Sie sind auf Seite 1von 32

Quantum Optics - lecture notes

R. J. C. Spreeuw
(Dated: 16 Nov. 2016 version 4.7)
2

CONTENTS

Preamble 3
What is quantum optics? 3
Examples of quantum optics experiments 3

I. Light, classical and quantum 3


A. Classical electromagnetic plane waves 3
B. Polarization, Jones matrices, and the Lie group SU(2) 5
C. Poincar and Bloch sphere 6
D. Quantization procedure 7
E. Expectation value of the field 8
F. Coherent or quasi-classical states 9
G. Superpositions of modes 9

II. Atom-light interaction 10


A. Time evolution driven by a perturbation; interaction picture 10
B. Time dependent perturbation theory, sinusoidal perturbation 11
C. Rotating wave approximation 11
D. Electric dipole coupling 12
E. Rabi oscillation 14
F. The density matrix 15
G. Optical Bloch equations: Schrdinger equation plus dissipation 17
H. Bloch vector 17
I. Steady state, polarizability, susceptibility, etc. 18
1. Steady state, and saturation 18
2. Lineshape and power broadening 19
3. Induced dipole moment 20
4. Linear polarizability of the atom 20
5. Linear susceptibility of the medium 21

III. Interaction of atoms with quantized fields 21


A. Jaynes-Cummings model (JCM) 21
B. Dressed states and light shifts 22

IV. Coherence functions and interferometry 23


A. First order coherence 23
B. Second order coherence 24

V. Entanglement 25
A. Entanglement 25
B. Reduced density matrix 25
C. Quantifying entanglement 25

A. Appendix 27
1. Real frequencies vs. angular frequencies 27
2. Circular polarization 27
a. Spin and helicity 27
3. Operators as matrices 28
4. Projection operators 28
5. Heisenberg picture 29
6. Interaction picture 29
7. Special (square) matrices/operators 30

Index 31

References 32
A Classical electromagnetic plane waves 3

PREAMBLE

Make sure you are formally registered for the course, to prevent issues with registering
your grade later.
Check your access to Blackboard. We will use Bb for announcing assignments, publishing
notes, to-dos, and various other course materials.
We will use the book by C.C. Gerry and P.L. Knight (to be abbreviated as GK), Intro-
ductory Quantum Optics (Cambridge Univ. Press, third printing, 2008). On Bb you will
find a list of chapters to study for the exam. There are also many other good books on the
subject [1]. In addition there are these supplementary notes.

What is quantum optics?

Narrow definition: The quantum theory of light. It includes all experiments, phenomena
where the discrete nature of light (photons) plays an essential role. The observation of any
such phenomenon (experiment) usually involves photodetection, and so the quantum theory
of photodetection is an important part of quantum optics.
Generalized definition: The description of light-matter interaction, with photodetection as a
special case. This usually also includes the case where the light can be described classically
and only the matter (atoms) is described quantum mechanically, a situation that is often called
semiclassical.

Examples of quantum optics experiments

1. The 2012 Nobel prize in physics was awarded to Serge Haroche and David J. Wineland,
two quantum optics pioneers. Haroches experiments involve a high-Q microwave cavity
allowing the measurement and manipulation of single microwave photons. Wineland
traps single ions and cools them to the vibrational ground state using lasers. While the
quantum nature of the laser light is usually not important, the quantization of vibrational
motion is. There are many analogies between the experiments of Haroche and Wineland,
and both are generally considered quantum opticians. For Haroches work the narrow
definition would apply, for Winelands the generalized one.
2. Hong-Ou-Mandel experiment [2]. If two photons impinge on a beam splitter, each enter-
ing from a different side, they always come out on the same output. This remarkable
quantum interference effect got a sequel in 2015. Several groups have now demonstrated
the same effect with atoms instead of photons [3].
3. The list of other examples of quantum optics experiments is endless. To mention a few:
single photon sources, in particular for quantum cryptography; Hanbury-Brown-Twiss
experiment; photon correlations using spontaneous parametric downconversion (SPDC)
sources, in particular Bell inequality measurements, teleportation, entanglement swap-
ping, etc.; light-matter interfaces to convert flying qubits to standing qubits; quantized
motion in optical lattices; cavity opto-mechanics; etc., etc.

I. LIGHT, CLASSICAL AND QUANTUM

A. Classical electromagnetic plane waves

In this course we are typically concerned with light (laser) beams and their interacation with
matter. A monochromatic plane wave with real-valued amplitude E0 and linear polarization e
4 I LIGHT, CLASSICAL AND QUANTUM

can be described by

E(r, t) = E0 e cos(krt + ) (I.1)

It will be convenient to split the cosine into two exponentials, as follows [see also GK (2.124)]

1
E(r, t) = E0 e ei(krt) + c.c. = E(+) (r, t) + E() (r, t) (I.2)
2

where c.c. stands for complex conjugate. A few remarks about this complex notation:

Note that in this notation the phase can simply be absorbed in the amplitude by allow-
ing it to be complex, E0 E0 ei . From now on we will drop the phase.

In the complex notation it is straightforward to express circular and elliptical polarization,


by using a complex polarization vector e. We will choose it to be normalized: e e = 1.
[N.B.: why cant Eq. (I.1) express circular polarization?]

By convention, the term E(+) (r, t) eit is called the positive frequency component, and
the eit term the negative frequency component; E(+) (r, t) is also called the analytic signal.
 
Since E(+) (r, t) = E() (r, t), all the information is contained in the analytic signal.

Looking ahead at field quantization, the positive and negative frequency components
will be associated with photon annihilation and creation operators, respectively.

