nowadays called Cooper pairs).
Figure 25.1: Illustration of atomic Broglie waves. From the top to the bottom the temper-
ature of the atoms is decreasing.
943
944 CHAPTER 25. THERMODYNAMICS OF IDEAL QUANTUM GASES
in weakly interacting systems, in 1962 Blatt et al. proposed the investigation of the
BEC in gases of excitons [79]. Excitons are bound electron-hole pairs that can form a
weakly interacting gas in certain non-metallic crystals. They are interesting because
their small mass allows BEC at high temperatures and gas density can be controlled
over a wide range, by only modifying the intensity of the optical excitation. Being
quasi-particles, excitons can be created and annihilated, that is their number is not
conserved. Excitons were discovered in 1968, and the first evidence for Bose-Einstein
of biexciton molecules in a CuCl crystal dates back to 1979 [131].
The laser as coherence phenomenon between photons shares many analogies with
condensates. However, photons are quasi-particles as well, and again their number
is not conserved 1 . Hence, there is no phase transition: When an optical cavity
containing photonic modes is cooled, the photons prefer to disappear in the walls of
the cavity instead of condensing.
Hecht [328] suggested in 1959, followed by Stwalley and Nosanow [692] in 1976,
that an atomic hydrogen gas with polarized spins would be an appropriate candidate
for BEC. The advantage of this system is that interactions between atoms are weak
and only give rise to a negligible quantum depletion below 1%. In 1978 Greytak and
Kleppner started at the MIT intensive efforts to generate BECs in dilute hydrogen
gases. In the 1990s, important advances in the cooling of atoms using laser light
allowed to reach very low temperatures, and the invention of the magneto-optical trap
(MOT) for neutral atoms permitted their spatial confinement and the compression of
their density. These successes boosted efforts to try to create BEC in alkaline gases,
which have electronic level schemes that lend themselves to optical cooling. Later, it
was discovered that the phase space density in MOTs is limited by radiation trapping
effects. As a solution to this problem, scientists had to learn how to trap atoms
without the use of light in conservative traps, e.g. by their magnetic dipole moment,
and to replace optical cooling with evaporative cooling. This was the crucial step
that finally permitted to reach BEC in alkaline gases in 1995. Later, the hydrogen
experiment, which initially stimulated the alkaline experiments, now taking advantage
of their success, has been taken to BEC as well [257].
Why did it take so long to reach Bose-Einstein condensation, seven decades after
its prediction by Bose and Einstein? How can we see when we have a condensate?
What are the characteristics of a BEC accessible to observation and how to measure
them? These are the answers that we will answer in the following sections. Solve
Figure 25.2: Scheme of the Broglie wave of cold atoms. From top to bottom the temperature
decreases.
Exc. 25.1.5.1.
1 The chemical potential of photons is µ = 0.
25.1. QUANTUM STATISTICS OF AN IDEAL BOSE GAS 945
where ~ = h/2π is Planck’s constant. In a gas of density n, the mean distance between
particles is n−1/3 . So, quantum effects are expected to emerge when n−1/3 ∼ λdB (T ),
such that the limit for this regime is defined by,
2π~2 2/3
kB T (n) = n . (25.3)
m
For example, an atomic gas with density n ∼ 1016 cm−3 and temperature 900 K is
certainly in the classical regime, since n−1/3 ∼ 106 cm λdB = 10−9 cm. To observe
quantum effects, we need relatively dense and cold clouds of atoms. In most gases,
lowering the temperature or increasing the density promotes the system to liquidity
before the quantum regime is reached. Well-known exceptions are spin-polarized
hydrogen (H↑), which does not become liquid and helium, which exhibits quantum
degeneracy effects in the liquid phase, although these effects are quite complex due
to strong interparticle forces.
We have already seen that all particles in the quantum world are either bosons with
integer spin or fermions with semi-integer spin. Fermions do not share a quantum
state, because they must follow the Pauli’s exclusion principle. They obey a quantum
statistical distribution called Fermi-Dirac distribution (FD). In contrast, bosons enjoy
to share a quantum state and even encourage other bosons to join them in a process
called bosonic stimulation. Bosons obey a quantum statistical distribution called
Bose-Einstein distribution (BE). The basic difference between the MB-statistics on
one hand and the BE- or FD-quantum statistics on the other is that the former applies
to identical particles which, however, are distinguishable from each other, while the
946 CHAPTER 25. THERMODYNAMICS OF IDEAL QUANTUM GASES
second describes identical indistinguishable particles. For the BE/FD statistics one
can derive [?] the occupancy number for a non-degenerate quantum state having the
energy ε when the system is kept at temperature T ,
1 1
wT,µ = 3 β(ε−µ)
, (25.4)
(2π) e ∓1
where we used the abbreviation β ≡ 1/kB T . The upper sign refers to the BE statistics,
the lower sign to the FD statistics. The chemical potential µ is an important system
parameter, which helps to normalize the distribution (25.4) to the total number of
particles,
X
N= wT,µ (ε) . (25.5)
ε
Z ≡ eβµ , (25.7)
we may write, Z ' 1−1/N . It should be noted that the chemical potential in a bosonic
system must always be less than the ground state energy in order to guarantee non-
negative occupation wT,µ (ε) of any state. Z ' 1 denotes macroscopic occupation of
the ground state. We define the critical temperature for Bose-Einstein condensation
via the occupation of the ground state. Above this temperature the occupancy of
ground state is not macroscopic, below it is.
25.1. QUANTUM STATISTICS OF AN IDEAL BOSE GAS 947
For a Bose gas of N non-interacting particles with mass m confined inside a box
potential of volume V = L3 the critical temperature for BEC can be calculated
from equation (25.3). The boundary conditions require that the momenta satisfy
pj = 2π~lj /L, where j = x, y or z and j are integers. Each state is labeled by a set
of three integers (lx , ly , lz ). In the thermodynamic limit, the sum over all quantum
states can be converted into an integral over a continuum of states,
X N →∞ 1 Z Z
−→ 3 d3 rd3 p = d3 rd3 k . (25.8)
~
r,k
For a free gas with energy ε = p2 /2m we can derive, simplifying the calculation
using the density of states η defined by (24.19), which basically depends on the ge-
ometry of our system 2 . Using the occupation number wT,µ (ε) for the Bose-Einstein
distribution (25.4) in the thermodynamic limit and the density-of-states in a box
potential (24.22), let us calculate the total number of particles,
ZZ Z ∞
N = N0 + wT,µ (ε(r, k))d3 rd3 k = N0 + (2π)3 wT,µ (ε)η(ε)dε (25.9)
0
√ 3 Z ∞
2m ε1/2 dε
= N0 + V ,
(2π)2 ~3 0 eβ(ε−µ) − 1
where Γ(η) denotes the Gamma function. Analogically, we can define the Fermi
function via,
X∞ Z ∞
(−Z)t 1 xξ−1 dx
fξ (Z) = − ξ = . (25.11)
t=1
t Γ(ξ) 0 Z −1 ex + 1
(a) 3 (b) 6
cη
g3 5
2 f3
g3/2 4
gξ (z)
ζ(ξ)
f3/2
3
1
2
0 1
0 0.5 1 1 2 3 4
z ξ
Figure 25.3: (code) (a) Bose and Fermi functions for box potentials (g3 and f3 ) and for
harmonic potentials (g3/2 and f3/2 ). Also shown is the Boltzmann limit (25.12). (b) Riemann
function.
With this definition and the definition of the thermal Broglie wavelength (25.2),
Eq. (25.9) becomes,
V
N = N0 + g3/2 (eβµ ) . (25.14)
λ3th (T )
(3/2)
We can use Eq. (25.14) to calculate the critical temperature3 Tc defined for N0 % 0
(3/2)
and µ → 0. Above the phase transition, T > Tc , the population is distributed over
(3/2)
all states, each individual state being weakly populated. Below Tc the chemical
potential is fixed by µ = 0 and the number of the particles occupying the excited
states is,
3/2
V T
Nth = 3 g3/2 (1) = N (3/2)
, (25.15)
λth (T ) Tc
with g3/2 (1) = 2.612. Since N0 + Nth = N , the number of particles in the ground
state is given by,
!3/2 2/3
(3/2)
N0 min(T, Tc ) 2π~2 N
=1− (3/2)
with kB Tc(3/2) = ,
N Tc m V g3/2 (1)
(25.16)
N0 /N is the fraction of the atomic cloud which is condensed in the ground state. The
abrupt occurrence of a finite occupation in a single quantum state at temperature
(3/2)
below Tc indicates a spontaneous change in the system and a thermodynamic
phase transition. Solve Exc. 25.1.5.2.
2 We must, however, keep in mind that the state density approach is an approximation not valid
with g3 (1) = 1.202. Since N0 + Nth = N , the number of particles in the ground state
is,
!3 1/3
(3)
N0 min(T, Tc ) N
=1− (3)
with kB Tc(3) = ~ω̄ . (25.19)
N Tc g3 (1)
2.0
1.5
E / NkBTo
0.0
1.0
-0.2
∆
0.5 -0.4
0.0
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
T / To(N)
For a confined gas, volume and temperature are interdependent, and the concept of
pressure is somewhat vague. In this case, we can not refer to the heat capacity at
constant volume or pressure. However, one can define the heat capacity for a fixed
number of particles,
∂E(T )
C(T ) = . (25.21)
∂T
Recalling the implicit dependencies of the thermodynamic variables on temperature,
25.1. QUANTUM STATISTICS OF AN IDEAL BOSE GAS 951
(μK)
Nc
0
μ/kB
0 -10
0 0.5 1 1.5 0 0.5 1 1.5
T /Tc T /Tc
(c) 10 (d) 20
(μK)
C/NkB
5 10
E/kB
0 0
0 0.5 1 1.5 0 0.5 1 1.5
T /Tc T /Tc
Figure 25.5: (code) Numerical calculation of thermodynamic potentials for a Bose gas as a
function of temperature for a given trapping potential. (a) Chemical potential, (b) energy,
(c) heat capacity per particle, and (d) total heat capacity.
1
n(x) = g3/2 (e−β[U (x)−µ] )
λ3dB
bosonic distribution functions . (25.25)
a6trap 2
n(k) = 3 g3/2 (eβ(µ−p /2m) )
λdB
In the classical limit, we can calibrate the chemical potential by Eq. (25.14) for a
box potential or by (25.17) for a harmonic potential,
( N 3
V λth 3
for a box potential
g3/2 (eβµ ) → c3/2 (eβµ ) = eβµ = .
c3 (eβµ ) = N k~ω̄
BT
for a harmonic potential
(25.26)
Hence, we obtain for the classical density distribution,
1 −β[U (x)−µ] eβµ −βU (x)
n(x) = c3/2 (e ) = e (25.27)
λ3 λ3th
th
N e−βU (x)
V q for a box potential
= 2
3 2 2 .
N mω̄ −βmω̄ x /2
2πkB T e for a harmonic potential
n(p) (cm−3 )
n(x) (cm−3 )
6 1.5
4 1
2 0.5
0 0
-50 0 50 -50 0 50
x/atrp atrp p/h̄
Figure 25.6: (code) (a) Density and (b) momentum distribution of a Bose gas (red) and a
Boltzmann gas (green) at T = 1.1Tc (solid line) and at T = 2Tc (dotted line).
where we defined,
r
kB T
rrms ≡ ttof . (25.31)
m
This distribution does not directly depend on the potential U (r), that is, the expansion
is isotropic. However, the chemical potential does depend on the potential. For very
long flight times (usually several 10 ms) the density resembles a Gaussian distribution
[157]. In Exc. 25.1.5.4(b) we determine the time-of-flight density distribution of an
ultracold Bose gas.
In a time-of-flight experiment, any deviation observed between the results (25.30)
and (25.31) points towards an impact of quantum statistics. However, absorption
images only record projections of the time-of-flight distribution on a plane.
The temperature of a Bose condensate is given by the ratio of the numbers of con-
densed and thermal atoms. What about collective excitations in pure condensate?
Can they be cooled? In fact, an oscillating BEC is not in thermal equilibrium. How-
ever, in the presence of some dissipation mechanism the excitations may thermalize
evolving towards a steady state. Once the collective excitations have become thermal
excitations, they simply increase the thermal fraction.
954 CHAPTER 25. THERMODYNAMICS OF IDEAL QUANTUM GASES
25.1.5 Exercises
25.1.5.1 Ex: Boson or fermion?
Whether an atom is a fermion or boson solely depends on its total spin. Half-integer
spin particles are fermions, integer spin particles are bosons. For example, Rb atoms
have in the ground state J = 1/2, I = 7/2 and F integer, and therefore are bosons.
Ca+ ions have J = 1/2 and no hyperfine structure, and thus are fermions. 6 Li has
the half-integer F and is a boson.
Decide on the bosonic or fermionic nature of the following atoms/molecules:
85
Rb with I = 3/2 in the state 2 S1/2
88
Sr with I = 0 in the state 1 S0
88
Sr with I = 0 in the state 3 P2
87
Sr with I = 9/2 in the state 1 S0
172
Yb+ with I = 0 in the state 2 S1/2
171
Yb+ with I = 1/2 in the state 2 S1/2
bosonic Cooper-pairs, similarly to the phenomena known as superconductivity in some metals and
956 CHAPTER 25. THERMODYNAMICS OF IDEAL QUANTUM GASES
In the last line, we inserted the density of the states into a harmonic potential (24.21)
and used the definition of the Fermi functions (25.11).
For low temperatures, x ≡ µ 1, we can use the Sommerfeld expansion which in
first order gives fη (ex ) ' xη /Γ(η + 1). From this we immediately obtain the energy,
the Fermi radius and the momentum of free particles,
Z→1
In the limit of high temperatures, fη (Z) −→ Z
3 3
kB T βµ kB T
N= e = (1 + βµ + ...) , (25.35)
~ω̄ ~ω̄
3 3
~ω̄ 1 EF
µ ≈ kB T ln N = kB T ln .
kB T 6 kB T
This means that we recover the Boltzmann gas, which satisfies an ideal gas equation
similar to that of particles in a box potential,
3
kB T
N= (Boltzmann) . (25.36)
~ω̄
as superfluidity of the fermionic 3 He.
25.2. QUANTUM STATISTICS OF AN IDEAL FERMI GAS 957
25.2.2 Energy
R
The total energy E1 ≡ E/N ≡ N −1 wd3 xd3 k per particle is given by,
R
wT,µ (x, k)d3 xd3 k
E1 = R (25.38)
wT,µ (x, k)d3 xd3 k
R −1
η() eβ(−µ) + 1 dε f4 (Z)
= R −1 = 3kB T .
η() eβ(−µ) + 1 dε f3 (Z)
Again using the Sommerfeld approximation, we see that for T → 0 the energy is
limited by
3
E1 = f4 (eβµ ) (25.39)
N β(β~ω̄)3
3µ4 2π 2 T →0 3
= 3 1+ 2
+ ... −→ EF (Fermi) .
4EF (βµ) 4
because the atoms are forced to adopt a state of larger dynamics in the outermost
T →0
regions of the trap. This implies, E1 /EF −→ 3/4.
In the limit of high temperatures, T → ∞, a classical gas has the energy per
particle,
3 −1 (βEF )3
E1 = f4 f3 ≈ 3kB T (Boltzmann) , (25.40)
N β(β~ω̄)3 6
Z→0
which is seen by taking the high temperature limit fη (Z) −→ Z and extrapolating
T →∞
to all Z. This implies, E1 /EF −→ 3kB T /EF .
Comparing with bosons,
g4 (1)
E1 = 3kB T ≈ 2.7 (Bose) . (25.41)
g3 (1)
For fermions well below the Fermi temperature, T → 0, using the Sommerfeld
approximation, we calculate,
T →0 3π 2 kB T
C1 −→ (Fermi) . (25.44)
2 TF
For high temperature T
(μK)
2 (a) 5 (b)
(μK)
0
-2
E/kB
μ/kB
-4
0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
T /TF T /TF
×106
10 10
(c) (d)
C/NkB
C/kB
5 5
0 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
T /TF T /TF
Figure 25.7: (code) Numerical calculation of thermodynamic potentials for Bose (red) and
Fermi (green) gases as a function of temperature for a given harmonic trapping potential.
(a) Chemical potential, (b) energy, (c) heat capacity per particle, and (d) total heat capacity.
The dotted magenta line in (a) shows the chemical potential calculated from the Sommerfeld
approximation.
such that,
n(x) = λ−3
dB f3/2 (e
−β[U (x)−µ]
) (Fermi) . (25.47)
25.2. QUANTUM STATISTICS OF AN IDEAL FERMI GAS 959
n(x) = λ−3
dB f3/2 (e
−β[U (x)−µ]
) (25.49)
2 3/2
3 mβ ω̄ 2 2
+ωy2 y 2 +ωz2 z 2 )/2
≈ λ−3
dB N (β~ω̄) e
−βU (x)
= N e−βm(ωx x .
2π
R
It’s easy to check, n(x)d3 x = N . Introducing the peak density n0 , we obtain,
2 2
n(x) = n0 e−mω ρ /2kB T
(Boltzmann) . (25.50)
q
The rms-radius of the distribution is σj = kB T /mωj2 , which seems contrary to the
2
above results, m 2
2 ωj xj = kB T . In comparison,
h i
n(x) = λ−3
dB g3/2 e
β(µ−U (x))
(Bose gas above Tc ) . (25.51)
p p
where λdB = 2π~2 /mkB T e atr = ~/mω̄.
3/2 Z √ 3/2
1 2 tdt 1 2
= = Γ(3/2)f3/2 (eβ(µ−ε) ) ,
(2π)2 βmωr2 eβ[ε+t−µ] + 1 (2π)2 βmωr2
such that,
ñ(k) = λ−3 6
dB atr f3/2 (e
β(µ−ε)
) (Fermi) . (25.53)
At low temperatures, T → 0,
3/2
1 2 Γ(3/2)
ñ(k) ≈ (β [µ − ε])3/2 (25.54)
(2π)2 βmωr2 Γ(5/2)
3/2 3/2 3/2
1 2 2 ~2 k 2 8 N k2
≈ E F − = 1 − .
(2π)2 mωr2 3 2m π 2 KF3 KF2
960 CHAPTER 25. THERMODYNAMICS OF IDEAL QUANTUM GASES
(25.60)
s √ 2
~2 kj2 ωj2 txj 2m mωj2 x2j
=
(1 + ωj2 t2 ) + q +
2m 2 1 + ωj2 t2 2(1 + ωj2 t2 )
m
= ξj + ω̌j2 x2j .
2
−1/2 q
2
where we define ω̌i ≡ ωi 1 + ωi2 t2 . With the substitution dξj = dkj 2~ m ξj 1 + ωj t
2 2
we obtain
3/2 Z 3/2 −1/2
1 mkB T 1 β (ξx ξy ξz ) dξx dξy dξz
nT oF (x, t) = Q 2 t2 )
(2π)3 2~2 i (1 + ω i eβΣj [ξj + 2 ω̌j2 x2j ]
m
/Z + 1
(25.61)
3 Z 3/2 −3/2 2
1 1 ω̃ β ξ 4πξ dξ
= 3 3/2 3 3
,
2 π λdB ω̄ βΣ
e j [ ξ+ m 2 2
2 ω̌j j ] /Z + 1
x
1/3
where ω̄ ≡ (ωx ωy ωz )1/3 and ω̌ ≡ (ω̌x ω̌y ω̌z ) .
1 ω̌ 3
βµ− 12 βmΣj ω̌j2 x2j
nT oF (x, t) = f 3/2 e . (25.62)
λ3dB
ω̄ 3
1 1 m 3
β (µ−mx2 /2t2 )
nT oF (x, t) = f 3/2 e = ñ(mx/t) . (25.63)
λ3dB ω̄ 2 t2 ~t
×1014
2
1.5
n (cm−3 )
0.5
0
0 50 100 150 200
r (μm)
Figure 25.8: (code) Time-of-flight velocity distribution of (red) a Li Fermi gas at T = 0 with
vanishing initial spatial distribution [119] and (green) a thermal gas at T = TF .
962 CHAPTER 25. THERMODYNAMICS OF IDEAL QUANTUM GASES
At low temperatures,
m 3 N 8 (mx/~t)2
3/2
nT oF (x, t) = 1 − (25.64)
~t KF3 π 2 KF2
!2 3/2
m 3 R 3 RF mx/~t
F
= 1−
~t 6π 2 λ (48N λ)
1/3
At high temperatures,
1 1 2 2
nT oF (x, t) = f3/2 (eβ(µ−mx /2t ) ) (25.65)
λ3dB ω̄ 2 t2
1 1 β(µ−mx2 /2t2 )
≈ 3 e
λdB ω̄ t2
2
3/2 3 3/2
mkB T 1 ~ω̄ −βmx2 /2t2 ω̄ m 2 2
≈ 2 2 2
N e ≈N 2 e−βmx /2t .
2π~ ω̄ t kB T t 2πkB T
A rms-width is,
Z
2
rT oF = r2 nT oF (x, t)d3 x (25.66)
Z
1 ω̌ 3 2 βµ− 12 βmΣj ω̌j2 x2j
= 3 r f 3/2 e d3 x
λdB ω̄ 3
Z
2 εg(ε)dε kB T g4 (Z)
= = .
mω̌r2 N eβ(ε−µ) + 1 mω̌r2 g3 (Z)
Z
16 u4 v 2 dudv
= 3 (25.67)
πN β 4 (~ω) eu2 +v2 /Z + 1
Z R 2 2
1 ~2 k 2 d3 xd3 k ~ k wT,µ (x, k)d3 xd3 k
= 2 k 2 /2m+mω 2 r 2 /2−µ] = R = Ekin,1 .
3
(2π) N 2m e β[~ +1 2m wT,µ (x, k)d3 xd3 k
This confirms the equipartition theorem for confined particles, which postulates,
hk 2 i 3T f4 (Z)
= , (25.75)
kF2 TF f3 (Z)
964 CHAPTER 25. THERMODYNAMICS OF IDEAL QUANTUM GASES
hk 2 i n+1
2 = . (25.76)
kF βEF
Must for a single dimension the value be divided by three? ~2 hkj2 i = 2mkB T f4 (Z)/f3 (Z)
setting = ~2 k 2 /m.
For a harmonic potential g() ∝ 2 and for a linear potential g() ∝ 7/2 . Interme-
diate values are possible for non isotropic traps, which are linear in some directions
and harmonic in others, e.g. for a radially quadrupolar and axially harmonic trap, we
expect g() ∝ 3 and thus E = 4N kB T . In general, we may have more complicated
situations, where the trap becomes non-harmonic beyond a certain distance from the
origin. In those cases, the density of states may be approximated by series,
which in the classical limit gives rise to energies E = 3..4N kB T depending on the
value of η.
Such effects must be considered when the time-of-flight method is used for temper-
atures measurements. For example, if we underestimate g(ε) by assuming a harmonic
potential at all ε, although the potential is quadrupolar at large ε kB T , we get a
wrong estimate for the temperature Twrng = E/3N kB instead of Tcorr = E/4N kB .
X3 m 2 2 f4 (Z)
ω hx i = 3kB T . (25.79)
j=1 2 j j f3 (Z)
3/2 3 s 3
1 ~ωtr −βε ~2 2 2
ñT →∞ (k) = 2 N e =N e−~ k /2mkB T .
2πβmωtr kB T 2πmkB T
(25.85)
This is easy to integrate by dimensions, so that,
Z Z s
∞ ∞
~2 2 2
ñT →∞ (kz ) = ñT →∞ (k)dkx dky = N e−~ kz /2mkB T . (25.86)
−∞ −∞ 2πmkB T
hx̃2tof i N f4 (Z̃)
CT (N ) = = (25.88)
hx2tof i Nnorm f4 (Z)
N f4 f3−1 (Nnorm (β~ω̄)3 )
= .
Nnorm f4 f3−1 (N (β~ω̄)3 )
to calculate pressure and volume? [check thermodynamic derivations from Romero-Rochin [see
Freddy’s thesis].
25.2. QUANTUM STATISTICS OF AN IDEAL FERMI GAS 967
be bosons. A fermion pairing with a boson will be a fermion. Composite trimers may
be either bosonic or fermionic depending on the coupling scheme. Can the quantum
nature change with the tightness of the binding? What is the total spin of a deeply
bound molecule? [589, 102, 282], [?, 349, 561, ?]
kF3
nloc = (25.89)
3π 2
the optical density is at T = 0,
Z Z RF 3/2
8σ N x2 + y 2 z2
σndy = − 1− dy (25.90)
π 2 RF3
−RF RF2 ZF2
3/2 Z RF 3/2
8σ N x2 z2 y2
= 2 3 1− 2 − 2 1− 2 dy .
π RF RF ZF −RF RF − x2 − RF2 z 2 /ZF2
p
Writing a = RF / RF2 − x2 − RF2 z 2 /ZF2 ,
Z Z a
8σ N
σndy = 2 2 4 (1 − ỹ 2 )3/2 dỹ (25.91)
π RF a −a
2σ N p p
= 2 2 4 9a 1 − a2 − 2a3 1 − a2 + 3 arcsin a .
π RF a
In the center, a = 1, Z
3N σ 9mω 2 N
σndy = 2 = 2 r , (25.92)
πRF kL EF
such that for EF ' 1µK we expect nloc ' 4 × 1012 cm−3 .
(a) ×10−15
1 (b) 0.04 total atom number N = 200000
temperature T = 0.5 μ K
Fermi temperature T F = 2.446
0.03
n(k)
n(k)
0.5 0.02
0.01
0 0
0 1 2 -2 0 2
k/kF k/kF
Figure 25.9: (code) (a) Radial momentum distribution and (b) distribution of momentum
classes in the direction of kz for a Fermi gas at T /TF = 0.2 µK (solid), a classical gas
(dash-dotted), and a Fermi gas at T = 0 (dashed).
968 CHAPTER 25. THERMODYNAMICS OF IDEAL QUANTUM GASES
1 2
This gives for a harmonic trap where ε = 2m p +m 2 2
2 ω r and with the spin degeneracy
factor g = 1, n
kB T
N= fn (z) . (25.94)
~ω
The Fermi energy again follows from Sommerfeld’s expansion,
In 1D,
1.02
N= (kB T )3/4 f3/4 (z)
~b
EF ≈ 0.87(N ~b)4/3 .
25.2.11 Exercises
25.2.11.1 Ex: Li Fermi gas
Programs on Li Fermi gases.
Bose-Einstein condensation
The experimental verification of Bose and Einstein’s prediction was for a long time
a cherished dream of many physicists. On the one hand, several phenomena have
been related to BEC in the past, for example, the phenomenon of superfluidity in
liquid helium and superconductivity. On the other hand, these strongly interacting
systems are not pure enough to clearly identify the role of BEC. In 1995, however,
Bose-Einstein condensation of weakly interacting confined atomic gases was achieved
in several laboratories [17, 180, 107, 320]. This success gave rise to a revolution
in atom optics documented in an enormous amount of theoretical and experimental
work. While initial work focused on the equilibrium thermodynamics of condensates
near the phase transition, very soon the dynamic response of condensates to pertur-
bations was the subject of in-depth investigations, followed by the study of superfluid
characteristics, quantum transport phenomena, the interaction of condensates with
light, of condensed gas mixtures [530, 683], and the behavior of condensates in peri-
odic potentials. To name only a few landmarks, we mention the creation of vortices
[479, 464] and quantum turbulence [339], the realization of various types of atom lasers
[492, 16, 84, 313] and atom interferometers with condensates [316, 422], the coherent
amplification of matter waves [365, 425, 367, 194], the creation of the Mott insulat-
ing states in optical lattices [293], the study of condensates in reduced dimensions
[520], the Anderson localization of atomic matter waves [130, 597], the observation
of Feshbach type collision resonances [160, 364, 724] and Efimov states [426, 53], the
creation of homonuclear molecular [294, 387, 165, 775, 759, 341] and heteronuclear
condensates [547] and degenerate Fermi gases [188], the observation of BCS type pair-
ing [350, 295], the observation of matter wave superradiance [365] and the interaction
of condensates with optical cavities [660, 121] and with surfaces [62].
It is clearly unthinkable to discuss all matters in this course. Let us, however, give
a basic and practical introduction to atomic optics with condensates.
971
972 CHAPTER 26. BOSE-EINSTEIN CONDENSATION
slower, which is a device that decelerates the fast atoms of the beam by means of
a counterpropagating laser exerting a radiative pressure force. A position-dependent
magnetic field applied along the trajectory of the atoms is calculated in order to
compensate for the Doppler decreasing shift of the decelerated atoms and to ensure
that the laser always stays in resonance with an atomic transition (see Exc. 14.4.6.3).
Velocities around 30 m/s are realistic and low enough to allow the capture of the
atoms by a magneto-optical (MOT) trap. Usually, some 109 atoms are captured in a
few seconds.
MOTs do not only trap atoms, but simultaneously cool them down to the Doppler
limit of typically some 10 µK. Quasi-resonant optical traps, such as MOT, are afflicted
by the problem of radiation trapping (see Sec. 24.3.1), which limits the densities of
atomic clouds to typically 1011 cm3 . This corresponds (at temperatures close to the
Doppler limit) to phase space densities several orders of magnitude away from the
threshold to Bose-Einstein condensation. For this reason, the atoms are transferred
from the MOT to a potential exempt of radiative pressure force, for example, an
optical dipole potential or a magnetic trap.
Once the atoms are confined in such a conservative potential, all the light beams
are turned off, and the technique of evaporative cooling is activated (see Sec. 24.4.3).
That is, the effective potential is deformed (for example, by an irradiated radiofre-
quency) in a way to skim out hot atoms and leave behind a cooler sample. This
however supposes that the atomic cloud finds back to thermal equilibrium afterwards.
As the rethermalization happens by elastic collisions, a high atomic density is neces-
sary, which is often achieved via a compression of the trapping potential. In general,
99.9% of the atoms must be sacrificed to condense the rest. Finally, the momentum
distribution is imaged after a time of free expansion. This is done by irradiation of a
probe laser pulse (see Fig. 24.31). The entire process of producing a condensed cloud
usually takes between 10 s and 60 s.
hydrogen is small, the critical temperature is high. The simplicity of its electronic
structure allows precise calculations of the interaction potentials based on fundamen-
tal principles, which can thus be tested experimentally.
Twenty years after having started the project of condensing dilute hydrogen gases,
Greytak and Kleppner [257] crossed the phase transition at a temperature of 50 µK
with an atomic density of 5 × 1015 cm−3 . The number of condensed atoms was 109 ,
which corresponds to a condensed fraction below 10% 1 . The condensed cloud had a
needle-like shape of 15 µm diameter and 5 mm length. It was detected by an in situ
measurement technique of the atomic velocity distribution.
processes predominate over the gain due to evaporative cooling of the thermal cloud [347].
974 CHAPTER 26. BOSE-EINSTEIN CONDENSATION
8πV E 2 dE
u(ν) = . (26.1)
c3 h2
It is the same as in the Debye model. Planck’s formula now follows from,
1
u()d = fBE ()ρ()d , (26.2)
V
where fBE is the Bose-Einstein distribution. Hence, the treatment of the photons
as indistinguishable particles following the Bose-Einstein distribution is equivalent to
assuming a Boltzmann distribution for waves with quantized energies.
Nevertheless, one can imagine a photonic gas in thermal equilibrium with an
atomic gas through Compton scattering [403]. In fact, photon condensation was
observed experimentally [412]. Also, there are theories about superfluid photon gases
[134] 2 .
2 Bose-Einstein condensation occurs in thermal equilibrium when entropy is maximized by putting
a macroscopic population of atoms into the ground state of the system. It might appear counter-
intuitive that an apparently highly ordered state as the Bose condensate maximizes entropy. How-
ever, only the particles in excited states contribute to the entropy. Their contribution is maximized
at a given total energy by forming a Bose condensate in the ground state and distributing the re-
maining atoms among higher energy states. A macroscopic population of atoms in the ground state
of the system is achieved simply by lowering the temperature of the sample. This is in contrast to
the optical laser where a non-equilibrium process is necessary to place a macroscopic population of
photons in a single mode of the electromagnetic field. This is due to the fact that, unlike photons,
the number of atoms is conserved. For bosonic atoms, the lowest entropy state below a certain tem-
perature includes a macroscopic population of the ground state. In contrast, when one cools down
a blackbody cavity, the cavity empties. Photons do not Bose condense into the ground state of the
cavity, but are absorbed by the walls. The absorbed energy leads to a larger entropy than forming a
Bose condensate. The laser phenomenon requires inversion of the active medium characterized by a
’negative’ temperature. In that sense, ’lasing’ of atoms is a simpler phenomenon than lasing of light
–all you need to do is cool a gas! However, if a photon gas would thermalize while the number of
photons is conserved, it would be described by a Bose-Einstein distribution with non-zero chemical
potential and could form a Bose condensate. Thermalization with number conservation is possible,
for example, by Compton scattering with a thermal electron gas [403].
26.2. QUANTUM THEORY 975
26.1.5 Exercises
26.1.5.1 Ex: Condensation of ions
Discuss the possibility of creating Bose-Einstein condensates from ionic clouds.
where Ψ0 is the vacuum state. The field operators ψ̂(r, t) and ψ̂ † (r, t) annihilate,
respectively create, an atom at position r and time t. This state represents a bosonic
Fock state and can be generated from the empty state by a sequence of individual
particle creation operators. Applying the notions and rules developed for the harmonic
oscillator in Sec. 3.6.1, we can define a coherent state of bosonic matter as,
X∞
N n/2
|Φ̂(t)i = √ |Ψ̂(r1 , ..., rn , t)i . (26.4)
n=0 n!
The field operators for particle creation and annihilation are normalized to the number
of atoms and satisfy the following commutation rules,
The last equation is the Heisenberg equation of motion and describes the dynamics of
a system whose many-body Hamiltonian H ≡ Hcm + Hself for N bosons interacting
within an external potential Vtrap in second quantization is defined by 3 ,
Z
3 † ~2 2
Hcm ≡ d rψ̂ (r) − ∇ + Vtrap (r, t) ψ̂(r)
2m
Z Z . (26.6)
Hself ≡ d3 r d3 r0 ψ̂ † (r)ψ̂ † (r0 )Vcoll (r − r0 )ψ̂(r0 )ψ̂(r)
The equations (26.5) and (26.6) represent the foundation of the theory of ultra-
cold bosonic gases. However, to solve the equations, we will need to apply some
simplifications, which will be discussed in the following sections. For example, we will
generally assume that the temperature of the sample is T = 0, and that all atoms are
condensed. Also, in a first time, we will neglect quantum fluctuations, replacing field
operators with complex numbers. And finally, we will need to handle the nonlinear
term appearing in Hself and which signs responsible for collisions between atoms.
eıkr
ψ(r) = eıkz + f (θ) , (26.8)
r
p
where k = 2mr E/~2 is the amplitude of the wavevector of the incident and scat-
tered waves and θ the angle between r and z. The function f (θ) is called scattering
amplitude and determines the scattering cross-section for s-wave collision through the
expression [see (12.123)],
dσ
= |f (θ)|2 . (26.9)
dΩ
3 Sometimes the Kamiltonian is used for canonical and macrocanonic ensembles. It defined by,
where dΩ = sin θdθdφ is an element of the solid angle. To calculate the scattering
amplitude, we expand the wavefunction (26.8) into spherical partial waves of orders
` of the angular momentum, as done in (12.136),
∞
1X
fk (θ) = (2` + 1)P` (cos θ)eıδ` sin δ` . (26.10)
k
`=0
The collision cross section has been shown in (12.148) to be given by,
∞
4π X
σ= (2` + 1) sin2 δ` . (26.11)
k2
`=0
For a potential with finite range, that is, a potential falling faster than r−3 with the
distance (interatomic potentials usually fall as r−6 or r−7 ), the phase shift satisfies,
δ` ∝ k 2`+1 (26.12)
for small k. In ultra-cold gases, the collision energy is very low and k → 0. Thus
the scattering will be dominated by terms with ` = 0. This is the so-called limit of
s-wave scattering. In this limit the Eqs. (26.10) can be approximated as,
1 ıδ0
fk (θ) = e sin δ0 . (26.13)
k
δ0
as = − and σ = 4πa2s . (26.15)
k
The scattering process can be interpreted as follows: During a collision, the sys-
tem’s wavefunction undergoes a phase shift, δ0 , which may be positive or negative,
depending on the sign of as . If as < 0, the phase is ’delayed’ by the collision. This
corresponds to an attractive interaction. In contrast, if as > 0, the phase is ’advanced’
and the interaction is repulsive. Of course, the intensity of the interaction is propor-
tional to the value of |as |. The expression for σ in Eq. (26.15) indicates, that the
atoms behave like hard spheres with radius |as |. The specific value of as will depend
on the interaction potential, however, the details of the potential are unimportant,
as all information about the collision is already contained in as . Consequently, in
the low-energy limit, we can assume that the collision is mediated by an effective
potential Vcoll (r), which has the particularity,
Z
4π~2
Vcoll (r)d3 r = as ≡ g . (26.16)
m
978 CHAPTER 26. BOSE-EINSTEIN CONDENSATION
This has been shown in Sec. 12.1.2. Consequently, the effective interaction between
two particles at positions r and r0 can be considered as contact interaction given by 5 ,
g
Vcoll (r, r0 ) =
δ(r − r0 ) . (26.17)
2
The interatomic interaction potential decides on the value of the scattering length
as : A repulsive potential corresponds to a positive as . For a purely attractive potential
that does not support bound states as is negative, and for an attractive potential that
supports bound states as may be positive or negative depending on the proximity of
the last bound vibrational state of the interaction potential below the dissociation
limit.
higher orders, for example, due to the renormalization of the scattering length.p It also neglects
√
quantum depletion due to the correlation effects of the order of Nout /N = 85 π a3 n0 . The the-
ory mean-field supposes the Rvalidity of the Born approximation, that is, two-body correlations are
neglected, ψ̂(r0 ) ≈ ψ̂(r) and d3 RVcoll (R) = g2 .
26.3. APPROXIMATE SOLUTIONS OF THE GROSS-PITAEVSKII EQUATION979
Remember that, despite their symbol ψ̂, which usually is associated to wavefunc-
tions, and the fact that they satisfy a non-linear Schrödinger equation the field op-
erators are represented by matrices acting of many-body states. That is, Eq. (26.20)
also represents a Heisenberg equation for the field operators.
A common approximation is the Bogolubov prescription, where the field operators
describing the condensate and its fluctuations are decomposed into a complex func-
tion, ψ0 (r) ≡ hψ̂(r)i called condensed wavefunction, which can be chosen as the order
parameter of the system, and a small perturbation, δ ψ̂(r) ≡ ψ̂(r) − ψ0 (r) describing
thermal excitations. At zero temperature, we can neglect the excitations [56], and our
system is completely described by a single wavefunction, ψ0 (r, t), obeying the famous
Gross-Pitaevskii equation (GPE),
−~2 2 2 ∂
∇ + Vtrap (r, t) + g|ψ0 (r, t)| ψ0 (r, t) = ı~ ψ0 (r, t) . (26.21)
2m ∂t
26.2.6 Exercises
26.2.6.1 Ex: Derivation of the non-linear Schrödinger equation
Derive the non-linear Schrödinger equation using (a) the commutator relation (26.5)(v)
and (b) the variational expression (26.18).
also called the Hartree solution. Inserted it into the equation of Gross-Pitaevskii
equation,
~2 k 2
E(k) = + gn , (26.25)
2m
we observe a gap in the energy spectrum due to the interaction.
In this limit, the GPE is nothing more than the usual Schrödinger equation, which
has the well-known spectrum (3.107) and the solutions (3.106).
m 2 2
In the case of an cylindrical harmonic oscillator potential, Vtrap (r, z) = 2 (ωr r +
ωz2 z 2 ), the chemical potential is,
2/5 3/5
15N g m 2
µ= (ωr ωz )2/3 . (26.32)
8π 2
n(0,0)
The radial size σ of the condensate follows from n(rhw , 0) = 2 ,
r
µ
rhw = . (26.33)
mωr2
Solve the Excs. 26.3.4.2 and 26.3.4.3.
1000
V /h (MHz)
500
0
-5 0 5
x (μm)
Figure 26.2: (code) Calculation of the BEC wavefunction by the steepest descent method
(red dots). Also shown are the ideal gas approximation (yellow line) and the Thomas-
Fermi approximation (magenta line), as well as the ground state energy (black line) and the
chemical potential (green line).
This procedure is repeated until the function converges. Since this method is derived
from a variational principle, it only serves to find the ground state of the trapping
potential. It can not be used for excited states (with the exception of vortex states,
where in cylindrical coordinates the Hamiltonian differs by an additional centrifugal
term).
×104
4
3
E[ψ0 ]/h̄ω
0
10−1 100 101
r̄/atrp
26.3.4 Exercises
984 CHAPTER 26. BOSE-EINSTEIN CONDENSATION
The Bogolubov prescription now consists in approximating the field operators of the
condensate by a sum of the equilibrium wavefunction, which is interpreted as a com-
plex number, and a small perturbation, which conserves its operator character,
This treatment assumes that most atoms are condensed, N − N0 N0 , and only a
few thermal atoms are left out.
Applying the Bogolubov transform,
X
δ ψ̂(r, t) ≡ uk (r)âk (t) + vk∗ (r)âk (t)† , (26.43)
k
into the Gross-Pitaevskii equation (26.41), we look for the lowest order,
~2 2 2
− ∇ + Vtrap + g|ψ0 | ψ0 = µψ0 , (26.46)
2m
986 CHAPTER 26. BOSE-EINSTEIN CONDENSATION
and we collect the terms of first order in e±iωk t , neglecting terms of order u2k , vk∗2 or
higher,
~2 2
− ∇ uk + Vtrap uk + 2gψ02 uk − gψ02 vk∗ = µuk + ~ωk uk (26.47)
2m
~2 2
∇ vk − Vtrap vk − 2gψ02 vk + gψ02 u∗k = −µvk + ~ωk vk .