In the laboratory we cannot directly measure the electric field amplitude of a laser beam, we
usually measure the power or the intensity. The latter can be obtained from the Poynting vector
S = EB/0 . We first calculate the magnetic field from the Maxwell equation E = B/t
which for a monochromatic plane wave simplifies to k E = B. The Poynting vector then
contains the double cross product E(k E) = (E E)k (E k)E. The second term vanishes
since another Maxwell equation in vacuum specifies that the electric field must be transverse,
through E = 0 or, for a plane wave, k E = 0. The Poynting vector can thus be written

1
S= (E(+) + E() ) (E(+) + E() ) k
0

Inspecting the various terms we find components oscillating as e2it , e2it , plus components
with no time dependence. In the optical regime the frequency is on the order 1015 s1 ,
much too fast for any photodetector. In the time-averaged Poynting vector the fast oscillating
terms disappear and we are left with only the E() E() terms. Finally we note that = c|k| =
ck and obtain

2  () k 1 k
hSi = E E(+) = |E0 |2 (I.3)
0 c k 2Z0 k
p
where Z0 = 0 /0 = 377 is called the impedance of the vacuum. The Poynting vector has
dimensions of an intensity: W/m2 and points in the direction of the wave vector; hSi is the
power that flows through a unit area perpendicular to the wave vector. The intensity of the
plane wave is thus proportional to the square of the electric field amplitude,

1 1
I = |hSi| = |E0 |2 = 0 c |E0 |2 (I.4)
2Z0 2

Note that the intensity is the product of the energy density 12 0 hE Ei+ 21 0 hB Bi = 0 hE Ei
(J/m3 ) and the speed of light c (m/s)[4].
B Polarization, Jones matrices, and the Lie group SU(2) 5

B. Polarization, Jones matrices, and the Lie group SU(2)

Lets consider a plane wave propagating in the z direction. As noted before E must be trans-
verse so it can have only x and y components. We can express the polarization vector as a
complex unit vector, a so-called Jones vector,

 
ex
e= (I.5)
ey

For example (ex , ey ) = (cos , sin ) describes linear polarization at an angle with respect to
the x axis, whereas (1, i)/ 2 describes circular polarization, , where the x and y compo-
nents are out of phase by /2.
We take the polarization of light, as expressed by a complex unit vector, as an opportunity to
discuss a few properties of the Lie group SU(2). Understanding some elementary properties of
SU(2) will let us describe several phenomena using the same tools. These phenomena include
(i) polarization of light, (ii) the dynamics of quantum two-level systems (qubits, two-level
atoms, spin-1/2 systems), and (iii) properties of a lossless beam splitter, interferometers, etc.
Lets consider a class of linear optical elements that can change the polarization. This class
contains waveplates, such as quarter and half wave plates and optical rotators. By linear we
mean that input and output are related by a linear function, and can therefore be described by a
matrix, eout = M ein . These matrices are also called Jones matrices. Many of these components
are to a good approximation lossless so that the Jones matrix preserves the norm of e for any
input polarization, i.e. M must be a 2 2 unitary matrix [5].
A light beam traversing a sequence of such lossless components undergoes a sequence of
polarization changes, described by the matrix product of the corresponding Jones matrices.
Since the sequence is itself lossless, its Jones matrix is also unitary. We say that the 2 2
unitary matrices form a continuous group, or Lie group, denoted U(2). The group criteria are
easily checked: (i) the product of unitary matrices is itself unitary, (ii) matrix multiplication
is associative, M1 (M2 M3 ) = (M1 M2 )M3 , (iii) the identity is unitary, I U(2), and (iv) each
element M has an inverse M 1 = M U(2).
It is often convenient to split off a phase factor, M = ei S, such that det S = 1. Such unitary
matrices with unit determinant form the Lie group SU(2). Since the global phase factor is
usually not significant, SU(2) is usually sufficient to describe the physics. A matrix in SU(2)
can be parametrized as
 

S= (I.6)

with ||2 + ||2 = 1.


A matrix in SU(2) can also be written as S = exp(iH) where H is hermitian and traceless. The
traceless property follows from the relation det eM = etr M . Since H is traceless 2 2 hermitian
it can be written as a linear combination of the Pauli matrices, H = h1 1 + h2 2 + h3 3 , with
real-valued hi and
     
0 1 0 i 1 0
1 = , 2 = , 3 = (I.7)
1 0 i 0 0 1

Introducing the vectors = (1 , 2 , 3 ) and h = (h1 , h2 , h3 ) the decomposition of H can be


abbreviated as H = h . Beware however that :
eiH 6= eih1 1 eih2 2 eih3 3 (I.8)
This factorization is incorrect because the Pauli matrices do not commute,
[1 , 2 ] = 2i3 (cyclic) or : [i , j ] = 2iijk k (I.9)
where summation over repeated indices is assumed, and ijk is the fully antisymmetric (Levi-
Civita) tensor: it equals +1 if (ijk) is an even permutation of (123), 1 for an odd permutation,
and zero otherwise.
6 I LIGHT, CLASSICAL AND QUANTUM

Figure I.1. Poincar sphere in the standard basis, with circular polarizations at the poles, linear polariza-
tions at the equator. In our definition, Eq. (I.11), the x and y polarizations are located at the poles, while
the circular ones are along the equatorial (0, 1, 0) directions. See also GK: Fig. A.1

For a traceless 2 2 matrix H = h we can correctly write (see also exercise):


 
iH sin h
e = (cos h) I + iH (I.10)
h
p
with h = |h| = det(iH) = det H and I the identity matrix. (If H is not traceless it can be
written as H = h0 I + h and the above expression for eiH gets multiplied by a scalar factor
eih0 .)
Eq. (I.10) is useful to obtain an explicit expression for the unitary evolution operator U =
eiHt/~ of a two-level system (with time-independent Hamiltonian), by substituting H
Ht/~.

C. Poincar and Bloch sphere

As noted above, the polarization can be represented by a two-component complex vector


(spinor) e = (ex , ey ). Alternatively, we can also form a three-component real vector s by taking
the expectation values of the three Pauli matrices. We define

ex ey + ey ex  2 Re(ex ey )

s1
s = s2 = hi = i ex ey ey ex = 2 Im(ex ey ) (I.11)
2 2 2 2
s3 |ex | |ey | |ex | |ey |

You may compare this to the definition by GK, in Appendix  A.2, given in 2termsofthe density
ex |ex | ex ey
matrix, which in this case would be = e e = (ex , ey ) = . Note that
ey ey ex |ey |2
one finds subtle variations on the definition of s in the literature [6].
If e is a unit vector, e e = 1, then s s = 1 (no complex conjugation needed, since s is real).
This means that we can graphically represent any polarization state as a real vector on a unit
sphere in the 3D (s1 , s2 , s3 ) space. This sphere is called the Poincar sphere, the real numbers
s1 , s2 , s3 are also called Stokes parameters. In the above definition the north and south pole
represent the linear x and y polarizations respectively. This is slightly different from the more
common definition where the two circular polarizations are depicted at the poles. The two
representations are of course related by a simple rotation of the Poincar sphere, i.e. a SU(2)
basis transformation on e, expressing each polarization as a superposition of right- and left-
handed circular polarizations[7].
The reason to stick to the above definition here is that it conforms to a similar definition for
the wave function (state vector) of a quantum two-level system, c1 |1i + c2 |2i. Writing (c1 , c2 )
as a column vector, the same definition for (s1 , s2 , s3 ) is called the Bloch vector which lives on
D Quantization procedure 7

the Bloch sphere, see also GK: Appendix A.2. In this case the north and south pole correspond
to the basis states |1i and |2i, and all other states are superpositions of these.
The Poincar (Bloch) spheres are powerful representations because it allows us to visualize
states and their changes by waving our arms. Any change of polarization (state) by a lossless
optical element now corresponds to a rotation of the entire Poincar sphere. For example,
sending a light beam through a quarter-wave plate corresponds to rotating the Poincar sphere
by 90. Similarly, the dynamical evolution of the quantum state of a two-level system is now
depicted as a path traced out on the Bloch sphere.
A more general definition of the Bloch sphere is in terms of the density matrix of the two-level
system. In that case the Bloch vector is no longer confined to the surface of the Bloch sphere.
The interior where |(s1 , s2 , s3 )| < 1 is then also accessible. We defer this discussion until we
discuss the density matrix. Analogously, the interior of the Poincar sphere corresponds to
light that is only partially polarized.

D. Quantization procedure

How does one go from a classical description of a physical system to the quantized version?
A very clear description of how one quantizes such a description in general can be found in
Ch. 15 of Basdevant & Dalibard[8]. The procedure goes back to Dirac (1925). GK are very
concise about quantizing the field. Here we sketch the more elaborate treatment given by
Grynberg-Aspect-Fabre (GAF):
1. First, a normal mode decomposition of the radiation field is made [GAF, Eqs. (4.63-65)].
By imposing periodic boundary conditions in some large, finite volume V = L3 , we have
gone from the continuous functions A(r, t), E(r, t), B(r, t) to a countably infinite set of
mode amplitudes l . Here l = (nx , ny , nz , s) is a set of discrete labels for the wave
vector k = 2L (nx , ny , nz ) and polarization s (s = 1, 2). Actually it is the transverse
vector potential A(r, t) (in the Coulomb gauge A = 0) that is decomposed as a Fourier
decomposition of plane waves. The electric and magnetic fields are obtained in the usual
way as E = A/t and B = A.
2. The Hamiltonian of the EM field can now be re-expressed in the mode amplitudes [GAF
(1)
(4.86-87)]. At this point the amplitude of a mode is still given as El l . This seems
inefficient, as one could lump this into a single symbol. Why it is useful to maintain this
(1)
notation becomes clear later on, where El will turn out to have a physical meaning: the
field of a single photon.
3. We recognize the real and imaginary parts of the normal mode amplitudes l as pairs of
conjugate variables, because they obey Hamiltons equations, [GAF (4.90-4.91)]

dQl H
=
dt Pl
dPl H
=
dt Ql
where Ql and Pl are proportional to Re(l ) and Im(l ) [GAF (4.92-93)].
4. Expressed in the Ql and Pl we recognize the Hamiltonian as being identical to the sum of
decoupled, simple harmonic oscillator (SHO) Hamiltonians, one SHO per normal mode,
[GAF, Eq. (4.105)].

5. With Ql (t) and Pl (t) we associate time-independent Hermitian operators Ql , Pl and im-
pose the canonical commutation relations
h i
Ql , Pl0 = i~ ll0 and [Ql , Ql0 ] = [Pl , Pl0 ] = 0

Together with Ql (t) and Pl (t), then also l is replaced by an operator, l al (not Her-
mitian).
8 I LIGHT, CLASSICAL AND QUANTUM

(1)
6. Finally, it is time to fix the value of El in such a way that the commutation relations
for al , al become equal to those of the raising and lowering operators of a harmonical
oscillator,
h i h i
al , al0 = ll0 and [al , al0 ] = al , al0 = 0

The operators themselves simplify to GAF, Eqs. (4.102-103).

7. The Hamiltonian H = H(Ql , Pl ) is recognized as a sum of SHOs. The operators al , al


raise or lower the energy of the harmonic oscillator by one quantum of energy ~l . They
are now interpreted as operators that create or annihilate a photon in mode l, respectively.
Note that Ql , Pl are Hermitian, but al , al are not.

The Hamiltonian of the radiation field can now be written as


 
X 1
HR = ~l al al +
2
l

The field operators take the form of [GAF (4.108-110)]. Note that the field operators in GAF are
independent of time, because they use the Schrdinger picture. The corresponding expressions
in GK are Eqs. (2.119-2.121), in this case given in the Heisenberg picture (see A 5), so they are
time-dependent. The time-dependent Heisenberg operators tend to be more common in the
literature.

E. Expectation value of the field

The electric field operator (Heisenberg picture) is given in GK: Eq. (2.120) as a Fourier series.
If we ask for the expectation value of the field thisDimmediately
E translates into a series with the
expectation values haks i Aks = |Aks | eiks and aks Aks as Fourier coefficients,

D E X  ~k 1/2 h i
E(r, t) = i eks Aks ei(krk t) Aks ei(krk t) (I.12)
20 V
ks
X  ~k 1/2
= 2 eks |Aks | sin(k r k t + ks ) (I.13)
20 V
ks

just a simple superposition of plane waves. Note that the polarization vector is here assumed
to be real, since it is outside the square brackets.
To get the coefficients Aks we need to calculate h |aks | i. Since we are working in the
Heisenberg picture
P the state vector is stationary, |i = |(0)i. For any given mode we could
write |i = n |ni. Then (omitting the index ks):
XX

XX

X


hai = m n hm |a| ni = m n n m,n1 = n1 n n A (I.14)
m n m n n

and of course a = A .