2m
−~2 ∇2
L≡ + Vtrap + 2gn − µ , (26.48)
2m
we can write
L − ~ωk gn uk L gn
= − ~ωk σ̂z φk = 0 , (26.49)
gn L + ~ωk vk gn L
with σ̂z being the third Pauli matrix. The solution of Eq. (26.49) is given by the
requirement, that the determinant of the matrix be 0,
p
~ωk = L2 − (gn)2 . (26.50)
The object
uk (r)
φk ≡ (26.51)
vk (r)
is called a normal mode of the condensate. The normal modes (26.50) are orthonormal,
Z
hφk |φk i = drφ†k (r)σ̂z φk0 (r) = δk,k0 ,
0 (26.52)
which means that the modes do not interact, that is, the quasi-particles do not collide.
For homogeneous systems or at the interior of a large condensate with strong
interactions, we can assume that the potential is approximately constant, Vtrap → 0,
and kinetic energy negligible compared to the self-energy, and Eq. (26.41) shows us,
µ → gn. Assuming plane waves,
The coefficients u and v describe the annihilation and u∗ and v ∗ the creation of quasi-
particles called fonons or elementary excitations. Two limits are interesting. In the
2 2
k
low energy limit, ~2m gn(r), we create phonon-like excitations. The Bogolubov
spectrum becomes,
r
ph ng
~ωk ' cs ~k with cs ≡ . (26.55)
m
Here is cs the sound velocity inside the condensate. The fact that, comparing (26.54)
and (26.55), we find,
for all k is the Landau criterion for the superfluidity of the condensate. It means that,
for an object being dragged through the condensate with a velocity less than cs , it does
not become energetically favorable to produce excitations, see Fig. 26.4. Therefore,
the object will move without dissipation, which is an important characteristic of
superfluids. Experiments demonstrated this manifestation of superfluidity by slowly
stirring a strongly focused blue-detuned laser through a condensate. We will come
back to this in Sec. 27.1.1.
2 2
k
In the high energy limit, ~2m gn(r), we create particle-like excitations. The
Bogolubov spectrum becomes,
~2 k 2
~ωkpa = . (26.57)
2m
In this limit, we recover the quadratic dispersion relation of free particles, as seen in
Fig. 26.4.
30
E/h (Hz)
20
10
0
0 0.5 1 1.5 2
k/k0
Figure 26.4: (code) Bogolubov spectrum (red), phonon limit (blue line) and particle limit
(green line).
∂
L − gn + f+ (r)e−ıωp t + f− (r)eıωp t ψ̂ = i~ ψ̂(r, t) . (26.58)
∂t
After the Bogolubov transform (26.45) we now have,
L gn u(r) f+ (r)ψ0 (r)
− ~ωk σ3 =− ∗ . (26.59)
gn L v(r) f− (r)ψ0∗ (r)
The solution of this equation is found by expanding the amplitudes u(r) and v(r) into
normal modes,
X X
u(r) f+ (r)ψ0 (r)
= ck φk (r) and ∗ = gk φk (r) , (26.60)
v(r) f− (r)ψ0∗ (r)
k k
The deviation, which is observed for resonant excitation, is due to the neglected losses
and non-linear effects.
bances render the dynamics chaotic, since the energy is coupled to many modes of excitation.
26.4. ELEMENTARY EXCITATIONS
Fig. 14. – Shape of low-lying collective excitations: 989 MIT), b) fast
a) slow m = 0 quadrupolar oscillation (JILA,
m = 0 radial oscillation (MIT), c) |m| = 2 oscillation (JILA).
where λ is the aspect ratio λ = ωz /ωr . Obviously, higher excitation orders depend
on the trap geometry. Moreover, we can easily see, that occasional degeneracies must
arise for specific aspect
Fig. 15. – m = 0ratios. For example,
p condensate
quadrupolar when viewed
oscillations ω+ (2,in0)time-of-flight
= 2ω(2,absorption
2), we derive
imaging. Oscillations in
from (26.63)
p the
the condition
aspect λ
ratio of the = 16/7,
expanding and when
condensate ω
are clearly
+ (2, 0)
visible. = 2ω
The (2, 0),
horizontal
− we
width derive
of each cloud is 1.2 mm.
1
√ taken from ref. [177].
Figure
λ = 6 77 + 5 145.
a) b) c)
350 µm
(d)
Shape of low-lying collective excitations:
Fig.a)16.
slow
– mm==0 0quadrupolar
quadrupolar oscillationoscillations
condensate (JILA, MIT), b)in-situ.
viewed fast Repeated phase-contrast images, taken at
Figure
al oscillation (MIT), c) |m| = 26.5 : Normal
2 oscillation modes
5 ms (JILA).
intervals, of large-amplitude
show a BEC. (a) oscillations
Shape oscillation, (b) breathing
of a low-temperature oscillation,
Bose-Einstein condensate. Figure taken
from ref. [30].
and (c) quadrupole oscillation. (d) Non-destructive measurements of quadrupolar
vibrations of a BEC [674].
lar modes observed at JILA, with out-of-phase oscillations along the axial and radial directions.
r frequency mode was primarily a radial breathing mode (fig. 14b). After locating the modes
selective “step” excitation, we used a five-cycle sinusoidal modulation of the trapping coils to
26.4.2.2 The
y excite the shape oscillations. Quantum
subsequent depletion
free oscillations were clearly visible as periodic
ns of the aspect ratio in time-of-flight (fig. 15) and in phase-contrast (fig. 16) as observed
Following Bogolubov’s theory the quantum depletion is given by,
Z
δN 1 X
= d3 r|vk (r)|2 . (26.64)
N N
k
26.4.3 Exercises
26.4.3.1 Ex: Sound velocity in the Thomas-Fermi regime
Consider a condensate of N = 105 87 Rb atoms confined to an isotropic harmonic
potential with secular frequency ωtrp = (2π) 50 Hz. Remember as = 110aB and use
the Thomas-Fermi approximation.
a. Evaluate the chemical potential.
b. Calculate the sound velocity at the center of the condensate.
large quantum depletion, which in 4 He is of the order of 10% and hides effects of quantum statistics.
Quantum depletion prevents any form of long-range order, as this requires delocalization.
9 Bogolubov’s theory assumes that the only impact of collisions is to deform the dispersion rela-
-200 0 200
Position (µm)
(26.66)
normal modes qR 1
macroscopic qa 1
beyond Bogolubov qa > 1
2
collision-less γcoll ~q
2m qlmf p 1
2
hydrodynamic γcoll > ~q
2m qlmf p > 1
(26.67)
992 CHAPTER 26. BOSE-EINSTEIN CONDENSATION
a = 0..1000aB = 0..50 nm
n−1/3 = 20..200 µm
2π/q = 0.2..∞ µm
√ −1 (26.68)
ξ = 8πna
p = 0.03..30 µm
aho = ~/mωtr = 0.1..3 µm
2π/kF = 2πaho (48N )−1/6 = 30..70 µm
The various regimes of sound are distinguished by frequency shifts and damping
rates of the collective modes, which depend on the density (determining the mean-field
energy and the mean free path) and the temperature (controlling the ratio between
condensed and normal fraction). The quantity
1
ξ≡√ , (26.69)
8πas n
2
2
1.5
3/2
phase
ψ/aho
1 0
0.5
-2
0
-1 0 1 -1 0 1
x/aho x/aho
Figure 26.7: Simulation of the propagation of sound toward the edges of the condensate.
Differently from (2.101) we chose here a normalization, which leaves the field operators
ak without unit.
26.6. MOMENTUM REPRESENTATION 993
1 X
âk = âk0 δ (3) (k − k0 ) . (26.71)
V 0
k
uk (r) = √ 1
eık·r , (26.73)
(2π)3 V
which is particularly adapted to the 3D box potential, Vtrap (r) ≡ ∞ se r > R. Also,
we calculate,
q Z Z
−ık·r 3
1
âk = (2π)3 V ψ̂(r)e d r = ψ̂(r)u∗k (r)d3 r . (26.74)
In Exc. 26.6.4.1 we will check the validity of the commutation rules (26.75), and in
Exc. 26.6.4.2 we derive the following representation of the Hamiltonian:
Z
3 † ~2 2
H = d rψ̂ (r) − ∇ + Vtrap (r) ψ̂(r)
2m
Z Z
g
+ d3 r d3 r0 ψ̂ † (r)ψ̂ † (r0 )δ(r − r0 )ψ̂(r0 )ψ̂(r) , (26.78)
2
X ~2 k 2 † X † g X † †
= âk âk + âk Vk,k0 âk0 + â â 0 âk00 âk+k0 −k00
2m 0
2 0 00 k k
k k,k k,k ,k
994 CHAPTER 26. BOSE-EINSTEIN CONDENSATION
Hence, just considering the ground state of the trap, that is, assuming that the atomic
motion be frozen, our Hamiltonian becomes:
g
Hcm + Hself = ~ωtrp ↠â + ↠↠ââ . (26.84)
2
26.6.4 Exercises
26.6.4.1 Ex: Commutators in momentum space
Demonstrate the validity of the commutation rules (26.75).
2
10
Aspect Ratio
4
2
1
4
2
0.1
2 4 6 2 4 6 2 4
1 10
Time (ms)
Figure 26.8: Inversion of aspect ratio during time-of-flight. The frequencies of the harmonic
trap were ωr = (2π) · 248 Hz and ωr = (2π) · 16 Hz [403].
This provides us with the energy eigenvalues of the excited trap states and their
eigenfunctions. One can f.e. guess an eigenvalue and numerically solve the Schrödinger
equation using Runge-Kutta integration routines. The eigenfunction will probably
diverge. In this case, we vary the eigenvalue until the solution converges.
Finally, we can calculate the thermal density distribution:,
−1
Nj ≡ [exp((E1,j − µ1 )/kB T1 ) − 1] (26.100)
X
nth (r1 ) = Nj |ψ1,j (r1 )|2 ,
j
that the condensate is only weakly disturbed by the thermal cloud. In contrast, the
presence of the condensate dramatically modifies the potential for the thermal atoms.
Therefore, we must first calculate the density of the condensate n0 (r), before calcu-
lating that of the thermal cloud. If this procedure is applied iteratively, it is called
Hartree-Fock method:
iii. assuming that collisions between atoms of the thermal cloud, the density of
which is weak, can be neglected, we can view the cloud as an ideal gas in a potential
modified by the strongly anharmonic mean field of the condensate, Vtrap (r)
+ 2gn(r),
Vtrap (r)−µ+2gn(r)
we calculate the thermal density n0th (r) = λ−3
dB g3/2 exp kB T ;
Z Z
2 2 2 2
U= d3 r ψ(r) −~2m∇ ψ(r) + Vtrap (r)n(r) + g2 [2n(r)2 − n0 (r)] + h−3 d3 k −~2mk n(r, k)
(26.101)
Z h i
2 2 Vtrap (r)−µ+2gn(r)
= d3 r ψ(r) −~2m∇ ψ(r) + Vtrap (r)n(r) + g2 [2n(r)2 − n0 (r)] + 3kB T
2λ 3 g3/2 exp kB T
dB
Consider the ideal gas in an isotropic harmonic potential. The excitation spectrum
then takes the form:
g −→ 0 (26.103)
g→0
µ −→ 32 ~ω
g→0
√
N0 r2
φ(r) −→ exp(− )
2a2trap
3/2
π 3/4 atrap
Z
−~2 ∆
Ekin = φ(r)d3 r = − 94 N0 ~ω
φ(r)
2m
Z h i
g→0 1 V (r)−µ
nth (r) −→ f (r · k)d3 k = 3 g3/2 exp trap kB T
λdB
Z 3 h i
g→0 kB T
Nth −→ nth (r)d3 r = g3 exp kBµT
~ω
Z 3 h i
g→0 kB T
3 3
U −→ εHF f (r · k)d r · d k = 3kB T g4 exp kBµT .
~ω
26.7.4 Exercises
26.7.4.1 Ex: Ballistic expansion of a condensate
∂ ı~ ∂ 2 ıV (x)
ψ(x, t) = ψ(x, t) − ψ(x, t) (26.106)
∂t 2m ∂x2 ~
is a parabolic partial differential equation. We usually seek a solution on an interval
x ∈ [a, b] and t > 0. The solution is uniquely determined from boundary conditions:
ψ(a, t) = ψ(b, t) = 0 and ψ(x, 0) = g(x). One method for numerical solution solves
for the values of the wavefunction on a regular grid of dimension h = (b − a)/Nx in x
and τ in t:,
ψjk = ψ(a + jh, kτ ) . (26.107)
The derivatives are replaced by simple finite differences. The r.h.s. of the equation at
the grid point (i, j) is then,
XN
ı~ k k k
iV (a + jh) k
ψ j+1 − 2ψj + ψj−1 − ψ j = iHjm ψjk , (26.108)
2mh2 ~ m=0
where H is a real symmetric tridiagonal matrix (provided V (x) is real). The l.h.s. of
the equation can be replaced either by a forward or backward difference,
This method is a second order algorithm in t, i.e. the discretization error decreases
as τ 2 . The finite difference representation of the second derivative d2 /dx2 is also good
to second order in h2 . The Crank-Nicholson Algorithm also gives a unitary evolution
in time. That is especially useful for quantum mechanics where unitarity assures that
the normalization of the wavefunction is unchanged over time. The algorithm steps
the solution forward in time by one time unit, starting from the initial wavefunction
at t = 0. According to the Crank-Nicholson scheme, the time stepping process is half
explicit and half implicit. The implicit part involves solving a tridiagonal system.
That solution is accomplished by Crout reduction, a direct method related to Gaussian
elimination and LU decomposition.
To simplify the algorithm we have chosen units in which the Planck constant ~ = 1,
time step τ = 1 and the spatial separation h = 1. This can always be arranged by an
appropriate redefinition of mass and potential: m = mSI h2 /τ ~ and V = VSI τ /h.
1002 CHAPTER 26. BOSE-EINSTEIN CONDENSATION
Introducing abbreviations,
B(ϕn+1
j −ϕnj ) = −A ϕn+1
j+1 − 2ϕj
n+1
+ ϕn+1 n n n
j−1 + ϕj+1 − 2ϕj + ϕj−1 +Cj ϕj
n n+1
+ ϕnj .
(26.113)
Can be rewritten as,
n+1 n
Aϕn+1
j+1 + −2A + B − Cj ϕj
n
+Aϕn+1 n n n
j−1 = −Aϕj+1 + 2A + B + Cj ϕj −Aϕj−1 = Xj .
n
(26.114)
The set of equations,
n
−2A − B − C1n A 0 ··· 0 ϕn+1 X1
1
A −2A − B − C2 n
A ··· 0 n+1 n
ϕ2 X2
.. n+1 n
0 A −2A − B − C3n . 0 ϕ3 = X3
. .
.. .. .. .. . . .
. . . . .. . .
0 0 ··· A −2A − B − Cjn ϕn+1
j Xjn
(26.115)
can be solved by inverting tridiagonal matrix. Set ϕn+1
0 = ϕn+1
J+1 = 0. Run a loop
n+1 n+1
j = 1, .., J. Assume given ϕ1 = ϕj ,
i.e. we discretize in space using x = a + j(b − a)/M and k = 2πl/(b − a) such that
k (x − a) = 2πlj/M . We propagate half the way, ∆t/2, in time,
ı 2 ∆t
ϕx ≡ e− ~ (V (x)+g1D |ψ(x,t)| ) 2
ψ(x, t) (26.119)
PM −1
for j = 0, .., M −1. Now we propagate the spectral components, (Fϕ)k ≡ j=0 ϕx e−2πilj/M ,
PM/2−1
in momentum space and transform back, (F −1 ϕ)x ≡ M −1 l=−M/2 ϕk e2πilj/M ,
h ~ 2 i
φx ≡ F −1 eı 2m k ∆t (Fϕ)k . (26.120)
for l = −M/2, .., M/2 − 1. Finally, we propagate the remaining time, ∆t/2,
ı 2 ∆t
ψ(x, t + dt) = e− ~ (V (x)+g1D |φx | ) 2
φx . (26.121)
for a < x < b. Choose periodic boundary conditions, ψ(a, t) = ψ(b, t) and ψx (a, t) =
ψx (b, t). We generalize the time-splitting spectral algorithm,
∂ψ1 (x, t) ı ı ı
= − V (x)ψ1 (x, t) − g1D |ψ1 (x, t)|2 ψ1 (x, t) − ı∆mw ψ1 (x, t) − Ωmw ψ2 (x, t)
∂t ~ ~ 2
∂ψ2 (x, t) ı ı ı
= − V (x)ψ2 (x, t) − g1D |ψ2 (x, t)|2 ψ2 (x, t) − Ωmw ψ1 (x, t)
∂t ~ ~ 2
∂ψ1 (x, t) ~
=i ψxx1 (x, t)
∂t 2m
∂ψ2 (x, t) ~
=i ψxx2 (x, t) , (26.123)
∂t 2m
i.e. we discretize in space and propagate half the way, ∆t/2, in time,
h ı ı ı i ∆t
ϕx1 ≡ ψ1 (x, t) − g1D |ψ1 (x, t)|2 + i∆mw ψ1 (x, t) − Ωmw ψ2 (x, t)
V (x) +
~ ~ 2 2
h ı ı ı i ∆t
ϕx2 ≡ ψ1 (x, t) − V (x) + g1D |ψ2 (x, t)|2 ψ2 (x, t) − Ωmw ψ1 (x, t) .
~ ~ 2 2
(26.124)
Since we here use the first order Taylor expansion of the exponential function, we
introduce an error that we have to keep low by renormalizing the wavefunction after
each step. Now we propagate the spectral components in momentum space and
transform back,
h ~ 2 i
φx1 ≡ F −1 eı 2m k ∆t (Fϕx1 )k (26.125)
h ~ 2 i
φx2 ≡ F −1 eı 2m k ∆t (Fϕx2 )k ,
for l = −M/2, .., M/2 − 1. Finally, we propagate the remaining time, ∆t/2,
h ı ı ı ı i ∆t
ψ1 (x, t + dt) = φx1 − g1D |φx1 |2 + ∆mw φx1 − Ωmw φx2
V (x) +
~ ~ 2 2 2
h ı ı ı i ∆t
2
ψ2 (x, t + dt) = φx1 − V (x) + g1D |φx2 | φx2 − Ωmw φx1 . (26.126)
~ ~ 2 2
~ t)
∂ ψ(x, ~ t) ,
= M ψ(x, (26.127)
∂t
with the matrix,
− ~ı V (x) − ~ı g1D |ψ1 (x, t)|2 − ı∆mw − 2ı Ωmw
M= , (26.128)
− 2ı Ωmw − ~ V (x) − ~ı g1D |ψ2 (x, t)|2
ı
26.8. NUMERICAL SIMULATIONS OF THE GP EQUATION 1005
This formula can easily be computed, because the block matrices of transform U are
diagonal in x.
∂ −~2 d2
ı~ ψ(z, t) = ψ(z, t) + V (z)ψ(z, t) . (26.134)
∂t 2m dz 2
Numerically the Schrödinger equation is integrated via,
26.8.3.2 2D simulations
We want to describe the two-dimensional evolution of an atomic wavepacket according
to the Schrödinger equation. The problem is fully described by,
−~2 d2 d2 ∂
2
+ 2
Ψ(x, z, t) + V (x, z)Ψ(x, z, t) = i~ Ψ(x, z, t) . (26.137)
2m dx dz ∂t
Since the potential is conservative, we separate the time-evolution,
Ψ(x, z, t) = ψ(x, z)e−ıEt/~ (26.138)
The initial shape of the wavefunction moving along the y coordinate could be a
Gaussian,
−1/2 −x2 /4σx
2
−z 2 /4σy2
ψ(x, z) = ψx (x, 0)ψz (0, z) = (2πσx σz ) e . (26.139)
We can assume σz σx and obtain an effectively one-dimensional model, E = Ekz .
Set,
ψ(x, z, 0) = φ(x, z)e−ıkz z (26.140)
and,
d2 d d2 2m 2m 2
− 2 φ(x, z) + 2ıkz φ(x, z) = 2 φ(x, z) + Ekz − 2 V (x, z) − kz φ(x, z) ,
dz dz dx ~2 ~
(26.141)
and,
d d2 2m
φ(x, z) = 2 φ(x, z) − 2 V (x, z)φ(x, z) (26.142)
dz dx ~
d 1 1 m
φ(x, z) = [φ(x − dx, z) − 2φ(x, z) + φ(x + dx, z)] − V (x, z)φ(x, z) .
dz 2ıkz dx2 ıkz ~2
Discretize in steps dz.
26.8.4 Exercises
26.8.4.1 Ex: Propagation of wavefunctions
Programs on the propagation of wavefunctions.
kB T 2πa2trp
= 2 . (27.1)
~ωtrp λtherm
For a typical experimental situation, kB T /~ωtrp ≈ N 1/3 ≈ 100. This shows that the
assumption of a locally homogeneous gas is generally a good approximation. With
the definition of the healing length (26.69), the degree of superfluidity of a condensate
with density n can be cast into the form,
gn a2trp
= 2 . (27.2)
~ωtrp ξ
1009
1010 CHAPTER 27. SUPERFLUID AND COHERENT PROPERTIES OF BOSE-EINSTEIN CO
For a typical experimental situation, a2trp /ξ 2 ≈ 100. A parameter that well charac-
terizes the importance of interatomic forces within a condensate is the gas dilution,
a2s
na3s = . (27.3)
8πξ 2
The typical numerical value na3s ≈ 10−5 shows, that atomic gases are usually very
dilute, although interatomic forces play an important role in the dynamics of con-
densates. In contrast, three-body collisions can be totally neglected, because the
probability to have three atoms close to each other is even lower than the probability
for two atoms.
There are several manifestations of phenomena linked to superfluidity, such as
the behavior of collective elementary excitations, superfluid flow of non-circulating
topological modes, quantized flux in vortices and matter wave solitons, which will be
discussed in the following sections.
(p + M v)2 p2
E = E(p) + = E(p) + p · v + 12 M v 2 + ' E(p) + p · v + 12 M v 2 , (27.4)
2M 2M
where M is the total mass of the superfluid. Since the term M v 2 /2 is the initial
kinetic energy of the fluid, E(p) + p · v represents the excitation energy. And since
the kinetic energy of the superfluid can only be diminished by the excitation, we need
E(p) + p · v < 0. Finally, as E(p) must be positive, the condition for generating
elementary excitations is,
E(p) ≤ pv , (27.5)
where p and v are antiparallel.
Therefore, there is a relative minimum velocity between the fluid and the obstacle,
called critical Landau velocity, for creating excitations,
E(p)
vc = min . (27.6)
p
the perturbation does not cross the Bogolubov spectrum. That is, the perturbation
can not be converted into excitations, which is only possible when v > cs .
In an ideal condensate without interactions, as = 0, the Bogolubov spectrum
reduces to the quadratic dispersion relation of free particles, which means that there
is no critical velocity, vc = cs . Consequently, it is always possible to excite the
condensate, that is, an ideal condensate can not be superfluid, and, as first pointed out
by Landau: superfluidity and Bose-Einstein condensation are different phenomena.
30
E/h (Hz)
20
10
0
0 0.5 1 1.5 2
k/k0
Figure 27.1: (code) Landau’s criterion for superfluidity. Slow perturbations (blue line)
do not cross the Bogolubov spectrum (red line) and do not generate excitations. Rapid
perturbations (green line) cross the spectrum and can be dissipated.
a microscopic object. Puzzled by this discrepancy, Chikkatur et al. [135] studied the
motion of microscopic atomic impurities through a condensate. The impurity was
created by transferring a few atoms out of the original BEC from the trapped state
|F = 1, mF = −1i to the free state |F = 1, mF = 0i via induced Raman transitions.
The initial velocity was adjusted by the laser beams exciting the Raman transition
(polarization, incident angle and relative detuning) to satisfy the Bragg condition (see
Sec. 28.2). After the free impurity passed through the BEC, the trap was switched
off, a Stern-Gerlach magnetic field gradient was pulsed to separate atoms in different
Zeeman states, and finally the atoms were detected via time-of-flight imaging (see
Sec. 24.6.1). When the initial velocity of the impurity was well above a critical value
given by the local velocity of sound, ultracold s-wave collisions between impurity
atoms and the stationary condensate distributed the momenta of the collision part-
ners uniformly. In TOF images this appeared as a circular halo centered around the
center-of-mass momentum of the collision partners (see Fig. 27.2). When, however,
the initial velocity of the impurity was reduced, its collision rate with the stationary
condensate was suppressed and the trajectory became superfluid.
Figure 27.2: (a) Scattering of impurities at velocities above the critical velocity. The presence
of a halo indicates the occurrence of collisions between the impurity and the condensate
dissipating the relative kinetic energy. (b) For velocities below the critical velocity the halo
disappears.
From the GPE we derive the continuity and the Navier-Stokes equations:
∂n
+ ∇(vs n) = 0 (27.10)
∂t
∂vs ~2 a √ m
m + ∇ Vtrp + gn − √ ∇2 n − µ + vs2 = 0 .
∂t 2m n 2
27.2. TOPOLOGICAL MODES 1013
We can see that the hydrodynamic behavior of a BEC depends greatly on the quantum
phase θ. When the kinetic pressure is small compared to the mean-field energy,
∂vs m
m + ∇ Vtrp + gn + vs2 = 0 . (27.11)
∂t 2
This is the Euler equation for a fluid with a potential flow. This equation and the
continuity equation have the typical structure of equations describing superfluids at
T = 0. This is due 1. to the presence of a Bose-Einstein condensate allowing us
to formulate an equation for a complex order parameter, and 2. to the presence of
interactions, included via the pressure term in the Euler equation, which are necessary
condition for superfluidity.
At zero temperature, the entire fluid is superfluid. Moreover, in the Gross-
Pitaevskii approximation, the whole fluid is condensed. Therefore, vs (r, t) is the
velocity flow of the superfluid 1 .
27.2.1 Vortices
From Eq. (27.9) it is easy to see, that the superfluid is non-rotational, that is,
∇ × vs (r, t) = 0 . (27.12)
This raises the question, how vortices are possible. The solution to this apparent
contradiction is, that Eq. (27.12) does not apply, when the phase exhibits a singularity.
Consider, for example, a closed loop C around the singularity. In a vortex, the
superfluid current is pulled by the phase gradient, v = m ~
∇θ. For the condensate
wavefunction to be well defined, the phase variation ∆θ around the loop must be a
multiple of 2π, that is, I
∆θ = ∇θ · dl = 2π` . (27.13)
C
1 This is not the case for strongly interacting superfluids, such as superfluid helium, where the
where ` is an integer. Therefore, we can calculate the circulation Γ around the closed
loop, that is, I
h
Γ= vs (r, t) · dl = ` . (27.14)
C m
Apparently, the superfluid circulation is quantized in units of ~/m. The parameter
` is called charge of the vortex and measures, in unit of h, the quantized angular
momentum of the vortex.
In addition, the flow must be persistent, because its circulation can only be
changed in a discontinuous way, by overcoming a discrete energy barrier, which re-
quires energy coming e.g. from thermal excitations. Clearly, the normal (thermal)
fraction of a gas can also have a circular flow. However, the disordered microscopic
motion of each individual particle causes a viscous drag that prevents the persistence
of the flow in the absence of a torque. This is in contrast to superfluid flow, which
persists even without external torque. The issues of vortex stability, formation, and
topology were addressed in recent experiments [479, 464, 465, 133, 15]. Solve the
Excs. 27.2.6.1 and 27.2.6.2.
The kinetic energy per unit vortex length can be estimated from a semi-classical
approach. First, we define as the mass density ρm of the superfluid. If n is the particle
density, ρm = nm. The kinetic energy of a flux line at the radius r is,
~2 `2 n
Ekin = 12 ρm vs2 = . (27.15)
2m r2
To obtain the kinetic energy per unit length, we integrate the expression (27.21) over
a plane perpendicular to the vortex axis. Note, however, that the velocity field is
vs ∝ r−1 and, therefore, can not be integrated from zero. Instead, we begin the
integration at a radius given by the healing length ξ, which represents a measure of
the vortex core size. Now, the kinetic energy per unit length is,
Z 2π Z R
~2 `2 R
Esemi = Ekin (r)rdrdθ = πn ln . (27.16)
0 ξ m ξ
Note, that a multiply charged vortex carrying the entire angular momentum ` = `0
of the superfluid is energetically less favorable than an ensemble of `0 vortices with
unit charge ` = 1. Therefore, a multiply charged vortex is unstable and may decay
to several single charge vortices.
To calculate exactly the energy of a vortex, we do the following ansatz,
Φ0 (r) = φ(r, z)ei`ϑ , (27.17)
we then solve the Gross-Piatevskii equation (26.21) numerically, and calculate the
expectation value of the energy of the vortex through the expression,
Z h 2 i
hΦ0 |Ĥ|Φ0 i = d3 r 2m~
|∇Φ0 (r, z)|2 + Vtrp |φ(r, z)|2 + g|φ(r, z)|4 , (27.18)
is a good approximation [559]. Note that the healing length ξ characterizes the size of
the vortex. For a superfluid, we can calculate the total energy of the vortex exactly,
because the fluid is confined in all directions,
Etot = 4πn 0 ~
3 m ln 0.671 Rr
ξ0 , (27.21)
where n0 and ξ0 are, respectively, the density and healing length at the center of the
fluid. Rz and Rr are, respectively, the extensions of the cloud along the axial and
radial directions within the Thomas-Fermi approximation.
27.2.1.2 Stability
In a topologically ’singly-connected’ trap, for example a potential harmonic, vortices
do not represent the lowest energy eigenstate, and they must decay to the ground
state. If the mean-field energy of the condensate is weak compared to the kinetic
energy, gn0 /~ωz 1, the healing length will exceed the size of the BEC, ξ atrp ,
and the vortex rapidly decays by dissipating its excess energy to thermal excitation.
Such BECs can not be considered superfluid. If the mean-field interaction is strong,
the vortex spontaneously breaks azimuthal symmetry, moves away from the center
and exits the condensate on a spiral-shaped trajectory [605]. Nevertheless, the decay
time may be quite long.
On the other hand, a vortex can be the ground state of a ’multiply-connected’
trap (for example, a toroidal potential). Such a potential can be realized as the
temporal average of a harmonic potential with a small rotating anisotropy [464]. In
such geometries vortices are extremely stable.
imbalances on the vortex surface. A radial force naturally arises, when the core is
displaced from the center, because local pressure gradients will force the vortex out
of the center toward regions with lower densities. Experimentally, a slow precession
(∼ 1 Hz) spiraling the vortex toward the rim of the condensate has been observed
[15] by taking a succession of nondestructive images.
These patterns of vortex lattices were observed in experiments using the stir-
ring method, which consists in brushing a far-detuned focused laser (’optical spoon’)
around the rim of a condensate [464] at a certain frequency Ω. In this experiment, the
ratio of the mean-field energy to the kinetic energy was gn0 /~ωz = a2trp /2ξ 2 > 100.
Beyond a certain critical rotation frequency, Ωc ≈ 2π × 150 Hz, a single central vortex
1018 CHAPTER 27. SUPERFLUID AND COHERENT PROPERTIES OF BOSE-EINSTEIN CO
27.2.3 Solitons
Work on solitons has been done by [769, 743, 479, 744, 518, 594, 372, 529, 118, 37,
224, 114, 196].
27.2.5 Turbulence
The issue of turbulence is one of the most important problems of classical physics
yet to be solved [245]. In superfluids, restrictions imposed by quantum mechanics
constrain the emergence of turbulence, which is then called quantum turbulence. On
the other hand, the study of quantum turbulence can improve our understanding of
classical turbulence [217]. Recently, the study of the dynamics of a ensemble of vortices
in a Bose-Einstein condensate allowed the identification of signatures of quantum
turbulence [339, 323].
27.2.6 Exercises
27.2.6.1 Ex: Comparison between the quantum and the classical vortex
Consider a vortex around a straight line along the z-axis at r = 0. Compare the
radial velocity variation of a quantum vortex with that of a classical one.
`h
∇ × vs (r, t) = ẑ δ(x)δ(y) .
m
3 The dynamics of atoms trapped in a harmonic potential is similar to the Jaynes-Cummings dy-
namics of an optical mode. However, the non-linear condensate self-interaction changes the situation
and makes the collisions being collective.
27.3. ATOM OPTICS 1021
The first vortex in a dilute gas was created at the JILA [480]. Study the paper and
explain in detail, how the vortex was created.
excitation. Moreover, in some interferometers, the de Broglie wave is not even split,
and one does interferometry with completely immobile atoms or ions [360].
The epitome of a coherent light source is the laser, and we may ask whether there
is a material analogue, which would be a source of coherent matter [749, 99, 669, 351,
526, 413] serving for coherent atom optics. In fact, we may already consider a trapped
condensate as a stationary atom laser pulse with the trapping potential playing the
role of the laser resonator. While most atom optical devices (including conventional
atomic interferometers) do not require mutual coherence of atoms, certain applications
take advantage of an intense, highly directional, monochromatic, and coherent atom
source. In this respect, atom lasers are much superior to thermal atomic beams.
While a thermal beam contains about 10−12 atoms per mode and a magneto-optical
trap about 10−6 , a condensed mode may contain more than 106 atoms. Condensates
offer the advantage of large de Broglie wave amplitudes and de Broglie wavelengths as
long as their actual size (which can be much longer than optical wavelengths). And for
an atom laser, the coherence length can even be longer than that of the condensate
from which it emerged. This obviously has a major impact on the sensitivity and
resolution of atom optical elements, in particular those, where atomic coherence is
important, for example, atom interferometers. Without doubt atom lasers will replace
conventional atomic beams in precision measurement of fundamental constants and
tests of fundamental symmetries ([Phys. World (mar,97) p.43]). Finally, atom lasers
are crucial for nonlinear optics.
Also of interest are the references [668, 519, 586, 436, 608, 43, 338, 221, 579, 195,
348, 595, 371, 630, 70, 209, 737].
The gain mechanism for optical lasers is photonic stimulation of atoms inciting them
to emit other photons into the stimulating mode. The atomic laser operates in a
similar way. Atoms trapped in a potential constitute a thermal reservoir. Binary
collisions redistribute the atoms among the vibrational energy levels of the potential.
If a vibrational level already contains an atomic population, Bose-Einstein quantum
4 Coherent reflection of atom laser beams has been demonstrated [92, 23, ?], and an atom laser
statistics will encourage atoms involved in a collision process to join this level. Ulti-
mately, this comes down to an irreversible pumping of a single level, where the atoms
accumulate to build a single degenerate quantum state. Bose condensation is always
the result of bosonic stimulation. However, the dynamics and time scale for the for-
mation of a condensate have been controversially discussed, until some experiments
[498] could directly visualize in vivo the process of nucleation and the exponential
amplification at the center of a thermal cloud (see Fig. 27.7).
0
0 100 200 300 400
(a) Time (ms) (b)
Figure 27.7: (a) Illustration of the idea of bosonic stimulation. (b) The curves show
the growth of a condensate toward thermal equilibrium for different initial numbers
of condensed atoms.
The very high concentration of population in a single phase space cell during the
creation of a condensate represents a dramatic reduction of entropy. Since the total
entropy can not decrease, the condensate must be coupled to a thermal reservoir
receiving the excess entropy. This coupling is necessarily irreversible. In practice, the
reservoir is the cloud of thermal atoms, whose energy is spread over many vibrational
states. Irreversibility is introduced by collisions and the subsequent removal of the
hottest atoms.
CCD
interfering BECs
optical pumping after 40ms TOF
beam
Figure 27.8: (a) Scheme for observation of matter wave interference. (b) Interference
fringes.
continuous and well-collimated atomic beam [313] and tunneling can give rise to a
pulsed mode-locked atom laser (mode-locked laser) [16].
A trapped condensate represents a finite reservoir from which an atom laser can
be fed. For a really cw atom laser, an incoherent pumping mechanism, continuously
feeding the BEC, while it delivers atoms to the atom laser, is still lacking [467].
Figure 27.9: Several types of atomic lasers realized, from left to right at the MIT in 1997,
in Munich in 1999, at Yale in 1998, and at NIST in 1999. The vertical sizes of the images
are, from left to right: 5, 2, 0.5, and 1 mm.
potentials, ϕ|±i (t) = µ|±i t. The accumulated phases are not directly observable, but
their difference, ∆ϕ(t), can be measured by Ramsey interferometry.
which in turn acts back onto the field. Higher-order processes, such as self-focusing,
second-harmonic generation, four-wave mixing, etc. are described by the non-linear
susceptibility χ(3) . These processes require the presence of a non-linear medium (the
vacuum polarization itself being too small [395, 303]).
The role of binary collisions in coherent matter wave optics, as described by the
nonlinear term in the Gross-Pitaevskii equation (26.21), is very similar to the role of
the third-order nonlinear susceptibility in quantum optics [443, 422, 194, 285, 711,
396, 325, 397, 445, 170, 327]. For example, if the atomic interaction is repulsive,
the nonlinear term tries to increase the size of the condensate as much as possible
within the limits imposed by the trapping potential. This behavior is analogous to
the phenomenon of self-defocusing known in non-linear optics.
Small-amplitude elementary oscillations are well described by the Bogolubov-de
Gennes equations, which are a linearized version of the Gross-Pitaevskii equation
[157]. On the other hand, large-amplitude oscillations, which are sensitive to the
nonlinear mean-field interaction [664], showed a splitting of the frequency for quasi-
particle excitation, in analogy with the generation of the second-harmonic (SHG) in
quantum optics [327].
Other phenomena, such as matter wave phase conjugation and four waves mixing
(4WM) [285] have been observed in experiments [194]. The three matter wave modes
for the nonlinear mixing were produced out of a single condensate by applying two
short consecutive sequences of Bragg diffraction pulses. The condensates are created
in the same spatial region, but at different times. The nonlinear mixing during the
process of spatial separation was observed by time-of-flight images.
The geometry of the laser beams is shown in Fig. 27.10(a) in the laboratory system.
A first standing wave light pulse is generated by lasers k1 and k2 detuned from each
other by an amount, such that the Bragg condition is satisfied and the momentum
p2 = ~k1 − ~k2 is transmitted to the diffracted atoms. Then, a second standing
wave light pulse formed by lasers k1 and k3 = −k1 transmits to the diffracted atoms
the momentum p3 = 2~k1 . The duration and intensity of the standing waves are
adjusted such as to create an approximately equal distribution of Nj0 atoms in all three
condensate momentum modes, each mode having a different momentum, p1 = 0, p2 ,
and p3 . Initially, the three wavepackets ψ̂j ≡ ψ̂0 (r)eipj r/~ overlap. The nonlinear
term in the Gross-Pitaevskii equation mixes the wavepackets while they separate to
27.3. ATOM OPTICS 1027
form other wavepackets, ψ̂4 ∼ g ψ̂j+ ψ̂m ψ̂n eip4 ·r/~ , where,
X3
N4 = −Nj + Nj0 = Nm − Nm 0
= Nn − Nn0 = (Nκ0 − Nκ ) (27.29)
κ=1
p4 = −pj + pm + pn
p24 = −p2j + p2m + p2n .
To produce the new momenta, p4 6= p1 , p2 , p3 , the corresponding wavepackets must
gather atoms from all three initial packets p1 6= p2 6= p3 .
The experimental observation of a newly emerging wavepacket ψ̂4 has two comple-
mentary interpretations: In the inertial system defined by p1 = −p3 [see Fig. 27.10(b)],
the wavepackets ψ̂1 and ψ̂3 suffer elastic collisions. The direction in which the con-
densates are scattered is, a priori, isotropic. The injection of a third condensate
ψ̂2 bosonically stimulates the formation of a forth one, ψ̂4 , in a predefined direction
p4 = −p2 . The laws of conservation for the particle numbers, momenta, and ener-
gies (27.29) only allow processes, which can be interpreted as degenerate 4WM in this
inertial system. Each of the wavepackets ψ̂1 and ψ̂3 sacrifice N4 atoms to create a
new wavepacket ψ̂4 and to amplify the wavepacket ψ̂2 . The redistribution of atoms
is a coherent process.
Figure 27.10: Matter wave 4WM can be illustrated (a) in the laboratory system, or (b) in
the inertial system defined by p1 = −p3 , or (c) in the system defined by p1 = −p2 . In each
system the process has a different physical interpretation (see text). The right image shows
the experimental result.
number states have been compared to coherent states. Superposition of the first ones exhibited no
interference while superposition of the second ones lead to observable interference.
6 If two Josephson-coupled BECs are initially in Fock-states, the current must initially be a quan-
27.4. QUANTUM ATOM OPTICS 1029
−~2 d2
Hψ = + v cos Kx + c|ψ|2 = µψ . (27.33)
2m dx2
We have a periodic potential, U (x) = U (x+X). In the following, we restrict ourselves
to 1D. Assume e.g. a sine wave potential,
∂2ψ
− 12 + (U cos x − U ) ψ + C|ψ|2 ψ = µψ . (27.35)
∂x2
The Bloch-theorem says that the Schrödinger equation be solved by any Bloch state.