We make a few observations:

Although the Fourier series may seem rather intuitive at first, this simplicity is deceptive.
The amplitudes Aks are not a simple measure for the amount of energy (photons) in
the mode. They depend on the n s that make up the state |i. One easily checks that
Aks vanishes if the mode ks is in a number state |nks i, irrespective of how large the
photon number nks is. Even though a number state may contain a lot of energy, its field
amplitude has zero expectation value... States with a well-defined photon number are
remarkably non-classical!
G Superpositions of modes 9

While a single photon state has zero field expectation value, the superposition
pstate |i =
1/ 2 (|0i + |1i) yields A = 1/2, so that the amplitude of the sine wave is just ~k /20 V .
The physical meaning of this quantity is therefore that it sets the scale on which the field
amplitudes of photons are defined. It is colloquially called field of a single photon,
although this name should clearly be taken with a grain of salt.

F. Coherent or quasi-classical states

First introduced by Roy Glauber (Nobel prize 2005), coherent states play a special role in
quantum optics. They can be defined in different, equivalent ways. One definition is as a
displaced vacuum state. The displacement in this case is in the EQ EP plane. Another
common definition is as the eigenket of the mode annihilation operator, al |i = |i, with
complex eigenvalue . The importance of coherent states is a result of the following properties:

A coherent state behaves in many ways similar to the corresponding classical field, hence
the name quasi-classical.

A coherent state has an amplitude || and phase arg , as well defined as the quantum
uncertainty permits. It is a minimal uncertainty state.

Robustness: a coherent state sent onto a beam splitter, with transmission and reflection
coefficients t, r comes out as another coherent state, with complex amplitudes t, r.

For a coherent state the calculation of expectation values of normally ordered field oper-
ators is trivial: one simply replaces a and a . This follows from the fact that
the state is an eigenket of a. Normal ordering means that all creation operators are to the
left of all annihilation operators. For example, a a aa is normally ordered, a aa a is not.

Classical oscillating currents produce coherent states. Laser light tends to be close to a
coherent state.

G. Superpositions of modes

A photon is an elementary excitation of a field mode, which behaves like a harmonic oscilla-
tor. The field modes we have encountered so far were plane waves, entirely smeared out over
all of space. This picture is strikingly different from the experimental fact that it makes a big
difference whether we place our photodetector inside or outside the laser beam. Obviously,
more photons hit the detector when it is inside the laser beam. How do we make the photon
picture consistent with this common sense experience?
At this point we must realize that modes are essentially basis functions. As with any basis
of a linear vector space, there is some freedom of choice. Using a basis transformation we can
define new mode functions as superpositions of the old modes. These new mode functions can
be constructed such that they are localized, e.g. where the laser beam is. Instead of eikr we
would then use a spatial function whose amplitude peaks at the laser beam and is essentially
zero outside. n o P
Given a set of modes {ai } we can define a new set bi , so that bi = j uij aj . It is easy to
show that if the transformation
h i is unitary, u u = uu = I, then this will preserve the commu-
tation relations, bi , bj = ij and preserve the total photon number, i bi bi = i ai ai . The
P P

vacuum |{0}i is also left unchanged. Some other properties do change, of course. For example,
while traveling plane wave modes have a well defined wave vector k, this is no longer true
for a superposition of plane waves. This means that the photons no longer have well-defined
momentum p = ~k.
As long as we only mix modes of the same frequency, we only change the spatial structure
of the modes. When we mix modes with different frequencies it becomes a little more compli-
cated. The Hamiltonian is then no longer a sum of decoupled harmonic oscillators, and will
10 II ATOM-LIGHT INTERACTION

contain cross
h terms.
i Nevertheless some properties are still conserved, such as the commutation
relation bi , bj = ij . It is also still meaningful to construct states with a total photon number,
it will be preserved.
Since there is now a range of different frequencies in the superposition, the resulting electric
field profile (e.g. the Heisenberg operator for E(r, t), or its expectation value) will be time
dependent. Such time-dependent superpositions are sometimes called wave packets.

II. ATOM-LIGHT INTERACTION

A. Time evolution driven by a perturbation; interaction picture

We write the Hamiltonian as a sum of a stationary part and perturbation, H(t) = H0 +


H1 (t). Assuming we know the eigenstates of H0 , so H0 |ni = En |ni we use it as a basis, and
investigate transitions between the |ni due to the perturbation by H1 (t). Since the eigenstates
evolve in time as exp(iEn t/~) |ni we can write [GK (4.17)]:
X
|(t)i = n (t) eiEn t/~ |ni (II.1)
n

so that the coefficients n (t) vary only slowly compared to the fast oscillating exponentials. We
now take the time derivative of this equation and set it equal to (i/~)H(t) |(t)i,
X iEn

iX  
n (t) n (t) eiEn t/~ |ni = n (t) eiEn t/~ En + H1 (t) |ni (II.2)
n
~ ~ n

Many terms cancel:


X iX
n (t) eiEn t/~ |ni = n (t) eiEn t/~ H1 (t) |ni (II.3)
n
~ n

We left multiply by hm| and use orthonormality, hm| ni = nm and obtain [GK (4.20)]

iX
m (t) = n (t) ei(Em En )t/~ hm| H1 (t) |ni (II.4)
~ n

We recognize the matrix elements of H1 (t) in the interaction picture (see also A 6):

(t)]
[H i(Em En )t/~
hm| H1 (t) |ni (II.5)
1 mn = e

so that finally Eq. (II.4) reduces to a matrix-vector product,

iX
m (t) = [H1 (t)]mn n (t) (II.6)
~ n

Lets make a few observations:

Eq. (II.6) is nothing but the Schrdinger equation. By choosing a discrete basis it takes
the form of a matrix equation v = M v, i.e. a set of coupled linear differential equations.

Eq. (II.6) shows that to drive a transition n m we need off-diagonal elements


hm| H1 (t) |ni.