Those are superpositions of plane-wave momentum states [?],
PN
Inserting the plane-wave expansions uk (x) = q=−N aq (k)eiqx into the Schrödinger
equation, we get,
N
X h
1
2
(q + k)2 aq eıqx + 21 U aq (eı(q+1)x + eı(q−1)x ) (27.38)
q=−N
N
#
X
+C am a∗n aq eı(m−n+q)x − [µ(k) + U ]aq e ıqx
=0
m,n=−N
XN
1
2
(q + k)2 aq + 21 U (aq−1 + aq+1 ) + Caq |an |2 − [µ(k) + U ]aq = 0 .
n=−N
0.6
0 1000
0.4
aq
-0.05 500
0.2
-0.1 0 0
-1 0 1 -1 0 1 -5 0 5
k k x
Figure 27.11: (code) Nonlinear interaction matrix finds just one minimum, dependence on
k, [752, 754].
and insert this into the Gross-Pitaevski equation Eq. (27.37), we find that the system
of differential equations Eq. (27.38) simplifies to,
!
~2 k 2 v
2m + c 2 + abc a
= µ
a
. (27.41)
2 2
v ~ (k−K) b b
2 + abc 2m +c
For −2ab = vc the off-diagonal elements vanish, and because the interacting BEC
shields the potential. Since |2ab| ≤ 1, this is only possible for c ≥ v.
27.4. QUANTUM ATOM OPTICS 1031
2 2 2 2
If we substitute µ = µ̃ + 12 ~2m
k
+ 12 ~ (k−K)
2m + 23 c, multiply the first row with b,
the second with a and add both rows, we obtain,
v
2µ̃ = . (27.42)
2ab
Example 119 (Estimations): Estimated radial BEC size rt ' 10 µm, esti-
mated number of atoms per antinode n1D ' 1000 µm−1 ,
~k2
ωr = = 2π · 3.8 kHz (27.43)
2m
4π~2 as
g1D = (27.44)
mrt2
g1D n1D = h · 77 Hz (27.45)
Since we need at least two Bloch bands to see interesting effects, we must arrange
for much stronger mean fields.
pearance and reappearance of coherence is reminiscent to the phenomenon of collapse and revival in
the Jaynes-Cummings model.
1032 CHAPTER 27. SUPERFLUID AND COHERENT PROPERTIES OF BOSE-EINSTEIN CO
The Bose-Hubbard model starts from the following Hamiltonian [546, 293],
X X X
Ĥ = −J â†i âj + εi n̂i + 12 U n̂i (n̂i − 1) , (27.46)
(i,j) i i
where Z
~ 2
J≡ d3 x w∗ (x − xi ) − 2m ∇ + V0 (x) w∗ (x − xi ) , (27.47)
and Z
4πas ~2
U≡ d3 x |w∗ (x)|4 , (27.48)
m
2 R
with εi ≡ 4πams ~ d3 x Vtrap (x)|w(x − xi )|2 ≈ Vtrap (x) and w(x) = w(x)w(y)w(z)
are the Wannier functions for an individual particle. We assume for the moment a
homogeneous condensate.
The Hilbert space dimension of the Bose-Hubbard model grows exponentially with
the number of atoms N and the number of sites L,
(Nb + L + 1)!
Db = . (27.49)
Nb !(L − 1)!
For the Fermi-Hubbard model, the Pauli exclusion principle leads to the Hilbert space
dimension,
L
Df = . (27.50)
Nf
In three dimensional lattices the Hilbert space grows even faster. Therefore, it is a
difficult computational task to model or simulate such systems, and generally not
possible fro more than 20 atoms and 20 lattice sites.
At zero temperature, the Bose-Hubbard model (in the absence of disorder) predicts
the atomic ensemble to be an a Mott insulating state (MI) when J U , a superfluid
state (SF) when J U , or a supersolid phase (SS), where both solid and superfluid
phases (diagonal and off-diagonal) coexist. Mott insulation phases are characterized
by integer site occupation numbers, by the existence of an energy gap for particle-
hole excitations, and zero compressibility. In the presence of disorder, a third phase,
the Bose glass exists. This phase is insulation because of the Anderson localization
effects. Bose glass is characterized by a finite compressibility, the absence of a gap,
and an infinite superfluid susceptibility [248]. See also (watch talk).
In the limit of strong tunneling and weak interactions, J U , the matter wave
function looks like a Bloch state,
M
!N
X
|ψSF i ∼ â†i |0i . (27.51)
i=1
√
The variance of the number of particles per site is Poissonian, σSF ∼ Ni , that is, the
wavefunction per lattice site is (quasi)-coherent. However, the total wavefunction is
delocalized over all lattice sites, the local wavefunctions have a rigid phase relationship
27.4. QUANTUM ATOM OPTICS 1033
27.4.4 Exercises
27.4.4.1 Ex: Superfluid to Mott insulator transition
Consider a condensate trapped in a 3D simple cubic optical lattice with lattice spacing
l. The Hamiltonian can be written as,
X (1) † X † †
Ĥ = −J δij âi âj + U2 âi âi âi âi ,
i,j i
where âi (â†i ) is the operator annihilating (creating) one boson in the lattice site i;
(1)
δij = 1 only when i and j are nearest neighbors, otherwise it is zero. Assume a total
of M 1 sites, periodic boundary conditions in the x, y, and z directions, and U > 0.
On this lattice we consider the construction operator b̂†s defined in the following way:
X †
b̂†s = √1M âi .
i
where SF stands for a name label (and not for quantum numbers). Show that for
N = M we have âα |N, SF i = |N − 1, SF i.
e. Define the density operator for the site i as n̂i = â†i âi . Calculate the density
fluctuation, q
∆ni = hn̂2i i − hn̂i i2
of the state |N, SF i with N = M .
f. Calculate the total energy hHi with the state |N, SF i.
g. Next consider the state, Y †
|N, M Ii ≡ âj |0i .
j
Calculate hâ†i âj i and the density fluctuations ∆ni for this state.
h. Calculate the total energy hHi with the state |N, M Ii.
i. Discuss the energy for the states |N, SF i and |N, M Ii. Which one is the ground
state? How to make a ground state change from |N, SF i to |N, M Ii?
V0 (Er)
again, and after 14 ms of ramp-down time the interference peaks
9
have narrowed to their steady-state value, proving that phase τ = 20 ms
coherence has been restored over the entire lattice. The timescale
for the restoration of coherence is comparable to the tunnelling time 0
8
ttunnel ~=J between two neighbouring lattice sites in the system,
b 125
z
75
x y
2 ᐜk 50
25
0
0 2 4
c d
Figure 1 Schematic three-dimensional interference pattern with measured absorption
images taken along two orthogonal directions. The absorption images were obtained after
ballistic expansion from a lattice with a potential depth of V 0 10E r and a time of ¯ight of
15 ms.
Interaction of Bose-Einstein
condensates with light
Light can essentially be used in two ways for the manipulation of matter waves:
1. Tuned far from resonance, light serves to create conservative optical dipole poten-
tials (see Secs. 24.3.2 and 27.4.2). In such circumstances the states of internal atomic
excitation can be adiabatically eliminated from the description of the center-of-mass
dynamics. 2. Close to resonance the situation is quite different, but even then, the
coherent excitation of internal states may allow for a controlled manipulation of the
atomic motion. Among the examples discussed in the following sections are the adia-
batic sweeps, the Raman output coupler, Bragg pulses, photoassociation, and matter
wave superradiance (see Fig. 28.1).
Figure 28.1: Different types of coherent coupling: (a) coherent coupling of hyperfine levels,
(b) coupling from a confined state to the continuum of free states, (c) mutual coupling of
different velocity states of an atom, and (d) coupling between a (collisional) state of two free
atoms with a bound molecular state.
1037
1038 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
Finally,
X X
ρ̂† (q) = â†k+q âk = |k + qihk| . (28.3)
k k
Hence, ρ̂† (q) describes the scattering of an atom with the momentum k to k + q. We
also find,
ρ̂† (q) = ρ̂(−q) . (28.4)
Understanding the fundamental state |gi as the state without excitations, we define
the static structure factor (19.5) now normalized to the number of particles [671],
1 †
S(q) ≡ N hg|ρ̂(q)ρ̂ (q)|gi , (28.5)
as a generalization of the classical structure factor. The static structure factor de-
scribes the probability to excite a condensate by creating a quasi-particle with mo-
mentum k. We will need these notions in the Sec. 28.2.3. R
The dynamic structure factor is obtained from S(q) = S(q, ω)dω and measures
the density of correlations in the ground state with 0 momentum. The formula de-
scribes, how an atom probes its environment by scattering quasi-particles back and
forth, X
S(q, ω) = N1 hg|ρ̂(q)|f ihf |ρ̂(q)|gi~δ(~ω − Ef + Eg ) . (28.6)
f
Thus, we can implement the second quantization by simply replacing eıq·r by ρ̂† (q).
Now, the Hamiltonian for the process of a photon absorption from the mode ĉki ,
taking into account the photonic recoil, was introduced in (18.6),
Going to second quantization, as done in (28.9), and doing the Fourier transform 1 ,
X †
H̃int = C ĉkf ĉki â†pf âpi δkf −ki +pf −pi . (28.12)
pf ,pi
We assume here, that the light modes ki and kf are predefined, but the velocities
pi are distributed. Otherwise, in the expression (28.12), we must also sum over light
modes 2 ,
X
H̃int = C ĉ†kf ĉki â†pf âpi δkf −ki +pf −pi , (28.13)
ki ,kf ,pi ,pf
momentum modes. We will discuss this general concept of 4WM in Secs. 28.3.4 and 27.3.4.
1040 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
~Γ2 ~Γ2 I
Vmod = I cos(q · r − ωt) = (eıq·r−ıωt + e−ıq·r+ıωt ) (28.16)
8∆ 8∆ 2Isat
V
→ V̂mod = [ρ̂† (q)e−ıωt + ρ̂† (−q)eıωt ] .
2
Now, the transition rate is [672],
2 X
W 2π V
= 2πΩ2R S(q, ω) = |hf |ρ̂† (q)|gi|2 δ(~ω − (Ef − Eg )) , (28.17)
N N~ 2
f
with V
2
= ~ΩR .
Assuming that all modes are not degenerate, such that [âpf , â†pi ] = δpf ,pi , that is,
â†pf and âpf only act on the mode |npf iat , etc.,
X X√
âp0i |...npi ...iat = npi |...npi − 1...iat (28.21)
p0i pi
X Xp
and â†p0 |...npf ...iat = npf + 1|...npf + 1...iat ,
f
p0f pf
28.1. SCATTERING OF LIGHT BY DEGENERATE GASES 1041
as well as,
X Xp
ĉk0i |...Nki ...irad = Nki |...Nki − 1...irad (28.22)
k0i ki
X Xq
and ĉ†k0 |...Nkf ...irad = Nkf + 1|...Nkf + 1...irad .
f
k0f kf
Let us now assume for a moment, that the photon is scattered to the vacuum,
that is, the final mode of light is initially empty, Nkf = 0 4 ,
hf |Ĥint |iiinel (28.23)
* +
...np00i ...np00f ... X p q √
...n − 1...n + 1...
pi pf
∝ Nki npf + 1 npi
...Nk00i ...Nk00f ... p ,p ,k ,k ...Nki − 1...1kf ...
f i f i
X p q √
= Nki npf + 1 npi δNk00 ,Nki −1 δNk00 ,Nkf +1 δnp00 ,npi −1 δnp00 ,npf +1
i f i f
pf ,pi ,kf ,ki
p q √
= Nki npf + 1 npi .
depends, in addition to the numbers of provided photons Nki and atoms npi in the
initial mode, also on the number of atoms in the final mode npf . This amplification
of the probability of the scattering process is called bosonic stimulation.
Now, we consider the degenerate case, where the initial atomic momentum mode
coincides with the final mode, npf = npi . In this case,
X † X † X
âp0 âp0i |...npi ...npf ...iat = âp0 âp0i |...npi ...npi ...iat = npi |...npi ...npi ...iat ,
f i
p0f =p0i p0i pi
(28.25)
and a calculus analogous to (28.23) yields,
* +
...np00i ...np00f ... X p ...n ...n ...
pi pi
hf |Ĥint |iiel ∝ Nki npi (28.26)
...Nk00 ...Nk00 ... ...Nki − 1...1kf ...
i f pi ,kf ,ki
X p
= Nki npi rad h...Nk00 ...Nk00 ...|...Nki − 1...1kf ...irad at h...np00 ...np00 ...|...npi ...npi ...iat
i f i f
pi ,kf ,ki
X p p
= Nki npi δNk00 ,Nki −1 δNk00 ,Nkf +1 δnp00 ,npi δnp00 ,npi = Nki npi .
i f i f
pi ,kf ,ki
photons already present in the final mode prior to the scattering process.
1042 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
only depends on the number of provided photons Nki and the number of atoms npi
in the initial mode.
The first term of (28.28) occurs when the momentum of the scattering atom does
not change, p = q, that is, when the populations of the momentum states np and nq
do not change. The corresponding term (28.27) describes elastic Rayleigh scattering.
This process is coherent, that is, the phase relationship between the incident wave
and the outgoing wave is fixed, because the photon emission is self-stimulated. I.e. it
decays to the original mode via forward scattering within the angle defined by the
phase matching condition ϑ < λ/d, where d is the size of the atomic sample [403].
This contribution is dispersive, reversible, and conservative, and it is at the origin of
the dipole force.
The second term of (28.28) is the inelastic part of Rayleigh scattering, where an
atom with the initial momentum np is scattered to the momentum state nq . This
term is absorptive, dissipative, and spontaneous. The frequency of the photons is
shifted, a momentum is imparted to the atom, such that the process is incoherent.
Hence, a suggestive way of expressing the differential scattering cross-section (28.19)
is, 2
dσ X X
∝ ni hi|Ĥint |ii + ni (1 + nf )|hi|Ĥint |f i|2 , (28.29)
dΩ i
i6=f
indicates some additional quantum number (for example, the vibrational state of a
loose trapping potential) helping us to enumerate fermionic states, and q indicates
the momentum (considered independent from the trapping potential for short enough
times). The operators respect the rules,
where upper signs hold for bosons and lower signs for fermions. Initially the N atoms
are in the bosonic, respectively, fermionic many-body state [516],
N
Y −1
(N ) † (N )
|Ψb (0)i = √1 â (0)N |0i
N! 0
resp. |Ψf (0)i = â†n (0)|0i . (28.32)
n=0
(N ) (N ) (N )
Note that |Ψb (0)i = |N i is a Fock state satisfying hΨb (0)|Ψb (0)i = 1, while
(N ) (N ) (N )
|Ψf (0)i is a product state being normalized as well, hΨf (0)|Ψf (0)i = 1.
Probing the number of atoms in a many-body state is done by,
Z X
N = hΨ|N̂ |Ψi where N̂ = ψ̂ † (r)ψ̂(r)d3 r = â†m (k)âm (k) , (28.33)
m,k
We will see in 28.1.6.9 that product states are normalized if the partial states are
normalized. Explicitly,
(N1 ) (N2 )1
|Ψb (0)i|Ψb ↠(0)N1 â†0 (q)N2 |0i
(q)i = √ (28.36)
N1 !N2 ! 0
1 −1
NY NY 2 −1
(N1 ) (N2 ) †
|Ψf (0)i|Ψf (q)i = ân (0) â†n (q)|0i .
n=0 n=0
Note, that product states in the same mode need renormalization when merged,
s
(N1 ) (N2 ) 1 † N1 +N2 N1 + N2 (N +N )
|Ψb (0)Ψb (0)i = √ â (0) |0i = |Ψb 1 2 (0)i ,
N1 !N2 ! 0 N1
(28.37)
(N ) (1) √ (N +1)
for example, |Ψb (0)Ψb (0)i = N + 1|Ψb (0)i.
A π/2-Bragg pulse has the faculty to transfer 50% of the atoms into a momen-
tum mode q, thus creating a new state where every single atom lives in a coherent
1044 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
Figure 28.2: Cartoon of (a,b) bosonic stimulation and (c,d) cooperative enhancement.
N
!
(brgg,N ) [â†0 (0) + â†0 (q)]N X N â†0 (0)N â†0 (q)N −n
|Ψb (0, q)i = √ |0i = √ |0i
2 N!
N
n=0
n 2N N !
. (28.38)
N −1
(brgg,N )
Y [â†n (0) + â†n (q)]
|Ψf (0, q)i = √ |0i
n=0 2N
In the following, we will study scattering processes, that is, fluctuations, in the
presence of the states (28.35) or (28.38).
which in fact is nothing else than the static structure factor. Applying this process
to bosonic or fermionic clouds, as defined in (28.32), we find bosonic stimulation only
in the case of bosons,
√
(N1 −1) (N2 ) (N1 ) (N2 −1) N1 N2 for bosons
hΨ (0)Ψ (q)|Ĥ|Ψ (0)Ψ (q)i = δk,q ,
1 for fermions
(28.40)
as will be shown in Exc. 28.1.6.10(a). The transition probabilities are then,
dσ 2 N1 N2 for bosons
= |hf |Ĥ|ii| = δk,q . (28.41)
dΩ 1 for fermions
Note, that these results does not change in the presence of a Bragg grating not
participating in the dynamics:
(M −1) (brgg,N +1) (M ) (brgg,N )
hΨb (0)Ψb (0, q)|Ĥ|Ψb (0)Ψb (0, q)i = 1 . (28.42)
28.1. SCATTERING OF LIGHT BY DEGENERATE GASES 1045
However, applying the same scattering process between a BEC with M atoms and
the Bragg state defined in (28.38) consisting of N atoms, we identify two interesting
possibilities for which we find,
where |n0 , nk , n−k i denotes the state with n0 atoms in the trap’s ground state and n±k
atoms in the momentum mode k, where vk2 is the average population of momentum
28.1. SCATTERING OF LIGHT BY DEGENERATE GASES 1047
28.1.6 Exercises
28.1.6.1 Ex: Sum rules for the dynamic structure factor
Derive the following sum rules for the norm, the kinetic energy, and the compressibil-
ity,
Z
S(q) = S(q, ω)dω
Z
~2 q 2
= ~ωS(q, ω)dω
2m
Z
κ2 S(q, ω)
= dω .
2 ~ω q→0
Exciting the Bragg resonance for Li, the Rb cloud would stay unaffected,the such that
around ω/2π = 295 kHz, S87,87 (q, ω) = 0, as well as, S6,87 (q, ω) = S87,6 (q, ω) = 0.
For Fermi gas S6,6 (q, δ) → 1, except if Bose gas induces Li-Li correlations, as in the
case of BCS-pairing of phonon-mediated Efimov states.
How does Rb influence the Bragg scattering of Li? Via a variation of the apparent
1048 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
âm (k)â†n (q)N = N δm,n δk,q â†n (q)N −1 + â†n (q)N âm (k) , (28.47)
b. For fermions do not support macroscopic populations of the type â†n (q)N . Per-
forming a number P of permutations of the operators and assuming that one fermion
of the product state is in the internal state n, we calculate,
N
Y −1 N
Y −1 N
Y −1
ân (k) â†m (q) = (−1)P δk,q â†m (q) − â†m (q)ân (k) . (28.49)
m=0 m6=n m=0
If none of the fermions of the product state is in state n the first part of the sum
simply vanishes. This expression can also be generalized to,
N
Y −1 N
Y −1
ân (k) [â†m (q1 ) + â†m (q2 )] = (−1)P [â†m (q1 ) + â†m (q2 )]ân (k)[â†n (q1 ) + â†n (q2 )]
m=0 m6=n
N
Y −1 N
Y −1
= (−1)P (δk,q1 + δk,q2 ) [â†m (q1 ) + â†m (q2 )] − [â†m (q1 ) + â†m (q2 )]ân (k) .
m6=n m=0
(28.50)
28.2. BRAGG DIFFRACTION 1049
Calculate the probability that the atom is scattered at a particular wavevector, i.e. de-
termine the cooperative enhancement factor.
d. Is cooperative enhancement possible with just one atom?
e. How about cooperative enhancement in a cavity, when cooperativity is ensured by
a single atom plus all its mirror images?
ulators.
1050 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
matter wave beamsplitters [422] and atomic lasers, or for the targeted excitation of
quasi-particles [422, 681, 671].
To implement Bragg diffraction, we consider two laser pulses with different fre-
quencies and propagation directions ω, kω and ω − ∆ω, kω−∆ω , detuned from an
atomic resonance and intersecting at the position of atoms under an angle ϑ, as
shown in Fig. 28.4(a). The superposition of the electric fields of the light beams,
E = E0 eı(kω ·r−ωt) + E0 eı[kω−∆ω ·r−(ω−∆ω)t] = E0 eı(kω ·r−ωt) 1 + e−ı(q·r−∆ωt)
ı
= 2E0 eı(kω ·r−ωt) e− 2 (q·r−∆ωt) cos q·r−∆ωt
2 , (28.51)
where q ≡ kω −kω−∆ω , produces a standing light wave with an intensity proportional
to,
|E|2 = 4E02 cos2 q·r−∆ωt
2 = 2E02 [1 + cos(q · r − ∆ωt) , (28.52)
with which the atoms interact.
The Bragg diffraction technique has proven extremely efficient: up to 100% of
the atoms can be transferred to well-defined momentum sidemodes. In general, the
components of a cloud with different momentum modes overlap during the time scale
of the Bragg pulses. They only separate spatially after a ballistic flight time, which
then allows their identification via absorption imaging.
Figure 28.4: Bragg scattering of atoms at a standing light wave. (a) Geometric layout: Short
pulses of two laser beams in Raman configuration enclosing an angle of ϑ and detuned from
each other by ∆ω are simultaneously irradiated into the cloud. (b) Bogolubov dispersion
relation for interacting condensates (blue line). The quadratic dispersion relation (cyan)
holds for free particles, and the linear dispersion relation (magenta) for phonon excitations.
pulsed laser beams 6 . We consider atoms initially at rest exposed to the standing
wave (28.51) with ∆ω = 0 and q ≡ qêz , such that the dipolar potential U ∝ |E|2 is,
with (28.52),
U (z) = U0 cos2 qz
2 . (28.53)
Assuming a given finite interaction time τ , the modulation of the local phase of the
atoms becomes 7 ,
ı
X
ψτ (z) = ψ0 (z)eiU (z)τ /~ = ψ0 (z)e 2~ U0 τ (cos qz+1) = ψ0 (z)eiU0 τ /2~ in Jn ( U2~
0τ
)einqz ,
n
(28.54)
where Jn are the Bessel functions of the first kind 8 .
The condensed wavefunction evolves into a superposition of motional modes, which
correspond to the diffraction orders of Bragg scattering and their amplitudes through
the Bessel functions Jn . The diffraction efficiency increases with laser intensity and
with time.
Now, we need to generalize this result to the propagating standing wave of the
expression (28.52). The intensity generates a dipole potential,
qz z−∆ωt
U (z, t) ≈ U0 cos2 2 , (28.55)
so that the condensate now evolves according to,
R t0 +τ R t0 +τ qz z−∆ωt
U0 cos2
ψ(z, t) = ψ0 (z)eı/~ t0 U (z,t)dt
= ψ0 (z)eı/~ t0 2 dt
. (28.56)
For short interaction times, τ 2π/∆ω, ...
diffraction is a coherent and thus reversible process, atoms initially in the recoil mode are transferred
back to the original matter wave mode.
7 Using the Jacobi-Anger expansion, eıβ cos x =
P n ınx .
n i Jn (β)e
8 Andreas got ψ (z)
P 2 inqz/2
0 n Jn (U0 τ /2~)e .
1052 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
Figure 28.5: Bragg scattering of atomic clouds. (a) Geometric layout as in 28.4. Here, we
assume pi = 0. (b) Experiment showing coherent splitting by a Bragg pulse.
Their is an obvious fundamental interest to study such states and the mechanisms
leading to their decoherence (see Sec. 16.1.1).
To clarify the situation, we first have to say, what we mean by Schrödinger cat.
A Schrödinger cat is a quantum superposition of many-body states. A perfect cat
made of N two-level atoms can be expressed as | + +...i ± | − −...i. The Einstein-
Podolsky-Rosen (EPR) and the Greenberger-Horne-Zeilinger states (GHz) belong to
this category. The degree of entanglement of the cat is measured by the information
entropy defined as S = −hlog2 ρ̂i, where ρ̂ is the density operator. The information
entropy measures the amount of classical information that can be encoded in the
quantum state. For example, the entropy of a perfect cat state is S = 1 bit, because
if we find one of the atoms in the state |+i, we know that all others are in the same
state.
Second, we have to explain, what we mean by a beamsplitter. Subject to a beam
splitting process, every individual atom has the choice between one of two output
ports. However, if the process is coherent, every atom will evolve into a coherent su-
perposition, but it does so independently from the other atoms. That is, we can write
the state of the atomic cloud as a product state of Schrödinger kittens (|+i ± |−i)N ,
but no real cat. The information entropy is S = N bit, as for independent atoms.
Clearly, the state (28.60) is NOT the state generated by a beamsplitter! Nevertheless,
the beamsplitter creates a certain correlation between the two modes (see Sec.15.6).
It may come as a surprise, that a condensate composed of totally delocalized
atoms coherently interacting with a homogeneous light field (the photons of the Bragg
beams are totally delocalized over the field mode including the condensate volume)
increases its entropy from 1 to N bits. This argument, however, would be the same
for a non-condensed thermal cloud being sufficiently cold that the thermal de Broglie
wavelengths exceeds the size of the cloud. That is, even in condensates the photons
are scattered by individual atoms. The interaction with the light localizes one atom
in the condensate before removing it from there by recoil. Nevertheless, cooperative
interaction of several atoms with a light mode is possible, e.g. in superradiance or when
the photons are recycled by means of an optical cavity, as in the Jaynes-Cummings
model.
using ktherm ≡ 2π/λtherm and (24.9). Because of the inhomogeneity of the distri-
bution, any evolution of atoms belonging to specific momentum classes caused by
velocity-selective radiation pulses, must be calculated with individual atoms. The
final momentum distribution (e.g. after a pulse sequence) can then be obtained by
weighing with the individual evolution with the initial momentum distribution 9 .
9 The procedure neglects interatomic interactions, which is always a good presumption, for exam-
ple, for an ultracold Fermi gas [188] or for 88 Sr atoms, which have a very small scattering length.
28.2. BRAGG DIFFRACTION 1055
Figure 28.7: (a) Geometry for Bragg diffraction. (b) Illustration of a Raman transition be-
tween two points of the free-particle dispersion curve, when the atoms are initially thermally
distributed. (c) Same as (a), but now the detuning ∆ω is adjusted for second-order Bragg
diffraction.
We describe the quantum state of a thermal atom |ψi as a plane wave, hz|ψi ∝
eıkz z , which in momentum space corresponds to a Dirac distribution, hkz |ψi ∝ δ(z −
z0 ). Without radiative coupling, we describe the evolution in momentum space by
the solution of the Schrödinger equation (or propagator),
where,
Ω1 Ω2 1 3λ2 Γ p 3πc2 I Γ
ΩR = = I1 I2 = (28.66)
2∆a ∆a 4π ~ω ~ω 3 ∆a
is the two-photon Rabi frequency. For very narrow momentum distributions, T
Trec , we may assume the atomic cloud to occupy only a single momentum state,
Ṅbrg ∝ ΩR D(kz ) . (28.67)
The transition rate per atom is then given by [672],
W 2π~ 2 X
= Ω |hf |ρ̂† (k)|gi|2 δ(~ν − Ef + Eg ) ≡ 2πΩ2R S(k, ν) . (28.68)
N N R
f
where ∆ = 2~q 2 /m − δ is the detuning of the Bragg lasers from the recoil shift. When
the Bragg light is switched off, the Hamiltonian simplifies to,
!
~2 2
k z 0
Ĥf ree = 2m ~2 02
. (28.73)
0 2m kz − δ
Concatenating temporal evolutions described by e−ıĤR t/~ and e−ıĤf ree t/~ , the phase
evolution of individual atoms in momentum state superpositions can be calculated
for arbitrary sequences of pulses separated by intervals of free evolution, for example,
Ramsey-type sequences.
28.2. BRAGG DIFFRACTION 1057
(a) 3
×10−8 (b) 1
1
#atoms
#atoms
2
2ρ22 − 1
0.5 0 (c)
1
-1
0 0 -1 0
0
-1 0 1 2 3 0 0.5 1 1 1
kz /q ωτ /2π Im ρ12
Re ρ12
Figure 28.8: (code) (a) Distribution of momentum classes in the direction of kz after irradia-
tion of a 2π Bragg pulse. The width of the structure is determined by the power broadening
of the Raman transition. (b) Temporal evolution of the populations of the momentum states
0 and 2q.
Beginning with the positions z and initial momenta kz , after a diffraction pulse trans-
ferring the recoil 2q to part of the atoms and being short enough not to change their
positions, the atoms follow classical trajectories. The atomic momenta (now depend-
ing on time) are simply the solution of the equation of motion ~k̇z = −mωz2 z with
the initial conditions z(0) = z and kz (0) = kz , respectively, kz0 (0) = kz + 2q,
mωz
k̃z (t) = kz cos ωz t − z sin ωz t (28.75)
~
mωz
k̃z0 (t) = (kz + 2q) cos ωz t − z sin ωz t ,
~
where the first expression holds for non-diffracted atoms and the second for diffracted
atoms. These momenta are substituted for kz and kz0 , respectively, in the Hamilto-
nians (28.72) and (28.73). As the Bragg pulses are short compared to the oscillation
period, τ 2π/ωz , the effect of the trapping potential may be neglected during
Bragg scattering. This means that the Hamiltonian ĤR depends on k̃z (t), but can be
1058 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
Since the amplitudes aj,z,kz now also depend on the initial atomic positions, the
final populations of the momentum states must be additionally weighted with the
initial spatial distribution. Therefore, the Eq. (28.64) for the expected the numbers
of diffracted and non-diffracted atoms must be replaced by,
ZZ
2
Nj (t) = D(kz )G(z) |aj,z,kz (t)| dkz dz . (28.77)
In Exc. 28.2.4.1 we will study the behavior of an ultracold atomic cloud subject
to a gravitational potential during a Ramsey-Bordé sequence.
Figure 28.9: (code) (a) Distribution of momentum classes in the direction of kz after irra-
diation of a π Bragg pulse. (b,c) Temporal evolution of the populations of the momentum
states nq.
all over the atomic cloud. At some point, however, the Rabi frequency becomes
comparable to the energy difference between adjacent momentum states, and Kapitza-
Dirac scattering sets in. According to [78] the scattering will stay two-state like as
long as the Rabi frequency fulfills the condition,
~
ΩR (4q 2 − 2qσkz ) . (28.84)
m
In the Kapitza-Dirac regime, the large energy uncertainty, connected with the
fast coupling rate, allows several momentum states to be coupled simultaneously. In
particular the Bragg Hamiltonian is replaced by,
.. ..
. .
.
.. ~(kz −2q)2
+δ ΩR
2m 2
ΩR ~kz2 ΩR
Ĥτ = . (28.85)
2 2m 2
~(kz +2q)2 ..
ΩR
−δ .
2 2m
.. ..
. .
Higher-order Bragg scattering is possible as well. Then the 2n-photon Rabi frequency
must be calculated and inserted in the second diagonal of (28.85).
28.2. BRAGG DIFFRACTION 1061
with Vx ≡ V eikx x . With this weak perturbation applied to the atoms, we make the
ansatz,
ψ(r, t) = e−ıµt/~ ψ0 (r, t) + u(t)eıkz z−ıωt − v(t)e−ıkz z+ıωt , (28.88)
where the amplitudes of the perturbation, u(t) and v(t), only vary slowly in time. As-
suming that ψ0 is fairly homogeneous, i.e. hψ0 |∇z |ψ0 i ' 0, the momentum transferred
to the BEC is,
hψ(r, t)| − ı~∇z |ψ(r, t)i = ~kz (|u|2 − |v|2 ) . (28.89)
We can now solve the Gross-Pitaevskii equation (26.20) in the same way as in
Sec. ?? inserting the ansatz (28.88). The terms proportional to eıkz z−ıωt are,
~2 k2z V 2 2 2
2 2 2 ∂
u+ ψ0 +g |ψ0 | + |u| + |v| u+gψ0 v+g|v| u+g|ψ0 | u = ı~ + ~ω + µ u ,
2m 2 ∂t
(28.90)
and the analogously for e−ıkz z+ıωt . Using µ = gn = g|ψ0 |2 and assuming that the
perturbation is weak, |u|, |v| |ψ0 |, we obtain,
2 2
~ k V ∂
+ gn u + ψ0 − gnv = ı~ + ~ω u (28.91)
2m 2 ∂t
2 2
~ k V ∂
− + gn v − ψ0 + gnu = −ı~ + ~ω v .
2m 2 ∂t
This effect is called Bragg diffraction of atoms by a standing light wave.
4
Δε/h (kHz)
Δε/h (kHz)
S(q)
2 0.5
0
−4
0 0
0 2 4 6 0 2 4 6 0 2 4 6
n0 (cm−3) n0 (cm−3) n0 (cm−3)
~2 k2 ∂
Using L ≡ 2m + Vtrap + 2gn − µ − ı~ ∂t , we can write,
L − ~ωk −gn u
=0. (28.92)
−gn L + ~ωk v∗
28.2.4 Exercises
28.2.4.1 Ex: Bragg diffraction in a Ramsey-Bordé interferometer
Simulate the behavior of a cold atomic cloud subject to a gravitational potential
during a Ramsey-Bordé sequence using the formalism developed in Sec. 28.2.2.
is random (except when bosonically stimulated). The atomic sample evolves, during
this time, to a state of coherent superposition, until all the atoms are deexcited. This
phenomenon is called Dicke superradiance. A similar effect exists for matter waves,
and will be discussed in the following sections [365].
Figure 28.12: Superradiance assumes small interatomic distances compared to the wave-
length of the excited dipole. Toroidal interference patterns interfering constructively in only
one direction.
and receives, in turn, the scattering atoms. In other words, the bosonic stimulation of
the scattering process by the Nj atoms already populating the recoil mode amplifies
the photon scattering rate by a factor of Nj + 1. The differential optical cross section
(power Pj scattered into the direction Ωj divided by the laser intensity I) is [365],
dσ Pj Ωj sin2 ϑj
= = σN0 (Nj + 1) . (28.94)
dΩj I 8π/3
Each scattering event of a photon into the mode Ωj transfers an atom to the recoil
mode Nj :
I Ωj sin2 ϑj
Ṅj = Pj /~ω ≡ Gj Nj where Gj = σ N0 . (28.95)
~ω 8π/3
I sin2 ϑj
Gj = σ n0 λ2 lj . (28.96)
~ω 8π/3
With the atomic density n0 = N0 /(Aj lj ), the factor n0 λ2 lj describes the optical
density of the BEC in the direction of the scattering. Therefore, we obtain exponential
gain for the atom number Nj , that is, the process is self-amplifying.
Using the terms of the dressed atom picture, we would say that the excited state
of our system is formed by the BEC and the laser mode. This state relaxes to a
state formed by the recoiled atoms and the scattered photons. The exponential gain
occurs, when this system exhibits inversion. The inversion is maintained, until the
BEC is completely transferred to the momentum recoil modes 11 .
waves in a four-wave mixing process (4WM) interpreted as scattering of light from a wave matter
stimulated by a wave matter mode. The process does not require nonlinearity due to collisions, but
it works with an ideal gas, g → 0. This distinguishes it from the 4WM of four matter waves, studied
in Sec. 27.3.4 and interpreted as a scattering of atoms by a matter wave stimulated by matter wave
mode, and 4WM in quantum optics involving four photons. But in all cases bosonic stimulation is
crucial.
28.3. MATTER WAVE SUPERRADIANCE 1065
It requires a sufficiently long coherence time for the atomic dipoles. On the other
hand, MWSR is a coherence effect between two states of the atomic center-of-mass
motion, i.e. the condensed state and a momentum sidemode. A long lifetime of this
coherence is necessary to allow for correlations between successive scattering events.
The coherence lifetime corresponds to a coherence length, which for a condensate
is equal to its physical size, whereas for a thermal cloud, it only corresponds to its
thermal Broglie wavelength. This explains why it is difficult to observe MWSR with
non-condensed atomic clouds.
Since the scattered photons quickly leave the BEC, there is no feedback or bosonic
stimulation by the optical output mode. This can be interpreted as a fast decaying
cavity mode, and the regime is called the bad cavity limit. Let us now imagine, that
the photons were recycled, for example, via an optical cavity constructed around the
BEC and reflecting back the photons emitted in the solid angle Ωj 12 . The mirrors of
the cavity create reflection images of the radiating atoms, which increase the density
of states, the scattering rate, and the gain Gj increase by a factor of 8F/π. An
interesting question now is, what happens in the limit F → ∞. If the decay of the
cavity mode, and therefore the removal of photons from the coherent interaction zone
(which is precisely the volume occupied by the BEC) is slow, the MWSR process
seems to be doubly stimulated, optically and atomically. However, in this case we
also expect the inverse process, where photons are scattered back to the original mode,
to occur and to be stimulated by the number of atoms N0 in the condensate. We see
that the simple picture of bosonic stimulation does not work in this case, and a more
complete model including the possibility of Rabi oscillations amplified by stimulated
emission is necessary 13 .
One of the interferometer arms consisted of the amplified seed condensate and the
other of a reference condensate (local oscillator) especially created by Bragg diffrac-
tion from the mother condensate. The observation of interference between these two
matter waves proved the coherence of the amplification process and the existence of a
well-defined phase relationship between the input and the output of the matter wave
amplifier.
Figure 28.15: Matter wave amplification of a ’seed condensate’ at the expense of a ’mother
condensate’. Time-of-flight imaging visualizes the condensate’s momentum distribution.
Figure 28.16: Schematic comparison of 4WM processes between light and matter wave
modes. (a) Spontaneous emission, (b) classical 4WM of light modes, (c) Bragg scattering,
(d) superradiant Rayleigh scattering, respectively, matter wave amplification, (e) optical
amplification, and (f) 4WM of matter waves.
the finite volume of the cloud inside the potential (Zeeman shift in magnetic traps or dynamic Stark
shift optical traps) or effects due to interatomic interactions. But we have to keep in mind, that
interatomic interaction can have a big impact on the dynamics, because the chemical potential of
each condensate depends on its number of atoms. Thus, in view of the inhomogeneity of the trapping
potential, the transfer of atoms can excite oscillations and sound waves in the condensates.
1068 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
a molecular and a state, where the same two atoms are free and involved in an elastic
collision.
Finally, in Sec. 28.4.3 we will show how to harness the forces induced by atom-light
coupling in order to design new geometries of trapping potentials.
where we allow for the fact that the trapping potentials can be different for the
two excitation states. We assume that atoms in different states of excitation are
distinguishable, such that their respective wavefunctions commute:
(28.100)
The semi-classical Hamiltonian (the light not being quantized) of individual particles
is, in the rotating wave approximation,
ˆ ˆ ˆ
Ĥf ield = ~ω↠â and Ĥint = −d̂ · E~ = −|2id̂+ · E~+ h1| − |1id̂− · E~− h2| . (28.102)
28.4. CONDENSATES IN ELECTRONICALLY EXCITED STATES 1069
For a condensate we still need to second-quantize the interaction part of the Hamil-
tonian which, disregarding collisions gives 15 ,
Z
ˆ ˆ
H = Ĥf ield + Hatom − d̂+ · E~+ ψ̂2† (r)ψ̂1 (r) + d̂− · E~− ψ̂1† (r)ψ̂2 (r) d3 r .
(28.103)
With the Hamiltonian (28.103) we derive the Heisenberg equations for the condensate
respecting the commutation rules (28.98) 16 ,
˙ ˙ ˆ
ψ̂2 = − ~ı (Ĥcm + ~ω0 )ψ̂2 + ~ı d̂+ · E~+ ψ̂1 (28.104)
˙ ˙ ˆ
ψ̂1 = − ~ı Ĥcm ψ̂1 + ~ı d̂− · E~− ψ̂2 .
˜ ˆ
We transform to the non-rotating coordinate system by ψ̃2 ≡ ψ2 eıωt and E~+ ≡ E~+ eıωt
introducing the detuning ∆ ≡ ω − ω0 :
˜
∂t ψ̃2 = ı(∆ − ~1 Ĥcm )ψ̃2 + ~ı d̂+ · E~+ ψ̃1 . (28.105)
Making the adiabatic approximation ∂t ψ̃2 = 0 and disregarding the external degree
of freedom, Ĥcm = 0, we obtain,
˜
ıd̂+ · E~+
ψ̃2 = ψ̃1 . (28.106)
~∆
(a) 50 (b) 1
ρ11
Δ (kHz) ρ22
ρ33
ρkk
0 0.5 ρ44
ρ55
-50 0
-10 -5 0 5 10 -10 -5 0 5 10
t (ms) t (ms)
Figure 28.18: (code) Adiabatic sweep through the ground state 87 Rb F = 2. The Rabi
t
frequency is Ω = (2π)8 kHz. The frequency ramp is ∆(t) = tm (2π)50 kHz with tm = 10 ms.
The red line shows the evolution of the population in the state |2, −2i, the green line of state
|2, +2i, and the blue line is the sum of the populations of all other states.
To discuss STIRAP, let us think of how to transfer atoms from the state |1i ≡ |2, 2i
to |2i ≡ |2, 1i. The general idea consists in applying two radiation fields with Rabi
frequencies Ω13 and Ω23 , tuned a bit out of resonance with an intermediate state |3i,
∆13 = ∆23 6= 0, in a counterintuitive pulse sequence. That is, the field Ω23 is applied
first and then adiabatically turned down, while the field Ω13 is ramped up. This
method is capable of transferring all atoms.
|1ñ=|2 2>
|2ñ=|2 1>
D23
n zm
n hf
n mw1 n mw2
D 13 |3ñ=|1 1>
Considering the example of the preceding section, we will discuss here another
option (see Fig. 28.19). We consider continuous microwave fields with fixed tunings
∆13 to the red of the state |3i ≡ |1, 1i state. Now, we apply a ramp to the second
microwave field ∆23 , such that the two radiation fields, at some point, cross the
Raman resonance, for example,
νmw1 = νhf + 3νzm + ∆13 and νmw2 (t) = νhf + 2νzm + ∆23 (t) . (28.114)
The evolution of the populations can be simulated by iterative solution of the Schrödinger
1072 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
ρkk
0 0.5
-50
-100 0
0 5 10 0 5 10
t (ms) t (ms)
Figure 28.20: (code) STIRAP ramp through a dark resonance. Here, the Rabi frequencies
are Ω13 = Ω23 = 2π10 kHz. The frequency of the first microwave is ∆13 = (2π)10 kHz and
the frequency ramp of the second microwave is given by ∆23 (t) = 2t−t
tm
m
(2π)100 kHz, where
tm = 10 ms.
and the abbreviation ~∆(r) ≡ µB gF B(r) − ~ω(r) we proceed to the second quantiza-
tion via,
XZ 2 2
−~ ∇
H = d3 r ψ̂k† + 21 ukk ψ̂k† ψ̂k + 12 uk,m ψm †
ψ̂m ψ̂k
2m
k6=m . (28.118)
h i
+ (−1)k ~2 ∆(r)ψ̂k† ψ̂k + ~2 Ω(r)(ψ̂k† ψ̂m + ψm †
ψ̂k )
The total energy of the system is the sum of the energies of the individual conden-
sates (kinetic, potential, and collisional interaction) plus the interspecies collisional
interaction energy. The coefficients ukl = 4π~2 akl /m with the scattering lengths akl
govern the impact of the collisions.
The interaction with the radiofrequency field leads to Rabi oscillations with fre-
quency Ω 17 The tuning of the radiofrequency depends on the position, due to the
non-homogeneous Zeeman shift ∆. Also, to simplify the model, we assume ukl = 0
for k 6= l,
XZ 2 2 h i
−~ ∇
Ĥ = d3 r ψ̂k† + 12 ukk ψ̂k† ψk ψk + (−1)k ~2 ∆(r)ψ̂k† ψ̂k + ~2 Ω ψ̂k† ψm + ψ̂m
†
ψ̂k .
2m
k6=m
(28.119)
From the Heisenberg equations of motion for the field operators of the matter wave
and the light, we obtain [773]DOI ,
2 2
˙ δ Ĥ −~ ∇ †
ıψ̂1 = − ~1 [Ĥ, ψ̂1 ] = = + u ψ̂
11 1 1 ψ̂ − ~
2 ∆(r) ψ̂1 + ~2 Ωψ̂2 (28.120)
~δ ψ̂1† 2m
2 2
˙ δ Ĥ −~ ∇ †
ıψ̂2 = − ~1 [Ĥ, ψ̂2 ] = = + u ψ̂
22 2 2 ψ̂ + ~
2 ∆(r) ψ̂2 + ~2 Ωψ̂1 .
~δ ψ̂2† 2m
The adiabatic potentials follow from a point-wise diagonalization of the Hamiltonian,
assuming that the atomic movement evolves in sufficiently small steps.
with, Z ∞
|φ̂1 (z)|2 + |φ̂2 (z)|2 dz = 1 . (28.124)
−∞
This set of equations can be solved numerically by first identifying the ground state
using the ’steepest descent’ method and then propagating it in real time while slowly
varying ∆(r, t) or Ω(t).
28.4.3.3 Damping
A heuristic way of introducing damping could be as follows: We write down the
condensate velocity field as,
Ef r = −γrv . (28.126)
The problem with the coupled GPE approach is that in the absence of damping
any even small modification of the coupling triggers collective oscillations, so that
the ground state is not found. A numerical method to minimize the total energy
cannot be used, because the minimum is the untrapped state, so that all atoms are
lost. It seems that a master equation approach is necessary to introduce damping.
An alternative (but cheap) way is the following. The coupled GPEs were written in
the dressed states basis. Thus, if the Rabi frequency is strong enough to completely
28.5. INTERACTION BETWEEN CONDENSATES AND OPTICAL CAVITIES1075
decouple the dressed states, the dressed states represent good quantum numbers, and
the GPEs are appropriately written in a diagonal basis, i.e. the GPEs decouple on
adiabatic potentials 18 ,
1p 2 2
Vad,k (z) = −(−1)k ~ Ω + [µB gF B(z) − ~ω]2 , (28.127)
2
and we obtain,
˙ −∂z2 †
ıφ̂k = + Vad,k (z) + gkk φ̂k φ̂k φ̂k . (28.128)
2m
The system under consideration is a BEC of two-level atoms with transition res-
onance frequency ω0 trapped in an external magnetic trap, located inside the mode
volume of a ring cavity and interacting with its light modes. The atomic field opera-
tors obey the commutation relations,
†
[ψ̂k (r), ψ̂m (r0 )] = δkm δ (3) (r, r0 ) (28.129)
0
[ψ̂k (r), ψ̂m (r )] = 0 = [ψ̂k† (r), ψ̂m
†
(r0 )] ,
where k, m = 1, 2 label the internal state of the atoms. In general, the light field
consists of only two counterpropagating monochromatic modes with Rabi frequencies
2g↱ Ⱡ. Choosing the coordinate system properly, we can set the wavevectors of
the modes as k+ · êr = kẑ = −k− · êr . The second-quantized Hamiltonian is a
generalization of (22.4),
where,
Z 2
p̂
Hatom = d3 rψ̂1† (r) + V1 (r) ψ̂1 (r)
2m
Z 2
3 † p̂
+ d rψ̂2 (r) + V2 (r) − ∆a ψ̂2 (r)
2m
Z
P
Hatom−atom = i=1,2 d3 rψ̂i† (r)ψ̂i† (r) u2ii ψ̂i (r)ψ̂i (r)
Z
, (28.131)
+ d3 rψ̂1† (r)ψ̂2† (r) u212 ψ̂2 (r)ψ̂1 (r)
Z
P
Hatom−cav = ± gâ± d3 rψ̂2† (r)e±ıkz ψ̂1 (r) + h.c.
P
Ĥcav = − ± ∆c ↱ â±
P
Ĥlaser−cav = − ± ıη± (â± − ↱ )
(22.26),
Z X †
3 † p̂2 u
H ' d rψ̂ (r) + Vtrap (r) + N̂ (r) ψ̂(r) + (N̂ U0 − ∆c ) â± â±
2m 2 ±
Z X ,
+ U0 â†+ â− d3 rψ̂ † (r)e−2ıkz ψ̂(r) + h.c. − ı η± (â± − ↱ )
±
(28.132)
where ψ̂(r) is now the ground-state wave function of the matter wave, u = 4π~2 as /m
in the mean field approximation and,
g2
U0 = , (28.133)
∆a
when the mode to mode coupling strength. This adiabatically approximated Hamil-
tonian looks very similar to (22.26).
The conversion of that Hamiltonian part Hbec to momentum space has been demon-
strated in (26.78),
X q X u X † 0† 00
Hbec = ĉ†q ĉq + ĉ0†
q Vq0 ,q ĉq + ĉq ĉq ĉq ĉq + q0 − q00 . (28.135)
q
2m 0
2 0 00
q,q q,q ,q
The light field part Ĥcav remains unchanged and the atom-light interaction becomes
using (28.3),
Z
Hbec−cav = U0 ψ̂ † (r)â†+ â− e−ı(k+ −k− )·r ψ̂(r)d3 r + h.c. (28.136)
Z Z X †
= ı~U0 â†+ â− V ĉ† (q0 )ĉ(q)δ 3 (q − q0 − 2kêz )d3 qd3 q 0 = U0 â†+ â− ĉq−2k ĉq .
q
All in all,
X q2 X † u X † †
H = ĉ†q ĉq + ĉq0 Vq0 ,q ĉq + ĉq ĉq0 ĉq00 ĉq+q0 −q00
q
2m 0
2 0 00
q,q q,q ,q
X † X † X .
†
+ (N̂ U0 − ∆c ) â± â± + U0 â+ â− ĉq−2k ĉq − ı η± (â± − ↱ )
± q ±
(28.137)
1078 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
The BEC-CARL equations of motion are readily obtained from the Heisenberg
equations,
dĉp q2 X X
ı = [ĉp , H] = ĉp + Vp,q ĉq + u ĉ†q ĉq0 ĉq−q0 +p + U0 â†+ â− ĉp+2k + h.c.
dt 2m q 0 q,q
dâ± X †
ı = [â± , H] = (N̂ U0 − ∆c )â± + U0 â∓ ĉq∓2k ĉq + ıη± . (28.138)
dt q
Neglecting the external trapping potential and collisions the Heisenberg equations for
the field and the atomic motion now read,
dĉn n2 k 2
= −ı[ĉn , H] = −ı ĉn + U0 â†+ â− ĉn+1 + h.c. (28.140)
dt 2m
dâ± X †
= −ı[â± , H] = (N̂ U0 − ∆c )â± + U0 â∓ ĉn∓1 ĉn + ıη± .
dt n
e |2,2ñ
Dlf 1
F=2
0
-1
S 1/2 -2
g1 Dmw |1,-1ñ
g2 Dcv F=1 0
1
2 MHz 6.8 GHz
Figure 28.22: (code) (a) Level scheme with a microwave-driven hyperfine structure, a laser-
driven optical transition, and a cavity-enhanced Raman deexcitation. (b) Ring cavity used
for the experiment. (c) Scheme for coupling the Zeeman states |2, 1i and |1, −1i with a
two-photon transition in a magnetic trap.
The BEC energy consists of the two energies (kinetic, potential, self) of the individual
BECs and the cross-species interaction proportional to ukl = 4π~2 akl /m. The interac-
tion energy with the classical microwave frequency (mostly we will assume ∆mw = 0)
leads to Rabi flopping with frequency Ωmw . Finally, the incident light field is treated
classically with a Rabi frequency Ωlf and a detuning ∆lf , while the cavity field is
treated quantum mechanically with a coupling constant Ωcv and a detuning ∆cv . The
irreversible coupling leads to an energy contribution where gcv ≡ Ωlf Ωcv /2∆lf is the
two-photon Rabi-frequency if the auxiliary state is adiabatically eliminated. From the
Heisenberg equations of motion for the matter wave and the optical field operators
we obtain,
˙ δH
ψ̂1 = ı[H, ψ̂1 ] = −ı † (28.142)
δ ψ̂1
−∇ 2
= −ı + V + u11 |ψ̂1 |2 + u12 |ψ̂2 |2 ψ̂1 − ı Ωmw
2 + gcv â ψ̂2 − ı∆mw ψ̂1 ,
2m
˙ δH
ψ̂2 = ı[Ĥ, ψ̂2 ] = −ı †
δ ψ̂2
−∇ 2
= −ı 2 2
+ V + u12 |ψ̂1 | + u22 |ψ̂2 | ψ̂2 − ı Ωmw
2 + gcv ↠ψ̂1 ,
2m
δH
â˙ = ı[Ĥ, â] − κ2 â = −ı † − κ2 â
Z δâ
= −ı d3 r gcv ψ̂1 ψ̂2† − ı∆cv + κ2 â .
ĉ˙1 = −ı Ωmw
2
+ gcv â ĉ2 − ı∆mw ĉ1 (28.144)
ĉ˙2 = −ı Ωmw 2
+ gcv ↠ĉ1
Obviously, since the matter wave degree of freedom is ignored, the dynamics
only concerns the Bloch vector. Simulations of these equations are exhibited in
Fig. 28.23 and in movies that can be watched under (watch movie) and (watch
movie).
3
0.4
φe − φg
ae /ag
|ak |2
0.5 2
0.2
1
0 0 0
0 1 2 0 1 2 0 1 2
t (ms) t (ms) t (ms)
Figure 28.23: (code) Bloch vector phase stabilization in the cavity. (a) Populations and (b)
coherences.
28.5.4 Exercises
28.5.4.1 Ex: BEC damping in cavities
Y. Yoshikawa et al., Superradiant light scattering from thermal atomic vapors [DOI]
M.M. Cola et al., Theory of Collective Raman Scattering from a Bose-Einstein Con-
densate [DOI]
1082 CHAPTER 28. INTERACTION OF BOSE-EINSTEIN CONDENSATES WITH LIGHT
M.G. Moore et al., Atomic Four-Wave Mixing Fermions versus Bosons [DOI]
M.G. Moore et al., Optical control and entanglement of atomic Schroedinger fields
[DOI]
W. Ketterle et al., Does Matter Wave Amplification Work for Fermions [DOI]
Instrumentation of a
Quantum Optics Lab
1083
Chapter 29
R
Spectral densities are denoted by an index ν, e.g. P = Pν (ν)dν. In (29.1), dAn ≡
n̂ · dA = dA cos θ is the projection of the surface element onto the surface normal.
The quantity, Z
I∗ = L(Ω)dΩ (29.2)
1085
1086 CHAPTER 29. GAUSSIAN OPTICS AND THE POLARIZATION OF LIGHT
Radiation passing under an angle θ through a hole of size dA = dAn / cos θ into
a specific solid angle element dΩ generates the radiance,
∂2P
= L(Ω) cos θ . (29.4)
∂Ω∂A
In any direction we get,
Z Z 2π Z π/2
∂Pν c cū
= Lν (Ω) cos θdΩ = ū dφ cos θ sin θdθ = .
∂A half sphere 4π 0 0 4
(29.5)
Summing up over all surface elements of a sphere of radius R,
cū 2 2π
Z Z Z π
∂P
P = dA = R dφ sin θdθ = πR2 cū . (29.6)
∂A 4 0 0
8πhν 3 1 8π 2 R2 hν 3 1
Pν = P 3
= . (29.7)
c eβ~ω −1 c2 eβ~ω − 1
dA0 cos θ 0
A detector covers itself a finite solid angle dΩ = r2 . The radiant flux for
r2 dA, dA0 can then be expressed as,
dA0 cos θ0
dΦ = L(Ω)dA cos θdΩ = L(Ω)dA cos θ (29.8)
Z Z r2
L(Ω)
that is Φ= 2
cos θ cos θ0 dAdA0 .
A A 0 r
Note that for isotropic sources (29.8) is symmetric upon interchanging emitter and de-
tector with regard to θ and θ0 or dA and dA0 . Furthermore, the formula demonstrates
that the radiant flux emitted into the unit solid angle is proportional to cos θ (Lam-
bert’s law). An example for such a source is a hole with the area dA in a blackbody
radiation cavity (see Fig. 29.1). Solve Excs. 29.1.2.1 and 29.1.2.2.
from the image dA∗ is emitted is also increased by the same factor. Therefore, the
radiance does not increase. In fact, because of inevitable reflection, scattering, and
absorption losses of the imaging optics, the radiance of the image dA∗ is, in practice,
always less than that of the source. A strictly parallel light beam would be emitted
into the solid angle dΩ = 0. With a finite radiant power this would imply an infinite
radiance L, which is impossible. This illustrates that such a light beam cannot be
realized. The radiation source for a strictly parallel beam anyway has to be a point
source in the focal plane of a lens. Such a point source with zero surface cannot emit
any power.
Example 129 (Radiance of the sun): An area A = 1 m2 of the Earth’s
surface receives at normal incidence from the sun about P/A = 1.35 kW/m2 of
intensity. Since the sun covers an angle of θ = 0.53◦ seen from the Earth, we
can estimate the sun’s radiance as,
P P P
L= = R 2π R θ ' ≈ 2 · 104 kW m−2 ster−1 .
AΩ A 0 0 sin θdθdφ Aπθ2
P 2
The total power received by the Earth is A
πREarth . The total power emitted
by the sun is P
A
4πd2sun−Earth .
Example 130 (Radiance of a HeNe laser ): We consider a HeNe laser emit-
ting P = 1 mW of power from a w0 = 1 mm beam waist into an angle of
2θ = 0.067◦ . With a typical emission bandwidth of ∆ν = 1 MHz the spectral
radiance is,
Pν P
Lν = ' ≈ 1 kW s m−2 ster−1 .
AΩ πw02 πθ2 ∆ν
For comparison, the sun’s radiance at it’s surface (T = 6000 K) at the same
wavelength as the HeNe laser is,
Luν
Lν = ≈ 5 nW s m−2 ster−1 .
ū
29.1.2 Exercises
29.1.2.1 Ex: Emission of an argon laser
The angular divergence of the output from a P = 1 W argon laser is assumed to be
α = 4 · 10−3 rad.
a. Calculate the radiance L and the radiant intensity I1 of the laser beam and the
irradiance I (intensity) at a surface 1 m away from the output mirror, when the laser
beam diameter at the mirror is 2ws = 2 mm.
b. What is the spectral power density u(ν) if the laser bandwidth is 1 MHz?
where λ1 = 400 nm and λ2 = 700 nm delimit the range, where photosynthesis takes
place. Calculate both quantities for a blackbody at T = 5800 K temperature. Plot
them as a function of temperature in the range T = 300..6000 K.
Normalizing the intensity to the total power, we may write the radial intensity dis-
tributions as,
2P 2 2 2
Iz (x, y) = e−2(x +y )/w(z) . (29.17)
πw(z)2
Figure 29.2: (Left) Propagation of the beam along the optical axis. (Right) Cross section of
a Gaussian laser beam.
Figure 29.3: Coupling a Gaussian beam of light into cavity requires matching of the phase
fronts.
29.2.3 Exercises
29.2.3.1 Ex: Imaging through a thin lens
In classical ray optics the equations describing the focusing of a thin lens are given
by,
1 1 1 b B
= + and = ,
f g b g G
where f is the focal distance of the lens, g the distance between the object and the
lens, b the distance between the image and the lens, G the size of the object, and B
the size of the image.
At what distance from an object of size G = 1 mm do you have to place a thin lens
with focal distance f = 100 mm in order to obtain a ten times larger image? Test
your result in practice.
29.2.3.5
Ex: The
eye !"#%$'&(*)
+(+
The effective distance ,,.-%/ between the cornea and the lens of the eye is 2.5 cm, and the
lens0 is1'23
in# 5 4
687:be9with
contact ;=<>ontwo
@the?BACnon-adjacent
;D<C"To
eye. FE resolve two very close points, their images on
the retina G HIJ!K L>*MNPOJO
must cone cells (i.e. there must be at least one
non-activated cone cell between the images). The cone cells are about 1 µm apart.
a. What is the smallest angle under which two points are still separately perceptible?
HIJ!K QO Assume that the direct beams P, P’ are not refracted.
b. How close can the two points P and P be, at a distance of 20 m from the eye, so
1 2
VUWYX[Z]\(Z_^a`bcX[ZedYZ_`@that ZFfZ_ghthey
^*ikj[ZlU
are\nm*^*b[still
opdBdYseparately
ZImrqhsut perceptible?
ISvm'wJx[`opynb[WY^tJz3i{onmaZj[x|odYZ_`>}JZ_^a~Iypb[WY^h_S
dpikZBz3i
opmaZma^aZIyYZi{oCx|ou^*bcgh^Svik^dYZIS UWt
%b[Svik^~_MZIiLLWpoYgh^aX[Z_X[ZIopma^b[ opdYZ[dpikZmaZIyY`opb[yYZ P 1
{ZIi{opb[opdYZ_`"ma^aZIyYZIoX[Z_^a`*ZIopou^Mb[yY`~IWpoYZIypSZIom*i{opd
2 P’
pm*maZIoi{yY`ZLi{{dYZ_`b[WYndYZ_`}JZ_^a~Iypb[WY^b[WYp~_MZIiopi{yu^ P ε 1 P’
nb[yu\nbc`*^aZbc pyYZIob[{kZIo m%SWY¡b[
maxSi{opdYZFt 2
IopmZIi{o¢opi{"y£^*bcgh^*ikjikZ_`*^aZIm
bc n¤"yYZIo~_
i{m*"yYZIo@dYZIo¢i{{dYZ_`o{ikZ_X[ZIoh%ikZ'bc pypZIo@m*i{opdKZ_^¥b
S¨j[x|oYZIi{onb[opdYZ_`©ZIou^aªZ_`ou^_
(¯® ikZKX[`x[¡i
ma^°dYZ_29.2.3.6 `gkZIi{opma^aZ ® i
oYg[Ex: ZIM±IMicroscope
Wno£^aZ_`dYZIS²dpikZK\ZIi
dYZIoLWpoYgh^aZKoYxyeX[Z_^a`*ZIopou^
b[yY`oYZIypS°\nbc`m*i{AondN³simple ´ZIyYZIohomemade
qikZdpbj[x|ob[Wpmicroscope
m_cdpbc¡CdpikZ©µ¶ik^a^aZI{nWpoYgh^*m*ofma^a`b[two
ynkZIo'convex
·[·C¸hopi{lenses,
"y£^ each one with a diffrac-
X[Z_\p`*xypZIo¯MZ_`dYZItiveo power of 20 dpt, fixed at theconsists
¯® ikZopb[yºgx| opoYZIoa.dpikWie Z\(ZIi{groß
dYZIoºistWnoYgudie ^aZ\(Tubuslänge
ZIiZIi{opZIS»U
\Nma^*dieses
b[opdj[x|oeMikroskops?
c¼Sj[x|SQUWYX[Zv\(ZIi½t
extremities of a 30 cm long tube.
ZIi{opb[ondYZ_`ma^aZIyYZIopb.dpb[SvWie
ik^©m*ikZ>groß
opxh"y@istX[Z_^ader
`*ZIopouAbbildungsmaßstab
^
b[yY`oYZIynS'\nbc`!mi{opdN³ des Mikroskops?
c. Welche Vergrößerung erreicht das Mikroskop? Gehen Sie davon aus, dass die deut-
liche Sehweite 25 cm beträgt.
HIJ!K ¾4 d. Wie weit muss sich der Gegenstand vor dem Objektiv befinden, damit er im Auge
des Betrachters scharf abgebildet wird?
_MZIi{maZIoqikZ[[dpbc¡C\(ZIihZIi{oYZ_`qb[SvSZI
{i{opmaZMdYZ_`M`*ZIopouLZIi{^aZ
¿KdpZ_`ghkZIi
opma^aZU
\nm*^*b[opd~_
i{m*"yYZIo
dpZIo ® Z_`*^À%ÁÃÂEx:
_X[ZIopma^*b[opdrWpopd@i{{d@29.2.3.7 ¿ypbcClassical
\(ZIo@SWY¡Y cloaking with four lenses
HIJ!K ÅÄ Can you design a system of four lenses (focal distances f , f , f = f , and f = f )
separated by three distances t = f + f , t , and t = t , such that the system
1 1 2 2 3
1
1
2 3 2 4 1
ÆZIi{op¤b["yYZImvmaZIk\Nma^aX[Z_\nb[WY^aZImµ¶ikgh`*x|mag[x[\ZIm*^aZIyYZ¶b[Wpm~_MZIiJqb[SSZI{{i{onmaZIoBÇÈÉZIdYZ¯Sik^vdYZ_`
appears to be invisible for an observer looking through the lenses [136]?
IyYgh`bc¤^Jc¼dYp^"Ê"dpik29.2.3.8
ZCb[o¯dYZIo¯ LopdYZIo¢Ex: ZIi{opZ_Diameter
`JËc¼_SÌ{b[opX[ZIofo¯Í°ax| yYGaussian`ZC\Z_¤ZIma^*ikX[^Jm*beam
i{opd
(¯® ikZ>X[`*x[¡i
ma^!dpikZCYou ÎWY\nWparem*Ïb[ oYblocking
X[ZCdpikZImaZIm
µrpart ikgh`*x|maofg[x[namг laser beam with a razor blade mounted on a translation
¯® ikZ>X[`*x[¡i
ma^!dYZ_`
stage UJ\p\ni{
allowing
dpWpoYX|m*Sbc¡|you ma^*bc\todpZImJvary Ñ\ÈaZ_the gu^*i{jhm"horizontal
³ position. At the same time, you observe the
¶® ZI{"yYZ'ÓZ_`*X[`Ôx[ ¡[Z_you `WpoYXKneed
transmitted Z_`*`*ZIi{"toy£^!dpvary
b[mµ¶the
power i{gu`*x|m*translation
P . g[x[³°´%ZIyYZIo¯m*byikZ>dn140
You observe that,bIj[x|oµm.
for b[Wpvarying
m_pdpWhat
bc¡dnikZCtheisdYZIthe
WYpower
^Ét diameter
between 16% and 84%,
{i{"yYZ°qhZIyuLZIi{^aZ°[beam?
_S¨\(Z_^a`:bc X[^_ of the Gaussian
¯® ikZ>MZIik^!S°Wp¡vm*i{y¶dYZ_`J´Z_X[ZIopma^*b[ond¯j[x[`JdpZISÖÑ\ÈaZ_gu^*i{j\(Z_×nopdYZIohdpb[Sik^©Z_`
i{SPUWYX[Z
dYZIm!MZ_^a`b["y£^aZ_`"m29.2.3.9
mypbc`*bc\pX[Z_\ni
{dYEx: Z_^©
ik`"Diffraction
dN³ of a Gaussian beam at a slit
a. Determine the power loss suffered by a Gaussian beam passing through a one-
dimensional slit, assuming that the beam hits the slit in its center.
b. Calculate the diffraction pattern produced by the slit.
c. A laser beam (λ = 633 nm) looses 50% of its power after being passed through a
1092 CHAPTER 29. GAUSSIAN OPTICS AND THE POLARIZATION OF LIGHT
Figure 29.5: (Left) Anamorphic prisms (Thorlabs, PS870). (Right) Geometry of the
beam transformation. The entrance side of a prism is often chosen close to the
Brewster angle, while the exit side treated carries an anti-reflexion coating.
prism pair does not allow to correct for astigmatism. This requires cylindrical lenses.
1094 CHAPTER 29. GAUSSIAN OPTICS AND THE POLARIZATION OF LIGHT
Figure 29.6: Beam steering with two rotatable wedged substrates (see Thorlabs, PS810).
In practice, the degree of freedom of polarization is often used for separating coun-
terpropagating light fields, e.g. in ring lasers, by means of elements called optical diode
or optical isolator, which consist of a Faraday rotator and λ/2 waveplate. Another
practical example is the use of λ/4 in double passage. An incoming beam can be
separated from a returning beam by using a λ/4 waveplate and a polarizing beam
splitter.
1096 CHAPTER 29. GAUSSIAN OPTICS AND THE POLARIZATION OF LIGHT
where φ is the rotation angle. For the Faraday rotator the sign of the rotation angle
depends on the propagation direction of the laser beam. A polarizer projects the
polarization onto a specific axis. In the case of the x-axis Jones matrix is,
1 0
Mpolarizer = . (29.26)
0 0
For θ = 2π/n we obtain a λ/n-waveplate. When we rotate the waveplate and therefore
the optically inactive axis to an angle φ, the Jones matrices are,
−1
cos φ sin φ 1 0 cos φ sin φ
Mθ-waveplate (φ) = (29.29)
− sin φ cos φ 0 eiθ − sin φ cos φ
cos2 φ + eiθ sin2 φ − sin φ cos φ + eiθ sin φ cos φ
= .
− sin φ cos φ + eiθ sin φ cos φ sin2 φ + eiθ cos2 φ
Figure 29.8: (a) Rotation of the polarization by a birefringent waveplate. (b) Illustration of
a circularly polarized light wave.
Combinations of λ/2 waveplates and Faraday rotators are used as optical isolator,
also called optical diode.
The angles of incidence and transmission are related by Snell’s law: n1 sin θi =
n2 sin θt .
The Brewster angle θi,B is reached, when θi,B + θt = 90◦ , i.e., when following
Snell’s law,
n1 sin θi,B = n2 sin(90◦ − θi,B ) = n2 cos θi,B . (29.33)
Hence, the Brewster angle is given by,
n2
tan θi,B = . (29.34)
n1
Rp , Rs , Tp , Tr 1
0.5
0
40 60 80 0 20 40 60 80
θi θi
1
Rp , Rs , Tp , Tr
Figure 29.9: (code) (Left) Fresnel formulae showing the angular dependence of ts
(red solid), rs (green dash-dotted), tp (blue dashed), and rp (cyan dotted) for reflec-
tion0.5
from and transmission through a piece of glass-air interface. (Right) Interfaces
between optical media with different reflection indices can act like polarizers: Light
reflected from a glas surface under the Brewster angle is completely s-polarized, while
the transmitted
0 light is partially p-polarized. The notation s comes from senkrecht,
40 60 80 0 20
i.e. perpendicular40 to the
60 plane
80 spanned by the incident and reflected light beams,
α θ
while p means paralleli to this plane.
The Stokes parameters exhaustively describe the polarization state of a light beam.
It is interesting, that the vector Ŝ with components Sj with j = 1, 2, 3 satisfies the
SU(2) spin algebra,
[Ŝk , Ŝm ] = 2ıkmn Ŝn and Ŝ2 = Ŝ02 + 2Ŝ0 , (29.37)
as will be shown in Exc. 29.3.4.6. It is conveniently pictured on a Poincaré sphere or
as a polarization ellipse. In the classical limit we get,
Ŝ0 = I (29.38)
Ŝ1 = Ip cos 2ψ cos 2χ
Ŝ2 = Ip sin 2ψ cos 2χ
Ŝ3 = Ip sin 2χ ,
29.3. INTRODUCTION TO POLARIZATION OPTICS 1099
with I the light intensity (eventually normalized to the single-photon light intensity),
the degree of polarization p. Obviously,
29.3.4 Exercises
29.3.4.1 Ex: Light power control using polarization optics
The power of a laser beam can be regulated by a combination of a half-wave plate and
a polarizing beam splitter. By how many degrees do you have to rotate the waveplate
in order to reduce the light power by a factor of 2? Use the Jones matrices to justify
your response. Advice: Look up the Jones matrices (29.31) and (29.26). Test your
result in practice.
1. Pass the beam of a helium-neon laser through a polarizer and a quarter- resp.
half-waveplate. Analyze the polarization using a rotatable second polarizer for
various rotation angles of the waveplate. Sketch the transmitted intensity as a
function of the rotation angle of the beamsplitter in a polar diagram. How good
can you achieve linear and circular polarization?
Figure 29.10: (a) Power control of a light beam and (b) separation of counterpropagating
beams through polarization optics.
2. Mirrors can change the polarization of a light beam and, for example, transform
a linear polarization into elliptical. Determine the degree of ellipticity for a given
mirror.
3. How does a mirror transform the polarization and the helicity of a reflected
laser beam?
Figure 29.11: Measuring the Brewster angle by varying the tilt of a glas plate.
2. Use the EOM as a Pockels cell by placing it between the half-waveplate and
the second polarizer. Supply a voltage between 0 V and 350 V to the EOM.
Measures the intensity of reflected and transmitted light as a function of the
supplied voltage and prepare a diagram. What do you observe when you rotate
either the half-waveplate or the EOM?
For a given zone plate there are other focuses at smaller distances,
d2n − k 2 n2 λ2 d2n f1
fk = ' = . (29.42)
2knλ 2knλ k
In order to separate the beams diffracted by the zone plate into a given focus from
those diffracted into other focuses or not being diffracted at all, we pass the beam
through an iris diaphragm localized at the desired focus and recollimate the beam by
means of a lens, as shown in Fig. 29.13.
The phase profile of the beam can be viewed interferometrically (see Fig. 29.13)
by overlapping a plane wave laser beam. With a neutral density filter the intensities
of the overlapping beams can be adjusted to maximize the contrast.
f 50 f 250 f 100
Laser
Zonenplatte Lochblende
Schirm
Graukeil
f 50 f 100
Now, for realizing Laguerre-Gaussian light modes, we use Fresnel zone plates
with spiral patterns, instead of concentric rings. In contrast to the Gaussian mode,
the Laguerre-Gauß modes exhibit an intensity minimum on the optical axis (doonat
mode). Their phase profiles can be viewed by interferometry [12].
1. Construct the interferometer sketched in Fig. 29.13 using adequate Fresnel zone
plates. What do you observe in the diffracted beam and in the interferogram,
when instead of filtering the principal focus f1 you filter a higher order focus?
2. Pass a Laguerre-Gauß laser beam through a λ/2 waveplate. How does the an-
gular orbital momentum change when you change the rotation? What happens
upon reflection from a mirror?
3. Slightly misalign the mode-matching between the Laguerre-Gauß beam and the
Gaussian reference beam until you observe multiple fringes. What do you ob-
serve?
Electronics and
radiofrequency
For the control and regulation of important quantum optical devices, such as EOMs,
AOMs, laser diodes, photodiodes, piezos etc., electronic circuits are necessary. The
aim of this chapter is to provide practical know-how in the basics of electronics.
1105
1106 CHAPTER 30. ELECTRONICS AND RADIOFREQUENCY
black 0 X1 0 0 X1
brown 1 1 x10 1 1 1 x10
red 2 2 x100 2 2 2 x100
orange 3 3 x1000 3 3 3 x1000
yellow 4 4 x10000 4 4 4 x10000
green 5 5 x100000 5 5 5 /10
blue 6 6 x1000000 6 6 6 /100
violet 7 7 /10 7 7 7
gray 8 8 /100 8 8 8
white 9 9 9 9 9 9 9
Figure 30.2: (Left) Electrolyte aluminum capacitors. (Center) Polyester capacitor and color
code. (Right) Ceramic capacitor.
Figure 30.3: (Left) Schematic symbol of an OpAmp; (Right) Pin layout of a standard
OpAmp.
Figure 30.4: (Left) Principle scheme of a standard OpAmp. (Center) Non-inverting amplifier.
(Right) Inverting amplifier.
This becomes clear noting that, since the input (-) does not drag current, the currents
traversing the resistances Rn and R1 must cancel each other.
1108 CHAPTER 30. ELECTRONICS AND RADIOFREQUENCY
1
ZL = iLω , ZC = . (30.3)
iCω
For the calculation of the amplification wit complex impedances, we just take the
absolute value of the gain G.
Figure 30.5:
30.1. INTRODUCTION TO ELECTRONIC CIRCUITS 1109
Q̇ = (ΠA − ΠB )I , (30.7)
where ΠA and ΠB are the Peltier coefficients of conductors A and B, and I is the
electric current (from A to B). The total heat generated is not determined by the
Peltier effect alone, as it may also be influenced by Joule heating and thermal-gradient
effects (see below).
The Peltier coefficients represent how much heat is carried per unit charge. Since
charge current must be continuous across a junction, the associated heat flow will
develop a discontinuity if ΠA and ΠB are different. The Peltier effect can be considered
as the back-action counterpart to the Seebeck effect (analogous to the back-EMF in
magnetic induction): if a simple thermoelectric circuit is closed, then the Seebeck
effect will drive a current, which in turn (by the Peltier effect) will always transfer heat
from the hot to the cold junction. The close relationship between Peltier and Seebeck
effects can be seen in the direct connection between their coefficients [357, 707]:
Π = TS . (30.8)
Figure 30.6: (a) Peltier cooler, (b) voltage generator or heat sensor, and (c) heat or radiation
sensor with thermocouples connected in parallel.
30.1.5 Exercises
30.1.5.1 Ex: Integrator with operational amplifier
Based on the two golden rules for operational amplifiers, (1) I+ = I− = 0 A and (2)
U+ = U− , show thatR the output voltage Ua at the integrated circuit shown in the
1
figure is: Ua = RC Ue dt.
1110 CHAPTER 30. ELECTRONICS AND RADIOFREQUENCY
IC C
R IR
UR UR
Using an inverting operational amplifier design a simple low-pass filter with a constant
amplification of 10 at low frequencies and diminishing gain above 10 kHz. Calculate
the Bode diagram (i.e. the frequency-dependent gain and phase-shift of your circuit.
What is the gain reduction per octave?
3. Modify the circuit such as to obtain a low-pass filter with fg = 50 kHz bandwidth
and test the circuit again.
1. Connect a Peltier element to a 1 A current source and bring one of the two
surfaces of the Peltier element into thermal contact with a heat sink. What do
you observe?
2. Bring a 10kΩ thermistor in thermal contact with either of the two surfaces of
the Peltier element and measure its resistance. What do you observe?
3. Can you imagine a feedback logic evaluating the measured resistance in order to
control the current applied to the Peltier element such as to maintain constant
the temperature of the surface not connected to the heat sink.
30.2. DETECTORS 1111
Figure 30.7: Pictures of Peltier elements, a thermistor, and an AD590 temperature trans-
ducer.
10
(kΩ)
5
Rt
0
0 20 40 60
T (◦ C)
30.2 Detectors
30.2.1 Photodiodes
Most active components are many of semiconductor characterized by a relatively large
band gap between the valence band and the conduction band. By appropriate doping
of the material with donors (p-type) or acceptors (n-type) a semiconductor can be
made conductive. The most basic semiconductor element, which is the diode consists
of a junction of two types of semiconductors, as shown in Fig. 30.9.
Our first task will be to construct a photo detector. The central part of a pho-
todetector is the photodiode. We have at our disposal silicium pin-photodiodes of the
type C30822E of the company Perkin Elmer and of the type FFD100.
PST l/4
laser AOM
Regler
Anwendung
Figure 30.9: Joining a p-type and a n-type doped semiconductor (left) one observes a charge
carrier redistribution across the pn transition (right).
Schaltung 5604
C30902
FFD100
C30822E 1k
n
i R 16.7k
p 220
+190 V
+8..18 V
2.2n
BZX200
e -
Figure 30.10: Circuit with photodiode.
p n
gives rise to an electric field across the junction. The energy liberated during the
recombination process can be dissipated via emission of light.
e -
The reverse
p process is also possible: Via the intrinsic photoeffect, light irradiated
into the pn-junction can lift electrons from the valence into the conduction band, thus
n
generating pairs of charge carriers. Under the influence of the electric field across the
junction, the holes flow to the edge of the p domain and the electrons flow to the n
domain. This part of Ithe current is called drift current. A smaller part, called the
diffusion current, has its origin in the diffusion of the electron-hole pairs formed in
the edge regions. Since these minority charge carriers have only a limited lifetime
before they recombine, onlyU the part of the current generated within a few units
of the1 mW
diffusion lengths near the charge carrier zone contributes. This results in an
external
10 mWphotovoltaic voltage at the electrodes of the photodiode. If the photodiode
is connected to a load, a photocurrent will flow, which is composed, as mentioned
above, by the drift current of the charge carrier zone and the diffusion current from
its edges.
The principal scheme of a pin diode is illustrated in Fig. 30.11(left): A weakly
30.2. DETECTORS 1113
doped intrinsic layer separates the p and the n conductor. This reduces the capacity
of the barrier. The current at short circuit is proportional to the light power. A
photodiode is always operated in blocking direction. A negative offset voltage reduces
the capacity of the pn-junction.
Despite all measures the pn-junction capacity remains finite. One can model the
impact of the pn-junction capacity via a replacement diagram. The voltage drop is
1 i R ||R
L
U (ω) RL || iωC Ri
= 1
= . (30.9)
U0 Ri + RL || iωC 1 + iωC(Ri ||RL )
For high load resistances the frequency response obviously becomes load-independent.
For small loads, RL < Ri , the band width of the photodiode is dramatically increased
to ωg = 1/RL C. In the same time, however, the amplification drops to V = RL /Ri .
30.2.2 Exercises
30.2.2.1 Ex: Photomultiplier
The anode of a photomultiplier tube is connected by a resistor of R = 1 kΩ to ground.
The stray capacitance is 10 pF, the current amplification 106 , and the anode rise time
1.5 ns. What is the peak amplitude and the halfwidth of the anode output pulse
produced by a single photoelectron? What is the dc output current produced by
10−12 W cw radiation at λ = 500 nm, if the quantum efficiency of the cathode is η =
0.2 and the anode resistor R = 106 Ω? Estimate the necessary voltage amplification
of a preamplifier (a) to produce 1 V pulses for single-photon counting; and (b) to read
1 V on a dc meter of the cw radiation?
data sheet for the Photodiode C30822E see appendix Fig. 34.21.
30.3. INTRODUCTION TO RADIOFREQUENCY COMPONENTS 1115
The modulation of the carrier wave generates sidebands. This can be seen by
expanding the signal which carries the phase modulation into a Fourier series,
∞
X
Aeiωt+iβ sin Ωt = Aeiωt Jk (β)eikΩt ' Aeiωt + J1 (β)Aeiωt+iΩt + J−1 (β)Aeiωt−iΩt
k=−∞
(30.10)
when the modulation index β is small. Here, J−k (β) = (−1)k Jk (β) are the Bessel
functions. This is in contrast to amplitude modulation, which is described by only
two symmetric sidebands,
iωt iωt β iΩt −iΩt
A (1 + β sin Ωt) e = Ae 1+ e −e . (30.11)
2i
For amplitude modulation (AM) the beat signals between the carrier frequency
and the two sidebands are in phase, i.e.,
2
iωt
e + ei(ω±Ω)t = 2 + eiΩt + e−iΩt . (30.12)
For phase modulation (PM) the beat signal are in counter-phase, i.e.,
2
iωt
e + ei(ω±Ω)t+iπ/2 = 2 + ie±iΩt − ie∓iΩt . (30.13)
In the case of AM, the amplitude is blurred, but the phase at zero-crossing is well
defined. In the case of PM, the amplitude in the antinode is sharp, but the phase of
the zero-crossing is blurred.
It is not easy to transform AM into PM, and vice versa. In fact, the phase
between carrier and sidebands can be varied, for example by adding an AC voltage,
√ iωt+3iπ/4
2e to the signal; however, it is not easy to transform synchronized phases
into opposite phases.