Due to the fast oscillating term ei(Em En )t/~ significant transition amplitude will only
build up if hm| H1 (t) |ni contains a frequency component near mn = (Em En ) /~.
A Time evolution driven by a perturbation; interaction picture 11

B. Time dependent perturbation theory, sinusoidal perturbation

Extensive accounts of perturbation theory are given in many text books. In these notes we
concentrate on the aspects that are most relevant for the rest of the course. We consider the
situation that the initial state is |ii, so that i (0) = 1 and all others are 0, so k (0) = ik . For
small times we can directly integrate Eq. (II.6) by assuming i (t) 1 and k (t)  1 for k 6= i.
This leads to the following expression for the transition amplitude,
t
1 0
Ski (t) = k (t) hk| H1 (t0 ) |ii eiki t dt0 (II.7)
i~ 0

2
and for the transition probability: Pik = |Ski |  1.
The integral can be done analytically for a sinusoidal perturbation,

W  i(t+) 
H1 (t) = W cos(t + ) = e + ei(t+) (II.8)
2
resulting in:
 
Wki sin(ki )t/2 i(ki )t/2i sin(ki + )t/2 i(ki +)t/2+i
Ski (t) e + e (II.9)
2~ (ki )/2 (ki + )/2

Significant excitation occurs only near resonance, ki . Inspection of the denominators in


above expression then shows that the second term must be expected to be much smaller than
the first. We can usually neglect it. This approximation is called the Rotating Wave Approxima-
tion (RWA). For short times, t  1/| ki | we can furthermore approximate the sin function,
yielding

|Wki |t |Wki |2 t2
|Ski (t)| , Pik (t) (II.10)
2~ 4~2
So we find that for short times the excitation probability grows as t2 . Note that this is
different from Fermis Golden Rule (FGR), which yields Pik (t) t. The difference in be-
haviour is a consequence of the spectral purity of the perturbation, which we assumed here to
be monochromatic. We can retrieve the FGR result by assuming a finite frequency bandwidth
in the perturbation. To see this we can plot the sinc function, see Fig. (II.1). The function has a
peak height of t2 . The first zero occurs at |ki | = 2/t. Thus we see that the t2 behaviour
occurs for detunings small compared to 2/t. If the excitation is spectrally broad compared to
2/t, we must effectively integrate over this sinc function, yielding a result t, so we retrieve
the FGR behaviour. For longer times the function approaches a delta function and effectively
only the spectral energy density around zero detuning is filtered out.
The t2 behaviour has a fascinating consequence. If one performs a quantum measurement
one will likely find the quantum system in the initial state. The measurement then projects the
system back on the initial state. If such measurements are repeated frequently, the system is
thereby prevented from leaving its initial state. This phenomenon whereby the quantum evo-
lution is frozen is called the quantum Zeno effect and was experimentally observed in 1990.[9]

C. Rotating wave approximation

The so-called rotating wave approximation (RWA) led us to neglect the second term in Eq.
(II.9). This boils down to selecting only the positive frequency component in hk| H1 (t) |ii. By
hermiticity we must then take the negative frequency component in hi| H1 (t) |ki. This can also
be expressed (see also A 3) by rewriting Eq. (II.8) as

Wki ei(t+)
 
Wki i(t+) Wik i(t+) 1 0
H1 (t) e |ki hi| + e |ii hk| = (II.11)
2 2 2 Wik ei(t+) 0
12 II ATOM-LIGHT INTERACTION

 2
sin(ki )t/2
Figure II.1. The function (ki )/2


with the matrix defined in the subspace spanned by {|ki , |ii}, and Wki = Wik for hermiticity.
Similar terms would appear for different |ki states in corresponding subspaces. Assuming that
|ki is at a higher energy than |ii, i.e. ki > 0 this means that excitation i % k is driven by the
positive frequency component ( eit ) of the perturbation. We have already seen [GK, Ch. 2]
that the positive frequency component of a laser field is associated with the photon annihilation
operator. This makes intuitive sense: a photon disappears when the atom is excited. The
opposite process, de-excitation k & i is associated with the negative frequency component
( e+it ) and a photon creation operator: a photon is emitted when the atom is de-excited to a
lower state.
The terms that have been neglected may seem energy nonconserving and therefore coun-
terintuitive: excitation of an atom while at the same time emitting a photon, or de-excitation
while absorbing a photon. These terms are not strictly forbidden, they are just extremely non-
resonant. The apparent violation of energy conservation occurs if we count only the energy of
the unperturbed Hamiltonian H0 . The full Hamiltonian, including resonant and non-resonant
interactions, strictly conserves energy.
The RWA is usually extremely well justified for optical transitions and therefore in practice
this approximation is almost always assumed. Conditions where deviations from the RWA
appear can, for example, be reached in the microwave or radio-frequency regimes.

D. Electric dipole coupling

A very clear account of the form of the Hamiltonian of the electric dipole interaction is given
in GAF (secs. 2.2.1-2.2.4). The p A and d E forms are essentially equivalent and related by a
gauge transformation. Here we will use the latter (as in GK),

HI = d E(r0 , t) (II.12)

Remarks:
d = q(r r0 ) = (dx , dy , dz ) is a vector operator; E is also a vector but not an opera-
tor (for now; until we replace it by the quantized field). This is called the semiclassical
description.
r is the position operator for the electron.

r0 is the position (not an operator, or strictly r0 1 ) of the core/nucleus of the atom (we
may set it to zero). The electric field is taken as uniform over the size of the atom. This
is called the long wavelength approximation. This is usually well satisfied, because light
in the visible spectral region has a wavelength of 1 m, while the extent of the electron
wavefunction is 0.1 nm. This is called the dipole approximation.
Since atomic Hamiltonians commute with the parity operator, unperturbed atomic states have
definite parity: the electron wavefunction is either even or odd. Such definite parity states
D Electric dipole coupling 13

Figure II.2. Probability density |(r, t)|2 for a superposition of an s-like (L = 0) and a p-like (L = 1)
i0 t
electron wavefunction, if the s and p state have E energy difference 0 , so |(r, t)i = |si + e
D an |pi.
The expectation value of the dipole moment d oscillates at a frequency 0 . In the visible domain:
0 1015 s1 .