0
spectrum (dB)
-50
-100
-20 -10 0 10 20
frequency (MHz)
As shown in Eq. (30.10), the spectrum of a signal with phase modulation (PM)
consists of discrete lines, called sidebands, whose amplitudes are given by Bessel
1116 CHAPTER 30. ELECTRONICS AND RADIOFREQUENCY
functions,
∞
X
S(ω) = |AJk (β)|2 δ(ω + kΩ) . (30.14)
k=−∞
In real systems, the sidebands have finite widths γ due to frequency noise or the finite
resolution of the detectors. In the case of Lorentzian line profiles, we have,
∞
X β2
S(ω) = |AJk (β)|2 . (30.15)
(ω − kΩ)2 + β 2
k=−∞
30.3.2 Mixers
A frequency mixer is a nonlinear electrical circuit that creates new frequencies from
two input signals, e.g. the sum and difference of the input frequencies. A device that
has a non-linear (e.g. exponential) characteristic can act as a mixer. Passive mixers
use one or more diodes and rely on their non-linear relation between voltage and
current. Active mixers use an amplifying device (such as a transistor) to increase the
strength of the product signal.
Mixers may be classified by their topology: An unbalanced mixer, in addition to
producing a product signal, allows both input signals to pass through and appear as
components in the output. A single-balanced mixer is arranged with one of its inputs
applied to a balanced (differential) circuit so that either the local oscillator (LO) or
signal input (RF) is suppressed at the output, but not both. A double-balanced mixer
has both its inputs applied to differential circuits, so that neither of the input signals
and only the product signal appears at the output. Double balanced mixers are more
complex and require higher drive levels than unbalanced and single-balanced designs.
In practice, mixers are widely used to shift signals from one frequency range to
another, a process known as heterodyning, in order to facilitate signal transmission or
further signal processing. Frequency mixers are also used to modulate a carrier signal
in radio transmitters, as product detectors, phase detectors, or frequency multipliers.
30.3.2.1 Diode
The non-linearity (or non-Ohmic behavior) of a diode can be used to create a simple
unbalanced mixer producing the original frequencies as well as their sum and their
difference. The current I through an ideal diode as a function of the voltage U across
it is given by an exponential function,
I = I0 (eqU/kB T − 1) . (30.16)
2
The exponential can be expanded as ex −1 ' x+ x2 . Suppose that the sum of the two
input signals U1 + U2 is applied to a diode, and that an output voltage is generated
that is proportional to the current through the diode [e.g. by providing the voltage
that is present across a resistor in series with the diode, as shown in Fig. 30.13(a)].
Then, disregarding the constants in the diode equation, the output voltage will have
the form,
U0 = (U1 + U2 ) + 12 (U1 + U2 )2 + ... (30.17)
2
= (U1 + U2 ) + 1
2 (U1 + 2U1 U2 + U22 ) + ... .
30.3. INTRODUCTION TO RADIOFREQUENCY COMPONENTS 1117
The ellipsis represents all the higher powers of the sum which we assume to be negli-
gible for small signals.
Figure 30.13: (a) Mixing two signal at a diode. (b) Schematic diagram of a double-balanced
passive diode mixer (also known as a ring modulator). There is no output unless both f1
and f2 inputs are present, though f2 (but not f1 ) can be DC.
Ignoring all terms except for the sin ω1 t sin ω2 t term we get,
30.3.2.2 Switching
Another form of mixer operates by switching, with the smaller input signal being
passed inverted or non-inverted according to the phase of the local oscillator (LO).
This would be typical of the normal operating mode of a packaged double balanced
mixer, with the local oscillator drive considerably higher than the signal amplitude.
The aim of a switching mixer is to achieve linear operation over the signal level by
means of hard switching, driven by the local oscillator. Mathematically, the switching
mixer is not much different from a multiplying mixer. Instead of the LO sine wave
term, we would use the signum function. In the frequency domain, the switching mixer
operation leads to the usual sum and difference frequencies, but also to further terms,
e.g. ±3fLO , ±5fLO , etc.. The advantage of a switching mixer is that it can achieve
(with the same effort) a lower noise figure and larger conversion gain. This is because
the switching diodes or transistors act either like a small resistor (switch closed) or
large resistor (switch open), and in both cases only a minimal noise is added. From
the circuit perspective, many multiplying mixers can be used as switching mixers,
just by increasing the LO amplitude.
we have a reference signal which we want to transport somewhere else, Uref = cos Ωt.
Used as a modulator the mixer will simple multiply,
1 1
Umod = Ucarrier Uref = cos ωt cos Ωt = 2 cos[(ω − Ω)t] + 2 cos[ω + Ω)t] . (30.20)
Thus the mixer output contains two frequencies, the sum and the difference. Sup-
posing that ω is a frequency in a range that can be radiated by antenna and Ω an
acoustic frequency, both frequency components ω ± Ω ' ω will be radiated.
On the side of the receiver, who also has access to a synthesizer generating a signal
Ucarrier = cos ωt, we will use the mixer as a demodulator,
Udemod = Ucarrier Umod = cos2 ωt cos Ωt = 12 (cos 2ωt + 1) cos Ωt . (30.21)
If ω Ω, the carrier oscillation can easily be removed by a low-pass filter,
1
Uf iltered = 2 cos Ωt ∝ Uref . (30.22)
That is, we recover the original information. And this holds even when the reference
signal is (slowly) varying in amplitude or frequency, such as in the case of an acoustic
signal 2 . Resolve Exc. 30.3.3.1.
Such techniques are widely used in lock-in amplifiers (see Sec. 30.4.3).
30.3.3 Exercises
30.3.3.1 Ex: Phase modulation
a. Show that it is not possible to construct a periodic phase modulation function such
that the signal has only two
P∞sidebands. P∞
b. Compare the spectra k=−∞ Jk (M )eıkΩt = eıM sin Ωt and k=−∞ |Jk (M )|eikΩt .
Can you detect phase modulation of a photodetector signal on a spectrum analyzer?
How about amplitude modulation?
1. Take a VCO, for example, ZOS-100+ from MiniCircuits. Study the datasheet
and drive the VCO with an AC voltage. Vary the amplitude and the frequency
of the voltage and observe the output signal of the VCO on a spectrum analyzer.
data sheet for the variable attenuator see appendix Fig. 34.16,
data sheet for the mixer see appendix data sheet Fig. 34.18.
1120 CHAPTER 30. ELECTRONICS AND RADIOFREQUENCY
Figure 30.15: (a) Principal scheme of a lock-in amplifier. (b) Mode of operation.
1. Create the circuit sketched in Fig. 30.15(a) on a circuit board and test it by
varying the phase between the modulated output signal and the TTL signal
provided by a function generator.
where the 2 × 2 element is the beam splitter matrix. r and t are the reflectance and
transmittance along a particular path through the beam splitter, that path being
indicated by the subscripts.
1121
1122 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
Assuming the beam splitter removes no energy from the light beams, the total
output energy can be equated with the total input energy, reading
|Ec |2 + |Ed |2 = (|rac ||Ea | + |tbc |Eb |)2 + (|tad ||Ea | + |rbd ||Eb |)2 = |Ea |2 + |Eb |2 . (31.2)
This can only be true for any field amplitudes, if the following relationships between
reflectance and transmittance are satisfied,
We write each r and t as a complex number having an amplitude and phase factor
accounting for possible phase shifts of a beam as it reflects or transmits at the beam
splitting surface. From the second equation (31.3) we obtain,
which is true when φad − φbd + φbc − φac = π. Squaring both sides of the expression
(31.4) and comparing this with the first equation (31.3) we obtain the result
√ √
|tad | = |tbc | ≡ T and |rac | = |rbd | ≡ R . (31.5)
It follows that
R+T =1 . (31.6)
The above result holds for any type of beam splitting device. For beam splitting
at dielectric interfaces in particular we know that the electric field amplitude does
not suffer phase shifts upon transmission and no phase shifts upon reflection at an
optical more dilute medium. But it suffers a 180◦ phase shift upon reflection at an
optical denser medium. For the situation depicted in Fig. 31.1, we conclude φac =
φbd = φbc = 0 and φad = π. With this constraints the matrix describing a lossless
beam splitter reads,
√ √
Ec √R √T Ea
= . (31.7)
Ed − T R Eb
only on the speed of sound (which is a constant of the material) and the dimensions
of the actuator. Therefore, actuators are also suitable for realizing oscillators (for
example, quartz crystals). The piezo-electric effect can only occur in non-conductive
materials (e.g., lead titanate zirconate).
When a voltage is applied to the piezo-ceramic in the direction of polarization, we
observe an expansion in this direction and a perpendicular contraction. Depending
on the employed material and the coefficient for piezo-electric strain d, stretches up
to ∆l/l = 0.15% can be obtained:
∆l = dEl0 , (31.8)
where l0 is the length of the actuator and E = U/l0 the amplitude of the electric
field. The elongation effect is therefore proportional to the field strength and the
overall length of the actuator. To achieve large stretches with manageable electrical
voltages, actuator discs are often stacked (mechanical circuit in series and electric
circuit in parallel).
mirror mirror
piezo piezo
l l
U U
Figure 31.2: (Left) Scheme of mirrors mounted on a piezo actuator having the shape
of a disc or a ring. (Right) Photos of a ring piezo and a piezo stack.
Negative voltages with respect to the orientation of the discs cause a contraction.
However, negative voltages can also cause a change in the polarization state of the
piezo and should therefore be avoided! In electrical circuits, piezoelectric actuators
introduce a capacitance with a relative dielectric constant between 600 and 5000 and
an internal resistance of about 108 Ω depending on the material. See Exc. 31.1.10.1.
Figure 31.3: Principle of a two-beam interferometer: (a) Michelson interferometer and (b)
Mach-Zehnder interferometer using non-polarizing beamsplitters. (c) Michelson interferom-
eter using a polarizing beamsplitter.
We consider the Mach-Zehnder interferometer sketched in Fig. 31.3 with one arm
of length Lr , which can be varied by a piezo, and the other arm of length Lt ,
The beams are recombined on a second beam splitter and sent to a photodetector,
whose signal is,
p p
I ∝ | R2 Er0 + 1 − R2 Et0 |2 (31.11)
p p p p
= | R2 R1 e2ıkLr + 1 − R2 1 − R1 e2ıkLt |2 Ei2 .
Hence,
p p p p
I ∝ R2 R1 +(1−R2 )(1−R1 )+2 R2 R1 1 − R2 1 − R1 cos[2k(Lt −Lr )] . (31.12)
1 1
I∝ 2 + 2 cos[2k(Lt − Lr )] . (31.13)
while the superposition of two plane waves with equal frequency but different polar-
izations does not,
2
E1 0
2 2
0 + E2 = |E1 | + |E2 | . (31.15)
hE − (t)E + (t + τ )i
g (1) (τ ) ≡ and SE (ω) ≡ F[g (1) (τ )] . (31.16)
hE − (t)E + (t)i
As an example, Fig. 31.4 shows the aurocorrelation function and spectrum of a si-
multaneously amplitude- and phase-modulated laser beam, E(t) = (1+n cos ωn t)eım sin ωm t .
1 0.8
gE (τ ) (Hz−1 )
0.5
0.6
SE (ω)
0.5
0 5 10 0
(b) 1 0.4
φ (π)
(1)
-0.5
0 0.2
-1 -1 0
0 5 10 -10 0 10 -5 0 5
ωt/2π ωτ 2πf /ω
Figure 31.4: (code) (a) Amplitude and (b) phase of the electric field, with n = 0.2, m = 1,
and ωn = ωm = (2π) 1 Hz. (c) Aurocorrelation function and (d) spectrum of the light field.
Such that,
Eout E eık0 L cos2 φ + eıke L sin2 φ Ein
= M in = . (31.18)
0 0 0
In practice Lyot filters are often used, placed under the Brewster angle inside ring cav-
ity lasers, as frequency selective elements. Frequently, birefringent plates of different
thicknesses are stacked,
T (λ) = T1 (λ)T2 (λ) , (31.21)
in order to increase the frequency selectivity. Furthermore, when the axis of the
birefringent plates are rotated by fixed angles ∆α with respect to each other,
the frequency of maximum transmission can be tuned by simply rotating the stack as
a whole by an angle α. See 31.1.10.5.
A Lyot filter can be considered an interferometer, because it splits and recombines
the polarization vector of a light beam in two parts following the ordinary and the
extraordinary axis of the birefringent crystal. This is somewhat analogous to the
Michelson interferometer depicted in Fig. 31.3(c).
Cavities are characterized on one hand by their geometry, i.e. the curvature and
the distance of their mirrors, and on the other hand by their finesse, which is given by
the reflectivity of their mirrors. Let us first study the finesse. Regarding the cavity
as a multipass interferometer [191], we can derive expressions for the reflected and
31.1. INTRODUCTION TO INTERFEROMETRY 1127
100 100
transmission (%)
reflection (%)
50 50
0 0
-2 0 2 4 -2 0 2 4
phase (π) phase (π)
Note that factors others than the finite reflectivity may degrade the finesse of a
cavity. For example, an imperfect mirror flatness (commonly specified as λ/Fsurf )
reduces the finesse to [191],
−1
1 1
Ftot = + . (31.27)
F Fsurf
.
The geometry of a cavity must satisfy certain conditions, in order to be stable
[415]. Besides the main longitudinal modes a cavity possesses transverse modes of the
order TEMmn , whose frequencies are given by 2 ,
s
m+n+1 L L
ν/δf sr = (q + 1) + arccos 1− 1− . (31.28)
π ρ1 ρ2
1 See script on Electrodynamics (2020), Exc. 7.3.6.16.
2 See script on Electrodynamics (2020), Sec. 7.4.2.2.
1128 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
For an optimal coupling of the light into the cavity the Gaussian laser beam must be
matched to the cavity’s geometry of the cavity, i.e. diameter and divergence of the
laser beam must be adapted to the cavity mode with a suitable arrangement of lenses
[192, 415]. See Excs. 31.1.10.6 to 31.1.10.11.
0.5
R
0
500 600 700 800
λ (nm)
Figure 31.7: (code) Reflection by a high reflecting mirror made of 10 layers with n1 =
2.4 and ∆z1 = 80 nm alternating with 10 layers with n2 = 1.5 and ∆z2 = 500 nm.
The absorption coefficient for each layer is supposed to be α = 0.2%. The beam
impinges from vacuum, n0 = 1.
is greater than the physical diameter of the fiber core, which means that some optical
power is always guided by the fiber cladding as an evanescent wave.
The cut-off wavelength λco of an optical fiber is the wavelength above which a
guided mode of a waveguide ceases to exist. For wavelength longer than λco an
optical fiber becomes single-mode. At wavelengths shorter than λco several optical
modes may propagate and the fiber becomes multi-mode. The cut-off wavelength is
directly related to the core diameter of the fiber λco ∝ ∅. For
λ
2 <∅<λ or equivalently ∅ < λ < 2∅ (31.30)
the fiber is single-mode. For λ > 2∅ no guided mode exists and for ∅ > λ the fiber
becomes multimode.
where δn is the index difference between the core and cladding. An optical fiber with
’high’ numerical aperture will confine light more strongly in the core, and so support
guidance further above cut-off. This attribute has two important effects: (a) it will
be single-mode over a greater range of wavelengths than is possible with a fiber with
a ’low’ numerical aperture fiber and (b) it will still guide a single-mode when coiled
or bent to a smaller diameter.
Figure 31.9: (Left) Fiber patch cord, (center) cross section of a bow-tie polarization main-
taining fiber, and (right) cross section of a photonic crystal fiber.
Figure 31.10: (a) Principle scheme of a fiber-based Sagnac interferometer. (b,c) Laser
gyroscope realized with a HeNe gain tube.
that is,
4πr2 Ω 4πr2 Ω 4AΩ
∆t ≡ t+ − t− = ' ≡ 2 , (31.33)
c2−Ω r2 2 c2 c
where A is the area enclosed by the path. This formula can be generalized to arbitrary
paths.
For example, assuming an interferometer with A = 1 m2 at rest in an Earth-
based system, Ω ' 2π/24 h, the time difference for light propagating along the two
directions is ∆t ≈ 3 · 10−21 s and the path difference ∆L = c∆t ≈ 100 fm. The
frequency shift is,
√
4πAΩ
∆v = k(v+ − v− ) = k(2Ωr) = , (31.34)
λ
31.1.10 Exercises
31.1. INTRODUCTION TO INTERFEROMETRY 1131
The figure 31.11 shows a Michelson interferometer containing in one arm an airtight
5 cm long cell with glass windows. Light with wavelength λ = 500 nm is used. After
the cell has been evacuated, the interference pattern shifts by 60 fringes. Use this
information to calculate the refractive index of air at atmospheric pressure. With
J!K Ì9 what accuracy can you determine the refractive index with this method?
\p\Ni{{dpWpopX
maZIyYZIoqi{ZMZIi
ovµ¶i{"yYZI{max|oYtDRTou^aZ_`*¤Z_`*x|SZ_^aZ_`I
IoZIi{oYZIoUJ`S5ZIi{oYZ{WY¤^*dpi{yu^aZhؼ_S»{b[oYX[Z@WY j[Z_^Ét
´%{b[ma¤ZIopma^aZ_`"oÆX[Z_\p`b[yu^JWY`dYZ[ m
i{`dÆz3i{"y£^vSvi{^
[¼opSQj[Z_`*MZIopdYZ_^_}Jb["ypdYZISdnikZv WY j[Z_^a^aZKZ_jÔbcgWpikZ_`*^
Ôj[Z_`m*"ypikZ_\p^m*i{"ydpb[mRTou^aZ_`*¤Z_`oYZ_~ISWpma^aZ_`WpSÚc¼'qh^a`*ZIikt
MZ_`*ZIynoYZIo¢qYikZ!b[WpmdnikZImaZIovUoYX|bc\(ZIo£dYZIoL`ZIyhWpoYX|mÉt
pdpZFÛKj[x|o¶zÜWY¤^©\ZIiU^*Svx|many3bc `*ZIondY`WpgN
rik^MZI{ypZ_`!´%ZIopb[WpikX[g[ZIik^LgÔb[onoKSvb[odYZIoKL`ZIyhWpoYX|mÉt
pdpZFÛSvik^!dnikZImaZ_`
µ¶Z_^*yYxhdpZ>\(ZIma^*i{SvSZIo³
i{m*"yYZIo¯dYZIo@L{bc^a^aZIo¯pattern
theJik`*ghminimum
^b[
m!dWp opopZ>path qYypi{"y£length
^_ ∆L that with one continuously moving mirror. Estimate
the mirror has to travel in order to reach an
ikZºjhi{ZIkZrRÉou^aZ_`*ªZ_`ZIoYaccuracy
~Ima^a`*ZIik¤ZIom*ofikZIyu^¢∆λZIi{o=L10Z_x[\Nb[yunm
−4
^aZ_`IJindYZ_the
` measurement of a laser wavelength at λ = 600 nm.
|o¯x[\(ZIo¯b[Wpdpi{ZImaZ>UJopx[`dpohWpoYXm*ynb[WY^"³
IorqikZ>ohWpo@b[opdpb[mm©31.1.10.4ZIi{oYb[{kZIopdpZzÜi{yu^
Ex: maZIiMRotating
ZIik¡Y the polarization with a Mach-Zehnder interfer-
"yp{ikZ_¡[^!dpb[mJµ¶Wpm*^aZ_`J{i{opghmSvik^©ZIi
oYZISPyYZI{kZIo¢xdYZ_`!ZIi
oYZISPdpWpoYgkZIo¯qh^a`*ZIi{ªZIo@bc\³
ometer
x|o@%ikZImaZISPÍJb[opd¢b[WpUsing mMX[ZImaZIyYthe
ZIo¢¤x|Jones
kX[ZIoRToumatrix
^aZ_`*¤Z_`*ZIoY~Iformalism
Svi{oni{Svbfigure
udpZ_`*demonstrate
ZIo¢z3bcX[Z%j[x|o@dYthe ` ® ZI½tot use theofMach-Zehnder
Z_how inter-
bc\ny b[ oYX[^_ ® setup
Iop b[ oYX[Z°dYZIm%zÜi{yu^aZIm%ferometer ZI{ypZIm>sketched
µ¶i{oni{S°WnS¨in{ikZ_the X[^dYZISÅÍ
b[optodorotateb[ yYZ_`I|ÈÉZIopZImJpolarization
WY ` a linearly polarized
pb[m
\n{b[WYZ%xdYZ_`ÈaZIoYZIm©laser "WY `!dpb[beam
m©`*x[^aZ°in opandYZ>dpelectronically
ZIm!m*i{"y£^a\nbc`ZIorqhcontrolled
(Z_gu^a`"WpSvm"³ way using a piezo actuator mounted in ones
of the interferometer arms.
1132 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
Figure 31.12:
Ioqi{ZMZIi
ovµ¶i{"yYZI{max|oYtDRTou^aZ_`*¤Z_`*x|SZ_^aZ_`I
31.1.10.5 Ex: Lyot filter
Ii{oYZ{WY¤^*dpi{yu^aZhؼ_S»{b[oYX[Z@WY j[Z_^Ét
Consider a Lyot filter with two plates (L1 = 1 mm and L2 = 4 mm) with the refraction
[Z_\p`b[yu^JWY`dYZ[ m
i{`dÆz3i{"y£^vSvi{^
indices no = 1.40 in the fast axis and ne = 1.45 in the slow axis.
dpb[mRTou^aZ_`*¤Z_`oYZ_~ISWpma^aZ_`WpSÚc¼'qh^a`*ZIikt
angle φ = 45◦ .
b. Determine the transmitted intensity I(φ) as a function of the rotation angle φ for
a fixed wavelength λ. What is the contrast of the transmitted intensity for arbitrary
pmdnikZImaZIovUoYX|bc\(ZIo£dYZIoL`ZIyhWpoYX|mÉt
values of λ if the absorption losses are 2%?
ZIiU^*Svx|many3bc `*ZIondY`WpgN
31.1.10.6 Ex: Wedge-shaped etalon
pb[WpikX[g[ZIik^LgÔb[onoKSvb[odYZIoKL`ZIyhWpoYX|mÉt
A beam of light of wavelength λ = 683 nm with large diameter is incident perpendic-
µ¶Z_^*yYxhdpZ>\(ZIma^*i{SvSZIo³
ularly on the first of two quadratic plates. Each plate has the edge length 120 mm;
at the left edge the plates touch each other, at the right edge they are separated by a
wire of dw = 0.048 mm in diameter. The air between the plates acts as a thin layer.
a. How many interference fringes does an observer see from above this arrangement?
b. Now suppose that the incident light be white. Will the interference pattern at the
far left be bright or dark?
c. Starting from the left edge there will be a series of interference minima, whose
u^¯b[WY°dnikZrx[\(Z_`*Zj[x|o~_MZIiWY \(Z_`*ZIi{opb[ondYZ_`
the minimum be closer to the edge?
%ikZ¯L{bc^a^aZIom*i{opd¶ÈaZ ¦ c¼SvSà{b[oYXYáb[S
m*i{ZZIi{opb[opdYZ_`(b[S»`*ZIyu^aZIo`"b[opdºm*i{opdm*i{Z
dpi{*g[ZIoK`b[y£^Lj[x|oYZIi{opb[ondYZ_`X[Z_^a`*ZIopou^_i{Z
c^a^aZIo¯Jik`*gh^b[
m!dWp opopZ>qYypi{"y£^_
ªZ_`ZIoY~Ima^a`*ZIik¤ZIom*ikZIyu^¢ZIi{oLZ_x[\Nb[yu^aZ_`IJdYZ_`
ImaZ>UJopx[`dpohWpoYXm*ynb[WY^"³
pb[mm©ZIi{oYb[{kZIopdpZzÜi{yu^
maZIiMZIik¡Y Figure 31.13:
pm*^aZ_`J{i{opghmSvik^©ZIi
oYZISPyYZI{kZIo¢xdYZ_`!ZIi
oYZISPdpWpoYgkZIo¯qh^a`*ZIi{ªZIo@bc\³
pd¢b[WpmMX[ZImaZIyYZIo¢¤x|kX[ZIoRTou^aZ_`*¤Z_`*ZIoY~ISvi{oni{SvbudpZ_`*ZIo¢z3bcX[Z%j[x|o@dYZ_` ® ZI½t
31.1.10.7 Ex: Fabry-Perot interferometer
a. Estimate the finesse from (31.24) and (31.27), the maximum transmission, and the
spectral resolution of the FPI for a plate separation of 5 mm.
b. Show that, for a given absorption, the transmitted intensity decreases with in-
creasing reflectivity. Explain why. Note: A trade-off between high finesse and high
transmission at a given absorption A > 0, called impedance matching, is reached by
maximizing the intracavity intensity. For a symmetric cavity, it can be shown that
impedance matching is reached for A = T .
a. Calculate the spectrum of longitudinal and transverse modes for (i) a confocal
cavity (ρa = ρb = L) and (ii) a concentric cavity (ρa = ρv = L/2). Interpret the
results.
b. Assuming radii of curvature ρa = ρb = 5 cm and a finesse of F = 500 for the
cavity, how precise must the length of the cavity be adjusted in order to observe only
longitudinal modes in the transmission spectrum?
A confocal FPI shall be used as optical spectrum analyzer, with a free spectral range
of δf sr = 3 GHz. Calculate the mirror separation L and the finesse that is necessary
to resolve spectral features in the laser output within ∆ν = 10 MHz. What is the
minimum reflectivity R of the mirrors, if the surface finesse is FS = 500?
1134 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
Figure 31.15:
2. Vary the length of one arm of the interferometer using the piezo. Measure
the contrast of the interference fringes and discuss from which parameters it
depends and how it can be maximized.
3. Rotate the first λ/2-waveplate (behind the laser). What do you observe? Ex-
plain the observation!
4. Remove the PBS in front of the photodetector. What do you observe? Explain
the observation!
5. If a piezo is used, vary the voltage applied to the piezo-electric actuator and
measure the voltage expansion coefficient d.
1136 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
Figure 31.17: Setup for (a) a Mach-Zehnder interferometer or (b) a Michelson in-
terferometer. (BS) non-polarizing beamsplitter, (PBS) polarizing beamsplitter. (c)
Signal on the photodetector as a function of the length variation of an arm of the
interferometer.
2. If an ECDL is used, vary the current and the temperature of the laser diode.
What do you observe? Vary the frequency of the diode laser by scanning the
piezo transducer of the laser cavity. Observe the mode spectrum of the laser
in the transmission signal of the cavity. Measure its free spectral range, the
transmission linewidth, and the finesse of the cavity.
Optical cavities are frequently used as optical spectrum analyzers. For this application,
it is helpful to simplify the intrinsic mode spectrum of the cavity by using a confocal
design, where all transerse modes are degenerated. We will now set up an optical
cavity and characterize it by its free spectral range and its finesse. Then we will
analyze its mode spectrum and modify its geometry to make it confocal.
1. Couple a laser beam into a cavity as shown in Fig. 31.18. The cavity provided
by this tinker course consists of a plane incoupler (ρ1 = ∞, R1 = 98%) and a
high reflector (ρ2 = 25 mm, R2 = 99.8%). Position the mirror at a distance
L, where the cavity is stable. Calculate the free spectral range, the finesse, the
diameter of the beam waist.
Figure 31.18: (a) Setup for aligning a confocal resonator. (b) Transmission spectrum of the
cavity for non-confocal alignment. (c) Same as (b) but for the case of confocal alignment.
which can be used to focus down to the beam waist of the cavity. How does the
transmission spectrum change upon the beam matching?
1. Redirect the light of a HeNe laser via two adjustable mirrors into a fiber colli-
mator in such a way that the beam is not deviated from the optical axis by the
collimator.
2. Now connect (a) a multimode fiber and (b) a single mode fiber to the collimator.
Optimize the coupling by walking the laser beam and by adjusting the focus
of the fiber collimator. What differences do you observe for multi- and single
mode fibers?
John von Neumann, and Isidor Rabi, argued that the maser violates Heisenberg’s
uncertainty principle and therefore could not work, in 1964, Charles Townes, Nikolay
Basov and Aleksandr Prokhorov shared the Nobel Prize in Physics for fundamental
work in the field of quantum electronics that led to the realization of oscillators and
amplifiers based on the maser principle.
In 1957, Charles Townes and Arthur Schawlow, from the Bell labs, began to seri-
ously study feasibility of an ’optical maser’. In 1958, the Bell labs submitted a patent
proposing a scheme for optical radiation, and Schawlow and Townes presented a sci-
entific paper. Simultaneously, at the Columbia University, the PhD student Gordon
Gould was working on the energy levels of excited thallium. In 1957-8, Gould and in-
dependently Prokhorov, Schawlow and Townes proposed the use of an open resonator,
which later became an essential component of the laser. Gould also proposed several
possible applications for a laser, such as spectrometry, interferometry, the radar, and
nuclear fusion. He continued to develop the idea, and filed a patent application in
April 1959. The United States Patent Office dismissed his application, and granted
a patent to the Bell Labs in 1960. Gould won his first minor patent in 1977 after
a 28-year fight, and it took him until 1987 to win his first significant process in the
struggle, when a federal judge ordered the United States Patent Office to issue to
Gould patents for optical pumping and the invention of a laser based on the principle
of electrical gas discharge.
It was Theodore Maiman, who on May 16, 1960, operated the first working laser at
the Hughes Research Laboratories, Malibu, California, evincing several other research
teams, including the ones of Townes at Columbia University, of Schawlow at Bell Labs,
and Gould at the company TRG (Technical Research Group). Maiman’s laser used a
synthetic solid-state ruby crystal pumped by a flash lamp to produce red laser light
at 694 nm wavelength; however, the device was only capable of pulsed operation due
to its three-level pumping scheme. Later in 1960, the first gas laser was built, using
a helium-neon mixture, which was capable of continuous operation in the infrared
spectrum. Basov and Javan proposed the concept of a semiconductor laser diode.
In 1962, the first laser diode device, made of gallium arsenide, was realized emitting
near-infrared light. Nowadays, laser diodes are available in various spectral regimes
up to the UV.
Figure 31.20: First suggested application in 1964 of a (left) HeNe laser and (right) a
diode laser.
Interestingly, despite many attempts, it has not yet been possible to manufacture
yellow or green laser diodes.
1140 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
The emission rate depends on the number of atoms N2 in the excited state, such
that,
Rsp ∝ A21 N2 . (31.36)
As the excited state has more energy, it can decay by itself (i.e. spontaneously) to
a lower energy state. Einstein’s brilliant idea now was to postulate a third process,
which he called stimulated emission,
In this process, an incident photon stimulates an excited atom to transfer the electron
to a lower orbit. The released energy is then used to form a second photon, which
is in all respects identical to the first. This process is necessary to ensure that, in
thermal equilibrium, the population of the states follows Boltzmann’s law.
Figure 31.22: Einstein’s model of absorption and spontaneous and stimulated emission.
Obviously, absorption decreases the intensity of a light beam crossing the atomic
sample, while stimulated emission amplifies it. In order to amplify incident light,
31.2. CONVENTIONAL LIGHT SOURCES AND LASERS 1141
the gain in intensity must overcome the losses. Therefore, we need the absorption
processes to be less frequent than the stimulated emission processes, i.e. the number
of atoms in the excited state N2 must exceed the number of atoms in the ground state
N1 < N2 .
We can easily write the rate equation,
dN2 dN1
= −A21 N2 − B21 I(ν)N2 + B21 I(ν)N1 = − , (31.38)
dt dt
with N = N1 + N2 . It is easy to solve this equation. The result is,
I(ν)B21 N
N2 = [1 − e−(A21 +2B21 I(ν))t ] < N1 . (31.39)
A + 2B21 I(ν)
The graphical representation 31.23 shows the temporal behavior of the populations
1000
increasing I(ν)
N1 , N2
500
increasing I(ν)
0
0 1 2 3 4 5
time
N1 (in green in the figure) and N2 (in blue), reaching a state of equilibrium after a
certain time. By increasing the intensity of the incident light, we observe that the
curves approach each other but never cross. That is, in a two-level system, we always
get N1 > N2 and the populations are never inverted. Therefore, amplification of light
as in the laser does not happen.
Fortunately, we can resort to a trick by inserting a third level. Ensuring that the
decay rate of the (metastable) state E3 is much slower than the optical pumping to
this state via the driven transition E1 → E2 followed by a rapid decay E2 → E3 ,
we can reach the situation N3 > N1 . Now it is possible to amplify light, which is
resonant with the transition E2 → E3 .
Figure 31.24: Basics of a laser: (a) Level system and (b) principle scheme.
1142 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
What are the minimum requirements for the realization of a laser? The first
condition is that the pumping cycle is irreversible to ensure that the processes of
stimulated emission and absorption do not compensate. Spontaneous emission is
irreversible and can be incorporated into a three-level system.
The second condition is the existence of a stimulated emission process, because we
want the amplified photon to be an exact copy of the incident photon.
The third requirement is a feedback mechanism that synchronizes the amplification
processes by different atoms in a disordered environment, such as a gas. Mirrors are
ideal because they increase the effective gain path, i.e. the distance within which
inverted atoms can amplify light. Also, the mirrors define the phase of the light wave,
since the standing wave formed by the counterpropagating light fields must have nodes
on the surfaces of the mirrors.
g1
α(ν) = [N1 − g2 N2 ]σ(ν) , (31.41)
is determined by the absorption cross section σ(ν) for the transition and by the
inversion,
∆N ≡ gg21 N2 − N1 , (31.42)
which determines whether stimulated emission prevails or absorption: For ∆N > 0,
the absorption coefficient α(ν) becomes negative and the incident wave is amplified
instead of attenuated [191].
If the active medium is placed between two mirrors [Fig. ??(b)], the wave is
reflected back and forth, and traverses the amplifying medium many times, which
increases the total amplification. With the length L of the active medium the total
gain factor per single round-trip without losses is,
I(ν, 2L)
G(ν) = = e−2α(ν)L . (31.43)
I(ν, 0)
A mirror with reflectivity R reflects only the fraction R of the incident intensity.
The wave therefore suffers at each reflection a fractional reflection loss of (1 − R).
Furthermore, absorption in the windows of the cell containing the active medium,
diffraction by apertures, and scattering due to dust particles in the beam path or due
to imperfect surfaces introduce additional losses. When we summarize all these losses
by a loss coefficient γ, which gives the fractional energy loss ∆W/W per round-trip
time T , the intensity I decreases without an active medium per round-trip as,
Including the amplification by the active medium with length L, we obtain for the
intensity after a single round-trip through the resonator with length d, which may be
larger than L:
I(2d, ν) = I(0, ν)e−2α(ν)L−γ . (31.45)
The wave is amplified if the gain overcomes the losses per round-trip. This implies
that,
−2Lα(ν) = 2L∆N σ(ν) > γ , (31.46)
which yields the threshold condition for the population difference,
γ
∆N > ∆Nthr = . (31.47)
2Lσ(ν)
frequency combs, and atomic clocks (laser-based clocks are up to 1000 times more
stable than the best state-of-the-art microwave clocks).
We all know that light is a wave. With the invention of the laser we found a process
and a device to make this light coherent. On the other hand, since de Broglie’s
assertion we know that matter is a wave, as well. Is it conceivable to construct a
matter laser? Yes, it is! The first coherent matter wave was in fact observed in 1995.
This state of matter, also called Bose-Einstein condensate, was predicted by Bose
and Einstein in 1924. To create a Bose-Einstein condensate, we need, similarly to the
laser, that the matter waves interfere constructively in a way that they amplify each
other. For this, the Broglie wavelength of the particles, which constitute the matter,
must be longer than the distance between them. Assuming a typical average distance
on the order of µm, this corresponds to an average velocity of the particles of mm/s
or a temperature of some 100 nK.
Figure 31.25: (a) Construction scheme of a HeNe laser. (b) Gain and emission profile. (c)
Optical pumping scheme.
Fig. 31.25(a) shows the principle scheme of a commercial HeNe laser. The distance
between the high-reflecting mirror (Rhr ) and the output coupler (Roc ) determines the
free spectral range δf sr . Typically, a HeNe laser operates on two or three longitudinal
modes separated by δf sr . As illustrated in Fig. 31.25(b), the numbers of lasing modes
31.2. CONVENTIONAL LIGHT SOURCES AND LASERS 1145
above threshold depends on the ratio of gain-to-loss. Fig. 31.25(c) shows the optical
pumping scheme to reach inversion on three of the lasing transitions at 632.8 nm,
1.15 µm, and 3.39 µm.
Figure 31.26: (Left) Laser diode with protective housing removed, e.g. using a can
opener. (Right) Laser diode collimator.
Electrons and holes present in the same region may recombine or ’annihilate’ by spon-
taneous emission of photons with energy equal to the difference between the electron’s
original state and hole’s state (see Fig. 30.9). This is in contrast to a conventional
semiconductor junction diode, where the energy released from the recombination is
carried away as phonons, i.e. lattice vibrations. Spontaneous emission below the lasing
threshold is the operating mode of an LED. While spontaneous emission is necessary
to initiate laser oscillation, it contributes to reduce the efficiency of a laser operating
above threshold.
One condition for lasing is that, in the absence of stimulated emission, electrons
and holes may coexist in proximity to one another without recombining immediately.
For typical diode laser materials the ’upper-state lifetime’ or ’recombination time’ is
on the order of a nanosecond. A nearby photon with energy equal to the recombina-
tion energy can cause recombination by stimulated emission. This generates another
photon of the same frequency, polarization, and phase, traveling in the same direction
as the first photon. In this way stimulated emission will cause gain for an optical wave
in the injection region, and the gain increases as the number of electrons and holes
injected across the junction increases.
1146 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
Figure 31.27: Typical beam profile of a multimode laser diode (Thorlabs, L450P1600MM.
For applications requesting small focused beams the waveguide must be made
narrow, on the order of the optical wavelength, such that only a single transverse
mode is supported, and one ends up with a diffraction-limited beam. Such single
spatial mode devices are used for optical storage, laser pointers, and fiber optics.
Note that these lasers may still support multiple longitudinal modes, and thus can
lase simultaneously at multiple wavelengths. The wavelength emitted is a function of
the band-gap of the semiconductor material and the modes of the optical cavity. In
general, the maximum gain will occur for photons with energy slightly above the band-
gap energy, and the modes nearest the peak of the gain curve will lase most strongly.
The width of the gain curve will determine the number of additional ’side modes’
that may also lase, depending on the operating conditions. Single spatial mode lasers
that can support multiple longitudinal modes are called Fabry-Pérot (FP) lasers. A
FP laser will lase at multiple cavity modes within the gain bandwidth of the lasing
medium. The number of lasing modes in an FP laser is usually unstable, and can
fluctuate due to changes in current or temperature.
Single spatial mode diode lasers can be designed so as to operate on a single lon-
gitudinal mode. These single frequency diode lasers exhibit a high degree of stability,
and are used in spectroscopy and metrology, and as frequency references. Single
frequency diode lasers are classed as either distributed feedback (DFB) lasers or dis-
tributed Bragg reflector (DBR) lasers.
31.2. CONVENTIONAL LIGHT SOURCES AND LASERS 1147
Due to diffraction, the beam diverges (expands) rapidly after leaving the chip,
typically at 30 degrees vertically by 10 degrees laterally. A lens must be used in order
to form a collimated beam like that produced by a laser pointer. If a circular beam is
required, cylindrical lenses and other optics are used. For single spatial mode lasers,
using symmetrical lenses, the collimated beam ends up being elliptical in shape, due
to the difference in the vertical and lateral divergences.
31.2.3.4 ECDL
An extended-cavity diode laser (ECDL) is an optical setup based on a laser diode
chip, which typically has one end anti-reflection (AR) coated, and the laser resonator
is completed with a collimating lens and a mirror, as shown in Fig. 31.28(a). The
extended external laser resonator introduces various new features and options: Com-
pared to a standard laser diode, the longer resonator increases the damping time of the
intracavity light according to Eq. (31.26), and thus allows for lower phase noise and
a smaller emission linewidth (in single-frequency operation). Furthermore, it opens
the way for inserting frequency-selective optical components into the extended laser
resonator, such as narrow-band Fabry-Pérot etalons or diffraction gratings, which can
further reduce the linewidth and even allow to tune and control the frequency of the
laser.
Tunable ECDLs based on diffraction grating as the wavelength-selective element
are also called grating-stabilized diode lasers. The common Littrow configuration
Fig. 31.28(b) generates optical feedback to the laser diode chip by retro-reflecting the
first-order diffracted beam from the grating. The emission wavelength can be tuned
by slightly tilting the diffraction grating. A disadvantage of this configuration is, that
1148 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
Figure 31.28: (a) ECDL with an AR-coated laser diode and an external mirror. (b) Littrow
configuration. (c) Littmann configuration.
the tilt also changes the direction of the output beam, which is inconvenient for many
applications.
In the Littman-Metcalf configuration Fig. 31.28(c), the grating angle is held fixed,
and an additional mirror is used to reflect the first-order beam back into the laser
diode. The wavelength can be tuned by rotating that mirror. This configuration offers
a fixed direction of the output beam, and also tends to exhibit a smaller linewidth, as
the wavelength selectivity is stronger, because the wavelength-dependent diffraction
occurs twice per resonator round trip. A disadvantage is that the zero-order reflection
of the beam reflected by the tuning mirror is lost, so that the output power is lower
than that of a Littrow laser.
Figure 31.29: (Left) Side view of a ’free-running’ laser diode mounted in a collimator
(Thorlabs, LT110P-B) clamped in an aluminum block cooled by a Peltier element
(Thorlabs, TEC3-6) and whose temperature is measured by a thermistor (Thorlabs,
TH10K). (Right) Top view of a home-built ECDL laser in Littrow configuration. The
diode collimator is clamped into the left mount. a holographic grating (Newport,
10HG2000-475-1) is glued to the right mount, whose angle can be adjusted mechani-
cally and via a piezo. The whole setup is cooled by a Peltier element mounted on the
bottom of the base plate.