D E
have no dipole moment: d = q (r) r (r) d3 r = 0, because the integrand is always an
odd function if (r) has definite parity. If we express the dipole moment operator in the basis of
atomic eigenstates, diagonal moments must therefore vanish. Off-diagonal elements between
states of different parity can be nonzero. For example
D E D E D E
g d g = e d e = 0, e dz g = d (II.13)

We call such matrix elements dik transition dipole  matrix  elements. The dipole operator is

a vector operator, with three components, d = dx , dy , dz . For simplicity we will from here
on look at one component only, namely the component that interacts with the laser, i.e. the
component parallel to the laser electric field, say dz = d(|gi he| + |ei hg|), choosing d to be real
without loss of generality. We will suppress the subscript z from here onward.
Superposition states can have a nonvanishing dipole moment, as can be seen in Fig. II.2.
This dipole oscillates at a frequency 0 = (Ee Eg )/~. Writing |(t)i = g |gi + e ei0 t |ei,
the dipole moment is
D E
d = g e d ei0 t + c.c. = 2d |g e | cos(0 t ) (II.14)

We can recognize 2g e as s1 + is2 in the Bloch vector picture, cf. Eq. (II.32-II.33). A typical
value for the amplitude is d ea0 1029 Cm. With real-valued d the dipole operator can be
written as d = d1 , which can be split into a raising and a lowering part [GK (4.94)],

d = d1 = d(+ + )
   
1 0
Choosing, arbitrarily, |ei = , |gi = , we have
0 1
 
0 1
+ = =
0 0

Using Eq. (II.14) we recognize the lowering part = |gi he| as the term giving rise to the
ei0 t term. Similar to the separation we made for the electric field, we could identify this term
as the positive frequency part (and e+i0 t as the negative frequency part). So we see that the
lowering part of the dipole operator is associated with the positive frequency part. This is similar to
the situation with the electric field: the positive frequency component is associated with the
photon annihilation operator, which lowers the number of photons (GK Ch. 2, or these notes,
Sec. I).
The electric dipole coupling Hamiltonian now takes a simple form if we make the rotating
wave approximation (RWA). The RWA consists of putting the positive frequency part of the
electric field on the (e, g) position in the operator matrix,
d
E0 eit
 
0
HI = d it 2 = d( E () + + E (+) ) (II.15)
2 E0 e 0

In the fully quantized treatment, we will associate E () with photon creation [see GK, Ch. 2].
The RWA then says that lowering of the atomic state ( , e & g) is accompanied by the addition
14 II ATOM-LIGHT INTERACTION

of a photon to the field (a ), and vice versa. The two neglected processes, lowering the atomic
state while absorbing a photon, and raising the atomic state while emitting a photon, are also
allowed but usually highly nonresonant.
Finally, we introduce the Rabi frequency 1 = dE0 /~ [10] so the interaction Hamiltonian
becomes (cf. Eq. (II.11))
1 eit
 
~ 0
HI = (II.16)
2 1 eit 0

E. Rabi oscillation

Eq. (II.16) describes the interaction Hamiltonian for two-level system driven by a monochro-
matic field, under the RWA. We consider a two-level system with states |ei , |gi, with energies
Ee = ~0 /2 = Eg . The total Hamiltonian can then be written as
1 eit
 
~ 0
H(t) = (II.17)
2 1 eit 0
The time evolution for this case can be solved analytically. We do this in two steps: (i) a
time-dependent unitary basis transformation U (t) to arrive at a time independent transformed
Hamiltonian H, and (ii) explicit calculation of exp(iHt/~).
1: Eremoving the time dependence We consider a unitary basis transformation |(t)i =
Step
U (t) (t) . By inserting this into the Schrdinger equation,

i E E E
|i = HU = U + U (II.18)

t ~ t
E
and multiplying from the left by U we can obtain a Schrdinger equation for (t) :

E i E
= H (II.19)
t ~
with the transformed Hamiltonian:
H(t) = U HU i~U U (II.20)
We can use such a transformation to remove the time dependence from the Hamiltonian (II.17).
We write |(t)i = eit/2 e (t) |ei + eit/2 g (t) |gi , i.e. we choose
 it/2 
e 0
U (t) = (II.21)
0 eit/2
This leads to
   
~ 0 1 ~ 1
H = = (II.22)
2 1 0 2 1
which is no longer time dependent. Here the detuning has been defined as = 0 .
Step 2: calculation
E of the time evolution
E The time evolution can now be calculated analyti-
cally as (t) = exp(iHt/~) (0) . The unitary evolution operator exp(iHt/~) follows

explicitly from the identity presented above, Eq. (I.10), by substituting


E H Ht/~.
Lets look at the situation where the initial state is (0) = |gi and ask for the probability

that at time t the atom is found in |ei. That is, we need to calculate only a single matrix element,
and find
21
 
iHt/~ 2 R
D E
2
Pge (t) = e e g = 2 sin t (II.23)
R 2
p
with R = 21 + 2 .
F The density matrix 15

Remarks on the method

Alternatively, we could have solved the coupled linear differential equations of Eq. (II.4)
by reducing them to a single second order differential equation. It resembles a driven
harmonic oscillator. See also [GK: Sec. 4.4]

The first step to remove the time dependence will become very natural in terms of quan-
tized field states. It is then known as the dressed atom picture (to be discussed later).

In the literature it is also often referred to as the rotating frame. This term comes histor-
ically from the language of magnetic spin resonance (NMR, ESR). The rotating wave is
then the component that is stationary in the rotating frame.

Discussion of the result (II.23)


p
We see that the transition probability oscillates in time at the frequency R = 21 + 2
called the generalized Rabi frequency. The oscillation itself is known as Rabi oscillation or
Rabi flopping.

The maximum excitation is obtained for zero detuning ( = 0), i.e. right on resonance.
For increasing detuning (positive or negative) the oscillation frequency goes up, the max-
imum excitation goes down.

The Rabi frequency


scales with the field amplitude, i.e. the square root of the intensity:
1 E0 I.