New concepts have recently become popular, such as the so-called cat-eye laser [42,
31.2. CONVENTIONAL LIGHT SOURCES AND LASERS 1149
279, 64], where the frequency-selective element is an extremely narrow-band (0.3 nm)
optical filter 4 .
By adjusting the tilt angle of a grating or a narrow-band filter by means of a piezo
an extremely fine tuning of the emission frequency is possible, while coarse tuning of
the frequency over a range of several nanometers is typically achieved by changing
the temperature and the laser current. Typical linewidths of free-running ECDLs are
well below 5 MHz. Controlling the laser temperature, current, and piezo voltage by
active feedback circuits (e.g. within a Pound-Drever-Hall servo electronics) emission
bandwidths in the milliHertz range have been achieved, which corresponds to quality
factors of the laser oscillator of up to 1018 .
Figure 31.30: (Left) Construction plan of a home-built ECDL lasers in cat-eye config-
uration. (Right) Side view of the cat-eye laser.
In comparison to other laser types, a diode laser exhibits, the advantage of a very
small size and a compact design. They are, in general, easy to handle and can be
controlled conveniently via current and temperature. However, they also have the
disadvantage of a large beam divergence and a broad emission spectrum. The beam
divergence can be compensated by a collimation optics in front of the laser diode.
The temperature has an impact on the band structure of the pn-transition of
the laser diode and hence on the frequency. Therefore, it is stabilized via a Peltier
element, which is mounted underneath the laser diode holder. The degree of freedom
is used for tuning the laser frequency in wide steps.
the drive current is modulated. Passive mode locking relies on a saturable absorber in
the resonator. This can simply be an unpumped section of the device. It is common to
apply an electrical bias for adjusting the absorber properties. However, the recovery
time of that kind of absorber is fairly long. Shorter recovery times are achieved e.g.
by implanting nitrogen (N + or N2+ ) ions from one facet. This introduces crystal
defects, where carriers can recombine. The absorber is often placed at a resonator
end, but it can also be placed somewhere within the resonator so that different pulses
can meet in the absorber (colliding pulse mode locking).
For pulse repetition rates roughly below 10 GHz, an external cavity setup is usually
required, as a monolithic device would become too long. The extended cavity may be
an ECDL setup. Another technical approach is to incorporate the semiconductor chip
into a ring laser resonator made of optical single-mode fiber. In the latter case, the
resonator is typically much longer, and allows the use of fiber-optic components. The
semiconductor device may then be a fiber-coupled semiconductor optical amplifier
(SOA).
External-cavity lasers have various advantages: The pulse repetition rate can be
chosen in a wide range, and can easily be tuned e.g. by moving the end mirror, or with
a fiber resonator by stretching a piece of fiber with a piezo transducer. It is possible
to insert an optical filter for fixing the emission wavelength, or use a diffraction
grating as the end mirror (Littrow configuration; see the article on external-cavity
diode lasers). Even for higher pulse repetition rates, where harmonic mode locking is
required, external-cavity devices can be advantageous, because they have a potential
for lower laser noise, e.g. in the form of timing jitter. Therefore, mode-locked external-
cavity diode lasers sometimes compete with mode-locked fiber lasers in areas where
monolithic laser diodes would not be suitable. On the other hand, a monolithic setup
with fundamental mode locking can be very compact, much cheaper to manufacture,
and can exhibit very robust pulse emission.
On the other hand, mode-locked diode lasers are subject to various limitations,
which do not allow them to reach the full performance potential of, e.g., mode-locked
fiber lasers: The pulse energy is fairly limited often far below 1 pJ. Average output
powers are often below 1 mW. Due to the short upper-state lifetime, ultrafast semi-
conductor lasers are generally not suitable for lower repetition rates of e.g. well below
1 GHz, except with synchronous pumping. Although the gain bandwidth of semicon-
ductors would be compatible with pulse durations of a few tens of femtoseconds, the
pulse durations achieved are usually much longer at least hundreds of femtoseconds,
and often picoseconds. The pulse formation dynamics are relatively complicated,
e.g. due to nonlinear phase changes associated with gain saturation, and difficult to
optimize. The pulse quality is normally not as good as e.g. for mode-locked fiber
lasers. In particular, there are often additional satellite pulses, caused e.g. by im-
perfections of the anti-reflection coating. Also, the pulses are often chirped, i.e. they
are not bandwidth-limited. The timing jitter and the noise of other pulse parameters
are higher than for other mode-locked lasers. This is partly a consequence of the low
power level.
31.2. CONVENTIONAL LIGHT SOURCES AND LASERS 1151
Figure 31.31: (a) Tapered amplifiers are available e.g. from Eagleyard. (b) Principle of
injection locking.
The description presented here uses semi-classical laser rate equations [453]. As-
suming that the master and the slave laser field are given by, respectively,
Einj = Ainj e−ıωinj t eıφinj (t) and E = Ae−ıωs t eıφs (t) , (31.49)
The phase difference between the both fields is denoted by φ(t) = φs (t)−φm (t). Now,
Considering the semi-classical laser rate equations we can describe the impact of the
master laser field in the slave laser,
dA(t) 1
= g[N (t) − Nth ]A(t) + κAinj cos φ(t) (31.50)
dt 2
dφ(t) α Ainj
= [N (t) − Nth ] − κ sin φ(t) − ∆ω
dt 2 A(t)
dN (t)
= J − γN N (t) − [γp + g[N (t) − Nth ]A(t)2 ,
dt
where A(t) is the field amplitude normalized as A2 (t) = S(t), and S(t) is the photon
number. N (t) is the number of carriers in the slave laser, and the other parameters
are g laser gain coefficient, Nth threshold carrier number, κ coupling coefficient, α
linewidth enhancement factor, γp photon decay rate, N carrier recombination rate,
J pump current normalized by electron charge, ∆ω frequency difference between the
master and the free running slave ωm − ωs . The parameter κ describes the rate at
which the photons of the master laser enter into the slave laser cavity and is given in
terms of the cavity quality factor,
ωs
κ= . (31.51)
2Q
From the steady state solutions of above equations, we can obtain the frequency
locking range, r r
p 2 Pinj Pinj
−κ (1 + α ) < ∆ω < κ , (31.52)
Ps Ps
1152 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
where Pinj /Ps is the master laser fraction power used for the injection locking and Ps
is the power of slave laser. From equation 4.25 we can see that the locking range is
determined by the amplitude ratio between the fields and by the cavity quality factor
since κ ∝ Q−1 . Therefore, lasers with low Q are easier to lock. On the other hand,
this leads to increased laser linewidth that reduces the phase noise performance of the
injection locking systems. For higher injection ratio Pinj /Ps also results in a large
locking range, which also makes the lock easier to achieve.
31.2.4 Exercises
31.2.4.1 Ex: Conventional light sources and lasers
Compare the properties of an incandescent light and a laser.
1. Couple the light of a HeNe laser simultaneously into an optical spectrum ana-
lyzer and a radiofrequency spectrum analyzer. What do you observe when you
1154 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
slightly heat the laser housing? Calculate from your observations the length of
the laser cavity.
2. Pass the light through a λ/4-plate and then through a polarizing beam splitter.
What do you observe in the two output ports of the PBS?
2. Optimize the threshold. Analyze the emission spectrum with an optical spec-
trum analyzer.
Figure 31.32: (a) Principle of the acousto-optic modulator. (b) Scheme of the diffraction
in an acousto-optic modulator: A photon with wavevector k is scattered by a phonon with
wavevector k1 resulting in a photon with wavevector kf .
Since the laser beam is refracted when it enters the crystal, the relation between
the incidence and exit angle is given by Snell’s law, sin α = n sin θ. With this, the
Bragg condition can be written,
fλ
sin α = . (31.54)
2cf
The angle between the 0th and the 1st order is, hence, 2α.
In a corpuscular picture, the process can be understood as a four-wave mixing
(4WM) between photons and phonons. The deflection of the laser beam is a con-
sequence of momentum, k1 = k + kf . The frequency shift corresponds exactly to
the Doppler shift induced by the Brillouin scattering (absorption and reemission of a
phonon in reverse direction), and we obtain a relationship that is equivalent to the
Bragg condition,
cf sin θ
f = ν1 − ν = 2ν . (31.55)
c/n
From the Bragg condition, knowing the deflection angle and the (fixed) frequency
shift, we can calculate the sound velocity. A typical value is cf ' 4200 m/s 5
An AOM works best (highest diffraction efficiency in to the first Bragg order, which
may reach more than 90%) at a specific radiofrequency, which typically is located
somewhere in the range f = 40...800 MHz, the most common one being 80 MHz.
Deviations from this ’center frequency’ are possible within a range of typical ±10%
of the center frequency.
5 The result (31.55) can be derived from conservations laws for energy ω = ω +ω and momentum
1 f
k1 = k + kf . Defining ω ≡ 2πν, ω1 ≡ 2πν1 and ωf ≡ 2πf we find,
EOMs are often used to generate sidebands in a monochromatic laser beam. They
are also used as Pockels cell, i.e., as a voltage-controlled phase-plate. The Pockels
effect produces in a medium a birefringence, which depends linearly on the applied
electric field. This is in contrast to the Kerr effectKerr, in which the birefringence
depends in a quadratic form of the electric field.
Suppose the optically inactive axis is x. In this case, the influence of EOM on the
polarization of a laser beam is described by
1 0
MEOM (θ) = . (31.56)
0 eiθ
For operation as a Pockels cell, the EOM is inserted between two crossed polarizers
31.3. INTRODUCTION TO OPTICAL PHASE AND FREQUENCY MODULATION1157
oriented, e.g., along the x and y axis. The EOM itself is rotated by an angle φ,
0 0 cos φ sin φ 1 0 cos φ − sin φ 1 0
MP ockels (θ, φ) =
0 1 − sin φ cos φ 0 eiθ sin φ cos φ 0 0
0 0
= 2ieiθ/2 sin θ2 sin φ cos φ , (31.57)
1 0
That is, an incident beam of light, E = Eêx , linearly polarized in x-direction is rotated
into the y-direction and, depending on the phase shift θ, it is completely blocked or
transmitted through the Pockels cell [see Fig. 31.35(a)],
Itr = I0 sin2 θ
2 . (31.59)
(a) 1 (b) 1
Jk (β)
0.5 0
It
0 -1
0 0.5 1 0 2 4 6
φEOM /π β
90 1
Figure 31.35: (code) (a) 120Effect of60a Pockels cell. The solid line was calculated with the
Eq. (31.57), the dotted 150
was measured
0.5 30experimentally. (b) Lowest-order Bessel functions.
180 (c) 0
210 330
31.3.3 Optical phase
240 modulation
300
270
The frequency and the phase of a laser beam can be influenced and modulated sim-
ilarly to radiofrequency signals. We can therefore use the calculation of Sec. 30.3.1
completely, only changing the carrier frequency to be the frequency of the light: The
Fourier expansion of a complex-valued periodic function s(x) into a series is defined
as,
N
X Z
1
sN (x) = ck eikx where ck = s(x)e−ikx dx . (31.60)
2π 2π
k=−N
Applying this to the modulated phase shift factor s(Ωt) = eiθ(t) with θ(t) ≡ β sin Ωt,
we get,
X∞ Z π
1
sN (Ωt) = ck eikΩt where ck = eiβ sin Ωt e−ikΩt dΩt , (31.61)
2π −π
k=−∞
1158 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
but the Fourier coefficients are nothing else than the integral definition of the k-th
order Bessel function, Z π
1
Jk (β) ≡ ei(β sin τ −kτ ) dτ , (31.62)
2π −π
where J−k (β) = −Jk (β). Hence, we may write the electric field,
∞
X
E(t) = ei[kz−ωt+iθ(t)] = ei(kz−ωt) Jk (β)eikΩt
k=−∞ . (31.63)
' J0 (β)eiωt + J1 (β)eiωt+iΩt + J−1 (β)eiωt−iΩt
For small modulation indices β only the lowest-order Bessel function contribute no-
ticeable amplitudes, as illustrated in Fig. 31.35(b).
The interpretation of this is, that phase modulation imprints sidebands onto a
monochromatic laser beam. These sidebands are independent modes which can be
resolved, e.g. with an optical spectrum analyzer, as illustrated in Fig. 31.36.
10 30
fluorescence (a.u.)
(a) (b)
20
5
10
0 0
-10 0 10 -10 0 10
frequency (MHz) frequency (MHz)
Figure 31.36: (code) Phase modulation sidebands with high (left) and low (right) modulation
index resolved by a narrow (red) or broad (blue) filter.
Figure 31.37: (a) Scheme for phase modulation of a diode laser by modulating the drive cur-
rent. (b) Phase modulation with an external EOM followed by frequency-selective absorption
of the lower sideband.
31.3.4 Exercises
31.3.4.1 Ex: Response time of an AOM
A beam of light at 689 nm focused to a waist of 100 µm passes through the crystal
of an 80 MHz AOM, characterized by a sound velocity of cs = 4200 m/s.
a. By how much the first diffraction order is deflected by the AOM? Regarding the
beam divergence of the Gaussian beam, will it be possible to spatially separate the
diffraction orders?
b. An experimentalist ramps the driving frequency between 70 and 90 MHz by means
of a voltage-controlled oscillator. What is the range of diffraction angles covered?
c. Estimating the response time of the AOM by the time that the traveling sound
wave needs to cover a distance corresponding to the focus of the light beam, how fast
can the experimentalist switch off the light beam by suddenly interrupting the driving
signal? What is the modulation bandwidth of the AOM?
d. The light beam passes through the AOM at a distance of d = 2 mm from the piezo
transducer generating the sound wave. How will this fact limit response time?
data sheet for the power divider see appendix Fig. 34.17,
data sheet for the mixer see appendix Fig. 34.18
1162 CHAPTER 31. QUANTUM OPTICS AND OPTICAL INTERFEROMETRY
Figure 31.39: Analogy between radio transmission (a) and heterodyne techniques with a
laser (b).
In practice, laser beams are usually not plane waves, but have a finite diameter
and radius of curvature. In order to get a high contrast signal, a good phase-matching
of the beams is important in order to obtain a strong photodiode signal.
31.4. RADIOFREQUENCY TECHNIQUES AND THE TRANSFER OF INFORMATION1163
we see, that the photodetector signal will reproduce the modulation signal. In other
words, we encoded information on a laser beam, which carries it (e.g. through an
optical fiber) to another place.
Figure 31.41: Principle scheme of the (a) homodyning and (b) heterodyning technique at
the example of a Mach-Zehnder interferometer. The components in the yellow area of (b)
constitute a Lock-In amplifier.
The photodetector signal generated by the beams after their recombination at the
beam splitter is,
2
I ∝ (1 − η)Er eıkLr + ηEt eıkLt +ıΩt (31.69)
2
= −(1 − η)ηEi eıkLr + η(1 − η)Ei eıkLt +ıΩt
2 2
= |(1 − η)ηEi | −e2ıkLr + eıkLt +ıΩt .
Now, demodulating this signal simultaneously with the local oscillator frequencies
cos ωlo t and sin ωlo t, we get,
where the oscillating terms can be removed by a low-pass filtering. Finally, we calcu-
late, p
Uc2 + Us2 Us
|Esig | = and tan φ(t) = , (31.74)
|Elo | Uc
and obtain the electric field via,
p
ıφ(t) Uc2 + Us2 ı arctan UUs
Esig = |Esig |e = e c . (31.75)
|Elo |
31.4. RADIOFREQUENCY TECHNIQUES AND THE TRANSFER OF INFORMATION1165
31.4.5 Exercises
31.4.5.1 Ex: Pound-Drever-Hall signal
Consider the photodetector signal of Exc. 31.3.4.4. What signal do you observe when
demodulating the signal with an oscillation of frequency f ? Calculate the derivative
of the signal close to resonance; from which parameters does the slope depend, and
how must you choose the modulation index to maximize it?
Optical spectroscopy
Modern ’optics’ is to be understood as ’physics of light-matter interaction’ in the
optical energy regime. In this sense, this area of physics comprises quantum optics,
photonics, atomic physics, and atom optics. Since the invention of the laser the
field of optics has seen a huge technological progress leading to the development of
extremely powerful and precise tool for investigating and manipulating matter. The
femtosecond laser, the frequency comb, atomic interferometers and clocks, and Bose-
Einstein condensation are just a few examples.
Spectroscopy is the art of taking and interpreting spectra, i.e. frequency-dependent
response functions. The variety of spectroscopic techniques is so overwhelming that
a survey is hopeless. As the course also aims at familiarizing the student with ap-
plications, a major part of this course will concentrate on techniques employed and
available in quantum optics labs. These techniques are mostly oriented toward ultra-
high resolution spectroscopy and techniques of manipulating the motion of atoms.
In Sec. 32.1 to 32.4.3, we will try various spectroscopic techniques applied to
atomic or cavity resonances.
sin α sin 21 (θ + )
n= = . (32.1)
sin β sin 21
Hence,
dn 1 cos 12 (θ + )
= , (32.2)
dθ 2 sin 21
1167
1168 CHAPTER 32. OPTICAL SPECTROSCOPY
or
dθ sin 12
= . (32.3)
dn 1 − n2 sin 12
The angular dispersion is therefore,
dθ sin 12 dn
= . (32.4)
dλ 1 − n2 sin 21 dλ
Figure 32.1: Illustration of (a) the prism spectrometer and (b) the grating spectrometer.
1
dθ = p dn ≈ 0.8◦ .
1 − (n/2)2
The resolving power can be calculated, once we have expressed the limiting aper-
ture a = d cos α where d = g/(2 sin 2 ),
λ dθ g cos α dn dn
∆λ = a dλ = 1 − n2 sin 1 dλ = g dλ . (32.5)
2
2π d
δ= λ ∆s = −2π sin β . (32.7)
λ
32.1. SPECTROMETER AND MONOCHROMATOR 1169
The sum of the reflections from N grooves of an incident plane wave Ein = E0 eı(k·r−ωt
is therefore,
N
X
√ √ 1 − eiN δ
Er = RE0 eıkr ·r−ωt eimδ = RE0 eikr ·r−ωt . (32.8)
m=0
1 − eıδ
sin2 N2 δ
Ir = RI0 . (32.9)
sin2 12 δ
0.8
0.6
Ir /I0
0.4
0.2
0
-10 -5 0 5 10
δ/2π
32.1.3 Exercises
32.1.3.1 Ex: Resolution of a grating spectrometer
Calculate the spectral resolution of a grating spectrometer with an entrance slit width
of 10 µm, focal lengths f1 = f2 = 2 m of the mirrors M1 and M2, a grating with
1800 grooves/mm and an angle of incidence α = 45◦ . What is the useful minimum
slit width if the size of grating is 100 × 100 mm2 ?
1170 CHAPTER 32. OPTICAL SPECTROSCOPY
different components of the fluorescence spectrum are measured, thus yielding a curve
u(ν), which represents the spectral energy density of the radiator [see Fig. 32.5(a,d)].
Figure 32.5: (a) Taking a fluorescence spectrum, (b) an excitation spectrum, and (c) an ab-
sorption spectrum. Typical level schemes for (d) fluorescence spectroscopy and (e) excitation
or absorption spectroscopy.
Figure 32.6: (a) Experimental scheme for saturated absorption spectroscopy. (b) Spectral
hole burning by the counter-propagating saturation and probe beams for (red) detuned and
(blue) resonant light.
P
n(T ) = . (32.13)
kB T
We also assume a cell length of L = 10 cm.
The probe laser intensity is below saturation, such that the optical cross section
for an atom moving with velocity v, is,
6π Γ2
σ(v) = , (32.14)
k 4(ω − ω0 − kv)2 + Γ2
2
where we considered the fact that the atoms moving with the velocity v along the
optical axis perceive the probe laser beam as Doppler-shifted by an amount kv.
The saturating laser now has high intensity. Let us suppose here, Ω ≡ 10Γ, where
Ω is the Rabi frequency caused by the saturating laser. In this way, it creates a
population of Ne atoms in the excited state. Since this population is missing in the
ground state, Ng = N − Ne , the absorption is reduced for the probe beam by a factor,
Ne Ω2
= . (32.15)
N 4(ω − ω0 + kv)2 + 2Ω2 + Γ2
1.5 1
MHz
0.5
0.5
0 0
-500 0 500 -500 0 500
Δ (MHz) I/I0
Figure 32.7: (code) (a) Optical density and (b) absorption. (Blue) Integral formula and
(green) approximation for high temperature and high saturation.
p
where v̄ = kB T /m is the mean atomic velocity (or the rms width) of Maxwell’s
distribution. Since the spectral width of the probe laser spectrum is much smaller,
we can replace it by a δ-function,
Γ2 πΓ
2 2
−→ δ(∆ − kv) , (32.18)
4(∆ − kv) + Γ 2
which gives,
r Z ∞
P m 6π 2Ω2
OD (T, ω) ' L 1− × (32.19)
kB T 2πkB T k 3 −∞ 4(∆ + kv)2 + 2Ω2 + Γ2
πΓ 2
× δ(∆ − kv)e−mv /2kB T dkv
2
r
P m 6π πΓ 2Ω2 2
=L 1− e−m(∆/k) /2kB T .
kB T 2πkB T k 3 2 8∆2 + 2Ω2 + Γ2
The Lamb dip is the narrow (Doppler-free) feature in the center of the the spectrum
exhibited in Fig. 32.7. Lamb-dip spectra are commonly serve as frequency references
for laser frequency stabilization schemes.
32.2. FLUORESCENCE, EXCITATION, AND ABSORPTION SPECTROSCOPY1175
The advantages of both techniques is, that they generate dispersive Doppler-free
lineshapes. The FMS signal appears as a sharp feature on top of a large Doppler
background (similarly to the Lamb-dip). In contrast, the MTS signal is free from
Doppler background. The MTS signal recorded by the photodetector is given by,
X µ2 1 1
ab
I(∆) = − (32.20)
γj + iδ γab + i(∆ + δ/2) γab + i(∆ + δ)
a,b
1 1
+ − ,
γab − i(∆ − δ) γab − i(∆ − δ/2)
where a and b denote the lower and upper levels, µ2ab is the electric dipole moment,
γab is the optical relaxation rate, γj is the decay of the energy level j of the molecule,
and δ is the modulation frequency. Behind the mixer we see the electric signal,
where ϑ is the demodulation phase [462]. The curves are shown in Fig. 32.8.
(a) (b)
m = 0.1
0.3
1
arb. units
arb. units
3
10
Figure 32.9: (code) (a) Calculated FMS signals and (b) MTS signals as a function of detuning
for various modulation indices.
general form P (ω + Ω) = χ(3) (ω + Ω)Es (ωs )Ep (ωs )Es∗ (ωs − Ω), or similar. When
the beams ω and ωs are counter-propagating, and the atoms are moving, their
resonances shift toward ω∓k·v, respectively in the atomic rest system. Resonant
enhancement of 4WM occurs, when one of the intermediate levels coincides with
ω0 :
ω − k · v = ω0
(ω − k · v) − (ω + k · v − Ω) = 0
(ω − k · v) − (ω + k · v − Ω) + (ω + k · v) = ω0
(ω − k · v) − (ω + k · v − Ω) + (ω + k · v) − (ω − k · v + Ω) = 0 .
k · v = 0, ±Ω ,
and similarly for the other 4WM processes. This means, that the saturation
beam burns holes in the velocity distribution at k · v = 0, ± 21 Ω, ±Ω, which
modulate the probe beam.
32.2.4 Exercises
32.3. POLARIZATION SPECTROSCOPY 1177
Calculate the FMS spectrum for a rubidium gas (describing the atoms as a two-
level system) under the conditions specified in Sec. 32.2.2 and assuming a modulation
frequency of f = 10 MHz and a modulation index of β = 1.
2. Fig. 32.12 shows a typical spectrum recorded with a rubidium gas cell. Find an
interpretation for the various lines of the spectrum.
Figure 32.12: (Black) Lamb-dip spectroscopy of a rubidium gas (natural isotope mixture of
85
Rb and 87 Rb) showing the hyperfine structure transitions of the D2-line. (Red) Derivative
of the spectrum in (a).
(b)
resonance frequencies of the s-polarized and the p-polarized modes are slightly shifted
from one another due to the different penetration depth of the s- and p-polarized light
modes into the layers of the dielectric mirrors. For a moderate finesse of the cavity
(say F = 2000), the modes actually overlap. This leads to a birefringence used in the
(b)
signal (arb. units)
0.4
0.2
-0.2
-0.4
-50 0 50
ν (MHz)
Figure 32.14: (code) (a) Generating a Hänsch-Couillaud error signal by (a) reflec-
tion from a ring cavity and (b) transmission of a linear cavity containing a Brew-
ster plate. (b) Signals recorded by the two photodetectors (blue and green) and
their difference (red). (a) Reflection signals |Es |2 and |Ep |2 from a birefringent
cavity with δf sr = 8.2 GHz, ϕs = 0, Rhr,s = 99.97%, Ric,s = 99.74%, Fs =
2
π(Rhr,s Ric,s )1/6 /[1 − (Rhr,s
2
Ric,s )1/3 ], ϕp = 0.01, Rhr,p = 99.92%, Ric,p = 99.34%,
2
and Fp = π(Rhr,p Ric,p )1/6 /[1 − (Rhr,p
2
Ric,p )1/3 ]. (b) Difference |Es |2 − |Ep |2 .
famous Hänsch-Couillaud locking scheme. The detector signal in the scheme shown
in the figure may calculated via a concatenation of the Jones matrices for a λ/2-plate,
the reflective response of the ring cavity, another λ/2-plate, and finally a polarizing
1180 CHAPTER 32. OPTICAL SPECTROSCOPY
beam splitter,
Es 1
=M with (32.25)
Ep 0
1−e−2πiω/δf sr +iφs
cos β sin β 1−Rs e−2πiω/δ+iφs 0 cos α sin α
M≡ .
− sin β cos β 1−e−2πiω/δf sr +iφp − sin α cos α
0 1−Rp e−2πiω/δ+iφp
Calculating |Es |2 and |Ep |2 as a function of the laser frequency ω reproduces the
curves plotted in Fig. 32.14(a).
1. x
where n is an integer, frep is the comb tooth spacing (equal to the mode-locked laser’s
repetition rate or, alternatively, the modulation frequency), and fceo is the carrier
offset frequency, which is less than frep . Combs spanning an octave in frequency (i.e.,
a factor of two) can be used to directly measure (and correct for drifts in) fceo . Thus,
octave-spanning combs can be used to steer a piezoelectric mirror within a carrier-
envelope phase-correcting feedback loop. Any mechanism by which the combs’ two
degrees of freedom (frep and fceo ) are stabilized generates a comb that is useful for
mapping optical frequencies into the radio frequency for the direct measurement of
optical frequency.
(a) 1 (b) 1
E(t)
E(t)
0 0
-1 -1
0 10 20 30 40 50 3.999 3.9995 4 4.0005 4.001
t (ns) t (ns)
(c) ×10−3 (d) ×10−3
3 3
2 2
E(ν)
E(ν)
1 1
0 0
0 5 10 4.999 4.9995 5 5.0005 5.001
ν (THz) ν (THz)
The carrier under each pulse is phase-shifted with respect to the adjacent pulse, except
when the laser is mode-locked, that is, when,
1
cos 2πνt = cos 2πν(t + frep ) . (32.28)
we get,
√ 2 2 X
E(ω) = F[E(t)] = 21 T πe−T (ω−2πν) /4 e−i(ω−2πν)n/frep . (32.31)
n
where we set m ∈ N and call |fceo | < frep the carrier envelope offset. We also express
the Fourier frequency by,
ω ≡ 2π(ηfrep + fceo ) , (32.33)
firstly without specifying that η be an integer number. The spectrum is,
2
X
2 1 −T 2 (ω−2πν)2 /2 −2πin(η−m)
|E(ω)| = 4 πT e e , (32.34)
n
The δ-function comes from the fact that the sum of (32.35) over many oscillations
e−2πin(η−m) vanishes by destructive interference, except when they are in phase, which
is just the case when ω = nfrep + fceo .
Figure 32.16: (a) Scheme and (b) operation principle of a frequency comb with control of
the repetition rate and the carrier envelope offset.
or,
fceo = −∆ν . (32.41)
gives information about the vibrational modes in the system. Infrared spectroscopy
typically yields similar, complementary, information.
Figure 32.18: Energy-level diagram showing the states involved in Raman spectra.
L.S. Ma et al., Spectroscopy of Te2 with Modulation Transfer Reference Lines for
Precision Spectroscopy in Yb [DOI]
Chapter 33
Locking circuits
In a laboratory we are often confronted with the need to control the value of a physical
parameter, f.ex., room temperature, currents and voltages, or the frequency and in-
tensity of laser beams. The physical discipline dealing with the fundamental concepts
of this field is called control theory and its application to development of automatic
control systems is called control engineering.
In this chapter, after a brief introduction into control theory, we will design and
construct a few automatic control systems, which are common in quantum optics labs.
1189
1190 CHAPTER 33. LOCKING CIRCUITS
control action to give a process output the same as the ’reference input’ or ’set point’.
For this reason, closed loop controllers are also called feedback controllers.
The definition of a closed loop control system is a control system capable of can-
celing the deviation of a system variable from a reference value by means of a feedback
signal computed from a measured value of the variable and used to act on the system
in a controlled way [679]. Automatic feedback control has revolutionized all areas of
human activities.
To overcome the limitations of the open-loop controller, control theory introduces
feedback. A closed-loop controller uses feedback to control states or outputs of a
dynamical system. Its name comes from the information path in the system: process
inputs (e.g. voltage applied to an electric motor) have an effect on the process outputs
(e.g. speed or torque of the motor), which is measured with sensors and processed
by the controller; the result (the control signal) is ’fed back’ as input to the process,
closing the loop.
Closed-loop controllers have the following advantages over open-loop controllers:
disturbance rejection (such as hills in the cruise control example above)
guaranteed performance even with model uncertainties, when the model struc-
ture does not match perfectly the real process and the model parameters are
not exact
unstable processes can be stabilized
reduced sensitivity to parameter variations
improved reference tracking performance
In some systems, closed-loop and open-loop control are used simultaneously. In such
systems, the open-loop control is termed feedforward and serves to further improve
reference tracking performance.
s ≡ ıf (33.1)
Figure 33.1: (a) General schematic diagram of a feedback loop. (b) Possible implementation
for the frequency stabilization of a laser to a cavity resonance,
P (f )C(f )
Y (f ) = R(f ) ≡ H(f )R(f ) . (33.3)
1 + P (f )C(f )
H(s) is referred to as the closed-loop transfer function of the system. The numerator
V (f ) ≡ P (f )C(f ) is called the forward gain (open-loop gain) from R to Y , and the
denominator is one plus the gain in going around the feedback loop, the so-called
loop gain. If |P (f )C(f )| 1, i.e. it has a large norm with each value of f , and then
Y (f ) ' R(f ) and the output closely tracks the reference input.
or,
1 1
Y (f ) = [R(f ) + Zc (f )] + Zp (f ) . (33.5)
1 + V (f )−1 1 + V (f )
This shows that, provided the open loop gain is high enough, perturbations affecting
the plant can be efficiently be neutralized. In contrast, noise entering via the control
cannot be suppressed, and this fact is independent on the chosen controller transfer
function: Perturbations entering between the measurement point and the input of the
regulator are not eliminated ! Consequently, the detector (which generally works with
very low signals) should not introduce or let penetrate noise, because this affects the
variable to be controlled: Any variation of the steering variable at the regulator will
be transmitted 1 to 1.
In the following, we will discuss the most common controller called PID-servo.
Figure 33.2: A block diagram of a PID controller in a feedback loop, r(t) is the desired
process value or ’set point’, and y(t) is the measured process value. A proportional-integral-
derivative controller (PID controller) is a control loop feedback mechanism control technique
widely used in control systems.
The theoretical understanding and application dates from the 1920s, and they
are implemented in nearly all analogue control systems; originally in mechanical con-
trollers, and then using discrete electronics and latterly in industrial process comput-
ers. The PID controller is probably the most-used feedback control design.
If u(t) is the control signal sent to the system, y(t) is the measured output and
r(t) is the desired output, and tracking error e(t) = r(t) − y(t), a PID controller has
the general form,
Z
de(t)
u(t) = KP e(t) + KI e(τ )dτ + KD . (33.6)
dt
33.1. INTRODUCTION TO CONTROL THEORY 1193
The desired closed loop dynamics is obtained by adjusting the three parameters KP ,
KP , and KI , often iteratively by ’tuning’ and without specific knowledge of a plant
model. Stability can often be ensured using only the proportional term. The integral
term permits the rejection of a step disturbance (often a striking specification in
process control). The derivative term is used to provide damping or shaping of the
response 1 .
Applying Laplace transformation results in the transformed PID controller equa-
tion,
1
U (f ) = KP + KI + KD ıf E(f ) ≡ C(f )E(f ) , (33.7)
ıf
The first one is that of a proportional servo, C(f ) = KP . Then, Eq. (33.5)
reads,
KP 1 + ıf /fc
Y = (R + Zc ) + Zp
1 + ıf /fc + KP 1 + ıf /fc + KP
( s→0
KP 1
−→ 1+K (R + Zc ) + 1+K Zp
s→∞ KP fc
P P .
−→ ıf (R + Zc ) + Zp
We see that, for limited open-loop gain, noise affecting the plant Zp is not
eliminated at low frequencies. Additionally, at high frequencies, the gain
for the error signal R drops like −6dB/oct.
The second case is that of an integral servo C(f ) = KI /ıf . Then Eq. (33.5)
reads,
KI /ıf 1 + ıf /fc
Y = (R + Zc ) + Zp
1 + ıf /fc + KI /ıf 1 + ıf /fc + KI /ıf
f →0
(
ıf
−→ R + Zc + K Zp
→∞ KI fc
I .
−→ (ıf )2 (R + Zc ) + Zp
1 PID controllers are the most well established class of control systems: however, they cannot be
used in several more complicated cases, especially if MIMO systems are considered.
1194 CHAPTER 33. LOCKING CIRCUITS
The third case is that of a PI-servo, C(f ) = KP + KI /ıf . Then Eq. (33.5)
reads,
KP + KI /ıf 1 + ıf /fc
Y = (R + Zc ) + Zp
1 + ıf /fc + KP + KI /ıf 1 + ıf /fc + KP + KI /ıf
→0
f−→ ıf
R + Zc + K Zp
→∞ fc KP
I ,
f−→ (R + Zc ) + Zp
ıf
KD ıf + KP + KI /ıf 1 + ıf /fc
Y = (R + Zc ) + Zp
1 + ıf /fc + KD ıf + KP + KI /ıf 1 + KD ıf + KP + KI /ıf
→0
f−→ ıf
R + Zc + K Zp
f →∞
I .
1
−→ (R + Zc ) + 1 Zp
1+1/fc Kd fc KD
The low-frequency behavior remains still the same, but at high frequencies
the gain for R stays constant.
A time domain analysis shows that P regulators have little phase lag, but the
controlled variable can not be zeroed. On the other hand, I regulators have finite
control bandwidth, but the controlled variable can be zeroed. P I regulators (parallel
circuit of P and I regulators) have a reaction time Tn = KP /KI ; that is, the jump
response is advanced by Tn in comparison to the regulator I.
In the time domain we can summarize that regulators
D are characterized by the absence of memory, but they are very fast,
For practical PID controllers, a pure differentiator is neither physically realizable nor
desirable due to amplification of noise and resonant modes in the system. Therefore,
a phase-lead compensator type approach is used instead, or a differentiator with low-
pass roll-off [357, 253, 254, 707]. See Excs. 33.2.3.1 and 33.2.3.2.
Like any real system, the controlled device behaves as a low-pass for manipulations
or perturbations at high frequency. In other words, the device can only respond to
external perturbations with finite speed. This delay of the response leads to a phase
shift that can reverse the sign of the error signal E(f ) and transform a negative
feedback into a positive feedback. Now, in the case that there are high frequency
perturbations, for which the amplification of the closed control circuit is > 1, these
perturbations can be amplified to form oscillations. These oscillations, which occur
at the bandwidth of the closed loop gain are called servo oscillations.
33.1. INTRODUCTION TO CONTROL THEORY 1195
Fig. 33.3 illustrates the necessity of optimizing the gain and the frequency response
of the servo circuit: A proportional servo simply providing a frequency-independent
gain C(f ) = KP will lead to a forward gain R(f ) ∝ P (f ) exhibiting a low-pass
behavior, i.e. a phase-shift φ = −π/2 transforming negative to positive feedback.
The gain at the frequency where this happens need to be lower than 1, otherwise
the feedback servo will generate servo-oscillations, i.e. it will oscillate at the lowest
Fourier frequencies where noise is amplified. A PI-servo alleviates this problem by a
phase shift in opposite direction.
50
G (dB)
−6dB/oct
-50
fLP1 fP I fLP2
0
φ (π)
-0.5
-1
100 101 102 103 104 105 106
f (Hz)
Figure 33.3: (code) Bode diagram for a plant subject to low-pass behavior and for a PI-servo.
(blue solid) Low-pass filter with 100 Hz bandwidth; (blue dash-dotted) another low-pass filter
at 100 kHz; (cyan) same as blue, but amplified by 20 dB; (red) PI-servo; (magenta) open-loop
amplification.
This result can be applied to our formula (33.5) describing a feedback loop,
2 2
1 1
Sy (f ) =
Sc (f ) + Sp (f ) . (33.10)
1 + V (f )−1 1 + V (f )
1196 CHAPTER 33. LOCKING CIRCUITS
This formula describes how a servo control shapes the noise spectrum of a feedback-
controlled variable y(t).
Sp (f ) ≡ 2 · 10−24 /f .
This noise enters the feedback loop via the fluctuations zp (t) depicted in
10−26
(0)
Sy (f ) (Hz−1 )
10−28 (I)
(P)
−30
10 (PI)
10−32
104 106
f (Hz)
Figure 33.4: (code) Spectral density of frequency fluctuations of the servo system exhibited in
Fig. 33.2. Shown is the spectral noise density Sy (f ) of the laser light for 4 different feedback
gain curves: (black solid) no feedback control C(f ) = 0, (blue) P-regulator with C(f ) = 4,
(red) I-regulator with C(f ) = 5 · 106 /ıf , and (green) PI-regulator C(f ) = 4 + 5 · 106 /ıf .
Fig. 33.1(a). The black solid line in Fig. 33.4 traces the noise spectrum Sp (f ),
which is also the noise expected for the laser without feedback loop. Further-
more the optical cavity, which constitutes the essential part of the regulator,
is itself afflicted by acoustic noise and thermal drifts which, for simplicity, we
describe by a white Fourier frequency spectrum given by,
Sc (f ) ≡ 2 · 10−30 Hz−1 ,
and represented by the black dotted line in Fig. 33.4. As we have seen, servos
systems always have a finite bandwidth beyond which noise is fully coupled to
the system. For example in laser frequency locks, the weakest point of a servo
chain is often the small bandwidth of a piezo transducer used to correct the
length of the laser cavity. We describe this behavior by a low-pass filter for the
transfer function of the laser,
1
P (f ) =
1 + ıf /fc
with a bandwidth of fc = 10 kHz.
Fig. 33.4 demonstrates how the 1/f noise Sp (f ) can be efficiently suppressed at
frequencies below the low-pass filter cut-off fc to a value limited by the white
noise Sc (f ) entering through the regulator. As predicted by the formula (33.10),
the suppression of the noise Sp (f )becomes all the better, as the open-loop gain
33.2. AMPLITUDE STABILIZATION CIRCUITS 1197
V (f ) gets higher. However, with the amplification of the controller, also the
gain of the closed control circuit increases, and this necessarily causes phase
shifts, which lead to (servo-)oscillations in the circuit appearing as a strong
peak of strong noise around a certain Fourier frequency fbw characterizing the
bandwidth of the closed loop servo system. Note that high-frequency noise, to
which the cavity might be subject, is not coupled to the laser, because it is
filtered by the low-pass filter P (f ) to the same extend as the error signal itself.
The next example, exhibited in Fig. 33.5, shows the experimental characterization
of a dye laser locked via the Pound-Drever-Hall technique (see Sec. 33.3.3) to an
optical cavity.
Sy (f ) (dBm)
arb. units
-28
-30
arb. units
σy2 (τ )
-14
-16
-6 -4 -2 0
-20 0 20 0 50
τ (s)
f (kHz) f (kHz)
Figure 33.5: Characterization of a Pound-Drever-Hall stabilization of a dye laser. (a) PDH
error signal (red) and cavity transmission signal (blue), (b) spectral density of frequency
fluctuations, (c) Allan variance. (d-e) Beat signals of Mach-Zehnder interferometers. In (d)
one interferometer arm is passed through an optical fiber attached to a piezo transducer
to which a sinusoidal 5 kHz modulation of is applied. This demonstrates the sensitivity of
optical fibers to acoustic noise. In (e) one interferometer arm is send to another optical table
and back. This demonstrates how mechanical vibrations of optical components in the beam
path can broaden the spectrum of a laser field.
One way of stabilizing the light intensity of a laser beam consists in using the an
acousto-optic modulator, as shown in Fig. 33.6. The first Bragg diffraction order (see
Sec. 31.3.1) is focused onto a photodiode. Intensity fluctuations of the light recorded
by the photodiode are converted into voltage fluctuations, processed by an electronic
circuit fed back to the AOM. The intensity of light diffracted into the first order can
be controlled via the power of the radiofrequency alimenting the AOM. The control
circuit can now be conceived such as to neutralize the intensity fluctuations recorded
by the photodiode.