For small times, the excitation grows quadratically with time, Pge (t) 21 t2 /4, in agree-
ment with first-order perturbation theory, see also Eq. (II.10).

Full excitation is obtained for = 0 after a time such that R t = 1 t = . This can be
achieved using a laser pulse that is well tuned and precisely timed, a so-called -pulse.

We can also justify the RWA a posteriori by estimating what we would have obtained if,
in Eq. (II.16), we had chosen the other exponential, eit eit . Replacing 20
(approximately) in Eq. (II.23) and using the fact that in the optical domain usually 1 
0 , we find that the maximum excitation is approximately 21 /402  1. For example, for
1 1010 s1 (quite high) and 0 1015 s1 we get 21 /02 1010 : the RWA is usually
extremely well justified[11].

F. The density matrix

Not all states can be represented by a wavefunction. This is particularly relevant when deal-
ing with dissipation. In such cases we will represent the state of the atoms by a so-called density
matrix. Consider for example an atom in its excited state |ei that decays by spontaneous emis-
sion of a photon to its ground state |gi. The initial state |ei will evolve into an entangled state
of the atom and the emitted photon. We can write down the combined quantum state (wave
function) of the atom plus the radiation field.

X
|ei |eiatom |vacifield + k, |giatom |1k, ifield
k,

where |1k, ifield stands for a state of the radiation field containing one photon with wave vector
k and polarization . This expression contains very many coefficients k, : too many to keep
track of and they usually do not interest us. Instead we want a description for the atom alone,
without the radiation field. The problem is that the above state cannot be factorized in a state
16 II ATOM-LIGHT INTERACTION

for the atom and a state for the field. Such non-factorizable states are called entangled states,
X
|eiatom |vacifield + k, |giatom |1k, ifield 6=
k,

X
6= (b1 |gi + b2 |ei)atom d0 |vaci + dk, |1k, i (II.24)
k,
field

The way out of this is to represent the state of the atom not by a wavefunction but by a gener-
alized object: a density matrix.
For a pure state, i.e. a state that can be represented by a wave function, the corresponding
density matrix is = |i h| . For example,

|c1 |2 c1 c2
     
c1 c1
c1 c2

|i = = = (II.25)
c2 c2 c2 c1 |c2 |2

The diagonal elements ii are called populations, describing the probabilities to find the system
in that state. Off-diagonal elements ij are called coherences. If we compare Eqs. (II.14) and
(II.25) we notice that ij determines the expectation value of the dipole moment.
A more general case is a statistical mixture: |i = |i i with probability pi . In that case the
density matrix is
X
= pi |i i hi | . (II.26)
i

For example, |i = |1i with probability p1 and |i = |2i with probability p2 . Then:
 
p1 0
= p1 |1i h1| + p2 |2i h2| = (II.27)
0 p2

Note that the two s are different only in their coherences: in a statistical mixture ij may
vanish, even though both populations are nonzero. For the of Eq. (II.27) with zero coherences
the corresponding wave function |i for the atom alone does not exist, 6= |i h|. This reflects
the fact that we have incomplete information about the atomic state because we ignored its
entanglement with the environment.
Properties of the density matrix

For a normalized |i, its density matrix is its projector |i h|. In particular, |i = |i
and 2 = . This is no longer true for a statistical mixture: in that case 2 6= .
D E
The expectation value of an observable A is given by A = Tr A = Tr A.

Tr = 1, because the diagonal contains populations, which mustD addE up to one; Tr is



also the expectation value of the identity operator: Tr = TrI = I = 1.

can describe anything from a quantum superposition (full coherence) to a classical


statistical distribution (no coherence).

is hermitian: = , or ij = ji . This follows directly from the definition (II.26),


however there are additional requirements:

is positive semidefinite. This means that ii 0 in any basis (equivalent to h| |i 0


for arbitrary |i.) All eigenvalues must be non-negative, because they represent popula-
tions in the basis of eigenvectors.
H Bloch vector 17

G. Optical Bloch equations: Schrdinger equation plus dissipation

The time evolution as governed by the Schrdinger equation can now be re-expressed in
terms of the density matrix. It is straightforward to show that a pure state density matrix
= |i h| evolves as = ~i [H, ]. Merely rewriting the Schrdinger equation in this dif-
ferent form is not very usful by itself. The great advantage however comes from the fact that
the density matrix formulation allows us to include interaction with an environment that we
cannot or do not want to include in our description of the system (atom). The environment
will then become manifest as dissipation. A well known example is spontaneous emission.
The dissipation terms are added as an extra term:
 
d i
= [H, ] + (II.28)
dt ~ t diss

We give here the form for a closed two level system, meaning that level e can decay only to
level g. In this case the dissipation terms are as follows:

ee gg
= ee = (II.29)
t t
ge eg
= ge = eg = (II.30)
t t
The precise form of the dissipation terms can be easily obtained using the Lindblad superop-
erator formalism, see the exercises. Note that in this case (ee + gg )/t = 0, i.e. the total
population remains constant. The decay rate of the coherences must be at least half the rate for
the populations, /2. The equality holds if the decay e g is the only mechanism that
reduces the coherences, i.e. if no additional dephasing processes, such as collisions, occur.
For a two-level system under the action of the Hamiltonian H described in Eq.(II.22) the time
evolution of the density matrix takes the following form,

1
ee = gg = i (eg ge ) ee
2
1
ge = eg = ige i (ee gg ) ge (II.31)
2
These coupled, linear differential equations are called the optical Bloch equations (OBE). It
is easy to solve for the steady-state, = 0, which is left as an exercise. Keep in mind that the
Hamiltonian H is valid within the RWA and has already been transformed to the rotating
frame, to get rid of fast oscillating terms at the frequency of the driving field . Without
this we would have obtained rapidly oscillating terms in the OBE. The time evolution of these
terms is rather trivial and obscures the relatively slow transitions that we are more interested in.
Furthermore, if we wish to numerically solve the OBE on a computer, the fast oscillating terms
could force the integration routine to make very many small time steps. This can unnecessarily
slow the calculation down. (If desired, the oscillating terms can always be put back in, by doing
the inverse transformation that was done in Sec. II E)