Figure 33.6: (a) Layout of the intensity control. The variable attenuator controls the am-
plitude of the radiofrequency driving the AOM: low voltage (0 V) increases the attenuation,
high voltage (+16 V) reduces it. The sketched control circuit realizes a negative feedback,
when the photodetector produces a positive signal. (b-d) Signals recorded at the test points
of the circuits shown in (a). See text for explanations.
It is now important to realize, that the variable attenuator works best around a
given control voltage, which is provided by adding via the trimmer T3 (test point
5) and a third OpAmp an appropriate offset. Furthermore, we note that variable
attenuator reduces its attenuation with increasing control voltage. Thus, the control
signal (test point 6) works to enhance the efficiency of the AOM, when the photodiode
signals a power drop, and vice versa. As a result, the light power in the first diffraction
order is stabilized, however, at a level inferior to the unstabilized power [707, 254].
The trimmers of the servo circuits can be adjusted using the following procedure:
1. Observe the light intensity and its fluctuations at test point (1), set test point
(4) to ground (e.g. short-circuiting the trimmer T2 ), and adjust trimmer T3 until
the light intensity level is at bit lower than the lowest fluctuations.
2. Reconnect test point (4) to the circuit and adjust trimmer T1 until the voltage
at test point (4) cancels to zero.
RL
BC107
1k
1k
33.2.3 Exercises
33.2.3.1 Ex: Integrator
Determine the transfer function of the circuit depicted in Fig. 33.8. What kind of
control circuit is it?
C Rn C(w)
R1
w
Figure 33.8:
wc
1
F (f ) = ,
1 + ıf TF
A
P (f ) = .
1 + ıf TP
How do you have to choose the various time constants in order to let the closed-loop
transfer function be H(f ) = 1.
1. Realize the optical setup illustrated in Fig. 33.6. Optimize the alignment of the
AOM (in particular, the focus and the Bragg angle) in order to maximize the
efficiency of the AOM. Take care not to saturate the photodiode, if necessary
adapt the load resistance (see Sec. 30.1.). Study the data sheet of the variable
attenuator.
2. Derive and plot the transfer function for ac-signals of the electronic circuit.
What kind of control circuit is it?
3. Set up the electronic circuit exhibited in Fig. 33.6. Test it by observing the
signals at the six test points marked in the circuit diagram. Understand and in-
terprete the roles of the three adjustable parameters: input offset, amplification,
and output offset.
4. Incorporate the servo circuit into the optical setup as shown in Fig. 33.6. How
to make sure the circuit is operating properly? 2 .
data sheet for the variable attenuator see appendix Fig. 34.16,
data sheet for the amplifier see appendix Fig. 34.14.
33.3. FREQUENCY STABILIZATION CIRCUITS 1201
2. Now control the gate voltage via the voltage measured at the small resistor and
measure again the dependency voltage-to-current.
Figure 33.9: (a) Frequency stabilization to a cavity using the side-of-fringe method. (b) The
laser is tuned to rising (or falling) slope of a transmission curve of a Fabry-Pérot cavity. The
error signal is compared with reference voltage.
Figure 33.10: (a) Frequency stabilization to a cavity using the lock-in method. (b)
Frequency-modulated signals applied to a resonance suffer a period doubling, when the
signal frequency is close to resonance. By demodulating the signals discriminated at the
resonance profile, we obtain, after averaging over a period, a DC voltage that is proportional
to the frequency detuning. The yellow area denotes the components constituting the lock-in
amplifier.
From the latter expression, it can be seen that the spectrum of sidebands is formed
√ 1 − e−ıω/δf sr
Er (ω) = E0 (ω) R , (33.16)
1 − R e−ıω/δf sr
where the amplitude and the phase of the reflected light field follow from the relation
Er (ω) = |Er (ω)|eıφ(ω) . Obviously the field of the reflected light is strong only, when
the laser frequency is close to one mode of the resonator (when ω/δ is an integer
number). By inserting Eq. (33.16) into Eq. (33.15), we obtain the response of the
resonator to a field containing sidebands as a function of the frequency of light ω, of
the modulation frequency Ω, and of the cavity finesse,
|Etot |2 = |eıωt [J1 (N )Er (ω + Ω)eıΩt + J0 (N )Er (ω) − J1 (N )Er (ω − Ω)e−ıΩt ]|2
= J0 (N )J1 (N )Er (ω + Ω)eıΩt + J0 (N )J1 (N )Er∗ (ω − Ω)eıΩt + ... + c.c. . (33.17)
The contributions of the reflected field to the current in the photodetector, |Er |2 ,
oscillating with frequency Ω and extracted by the alternating current e−iΩt+iθ (θ is
3 Remember that we specify all frequencies except the free spectral range δf sr in radians.
1204 CHAPTER 33. LOCKING CIRCUITS
1 2 1
transmission signal
reflection signal
reflection signal
1
0 0
0
-1 -1 -1
-50 0 50 -50 0 50 -1 0 1
detuning (MHz) detuning (MHz) detuning (MHz)
Figure 33.12: (code) (Left) Pound-Drever-Hall reflection signal for θ = 0, π/2. (Right)
Transmission signal.
dφ
Urf (t) = 2B cos φ(t) with = ω0 + KUct (t) . (33.19)
dt
The local oscillator produces an ac-voltage, Ulo (t) = A sin φlo (t). A mixer multiplies
both signals,
Figure 33.14: (a) PLL to lock a VCO to a reference oscillator LO. (b) PLL to lock the
difference frequency of two lasers to a LO.
F (f ) = (1 + sRC)−1 . (33.21)
1206 CHAPTER 33. LOCKING CIRCUITS
The signal of the filter is Ulp (f ) ≡ F (f )Ud (f ). In time domain, which is obtained by
a Laplace transform, F (t) = (RC)−1 θ(t)e−t/RC , such that
Z ∞ Z t
e−t/RC
Ulp (t) = F ? Ud (t) = F (t − τ )Ud (τ )dτ = eτ /RC Ud (τ )dτ . (33.22)
−∞ RC −∞
In most cases the LO frequency varies slowly, so that we may assume φ˙lo = ωlo ,
d2 ψ 1 dψ KAB 1
+ + sin ψ = − (ωlo − ω0 ) . (33.25)
dt2 RC dt RC RC
Hence, a PLL generates a signal Urf (t) having approximately the same (time-dependent)
frequency as the local oscillator Ulo (t). The equation is identical to that of an over-
damped rotator or a resistively shunted Josephson junction [563].
We observe that the PLL is locking to servo oscillations. The spectrum of signal
produced by the VCO exhibits sidebands as soon as the loop is closed. Their ampli-
tude depends on the gain, their frequency varies with the offset voltage controlling
the VCO.
4 4000
Δψ/2π
Δω/2π
2 2000
0 0
0 5 10 0 5 10
t (ms) t (ms)
Figure 33.15: (code) Simulation of (a) the phase and (b) frequency difference in a PLL for
(red) ωlo − ω0 = (2π) 2 kHz and (blue) 4 kHz.
33.3. FREQUENCY STABILIZATION CIRCUITS 1207
Figure 33.16: Schematic view of a transfer cavity locking system involving three cascaded
servo systems for (i) the stabilization of the reference laser to a known frequency; (ii) of the
transfer cavity to the reference laser, and (iii) of the spectroscopy laser to the transfer cavity.
Two issues need to be considered when using the transfer cavity scheme:
The scheme does not permit tuning of the spectroscopy laser; this feature
must be included using, e.g. AOMs (see Sec. 31.3.1) of PLL offset locks (see
Sec. 33.3.5).
The use of piezo in the transfer cavity is incompatible with its high-level thermal
and mechanical stabilization. This means that special care must be taken in
the design of the reference laser in order to avoid degradation of the stability of
the transfer cavity via the servo lock.
The whole locking scheme consists of three cascaded servo loops (see Fig. 33.16).
To model the transfer of stability from the reference to the spectroscopy laser, we
write down the following relations,
We assume specific transfer functions for the closed-loop gains of the three servos
and describe the impact of noise by adding frequency deviations ∆ω entering at
various points.
33.3.7 Exercises
33.3.7.1 Ex: Schemes for laser tuning
Discuss the two PLL-setups shown in Fig. 33.17.
2. If you do not have a lock-in amplifier available, construct one following the
project Sec. 30.4.3.
3. Now, do the opposite, stabilizing the optical cavity to the laser frequency using
the resonator piezo as control element.
4. Vary the optical setup now modulating the frequency using an AOM (see Fig. 33.18).
Figure 33.18: Variations on the same theme: (a) Frequency stabilization of a cavity to a
laser frequency using the lock-in method. This method is often used for spectral filtering of a
laser beam by a transmission etalon. (b) Frequency stabilization of a laser to a cavity using
an AOM. The advantage of using an AOM compared to the scheme shown in Fig. 33.10 is,
that only the beam injected into the cavity is modulated, but not the beam used for the
main experiment.
1. Consider the reflected signal. To do this, separate the beam injected into the
resonator from the reflected beam by means of a λ/4 waveplate and a polarizing
beam splitter.
2. Now analyze the reflected signal with a fast photodetector at a spectrum ana-
lyzer.
3. Demodulate the signal with the modulation frequency. Vary the length of the
cables. Optimize the slope of the error signal by a suitable choice of frequency
and modulation excursion 4 .
J.L. Stewart, The power spectrum of a carrier frequency modulated by Gaussian noise
[DOI]
D.S. Elliott et al., Extracavity laser band-shape and bandwidth modification [DOI]
L.S. Cutler, Some Aspects of the Theorv and Measurement of Frequency Fluctuations
in Frequency Standards [DOI]
data sheet for the power splitter see appendix Fig. 34.17,
data sheet for the mixer see appendix Fig. 34.18.
1210 CHAPTER 33. LOCKING CIRCUITS
G. Di Domenico et al., Simple approach to the relation between laser frequency noise
and laser line shape [DOI]
J. Appel et al., A versatile digital GHz phase lock for external cavity diode lasers
[DOI]
L. Ricci et al., A compact grating-stabilized diode laser system for atomic physics
[DOI]
G. Ritt et al., Laser frequency offset locking using a side of filter technique [DOI]
E.C. Cook et al., High passive-stability diode-laser design for use in atomic-physics
experiments [DOI]
C.J. Hawthorn et al., Littrow configuration tunable external cavity diode laser with
fixed direction output beam [DOI]
Y. Shimada et al., A simplified 461-nm laser system using blue laser diodes and a
hollow cathode lamp for laser cooling of Sr [DOI]
TO: J. Barnes
FROM: W. Riley
Chapter
(1) Typo in J.A. 34
Barnes, "The Measurement of Linear Drift in Oscil-
lators'l, Proc. 15th PTTI Meetinq, 1983, p. 566 (p. TN-279 of NIST
Technical Note 1337):
See
Appendices to
The.expression
Pa 568 (TN-281)
for A is missing
for the correct
the term N after
expression.
the 2nd 3.
’Instrumentation of a
Quantum Optics Lab’
NON SEQUITUR by Wiley
1
wk ≡ for σk = σ0 (34.3)
σk2
1213
1214 CHAPTER 34. APPENDICES TO ’INSTRUMENTATION OF A QO LAB’
gives,
P xk
k σk
2 X xk σk =σ0 1 X 1 σ0
σk =σ0
x̄ ≡ P 1 = σx̄2 −→ xk with σx̄ ≡ qP −→ √ (34.4)
k σk
2 σk2 N 1 N
k k k σk2
or,
s
kσ =σ
0 1 X
σ̂x̄ ≡ σx̄ χv −→ (xk − x̄)2 (34.5)
N −1
k
s
1 X (xk − x̄)2
χv = .
N −1 σi2
k
34.1.1.1 χ2 -fit
The χ2 -fit of a constant of a function y = f (x) to a measured data set (xk , yk ) is,
1 X
χ2 = [f (xk ) − yk ]2 . (34.7)
N (N − 1)
k
x1
f=
x2 + a
∆x1 ∆(x2 + a) ∆x1 ∆x2
=⇒ ∆f = f +f =f +f .
x1 x2 + a x1 x2 + a
N N
1 X 1 X
x̄ ≡ xk , xy ≡ xk yk . (34.12)
N N
k=1 k=1
For a fit of a linear curve f (x) = ax + b to a data set {xk , yk }k∈[1,N ] , we calculate,
xy − x y (x − x̄)y y x2 − xy x
f (x) = ax + b , a= 2
= , b= = y − ax .
x2 − x (x − x̄)2 x2 − x2
(34.13)
The uncertainties are obtained via,
r s
N 1 ∆y
∆y = (ax + b − y)2 , ∆a = ∆y =√
(34.14)
N −2 N x2
N x − x̄−x 2
s s
x 2 x2
, ∆b = ∆y = ∆y .
N x2 − x 2 N x − x̄
To fit an exponential curve f (x) = βeαx , we simply convert the data set {xk , vk } ≡
{xk , lg yk }, calculate the mean values of the decadal logarithm using the recipe (34.12),
and fit a linear curve f˜(x) = lg f (x) = lnα10 x + lg β ≡ ax + b in a semi-logarithmic
scale to the data set {xk , vk }. This gives,
α x lg y − x lg y lg y x2 − x lg y x
f (x) = βeαx , =a= , lg β = b = .
ln 10 x2 − x2 x2 − x2
(34.15)
To fit a power law curve f (x) = βxα , we simply convert the data set {uk , vk } ≡
{lg xk , lg yk }, calculate the mean values of the decadal logarithm using the recipe
(34.12), and fit the linear curve f˜(x) = lg f (x) = α lg x + lg β ≡ ax + b in a double-
logarithmic scale to the data set {uk , vk }. This gives,
lg x lg y − lg x lg y lg y lg x2 − lg x lg y lg x
f (x) = βxα , α=a= 2 , lg β = b = 2 .
lg x2 − lg x lg x2 − lg x
(34.16)
1216 CHAPTER 34. APPENDICES TO ’INSTRUMENTATION OF A QO LAB’
xn = P −1 (ζn ) (34.19)
0.4
n(ζ)
n(ζ)
1
0.2
0 0
0 1 2 3 -2 0 2
ζ ζ
Figure 34.1: (code) (a) Boltzmann distribution simulated by random numbers (histogram).
The numerical derivative of this distribution is shown as a blue dotted line, and the proba-
bility density as a green solid line. (b) Error function probability distribution simulated by
random numbers (histogram). The numerical derivative of the error function is shown as a
blue dotted line, and the Gauss function as a green solid line.
While we have described above the transfer of (time-varying) signals, the same
feature can be treated in frequency domain via Fourier or Laplace-transforms. This
script is not the right place to recapitulate the mathematics of these transforms, and
we will restrict ourselves to reproducing some on the most fundamental results, as we
may need them in the following.
as functions of frequency. The input signal and the system’s transfer function are
converted from time functions to functions of frequency by a transform such as the
Fourier transform, Laplace transform, or Z transform. The advantage of this tech-
nique is that it results in a simplification of the mathematics; the differential equations
that represent the system are replaced by algebraic equations in the frequency domain
which are much simpler to solve. However, frequency domain techniques can only be
used with linear systems, as mentioned above.
Time-domain state space representation: In this type the values of the state vari-
ables are represented as functions of time. With this model the system being analyzed
is represented by one or more differential equations. Since frequency domain tech-
niques are limited to linear systems, time domain is widely used to analyze real-world
nonlinear systems. Although these are more difficult to solve, modern computer sim-
ulation techniques such as simulation languages have made their analysis routine.
In contrast to the frequency domain analysis of the classical control theory, mod-
ern control theory utilizes the time-domain state space representation, a mathematical
model of a physical system as a set of input, output and state variables related by
first-order differential equations. To abstract from the number of inputs, outputs and
states, the variables are expressed as vectors and the differential and algebraic equa-
tions are written in matrix form (the latter only being possible when the dynamical
system is linear). The state space representation (also known as the ’time-domain ap-
proach’) provides a convenient and compact way to model and analyze systems with
multiple inputs and outputs. With inputs and outputs, we would otherwise have to
write down Laplace transforms to encode all the information about a system. Un-
like the frequency domain approach, the use of the state-space representation is not
limited to systems with linear components and zero initial conditions. ’State space’
refers to the space whose axes are the state variables. The state of the system can be
represented as a point within that space.
later sections).
34.2. DEEPENING CONTROL THEORY 1219
as the reaction of a system T to a pulse δ(t). Now, it is easy to see, that the operator
P defined as,
P... ≡ h(t) ? ... , (34.28)
and which describes the convolution of an arbitrary input signal with the pulse re-
sponse, satisfies the above linearity condition. Now calculating,
Z 0
Pest = h(t) ? est = h(τ )es(t−τ ) dτ = L[h(t)] · est = (Lh)(s) · est , (34.29)
−∞
we find that the functions est are eigenfunctions of the operator P with the eigenvalues
L[h(t)], which are just the Laplace transforms of the pulse response.
We can now expand arbitrary functions x(t)θ(t) in a Laplace series and obtain,
Z 0 Z 0
L[h(t) ? x(t)] = h(t) ? est x(t)dt = (Lh)(s) est x(t)dt = (Lh)(s) · (Lx)(s) .
−∞ −∞
(34.30)
The convolution on the left-hand side is in time domain, while the product on the
right-hand side is in frequency domain.
and calculating
y(s)
F (s) = . (34.32)
x(s)
The transitory behavior F (t) can in practice be extracted via an adequate choice
of the test function, f.ex., the response to a pulse:
or to a sudden jump:
Here, θ(t) denotes the Heavyside function, which is 1 for t > 0 and 0 else. The time-
dependent function, which describes the pulse response is often used as a symbol for
a specific control loop element.
The pulse response works in a similar way as the Green’s function procedure:
Wanting to know how a loop control element F transforms a given input signal x(t)
into an output signal y(t), i.e., y(t) = F (t) ? x(t), we produce a rapid pulse leading
to the output,
yδ (t) = F (t) ? δ(t) = F (t) . (34.35)
Now, once we know F (t), the response to arbitrary input signals can be computed
via,
y(t) = yδ (t) ? x(t) . (34.36)
(s − a1 )(s − a2 )...(s − an )
F (s) = A . (34.37)
(s − b1 )(s − b2 )...(s − bn )
With this, F (s) is analytical and conform, i.e., multiple curves in the s-plane are
represented in an isogonal way in the F (s)-plane. In order to avoid that the eigen-
functions est oscillate and diverge, it is necessary that all the poles and zeros are in
the left halfplane.
34.2. DEEPENING CONTROL THEORY 1221
0
100
φ (π)
|F (ω)|
-0.2
-0.4
10−2
102 104 102 104
f (Hz) f (Hz)
0 90 1
120 60
150 0.5 30
-0.2
φ (π)
φ(π)
180 0
-0.4
210 330
240 300
-0.6 270
0 0.5 1
f (Hz) f (Hz)
Mathematically, the functions f (t) form a vector space and the operators F [f ] a
ring. The linear operators generally are defined implicitly by a system of differential
equations. The particular case of linear systems is considerably simpler. The general
circuit shown in Fig. 34.4(g) corresponds to the differential equation,
The linearity F [λf1 + µf2 ] = λF [f1 ] + µF [f2 ] warrants that this equation becomes,
Z Z
0 = [1 + ∂t + ... + dt + ...]yk = [1 + ∂t + ... + dt + ...]xj . (34.39)
1222 CHAPTER 34. APPENDICES TO ’INSTRUMENTATION OF A QO LAB’
Note that the multiplication, the derivation, and the integration are linear operators
in the same sense as the Fourier and the Laplace transformation.
Linear differential equations can be Laplace transformed. The corresponding
transfer function is,
Ly(t)
F (s) = F (−δ + ıω) ≡ . (34.40)
Lx(t)
In the Laplace-transformed space the operations multiplication, derivation, and inte-
gration are all replaced by multiplications:
Z
1
L[λ + ∂t + ... + dt + ...] = λ + s + ... + . (34.41)
s
With this, the control loop elements and the additive nodes can be used to completely
represent a control circuit.
The characteristic responses of components are frequently non-linear (e.g. transis-
tor). For small signal amplitudes, these response functions, and also multiplication
points (e.g. mixers) can be linearized by Taylor expansion up to first order,
34.2.2.1 Regulators
For many circuits, it is sufficient to restrict to combinations of resistive (proportional),
capacitive (integral), and inductive (differential) circuits. Then, the general case of a
control regulator is that of a P ID − T1 ...Tn -element, meaning that:
Z Z
1
T22 ÿ + T1 ẏ + y = KD ẋ + KP x + KI dtx = KP x + Tv ẋ + dtx , (34.45)
Tn
corresponding to the transfer function,
KD s + KP + KI /s (1 + Tv s + 1/Tn s)
F (s) = = . (34.46)
1 + T1 s + T22 s2 1 + T1 s + T22 s2
In literature, two notations are used for the constants. They are linked via: KD ≡
KP Tv and KI ≡ KP /Tn . The stationary behavior is obtained setting the delays to
1224 CHAPTER 34. APPENDICES TO ’INSTRUMENTATION OF A QO LAB’
zero: Tn ≡ 0.
Example 137 (PID regulators): For example, for a proportional regulator,
we have,
y = KP xw and FR = KP , (34.47)
for an integral regulator with time delay T1 , we have,
Z
KI
T1 ẏ + y = KI xw dt and FR = , (34.48)
s(1 + sT1 )
or for a PID regulator without delay, we have,
Z
KI
y = KP xw + KD ẋw + KI xw dt and FR = KP + + KD s . (34.49)
s
Since there are three basic operations (multiplication with 1, s and 1/s), in the
end, all rational circuit elements can be reduced to an addition and concatenation of
proportional F (s) = KP , integral F (s) = KI /s, and differentials elements, F (s) =
KD s. In particular, P ID−T1 ...Tn circuits can be constructed by putting in parallel P ,
I, and D regulators concatenated with delay elements T1 . The possibility of feedback
opens other possibilities [see Fig. 34.6(a)].
Figure 34.6: (a) Circuit with feedback, (b) low-pass filter circuit.
Example 138 (Low-pass filter and time delays): We consider the example
of a low-pass filter exhibited in Fig. 34.6(b) and described by the equation,
R + ıωL + 1/ıωC
F (ıω) = . (34.50)
Ri + R + ıωL + 1/ıωC
I.e., we have a P ID − T1 T2 circuit.
Another example, is the dead time circuit,
y(t) = x(t − Tt ) and F (s) = e−sTt . (34.51)
We have,
F (ıω) = e−ıωTt F0 and |F (ıω)| = F0 and ϕ(ıω) = −ωTt . (34.52)
Hence, dead time circuits produce phase shifts, which are proportional to the
dead time interval Tt .
Figure 34.7: Some examples for feedback regulators. The columns show from left to right:
the nomenclature, the circuit diagram, the behavior, and the constants of the LTI system.
34.2.5 Exercises
34.2.5.1 Ex: Step response
Calculate the step response of a low-pass filter using the Laplace transform formalism.
3.141 59 (12), then the accuracy is x − π and the precision 0.000 12.
1228 CHAPTER 34. APPENDICES TO ’INSTRUMENTATION OF A QO LAB’
The stability that an ideal frequency standard can achieve in principle is propor-
tional to quality factor of the resonance ω/(∆ωnat +∆ωbroaden ) and the signal-to-noise
ratio of the control signal S/N . Hence, the stability can be optimized by choosing
atomic transitions with small spontaneous decay rates ∆ωnat . Appropriate techniques
for the experimental preparation of the resonance limit the influence of line broaden-
ing mechanisms ∆ω. According to the Fourier theorem, the resolution of narrow lines
requires long observation times. Now, the control signal of a feedback regulator can
only be determined after a whole observation period. This means that the appara-
tus already must have an intrinsic stability good enough that the frequency does not
leave the control range during a period of observation. Furthermore, a better signal-
to-noise ratio can improve the precision with which the line center of the resonance
can be determined.
Ultimately, the most promising way to increase the accuracy of a frequency stan-
dard seems to be to choose the highest possible transition frequencies ω, provided that
the oscillations can still be counted electronically, or be linked in a phase-coherent
way to oscillators generating countable oscillations, e.g. using frequency combs. The
requirement of phase coherence at optical frequencies puts the laser in the focus of
interest in metrology.
however, is stationary:
Z T
1
ysto (t) = y(t) − ysys (t) = 0 defining y(t) ≡ lim y(t)dt (34.59)
T →∞ T 0
as the time average. In the following only stationary stochastic fluctuations are con-
sidered.
Let us take a look at the autocorrelation function of the phase defined as,
Z T
1
Ry (τ ) ≡ y ∗ (t)y(t + τ ) = lim y ∗ (t)y(t + τ )idτ (34.60)
T →∞ T 0
and the spectral density of phase fluctuations which, according to the Wiener-Khintchine
theorem, can be obtained as the Fourier transform of the autocorrelation function,
Z ∞
Sy (f ) ≡ FRy (τ ) = Ry (τ )e−2πıf τ dτ . (34.61)
−∞
Sẋ (f ) = f 2 Sx (f ) , (34.62)
Sax (f ) = a2 Sx (f ) , (34.63)
2
X
Sy (f ) = hβ f β . (34.65)
β=−2
Table 34.1: Spectral fluctuation density and Allan-variance for common noise pro-
cesses [51].
10−26
Sy (f ) (Hz−1 )
10−28
10−30
100 105
f (Hz)
perceived. In addition, every real data acquisition system has a low-pass behavior
with a cut-off frequency fmax , so that high frequencies also do not contribute to the
integral. These bandwidth constraints enforce the condition (34.64) for the five noise
processes assumed by (34.65) [446]. The measure for the noise power in any case has
R fmax
the form: fmin Sy (f )df .
Figure 34.10: (a) Noise can exhibit very different short and long time behavior. (b) Any
measurement needs a minimum integration time.
It turns out that the variance for 1/f noise and for stochastic drifts (1/f 2 noise)
diverges at the lower limit, i.e. this variance is not useful for practical applications.
The divergence comes from the fact that for longer and longer measurements (N →
∞), respectively, smaller and smaller Fourier frequencies (f → 0), longer and longer
periodic fluctuations can be identified as such, while for shorter measurements they
appear as linear drifts. One way out is to calculate the variance for a limited number
k of measurement data and to average the variances of M of such data sets of length
k. This variance converges for a larger number of noise processes. This so-called pair
variance (k = 2) or Allan variance is widely used:
M
1 X
σy2 (τ ) = (y2j − y2j−1 )2 . (34.69)
2M j=1
Like the normal variance, the Allan variance can also be related to the spectral density
of frequency fluctuations:
Z ∞
2
σy (τ ) = Sy (f )|Fζ2 (f )|2 df where ζ2 (t) ≡ √12 [ζ1 (t) − ζ1 (−t)] . (34.70)
0
1232 CHAPTER 34. APPENDICES TO ’INSTRUMENTATION OF A QO LAB’
The Fourier transforms of the step functions ζ1 (t) and ζ2 (t) will be calculated in
Exc. 34.3.3.2, as well as the variances for white noise.
Figure 34.11: (left) Setup for measuring the spectral density of frequency fluctuations Sy (f )
and the Allan variance σy2 (τ ) of laser 1 discriminating it at the slope of transmission signal of
an optical cavity. (right) The power spectral density SE (ω) can be found as the beat signal
between two lasers and either be exhibited on a spectrum analyzer SE (ω) or counted and
processed to an Allan variance σy2 (τ ).
The beat spectrum is the convolution of the emission bandwidth of the two oscil-
lators,
SE (ω) = SE,laser1 (ω) ∗ SE,laser2 (ω) . (34.73)
In particular, for the case that we have good reasons to believe that one laser is much
narrower than the other, the power spectral density SE (ω) will reflect the emission
spectrum of just the broader laser. Note that the power spectral density derived
4 Note that the first-order coherence is just the normalized autocorrelation,
hE ∗ (t)E(t + τ )i RE (τ )
g (1) (τ ) ≡ = .
hE ∗ (t)E(t)i RE (0)
.
34.3. CHARACTERIZATION OF STABILITY 1233
Z ∞
2h0 ω02
SE (ω) = E02 RE (τ )e−ı(ω−ω0 )τ dτ = , (34.80)
−∞ 4∆2 + (h0 ω02 /2)2
where ∆ ≡ ω − ω0 is the deviation of the oscillator frequency from the center fre-
quency ω0 . Thus, the the emission spectrum for the case of white Gaussian noise is a
Lorentzian profile with the laser emission bandwidth,
β = 12 h0 ω02 . (34.81)
4
noise amplitude
(a)
2
0
-2
-4
0 1000 2000 3000 4000 5000
t (s)
(b) (c)
10−10 10−15
S(ω)
σ(τ )
10−20
100 10−4 10−2
τ (s) ω (Hz)
Figure 34.12: (code) Allan variance (b) and spectral density of frequency fluctuations (c)
calculated from a randomly generated data set (a).
h0 1 11
σy2 (τ ) = = 2 . (34.82)
2τ Q Γτ
with the linewidth (FWHM) Γ = h0 ω02 /2 and Q = ω0 /Γ and indicates the maximum
achievable stability for a frequency standard that is coupled to the resonance ω0 .
5 Using
R∞ 2a
F [e−a|x| ] = −∞ e−a|t| e−ıωt dt = a2 +ω 2
.
34.4. DATA SHEETS 1235
The spectral noise power density consists of a discrete spectrum of sidebands, the
number and height of which is given by the modulation index N :
∞
X
SE (ω) = |Jn (N )|2 δ(ω − ω0 − Ω) . (34.84)
n=−∞
A full width at half maximum of the spectral noise power density cannot be specified.
However, as a measure of the emission bandwidth of the oscillator, the frequency
spacing of the sideband of the highest order can be understood, the height of which
corresponds to at least half the height of the carrier frequency ω0 .
34.3.3 Exercises
34.3.3.1 Ex: Spectral density of frequency fluctuations
Prove the relationship Sẋ (f ) = f 2 Sx (f ).
hA2n i hA2 in
= n and hA2n−1 i = 0 .
(2n)! 2 n!
HL6722G
AlGaInP Laser Diode
ODE-208-220E (Z)
Rev.5
Mar. 2005
Description
The HL6722G is a 0.67 µm band AlGaInP index-guided laser diode with a multi-quantum well (MQW)
structure. It is suitable as a light source for bercode scanner, and various other types of optical equipment.
Hermetic sealing of the package assures high reliability.
Features
• Visible light output at wavelengths up to 680 nm
• Single longitudinal mode
• Continuous operating output: 5 mW CW
• Low voltage operation: 2.7 V Max
• Low current operation: 32 mA Typ
• Built-in monitor photodiode
PD LD
Figure 34.13: Data sheet for the diode laser from Thorlabs, model Hitachi HL6722G.
34.4. DATA SHEETS 1237
Coaxial
Amplifier ZHL-2-8
50Ω Medium High Power 10 to 1000 MHz
Features
• wideband, 10 to 1000 MHz
• high IP3, +38 dBm typ.
ZHL-2-8X ZHL-2-8
• medium high power, 29 dBm min. SMA version shown
Electrical Specifications
MODEL FREQ. GAIN MAXIMUM POWER DYNAMIC VSWR DC
NO. (MHz) (dB) OUTPUT RANGE (:1) POWER
(dBm) Max.
NF IP3
Volt Current
Flatness (1 dB Compr.) (dB) (dBm)
Input (V) (A)
fL fU Min. Typ. Max. Min. (no damage) Typ. Typ. In Out Nom. Max.
ZHL-2-8 10 1000 31 35 ±1.0 +29 +5 10.0 +38 2.0 2.0 24 0.6
ZHL-2-8X* 10 1000 31 35 ±1.0 +29 +5 10.0 +38 2.0 2.0 24 0.6
* Heat sink not included To order without heat sink, add suffix X to model number. Alternative heat sinking and heat
Open load is not recommended, potentially can cause damage. removal must be provided by the user to limit maximum temperature to 65°C, in order to ensure
With no load derate max input power by 20 dB
proper performance. For reference, this requires thermal resistance of user’s external heat sink
to be 1.35°C/W Max.
Maximum Ratings
Operating Temperature -20°C to 65°C
Storage Temperature -55°C to 100°C
DC Voltage +25V Max.
Permanent damage may occur if any of these limits are exceeded.
Outline
OutlineDrawing
Dimensions
Mini-Circuits
®
For detailed performance specs
& shopping online see web site
ISO 9001 ISO 14001 AS 9100 CERTIFIED ®
P.O. Box 350166, Brooklyn, New York 11235-0003 (718) 934-4500 Fax (718) 332-4661 The Design Engineers Search Engine Provides ACTUAL Data Instantly at minicircuits.com
IF/RF MICROWAVE COMPONENTS REV. C
Notes: 1. Performance and quality attributes and conditions not expressly stated in this specification sheet are intended to be excluded and do not form a part of this specification sheet. 2. Electrical specifications M133857
and performance data contained herein are based on Mini-Circuit’s applicable established test performance criteria and measurement instructions. 3. The parts covered by this specification sheet are subject to ZHL-2-8
Mini-Circuits standard limited warranty and terms and conditions (collectively, “Standard Terms”); Purchasers of this part are entitled to the rights and benefits contained therein. For a full statement of the Standard 110929
Terms and the exclusive rights and remedies thereunder, please visit Mini-Circuits’ website at www.minicircuits.com/MCLStore/terms.jsp. Page 1 of 2
1238 CHAPTER 34. APPENDICES TO ’INSTRUMENTATION OF A QO LAB’
Coaxial
Voltage Controlled Oscillator ZOS-100+
Dual Output 50 to 100 MHz
Features
• octave bandwidth
• linear tuning, 4.5 MHz/V typ.
• excellent harmonic suppression, -29 dBc typ.
• rugged shielded case CASE STYLE: BR386
• protected by US Patent, 6,943,629 Connectors Model Price Qty.
SMA ZOS-100+ $119.95 (1-9)
Outline Drawing
ZOS-100+
FREQUENCY vs.TUNING VOLTAGE
120
FREQUENCY (MHz)
100
80
60
40
20
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
V TUNE(V)
Outline Dimensions ( inch
mm )
A B C D E F G H J K L M N P Q R wt
3.25 1.38 1.25 .71 1.13 .125 2.25 .71 .41 .98 1.28 2.950 .15 1.100 .14 .150 grams
82.55 35.05 31.75 18.03 28.70 3.18 57.15 18.03 10.41 24.89 32.51 74.93 3.81 27.94 3.56 3.81 180
Figure 34.15: Data sheet for the Voltage-Controlled Oscillator (VCO) from Minicir-
cuits, model ZOS-100+. Mini-Circuits
®
For detailed performance specs
& shopping online see web site
ISO 9001 ISO 14001 AS 9100 CERTIFIED TM
P.O. Box 350166, Brooklyn, New York 11235-0003 (718) 934-4500 Fax (718) 332-4661 The Design Engineers Search Engine Provides ACTUAL Data Instantly at minicircuits.com
IF/RF MICROWAVE COMPONENTS REV. B
Notes: 1. Performance and quality attributes and conditions not expressly stated in this specification sheet are intended to be excluded and do not form a part of this specification sheet. 2. Electrical specifications M113397
and performance data contained herein are based on Mini-Circuit’s applicable established test performance criteria and measurement instructions. 3. The parts covered by this specification sheet are subject to ZOS-100+
Mini-Circuits standard limited warranty and terms and conditions (collectively, “Standard Terms”); Purchasers of this part are entitled to the rights and benefits contained therein. For a full statement of the Standard SK/TD/CP/AM
Terms and the exclusive rights and remedies thereunder, please visit Mini-Circuits’ website at www.minicircuits.com/MCLStore/terms.jsp. 091223
34.4. DATA SHEETS 1239
%QCZKCN
8QNVCIG8CTKCDNG#VVGPWCVQT <:
:
/CZKOWO4CVKPIU (GCVWTGU
D/'
1*$-/'
0' E
E %&'()*(+
4
'$-/'
0' E
E %,
+!
(&-
./ ('/#.'5/#UJQYP
%(&
*0
1*2 "44$LCJ3
")6:0
940//:.:
F!G
%3(45
1*#0
/'
67'
")6:0
9*
'::
F<'
:G
*0
1*'* 4" **<
'6 (: '1< H
. 6
")6:0
9;,*/0
C7: !(&- %89
'*:)16*(;-
<=1* +0276 176276
'-**
(--.<<0'1 *.
=6:1-1
6'9<(( *
>'?'@01'( ('/#.' ('/#.'
4! I FG 3
%4=1:((<6 /#.' ('/#.'
4! I FG 3
'NGEVTKECN5RGEKſECVKQPU
6#/$%
(4'3 /+0+05'46+10 /#:#66'07#6+10 +0276 %10641. +2 4'6740 219'45722.;
/*\ .155F$
8 F$
8 219'4 8QNVCIG %WTTGPV
F$O .155 8QNVCIG %WTTGPV
F$O
8
O#
F$
8
O#
1* 9 $./ 9 $./ 1* 9 9 $./ $./ 9
!
!
!
!
!
!
!
!
!
8
6J
16#;::
1-JK4<#K4<$./
4>1
<=1*$1-+
0'**# JK4<#K4<$./
,-/'7(C661*#0
/' '-*<<*)<=17(
<'
1* '@0*<16).<=61*!)
>*!
!
1WVNKPG&KOGPUKQPU
KPEJ
-- 'SWKXCNGPV5EJGOCVKE
9
" & ; 3 N
&21752/92/7$*(9
Figure 34.16: Data sheet for the voltage-controlled variable attenuator from online see Minicir-
®
Mini-Circuits For detailed performance specs
& shopping web site
ISO 9001 ISO 14001 AS 9100 CERTIFIED
cuits, model ZX73-2500+.
P.O. Box 350166, Brooklyn, New York 11235-0003 (718) 934-4500 Fax (718) 332-4661 The Design Engineers Search Engine
®
Provides ACTUAL Data Instantly at minicircuits.com
IF/RF MICROWAVE COMPONENTS
Notes: 1. Performance and quality attributes and conditions not expressly stated in this specification sheet are intended to be excluded and do not form a part of this specification sheet. 2. Electrical specifications
!
and performance data contained herein are based on Mini-Circuit’s applicable established test performance criteria and measurement instructions. 3. The parts covered by this specification sheet are subject to "#$#
Mini-Circuits standard limited warranty and terms and conditions (collectively, “Standard Terms”); Purchasers of this part are entitled to the rights and benefits contained therein. For a full statement of the Standard
Terms and the exclusive rights and remedies thereunder, please visit Mini-Circuits’ website at www.minicircuits.com/MCLStore/terms.jsp.
Plug-In PSC-2-1+
Power Splitter/Combiner PSC-2-1
2 Way-0° 50Ω 0.1 to 400 MHz
Maximum Ratings Features
Operating Temperature -55°C to 100°C • wideband, 0.1 to 400 MHz
CASE STYLE: A01
Storage Temperature -55°C to 100°C • low insertion loss, 0.4 dB typ. PRICE: $14.20 ea. QTY. (1-9)
• rugged welded construction
Power Input (as a splitter) 1W max.
+ RoHS compliant in accordance
Internal Dissipation 0.125W max.
Applications with EU Directive (2002/95/EC)
• VHF/UHF The +Suffix identifies RoHS Compliance. See our web site
Pin Connections • federal & defense communications for RoHS Compliance methodologies and qualifications.
SUM PORT 1
PORT 1 5
PORT 2 6 Electrical Specifications
GROUND 2,3,4,7,8
FREQ. ISOLATION INSERTION LOSS (dB) PHASE AMPLITUDE
CASE GROUND 2,3,4,7,8 RANGE (dB) ABOVE 3.0 dB UNBALANCE UNBALANCE
(MHz) (Degrees) (dB)
L M U L M U L M U L M U
fL-fU Typ. Min Typ. Min Typ. Min Typ. Max. Typ. Max. Typ. Max. Max. Max. Max. Max. Max. Max.
0.1-400 20 15 25 20 25 20 0.2 0.6 0.4 0.75 0.6 1.0 2.0 3.0 4.0 0.15 0.2 0.3
Outline Drawing
L = low range [fL to 10 fL] M = mid range [10 fL to fU/2] U= upper range [fU/2 to fU]
40
3.6
35
3.4
30
3.2 25
3.0 20
0 80 160 240 320 400 0 80 160 240 320 400
FREQUENCY (MHz) FREQUENCY (MHz)
PSC-2-1
VSWR
1.5
#S-VSWR #1-VSWR #2-VSWR
electrical schematic
1.4
1.3
VSWR
1.2
1.1
1.0
0 80 160 240 320 400
FREQUENCY (MHz)
® W
ALL NE
Mini-Circuits
ISO 9001 ISO 14001 CERTIFIED
minicircuits.com
REV. A
M98898
P.O. Box 350166, Brooklyn, New York 11235-0003 (718) 934-4500 Fax (718) 332-4661 For detailed performance specs & shopping online see Mini-Circuits web site PSC-2-1
HY/TD/CP
The Design Engineers Search Engine Provides ACTUAL Data Instantly From MINI-CIRCUITS At: www.minicircuits.com 070202
RF/IF MICROWAVE COMPONENTS
Figure 34.17: Data sheet for the power divider from Minicircuits, model PSC-2-1.
34.4. DATA SHEETS 1241
Plug-In
Frequency Mixer SRA-2CM+
Level 7 (LO Power +7 dBm) 5 to 1000 MHz
Electrical Schematic
Mini-Circuits
® W
ALL NE
minicircuits.com
ISO 9001 ISO 14001 CERTIFIED REV. A
M98898
P.O. Box 350166, Brooklyn, New York 11235-0003 (718) 934-4500 Fax (718) 332-4661 For detailed performance specs & shopping online see Mini-Circuits web site
SRA-2CM+
The Design Engineers Search Engine Provides ACTUAL Data Instantly From MINI-CIRCUITS At: www.minicircuits.com DJ/TD/CP/AM
061211
RF/IF MICROWAVE COMPONENTS Page 1 of 2
Figure 34.18: Data sheet for the mixer from Minicircuits, model SRA-2CM+.
1242 CHAPTER 34. APPENDICES TO ’INSTRUMENTATION OF A QO LAB’
SPECIFICATIONS
Outline Drawing:
AO Medium TeO2
Acoustic Velocity 4.2 mm/µs
Detectors
Active Aperture* 2.5 mm 'L' X 1 mm 'H'
Center Frequency (Fc) 80 MHz
Silicon Epitaxial PIN Photodiodes - High Speed – 400 nm to 1100 nm
RF Bandwidth 20 MHz @ -10 dB Return Loss
The C30736 series of high-speed epitaxial silicon PIN photodetectors provide fast response and good quantum efficiency in the spectral
range between 400 nm andInput
1100Impedance
nm. These devices are optimized for high-speed, high volume 50and
Ohms Nominal
low cost applications. Standard sizes
include 0.25 mm, 0.5 mm, 1.0 mm, 1.5 x 1.5 mm, and custom sizes can be accomodated depending on volume required. Available in plastic
VSWR @ Fc 1.3 :1 Max
surface mount packages and in chip form.
Wavelength 442-633 nm
Silicon Epitaxial PIN Photodiodes Typical Characteristics @ T = 22º C
Insertion Loss 4 % Max
Part # Standard Photo Sens. Resp.
Reflectivity per @ 870 nm Dark Current Spect. Noise
Surface Cap. @ 100 kHz Response
1 % MaxTime NEP @ 870 nm Bias Volt
Package Diam. (mm) (A/W) @2V Id (nA) Curr. Dens. Cd (pF) tr (ns) (fW/√Hz) (V)
Anti-Reflection Coating In (fW/√Hz) MIL-C-48497
W/mm2
C30736-1 Chip form 0.20 0.55 0.05 6 0.75 0.3 11 2
Optical Power Density 250
C30736-2 Chip form 0.50 0.55 0.10 10 1.5 0.5 18 2
Contrast Ratio 1000 :1 Min
C30736-3 Chip form 1.5 x 1.5 0.55 0.50 50 14 0.3 91 2
Polarization 90 ° To Mounting Plane
PERFORMANCE VS WAVELENGTH
Silicon PIN Photodiodes - Standard
Wavelength (nm) N-Type – 400 nm to 1100 nm
442 488 515 633
The C308XX series devicesSaturation RF Power
are high-quality N-type(W) 0.27 sealed
Si PIN photodiodes in hermetically 0.33 0.36 designed
TO packages 0.55 for the 400 nm to 1100
nm wavelength region. Bragg Angle (mr) 4.2 4.6 4.9 6
Beam Separation (mr)
Si PINs – Standard N-Type 8.4 9.2 9.8 12
Notes:
Typical Characteristics @ T = 22º C
PERFORMANCE VS BEAM DIAMETER Optical Ghosting Due To Acoustic R
Part # Standard Photo Sens. Resp. @ 900 nm Dark Current Spect. Noise Cap. @ 100 kHz Response Time NEP @ 900 nm Bias Volt
Beam
Package Diam. Diameter
(mm) (µm)
(A/W) Id (nA) Curr. Dens. 200
Cd (pF) 300tr (ns) 500 (fW/√Hz) (V)
at Wavelength (nm) (fA/√Hz) 633 633 633
C30807E A Diffraction
1 Efficiency
0.6 (%) 1 18 2.5 80 83 3 85 30 45
C30808E B Rise
2.5 Time (nsec) 0.6 3 31 6 34 49 5 80 52 45 IS THE PROPERTY OF CR
THIS DOCUMENT
15.9 10.6 6.3 DISCLOSED IN WHOLE OR IN PART OTH
C30822E C 5 0.6 5 40 17 7 67 45 REPRESENTATIVES AND
CONTRACTED
10 5 1 CONSENT OF AN AUTHORIZED REPRESENT
C30809E C3 8 0.6 7 47 35 10 79 45
TOLERANCES: A. Camp
.XX ± .01 DR
Figure
C30810E : Data sheet for the acousto-optic modulator (AOM) from Crystal Tech-
34.19
D 11.4 0.6 30 98 70 12 163 45
.XXX ± .005
6/17/200
Typical Applications
nologies, model AOMO 3080-120. MATERIAL:
CHK
Laser detection systems, photometry, data transmission, instrumentation, and high-speed switching.
FINISH:
APP
Preamplifiers
Preamplifier modules incorporating these photodiodes are available on a custom basis.
Si PINs – Large Area, Fast Response
Typical Characteristics @ T = 22º C
Part # Standard Photo Sens. Resp. @ 900 nm Dark Current Spect. Noise Cap. @ 100 kHz Response Time NEP @ 900 nm Bias Volt
Package Diam. (mm) (A/W) Id (nA) Curr. Dens. Cd (pF) tr (ns) (fW/√Hz) (V)
In (fW/√Hz)
FFD-040B Y 1 0.58 1 18 1.8 2 31 15
FFD-100 B 2.5 0.58 2 25 8.5 3.5 44 15
6 Figure 34.20: Data sheet for the photo diode from Perkin Elmers FDD100.
34.4. DATA SHEETS 1243
Detectors
Silicon Epitaxial PIN Photodiodes - High Speed – 400 nm to 1100 nm
The C30736 series of high-speed epitaxial silicon PIN photodetectors provide fast response and good quantum efficiency in the spectral
range between 400 nm and 1100 nm. These devices are optimized for high-speed, high volume and low cost applications. Standard sizes
include 0.25 mm, 0.5 mm, 1.0 mm, 1.5 x 1.5 mm, and custom sizes can be accomodated depending on volume required. Available in plastic
surface mount packages and in chip form.
Part # Standard Photo Sens. Resp. @ 870 nm Dark Current Spect. Noise Cap. @ 100 kHz Response Time NEP @ 870 nm Bias Volt
Package Diam. (mm) (A/W) @2V Id (nA) Curr. Dens. Cd (pF) tr (ns) (fW/√Hz) (V)
In (fW/√Hz)
C30736-1 Chip form 0.20 0.55 0.05 6 0.75 0.3 11 2
Part # Standard Photo Sens. Resp. @ 900 nm Dark Current Spect. Noise Cap. @ 100 kHz Response Time NEP @ 900 nm Bias Volt
Package Diam. (mm) (A/W) Id (nA) Curr. Dens. Cd (pF) tr (ns) (fW/√Hz) (V)
(fA/√Hz)
C30807E A 1 0.6 1 18 2.5 3 30 45
C30808E B 2.5 0.6 3 31 6 5 52 45
C30822E C 5 0.6 5 40 17 7 67 45
C30809E C3 8 0.6 7 47 35 10 79 45
Silicon PIN Photodiodes - Large Area, Fast Response Time – 400 nm to 1100 nm
The FFD series devices are high-quality, large-area, high-speed, N-type Si PIN photodiodes in hermetically sealed TO packages designed for
the 400 nm to 1100 nm wavelength range. The FND-100Q has a quartz window to enhance UV responsivity.
Preamplifiers
Preamplifier modules incorporating these photodiodes are available on a custom basis.
Si PINs – Large Area, Fast Response
Typical Characteristics @ T = 22º C
Part # Standard Photo Sens. Resp. @ 900 nm Dark Current Spect. Noise Cap. @ 100 kHz Response Time NEP @ 900 nm Bias Volt
Package Diam. (mm) (A/W) Id (nA) Curr. Dens. Cd (pF) tr (ns) (fW/√Hz) (V)
In (fW/√Hz)
FFD-040B Y 1 0.58 1 18 1.8 2 31 15
FFD-100 B 2.5 0.58 2 25 8.5 3.5 44 15
FFD-200 C3 5.1 0.58 4 36 30 5 62 15
Figure 34.21: Data sheet for the photo diode from Perkin Elmers C30822E.
1244 CHAPTER 34. APPENDICES TO ’INSTRUMENTATION OF A QO LAB’
MAX320/MAX321/MAX322
The MAX320/MAX321/MAX322 are precision, dual, ♦ Low On-Resistance, 35Ω max (16Ω typical)
SPST analog switches designed to operate from ±3V to
♦ RON Matching Between Channels <2Ω
±8V dual supplies. The MAX320 has two normally open
(NO) switches and the MAX321 has two normally ♦ RON Flatness <4Ω
closed (NC) switches. The MAX322 has one NO and
♦ Guaranteed Charge Injection <5pC
one NC switch. Low power consumption (1.25mW)
makes these parts ideal for battery-powered equip- ♦ Bipolar Supply Operation (±3V to ±8V)
ment. They offer low leakage currents (100pA max) and ♦ Low Power Consumption, <1.25mW
fast switching speeds (tON = 150ns max, tOFF = 100ns
max). ♦ Low Leakage Current Over Temperature,
<2.5nA at +85°C
The MAX320 series, powered from ±5V supplies, offers
35Ω max on-resistance (R ON ), 2Ω max matching ♦ Fast Switching, tON <150ns, tOFF <100ns
between channels, and 4Ω max RON flatness.
♦ Guaranteed Break-Before-Make (MAX322 only)
These switches also offer 5pC max charge injection
and a minimum of 2000V ESD protection per Method ______________Ordering Information
3015.7.
For equivalent devices specified for single-supply oper- PART TEMP. RANGE PIN-PACKAGE
ation, see the MAX323/MAX324/MAX325 data sheet. MAX320CPA 0°C to +70°C 8 Plastic DIP
For quad versions of these switches, see the MAX320CSA 0°C to +70°C 8 SO
MAX391/MAX392/MAX393 data sheet.
MAX320CUA 0°C to +70°C 8 µMAX
MAX320C/D 0°C to +70°C Dice*
________________________Applications MAX320EPA -40°C to +85°C 8 Plastic DIP
MAX320ESA -40°C to +85°C 8 SO
Battery-Operated Systems Sample-and-Hold Circuits
MAX320EJA -40°C to +85°C 8 CERDIP**
Heads-Up Displays Guidance and Control Systems
MAX320MJA -55°C to +125°C 8 CERDIP**
Audio and Video Switching Military Radios Ordering Information continued at end of data sheet.
Test Equipment Communications Systems * Contact factory for dice specifications.
** Contact factory for availability.
±5V DACs and ADCs PBX, PABX
TOP VIEW
MAX320 MAX321 MAX322
Figure 34.22: Data sheet for the digital switch MAX322 from Maxim.
34.4. DATA SHEETS 1245
LM741
Operational Amplifier
General Description output, no latch-up when the common mode range is ex-
ceeded, as well as freedom from oscillations.
The LM741 series are general purpose operational amplifi-
The LM741C is identical to the LM741/LM741A except that
ers which feature improved performance over industry stan-
the LM741C has their performance guaranteed over a 0˚C to
dards like the LM709. They are direct, plug-in replacements
+70˚C temperature range, instead of −55˚C to +125˚C.
for the 709C, LM201, MC1439 and 748 in most applications.
The amplifiers offer many features which make their appli-
cation nearly foolproof: overload protection on the input and Features
Connection Diagrams
Metal Can Package Dual-In-Line or S.O. Package
00934102 00934103
Note 1: LM741H is available per JM38510/10101 Order Number LM741J, LM741J/883, LM741CN
Order Number LM741H, LM741H/883 (Note 1), See NS Package Number J08A, M08A or N08E
LM741AH/883 or LM741CH
See NS Package Number H08C
Ceramic Flatpak
00934106
Order Number LM741W/883
See NS Package Number W10A
Typical Application
Offset Nulling Circuit
00934107
Figure 34.23: Data sheet for the operational amplifier LM741 from National Semicon-
ductor.
1246 CHAPTER 34. APPENDICES TO ’INSTRUMENTATION OF A QO LAB’
a Low-Noise, Precision
Operational Amplifier
OP27
FEATURES PIN CONNECTIONS
Low Noise: 80 nV p-p (0.1 Hz to 10 Hz), 3 nV/√Hz
TO-99
Low Drift: 0.2 V/C
(J-Suffix)
High Speed: 2.8 V/s Slew Rate, 8 MHz Gain
Bandwidth BAL
Low VOS: 10 V
Excellent CMRR: 126 dB at V CM of ±11 V BAL 1
OP27
V+
High Open-Loop Gain: 1.8 Million
Fits 725, OP07, 5534A Sockets –IN 2 OUT
V+
R3 R4 C2
1 8
Q6
Q22 Q46
VOS ADJ. C1
R1* R2* R23 R24
Q21
Q23 Q24
R9
Q20 Q19
Q1A Q1B Q2B Q2A R12 OUTPUT
NONINVERTING
R5 C3 C4
INPUT (+)
Q3
INVERTING Q26
Q11 Q12 Q45
INPUT (–)
Q27 Q28
V–
[3] C. S. Adams, M. Sigel, and J. Mlynek, Atom optics, Phys. Rep. 240 (1994),
143.
[9] R. Alicki, The markov master equations and the fermi golden rule, International
journal of Theoretical Physics (1977).
[11] D. W. Allan, J. H. Shoaf, and D. Halford, Statistics of time and frequency data
analysis, NBS Monographs 140 (1974), 151.
1247
1248 BIBLIOGRAPHY
[22] M. Antezza and Y. Castin, Fano-hopfield model and photonic band gaps for an
arbitrary atomic lattice, Phys. Rev. A 80 (2009), 013816, .
[26] N. W. Ashkroft and D. N. Mermin, Solid state physics, Hartcourt College pub-
lishers, 1976, ISBN.
[28] Alain Aspect, J. Dalibard, and J. Roger, Experimental test of bell’s inequality
using time-varying analyzers, Phys. Rev. Lett. 49 (1982), 1804, .
[29] Alain Aspect, P. Grangier, and G. Roger, Opt. Lett. 8 (1981), 511.
BIBLIOGRAPHY 1249
[31] C. Audoin, Frequency metrology, Rendiconti della Scuola Intern. di Fisica ?En-
rico Fermi? LIVIII or LXVIII (1980), 169.
[82] Felix Bloch, Nuclear induction, Physical Review 70 (1946), no. 7, 460.
[83] I. Bloch, Ultracold quantum gases in optical lattices, Nature Physics 1 (2005),
23, .
[84] I. Bloch, T. W. Hänsch, and T. Esslinger, Atom laser with a cw output coupler,
Phys. Rev. Lett. 82 (1999), 3008, .
[85] I. Bloch, Th. W. Hänsch, and T. Esslinger, Measurement of the spatial coherence
of a trapped bose-gas at the phase transition, Nature 403 (2000), 166, .
[88] J. L. Bohn and P. S. Julienne, Prospects for influencing scattering lengths with
far-off-resonance light, Phys. Rev. A 56 (1997), 1486, .
[95] R. Bonifacio and L. De Salvo, Collective atomic recoil laser (carl) optical
gain without inversion by collective atomic recoil and self-bunching of two-level
atoms, Nucl. Instrum. Methods A 341 (1994), 360, DOI.
[97] R. Bonifacio and Ph. Verkerk, Opt. Commun. 124 (1996), 469, .
[98] Ch. J. Bordé, Atom interferometry with internal state labeling, Phys. Lett. A
140 (1989), 10, .
[104] E. Braaten and H.-W. Hammer, Universality in few-body systems with large
negative scattering length, Phys. Rep. 428 (2006), 259, .
[115] J. P. Jr. Burke, Theoretical investigation of cold slkali atom collisions, PhD
thesis (1999).
[119] D. A. Butts and D. S. Rokhsar, Trapped fermi gases, Phys. Rev. A 55 (1997),
4346, DOI.
[127] L. D. Carr and et al., Cold and ultracold molecules: science, technology and
applications, New J. Phys. 11 (2009), 055009, .
[132] W. Chen, D. Meiser, and P. Meystre, Cavity qed determination of atomic num-
ber statistics in optical lattices, Phys. Rev. A 75 (2007), 23812, DOI.
[134] R. Y. Chiao and J. Boyce, Bogoliubov dispersion relation and the possibility of
superfluidity for weakly-interacting photons in a 2d photon fluid, Phys. Rev. A
(1999), quant-ph/9905001.
[136] J. S. Choi and J. C. Howell, Paraxial ray optics cloaking, Opt. Exp. 22 (2014),
29465, DOI.
1256 BIBLIOGRAPHY
[138] J. I. Cirac, R. Blatt, P. Zoller, and W. D. Phillips, Laser cooling of trapped ions
in a standing wave, Phys. Rev. A 46 (1992), 2668, .
[142] , Quantum computation with cold trapped atoms, Phys. Rev. Lett. 74
(1995), 4091, DOI.
[148] M. M. Cola, Collective atomic recoil in ultracold atoms: Advances and applica-
tions, Dissertation Milano (2003).
[168] M. Ben Dahan, E. Peik, Y. Castin, and C. Salomon, Bloch oscillations of atoms
in an optical potential, Phys. Rev. Lett. 76 (1996), 4508, .
[169] J. P. Dahl and M. Springborg, The morse oscillator in position space, momen-
tum space, and phase space, J. Chem. Phys. 88 (1988), 4535.
[174] , Laser cooling below the doppler limit by polarization gradients: Simple
theoretical studies, J. Opt. Soc. Am. B 6 (1989), 2023.
[180] K. B. Davis, M.-O. Mewes, and W. Ketterle, An analytical model for evaporative
cooling of atoms, Appl. Phys. B 60 (1995), 155, .
[189] D. DeMille, Quantum computation with trapped polar molecules, Phys. Rev.
Lett. 88 (2002), 067901, .
[198] I. Deutsch, Barenco, and Ekert, Proc. Roy. Soc. Lond. x (1995), 669.
[199] I. H. Deutsch and P. S. Jessen, Quantum state control in optical lattices, Phys.
Rev. A 57 (1995), 1972, .
1260 BIBLIOGRAPHY
[217] R. Donelly and C. Swanson, Quantum turbulence, J. Fluid Mech. 173 (1986),
387, .
[223] O. Dulieu and P. S. Julienne, Coupled channel bound states calculations for
alkali dimers using the fourier grid method, J. Chem. Phys. 103 (1995), 60, .
[228] V. Efimov, Energy levels arising from resonant two-body forces in a three-body
system, Phys. Lett. B 33 (1970), 563, .
[233] D. S. Elliott, Rajarshi Roy, and S. J. Smith, Extracavity laser bandshape and
bandwidth modification, Phys. Rev. A 26 (1982), 12, .
[239] B. D. Esry, Ch. H. Greene, and H. Suno, Threshold laws for three-body recom-
bination, Phys. Rev. A 65 (2001), 10705(R), .
[249] P. T. Fisk, M. A. Lawn, and C. Coles, Laser cooling of 171yb+ ions a linear
paul trap, Appl. Phys. B 57 (1993), 287.
[252] DOIR. Bachelard and N. Piovella and Ph. W. Courteille, Cooperative scattering
and radiation pressure force in dense atomic clouds, Phys. Rev. A 84 (2011),
013821.
[255] C. J. Foot, H. Wu, E. Arimondo, and G. Morigi, Optical elements for interfer-
ometry of atoms with j-greater-than-1, J. de Physique 4 (1993), 1913.
[259] , Reply to: Comment on: Effects of including the counterrotating term
and virtual photons on the eigenfunctions and eigenvalues of a scalar photon
collective emission theory; phys. lett. a 372 (2008) 2514; phys. lett. a 372 (2008)
5732, Phys. Lett. A 372 (2008), 5734, DOI.
[266] , Cold atoms in a high-q ring cavity, Phys. Rev. A 61 (2000), 043405, .
[269] C. W. Gardiner and M. J. Collett, Input and output in damped quantum systems:
Quantum stochastic differential equations and the master equation, Phys. Rev.
A 31 (1985), 3761, .
[270] B. M. Garraway, The dicke model in quantum optics: Dicke model revisited,
Phil. Trans. R. Soc. A 369 (2011), 1137, DOI.
[272] J. D. Gaskill, Linear systems, fourier transforms and optics, John Wiley & sons,
1978, ISBN.
[277] Ch. Gerry and P. L. Knight, Introductory quantum optics, Cambridge, 2005,
ISBN.
BIBLIOGRAPHY 1265
[282] S. Giovanazzi, A. Görlitz, and T. Pfau, Tuning the dipolar interaction in quan-
tum gases, Phys. Rev. Lett. 89 (2002), 130401, .
[287] J. P. Gordon and A. Ashkin, Motion of atoms in a radiation trap, Phys. Rev.
A 21 (1980), 1606, DOI.
[290] I. S. Gradshteyn and I.M. Ryzhik, Table of integrals, series, and products, El-
sevier, 2007, ISBN.
[291] R. Graham and D. Walls, Spectrum of light scattered from a weakly interacting
bose-einstein condensed gas, Phys. Rev. Lett. 76 (1996), 1774.
[328] C. E. Hecht, The possible superfluid behaviour of hydrogen atom gases and liq-
uids, Physica 25 (1959), 1159.
[331] D. J. Heinzen and M. S. Feld, Vacuum radiative level shift and spontaneous-
emission linewidth of an atom in an optical resonator, Phys. Rev. Lett. 59
(1987), 2623, .
[340] K. Hepp and E. Lieb, On the superradiant phase transition for molecules in a
quantized radiation field: the dicke maser model, Annals of Phys. 76 (1973),
360, DOI.
[344] C. Herzog and M. Olshanii, Trapped bose gas: The canonical versus grand
canonical statistics, Phys. Rev. A 55 (1997), 3254, .
[346] J. Heurich, M. G. Moore, and P. Meystre, Cavity atom optics and the ’free atom
laser’, Opt. Commun. 179 (2000), 549, .
[357] P. Horowitz and W. Hill, The art of electronics, Cambridge Universty Press,
2001, ISBN.
[358] Ch. Hotter, D. Plankensteiner, and H. Ritsch, Continuous narrowband lasing
with coherently driven v-level atoms, New J. Phys. 22 (2020), 113021, DOI.
[359] K. Huang, Statistical mechanics, John Wiley and Sons, 1987, ISBN.
[360] R. Huesmann, Ch. Balzer, Ph. W. Courteille, W. Neuhauser, and P. E. Toschek,
Single-atom interferometry, Phys. Rev. Lett. 82 (1999), 1611.
[361] D. A. W. Hutchinson, Self-consistent effects of continuous wave output coupling
of atoms from a bose-einstein condensate, Phys. Rev. Lett. 82 (1999), 6, .
[362] D. A. W. Hutchinson, E. Zaremba, and A. Griffin, Finite temperature excitations
of a trapped bose gas, Phys. Rev. Lett. 78 (1997), 1842, .
[363] Tetsuya Ido, Yoshitomo Isoya, and Hidetoshi Katori, Optical-dipole trapping of
sr atoms at a high phase-space density, Phys. Rev. A 61 (2000), 061403(R), .
[364] S. Inouye, M. R. Andrews, J. Stenger, H.-J. Miesner, D. M. Stamper-Kurn, and
W. Ketterle, Observation of feshbach resonances in a bose-einstein condensate,
Nature 392 (1998), 151, DOI.
[365] S. Inouye, A. P. Chikkatur, D. M. Stamper-Kurn, J. Stenger, D. E. Pritchard,
and W. Ketterle, Superradiant rayleigh scattering from a bose-einstein conden-
sate, Science 285 (1999), 571, DOI.
[366] S. Inouye, R. F. Löw, S. Gupta, T. Pfau, A. Görlitz, T. L. Gustavson, D. E.
Pritchard, and W. Ketterle, Amplification of light and atoms in a bose-einstein
condensate, Phys. Rev. Lett. 85 (2000), 4225, .
[367] S. Inouye, T. Pfau, S. Gupta, A. P. Chikkatur, A. Görlitz, D. E. Pritchard, and
W. Ketterle, Phase-coherent amplification of atomic matter-waves, Nature 402
(1999), 641, DOI.
[368] W. M. Itano, J. C. Bergquist, J. J. Bollinger, J. M. Gilli-Gans, D. J. Heinzen,
F. L. Moore, M. G. Raizen, and D. J. Wineland, Quantum projection noise:
Population fluctuations in two-level systems, Phys. Rev. A 47 (1993), 3554,
DOI.
[369] W. M. Itano, J. J. Bollinger, and D. J. Wineland, Quantum zeno effect, Phys.
Rev. A 41 (1990), 2295, .
[370] W. M. Itano and D. J. Wineland, Laser cooling of ions stored in harmonic and
penning traps, Phys. Rev. A 25 (1982), 35, .
[371] H. Ito, T. Nakata, K.Sakaki, M. Ohtsu, K. I. Lee, and W. Jhe, Laser-
spectroscopy of atoms guided by evanescent waves in micron-sized hollow optical
fibers, Phys. Rev. Lett. 76 (1996), 4500.
[372] A. D. Jackson, G. M. Kavoulakis, and C. J. Pethick, Solitary waves in clouds
of bose-einstein condensates, Phys. Rev. A 58 (1998), 2417, .
BIBLIOGRAPHY 1271
[373] B. Jackson, J. F. McCann, and C. S. Adams, Vortex line and ring dynamics in
trapped bose-einstein condensates, Phys. Rev. A 61 (2000), 013604.
[374] J. D. Jackson, Classical electrodynamics, John Wiley and Sons, 1999, ISBN.
[378] J. Javaloyes, Ph. W. Courteille, M. Perrin, G.-L. Lippi, and A. Politi, Sponta-
neous ordering in cold atomic clouds, Opt. Phot. News 16, (12) and 22 (2005)
(2005).
[379] J. Javaloyes, G. L. Lippi, and A. Politi, Reduced model for the description of
radiation-matter interaction including atomic recoil, Phys. Rev. A 68 (2003),
033405, .
[381] J. Javaloyes, M. Perrin, and A. Politi, Collective atomic recoil laser as a syn-
chronization transition, Phys. Rev. E 78 (2008), 11108, .
[384] P. S. Jessen and I. H. Deutsch, Optical lattices, Adv. At. Mol. Opt. Phys. 37
(1996), 95, .
[391] R. Kaiser, Quantum multiple scattering, J. Mod. Opt. 56 (2009), 2082, DOI.
[393] P. L. Kapitza and P. A. M. Dirac, The deflection of electrons from standing light
waves, Proc. Roy. Soc. Lond. (1933), 297.
[395] R. Karplus and M. Neumann, The scattering of light by light, Phys. Rev. 83
(1950), 776.
[402] , Evaporative cooling of trapped atoms, Adv. At. Mol. Opt. Phys. 37
(1996), 181, .
BIBLIOGRAPHY 1273
[404] W. Ketterle and Shin Inouye, Does matter wave amplification work for
fermions?, Phys. Rev. Lett. 86 (2001), 4203, DOI.
[405] W. Ketterle and H.-J. Miesner, Coherence properties of bose condensates and
atom lasers, Phys. Rev. A 56 (1997), 3291, .
[410] M. Kitagawa and M. Ueda, Squeezed spin states, Phys. Rev. A 47 (1993), 5138,
DOI.
[411] C. Kittel, Elementary statistical physics, John Wiley and Sons, 1976, ISBN.
[414] A. F. Koenderink and W. L. Vos, Light exiting from real photonic band gap
crystals is diffuse and strongly directional, Phys. Rev. Lett. 91 (2003), 213902.
[415] H. Kogelnik and X. Y. Li, Laser beams and resonators, Appl. Opt. 5 (1966),
155, DOI.
[486] S. Y. T. Meerakker and et. al., Direct measurement of the radiative lifetime of
vibrationally excited oh radicals, Phys. Rev. Lett. 95 (2005), 013003, .
[487] D. Meiser, Jun Ye, D. R. Carlson, and M. J. Holland, Prospects for a millihertz-
linewidth laser, Phys. Rev. Lett. 102 (2009), 163601, DOI.
[490] I. V. Mekhov, C. Maschler, and H. Ritsch, Light scattering from ultracold atoms
in optical lattices as an optical probe of quantum statistics, Phys. Rev. A 76
(2007), 053618, DOI.
[516] , Atomic four-wave mixing: Fermions versus bosons, Phys. Rev. Lett.
86 (2001), 4199, DOI.
[522] N.F. Mott and H. S. W. Massey, The atomic theory of atomic collisions i and
ii, Oxford Scientific Publications at the Clarendon Press, 1965, ISBN.
[524] G. M. Moy, J. J. Hope, and C. M. Savage, Atom laser based on raman transi-
tions, Phys. Rev. A 55 (1997), 3631.
[525] , Born and markov approximations for atom lasers, Phys. Rev. A 59
(1999), 667.
[526] G. M. Moy and C. M. Savage, Output coupling for an atom laser by state change,
Phys. Rev. A 56 (1997), R1087.
[547] S. Ospelkaus, A. Pe’er, K.-K. Ni, J. J. Zirbel, and B. Neyenhuis, Ultracold dense
gas of deeply bound heteronuclear molecules, ePrints (2003), 0802.1093, .
[548] H. Ouerdane and M. J. Jamieson, Scattering parameters for cold lirb and narb
collisions derived from variable phase theory, Phys. Rev. A 70 (2004), 022712, .
[551] A. S. Parkins and D. F. Walls, The physics of trapped dilute gas bose-einstein
condensates, Phys. Rep. 303 (1998), 1.
[555] E. Peik, M. Ben Dahan, I. Bouchoule, Y. Castin, and C. Salomon, Bloch oscil-
lations of atoms, adiabatic rapid passage, and monokinetic atomic beams, Phys.
Rev. A 55 (1997), 2989, DOI.
[605] D. S. Rokhsar, Vortex stability and persistent currents in trapped bose gases,
Phys. Rev. Lett. 79 (1997), 2164, .
[615] P. L. Saldanha, Quantum analysis of the direct measurement of light waves, New
J. of Phys. 16 (2014), 013021.
[624] Th. Sauter, R. Blatt, W. Neuhauser, and P.E. Toschek, Observation of quantum
jumps, Phys. Rev. Lett. 57 (1986), 1696, .
[627] A. L. Schawlow and C.H. Townes, Infrared and optical masers, Phys. Rev. 112
(1958), 1940.
[633] S. Schiller, Hydrogenlike highly charged ions for tests of the time independence
of fundamental constants, Phys. Rev. Lett. 98 (2007), 180801, .
[638] M. Schubert, I. Siemers, and R. Blatt, Line shape of three-level ions in paul
traps, Phys. Rev. A 39 (1989), 5098.
[641] , Condensation of n bosons and the laser phase transition analogy, Phys.
Rev. Lett. 82 (1999), 3927, .
[643] , The effects of the n atom collective lamb shift on single photon super-
radiance, Phys. Lett. A 373 (2009), 1283, DOI.
[647] Jian Qi Shen, Negatively refracting atomic vapour, J. Mod. Opt. 53 (2006),
2195, .
[648] Jian Qi Shen, J. Almlöf, and S. He, Negative permeability in a λ-type three level
atomic vapor, Appl. Phys. A 87 (2006), 291, DOI.
[649] Jian-Qi Shen, Zhi-Chao Ruan, and Sailing He, How to realize a negative refrac-
tive index material at the atomic level in an optical frequency range, J. Zhejiang
University SCIENCE 5 (2004), 1322, DOI.
[650] Jiteng Sheng, Yuanxi Chao, Santosh Kumar, Haoquan Fan, J. Sedlacek, and
J. P. Shaffer, Intracavity rydberg-atom electromagnetically induced transparency
using a high-finesse optical cavity, Phys. Rev. A 96 (2017), 033813, DOI.
[651] V. B. Shenoy and Tin-Lun Ho, First and second sound modes of a bose-einstein
condensate in a harmonic trap, Phys. Rev. Lett. 80 (1998), 3895.
[677] J. Stare and G. G. Balint-Kurti, Fourier grid hamiltonian method for solving
the vibrational schrödinger equation in internal coordinates: Theory and test
applications, J. Phys. Chem. A 107 (2003), 7204, .
[684] S. Stenholm, The semiclassical theory of laser cooling, Rev. Mod. Phys. 58
(1986), 699, .
[688] H. T. C. Stoof, Atomic bose gas with a negative scattering length, Phys. Rev. A
49 (1995), 3824, .
[689] H. T. C. Stoof and M. Bijlsma, Degeneracy effects on the relaxation and recom-
bination of adsorbed doubly polarized atomic hydrogen, Phys. Rev. B 49 (1994),
422.
[692] William C. Stwalley and L. H. Nosanow, Possible new quantum systems, Phys.
Rev. Lett. 36 (1976), 910.
[694] A. A. Svidzinsky and Jun-Tao Chang, Comment on: Effects of including the
counterrotating term and virtual photons on the eigenfunctions and eigenvalues
of a scalar photon collective emission theory; phys. lett. a 372 (2008) 2514,
Phys. Lett. 372 (2008), 5732, .
[704] E. Tiesinga, B.J. Verhaar, and H. T. C. Stoof, Threshold and resonance phe-
nomena in ultracold ground-state collisions, Phys. Rev. A 47 (1993), 4114, .
[709] G. Tóth, C. Knapp, O. Gühne, and H. J. Briegel, Optimal spin squeezing in-
equalities detect bound entanglement in spin models, Phys. Rev. Lett. 99 (2007),
250405, DOI.
[714] D. V. van Coevorden, R. Sprik, A. Tip, and A. Lagendijk, Photonic band gap
structure of atomic lattices, Phys. Rev. Lett. 77 (1996), 2412, .
[715] R. van der Stam, Superradiant scattering of laser light from a bose-einstein
condensater, Dissertation (2006).
[716] J. Vanier and C. Au doin, The quantum physics of atomic frequency standards,
Adam Hilger, Bristol and Philadelphia, 1989, ISBN.
[736] J. Weiner and P.-T. Ho, Light-matter interaction, fundamentals and applica-
tions, John Wiley & Sons, Hokoken, New Jersey, 2003, ISBN.
[737] D. S. Weiss, B.C. Young, and S. Chu, Precision measurement of h/mcs based on
photon recoil using laser-cooled atoms and atomic interferometry, Appl. Phys.
B 59 (1994), 217.
[739] M. Weissbluth, Atoms and molecules, Students Edition Academic Press, San
Diego, 1978, ISBN.
[740] G. Wentzel, Eine verallgemeinerung der quantenbedingungen für die zwecke der
wellenmechanik, Zeitschrift für Physik 38 (1926), 518.
[742] E. P. Wigner, On the behavior of cross sections near thresholds, Phys. Rev. 73
(1948), 1002.
[748] H. M. Wiseman, Defining the (atom) laser, Phys. Rev. A 56 (1997), 2068.
BIBLIOGRAPHY 1295
[766] Jun Ye, H. Schnatz, and L. W. Hollberg, Optical frequency combs: From fre-
quency metrology to optical phase control, IEEE J. Sel. Top. QE 9 (2003), 1041.
[767] Yutaka Yoshikawa, Yoshio Torii, and Takahiro Kuga, Superradiant light scat-
tering from thermal atomic vapors, Phys. Rev. Lett. 94 (2005), 083602, .
88
[768] Deshui Yu, Photonic band structure of the three-dimensional sr atomic lattice,
Phys. Rev. A 84 (2011), 043833, DOI.
[769] V. I. Yukalov, E. P. Yukalova, and V. S. Bagnato, Non-ground-state bose-
einstein condensates of trapped atoms, Phys. Rev. A 56 (1997), 4845, .
[771] Francesca Zambelli and Sandro Stringari, Quantized vortices and collective os-
cillations of a trapped bose-einstein condensate, Phys. Rev. Lett. 81 (1998),
1754, .
[772] Bihui Zhu, J. Marino, N. Y. Yao, M. D. Lukin, and E. A. Demler, Dicke time
crystals in driven-dissipative quantum many-body systems, Phys. Rev. Lett. 21
(2019), 073028, DOI.
[773] O. Zobay and B. M. Garraway, Atom trapping and two-dimensional bose-
einstein condensates in field-induced adiabatic potentials, Phys. Rev. A 69
(2004), 023605, .
[774] P. Zoller, M. Marte, and D. F. Walls, Quantum jumps in atomic systems, Phys.
Rev. A 35 (1987), 198, .
[775] M. W. Zwierlein, C. A. Stan, C. H. Schunck, S. M. F. Raupach, S. Gupta,
Z. Hadzibabic, and W. Ketterle, Observation of bose-einstein condensation of
molecules, Phys. Rev. Lett. 91 (2003), 250401, .
Index
Movie: ultracold fusion, 1076 absorption imaging, 932
Movie: ultracold single-atom fusion, 1076 absorption rate, 23
Movie: gravimeter Bloch, 830 absorption spectrum, 26
Movie: gravimeter Bloch & CARL, 830 accuracy, 1225
Movie: gravimeter CARL, 830 acousto-optic modulator, 1152, 1200
Movie: optomechanical coupling, 801 active component, 1104
Movie: CARL quantized motion 1, 830 active device, 1062
Movie: CARL quantized motion 2, 830 actuator, 1187
Movie: coupled dipoles model, 668 adiabatic cooling, 898
Movie: Feshbach resonance, 401 adiabatic decompression, 918
Movie: multi-level Bloch equations, 477 adiabatic potential, 604, 606, 920
Movie: quantum jumps, 559 adiabatic sweep, 473, 1066
Movie: optical Schrödinger cats, 503 adiabatic transfer, 606
Movie: Bloch oscillations, 187 Aharonov-Bohm effect, 578
Movie: forced oscillator, 126 Airy formula, 1125, 1200
Talk: AMO for long range studies, 884 Allan variance, 1225, 1229
Talk: quantized CARL, 823 AM, 1113
Talk: techniques for BECs, 884 amplitude modulation, 1113
Talk: Bloch oscillations, 187 analyzers
Talk: Bose glasses, 1028 optical spectrum, 1134
Talk: Bragg scattering, 670 anamorphic prism, 1091
Talk: cavity characterization, 707 Anderson
Talk: classical CARL, 801 Philip Warren, 669
Talk: Carl in the CQED regime, 842 Anderson localization, 669
Talk: Kuramoto model, 860 angular momentum quantum number,
Talk: lasers, 1136 147
Talk: long-range interactions, 860 annihilation operator, 488
Talk: matter wave superradiance, 1060 ansatz, 35
Talk: mirror-assisted CBS, 669 anti-particle, 237
Talk: phase-locked loops, 1202 antibunching, 521
Talk: photonic bands, 670 antisymmetric order, 130
Talk: projection noise, 572 antivalence, 775
Talk: quantum gates, 773 AOM, 1152
Talk: quantum mechanics, 35 Aristotle, 4
Talk: inertial sensing, 830 atom laser, 883, 1018
Talk: quantum Zeno effect, 568 atom optics, 883
Talk: recoil on trapped atoms, 126 atomic orbital, 320
Talk: ultracold fusion, 1073 atomic units, 223
Autler-Townes splitting, 457, 473
Abel transform autocorrelation function, 1227
inverse, 935 avoided crossing, 388, 457, 679, 919
Abricosov lattice, 1013 azimuthal equation, 146
absorption, 407, 412
absorption coefficient, 425, 931 Bénard
1297
1298 INDEX
Kuramoto Leucippus, 3
modelo de, 861 Lie algebra, 48
Kuramoto equation, 864 light shift, 457, 473, 611
Kuramoto model, 864 Lindbladt operator, 566, 611, 691
Lindbladt term, 568, 993
Laguerre linear cavity, 788
Edmond, 153 linear combination of atomic orbitals,
Laguerre polynomials, 153, 227 320
Laguerre’s associated differential equa- linear momentum space, 54
tion, 153 linear operator, 1216
Laguerre-Gaussian mode, 1100 Liouville equation, 438
Lamb Liouville operator, 438, 470
Willis Eugene, Jr., 255 Lippmann-Schwinger equation, 372
Lamb dip, 1172 Littrow configuration, 1166
Lamb shift, 255 local causality, 756
cooperative, 704 local density approximation, 976
Lamb-Dicke parameter, 117, 897 local density of states, 680
Lamb-Dicke regime, 117, 617, 618 local oscillator, 1115
Lamb-dip, 461 local realism, 761
Lamb-dip spectroscopy, 1171 localization energy, 85, 337
Lamb-shift lock-in amplifier, 1118
collective, 635 lock-in method, 1199
Lambert-Beer law, 28, 931, 1140 locking, 1189
Landé factor, 272, 275, 604 loop, 1105
Landau criterion, 985 loop gain, 1189
Landau gauge, 279 Lorentz
Landau level, 279 Hendrik Antoon, 247
Landau velocity Lorentz boost, 75
critical, 1006 Lorentz distribution, 421
Landau-Zener, 388 Lorentz gauge, 580
Landau-Zener formula, 920 Lorentz model, 423, 607
Langevin equation, 454, 850, 851 Lorentz transform, 75
quantum, 792 low-pass filter, 1203
Laplace transform, 1217 lowering operator, 159
large component, 235 LS-coupling, 169, 310
Larmor frequency, 12, 449, 911 LTI, 1216
laser, 511 Luttinger liquid, 967
state of a, 133 Lyot filter, 1123
laser emission bandwidth, 1232
laser gyroscopes, 711 Mößbauer
laser without inversion, 596 Rudolf Ludwig, 618
Legendre Mößbauer effect, 118, 618, 897
Adrien-Marie, 147 Mach
Legendre operator, 146 Ernst Waldfried Josef Wenzel, 1121
Legendre polynomials, 147, 228 Mach-Zehnder interferometer, 1121
lepton, 263 macromotion, 926
LeRoy-Bernstein method, 338 magnetic bottle, 909, 912
INDEX 1305
Zeeman
Pieter, 271
Zeeman effect, 12
anomalous, 272
normal, 272
Zeeman slower, 461, 907, 969
Zeeman splitting, 271
Zeno
de Elea, 568
Zeno effect, 569
zero point energy, 85
zero sound, 988
zero temperature reservoir, 1075
Zitterbewegung, 237, 253