H. Bloch vector

For a two-state system, the density matrix can also be represented by a real, three-component
vector, called the Bloch vector (see also GK: A.2). Similar to the definition of the Poincar vector
(Sec. I C) we take the expectation values of the Pauli matrices, si = hi i = Tr (i ), obtaining

s1 = eg + ge = 2 Re eg (II.32)
s2 = i(eg ge ) = 2 Im eg = 2 Im ge (II.33)
s3 = ee gg (II.34)
18 II ATOM-LIGHT INTERACTION

It is left as an exercise to show that the definition s = hi 


is equivalent
 to writing
 =
1/2(I + s ) as in GK: Eq. (A25). Choosing, arbitrarily, |ei =
1 0
2 , |gi = , the Bloch
0 1
vector points upward for |ei, downward for |gi.
The OBE can also be expressed in terms of the Bloch vector:
s1 = s2 s1
s2 = s1 1 s3 s2
s3 = 1 s2 (1 + s3 )
In the absence of damping ( = = 0) we may recognize a simple precession motion:

1
s = R s, with R = 0 (II.35)

The Bloch vector s behaves like a fictitious spin


p vector, precessing about a fictitious magnetic
field R , at an angular frequency R = 21 + 2 . The damping terms drive the Bloch
vector towards the ground state, s = (0, 0, 1), the south pole on the Bloch sphere.
A few remarks
In all of this we have used the transformed hamiltonian, Eq. (II.22) in the so-called rotat-
ing frame. We can now better understand this nomenclature. Eq. (II.21) just corresponds
to a transformation to a rotating coordinate frame in s space. Without this transformation,
the atomic dipole would spin around in the s1 s2 plane at a very high frequency 0 . The
driving field 1 cos t = (1 /2)(eit + eit ) would contain two terms that rotate in the
s1 s2 plane in opposite directions at angular frequency . One of these would rotate in the
same sense as the spinning dipole, the other would counterrotate. The counterrotating
term is of course the one we neglect in the RWA. The rotating frame transformation is
done so that the corotating field appears stationary in the rotating frame.
The fictitious magnetic field of Eq. (II.35) gives intuitive insight into Eq. (II.23). On
resonance ( = 0), R points along the s1 direction, so s can precess all the way from the
south pole to the north pole. On the other hand, if 6= 0 the R is tilted up or down, so
that the precession cone, starting from |gi is narrower
p so that |ei cannot be reached. At
the same time the precession frequency R = 2 + 21 increases.

I. Steady state, polarizability, susceptibility, etc.

The optical Bloch equations (OBE) above, Eq. (II.31), were given for a two-level system. Gen-
eralizing this to larger systems of more than two levels is straightforward. The OBE provide a
general framework to solve for the time dependence of a quantum system with discrete states
in the presence of dissipation. The time-dependent solutions typically show initial transient be-
havior which is damped by the dissipation. An important case is where the dissipation is due
to spontaneous emission to lower energy levels. The transients then damp out in a few sponta-
neous emission lifetimes, typically 10-100 ns. Therefore we will now analyze the steady state of
the optical Bloch equations, to see what information can be obtained. We take the steady state
solution as a starting point, describing a two-level atom excited by a near-resonant, monochro-
matic laser field. Analyzing this as a function of the detuning and intensity of the exciting
laser we find the (Lorentzian) line shapes, as well as phenomena like saturation and power
broadening, see also exercises.

1. Steady state, and saturation

Obviously, the diagonal density matrix elements give us the populations in the various energy
levels. We take the steady state of the optical Bloch equations and set = /2 (i.e. no other
I Steady state, polarizability, susceptibility, etc. 19

dephasing apart from spontaneous emission). It is often convenient to define the saturation
parameter,

21 /2 I/I0
s= = (II.36)
2 + 2 /4 1 + 42 /2

where the second equality is obtained by defining the saturation intensity I0 by the relation
I/I0 = 221 /2 . The saturation parameter is thus proportional to the intensity of the laser:
21 E02 I = 21 0 cE02 .
The steady state of the OBE can now be expressed as
1 s
(ss)
ee =
2 1+s
1 2+s
(ss)
gg = = 1 (ss)
ee
2 1+s
p
1 i1 1 s/2
(ss)
eg = , |(ss)
eg | =
2 /2 i 1 + s 1+s
Lets make a few observations:
(ss)
1. At low saturation parameter the excited state population ee s/2 , i.e. proportional to
the laser intensity.
(ss)
2. At large saturation ee < s/2 so it grows less than linear with s. This phenomenon is
called saturation. In fact, the excited state population reaches an asymptotic value of 1/2.
Since the ground and excited state populations add up to one, the ground state also has
a population of 1/2. In other words, for strong saturation the populations become equal.
(ss) (ss)
3. At s = 1 the population difference gg ee is reduced to 1/2, for s it is reduced to
zero.

2. Lineshape and power broadening

The excitation lineshape is most easily seen from

1 21 /2
(ss)
ee =
2 2 + 2 /4 + 21 /2

for = /2. As the laser frequency is scanned, the excited state population shows resonance
behavior. The maximum is reached on resonance, = 0. The resonance has the lineshape of
a Lorentzian function ( 1/1+x2 ). At low laser intensity the width of this Lorentzian is . The
width is here measured by looking where the Lorentzian function has half its peak value (Full
Width at Half Maximum, FWHM).
At
p increasing laser intensity the lineshape stays Lorentzian. The width however increases

to 2 + 221 = 1 + s0 , where s0 is the saturation parameter on resonance. This linewidth
broadening with increasing laser intensity is called power broadening.
The width is called the natural linewidth: the minimum possible width assuming only
spontaneous emission and no other linewidth broadening mechanisms. In reality there are
usually reasons why the observed linewidth is larger than . For example if atoms in a gas
collide with other atoms, this tends to dephase the atomic dipoles, > /2. This is called
collisional broadening. Very common is also Doppler broadening. The atoms in a gas move
in random directions with thermal velocities. In the rest frame of a moving atom the laser fre-
quency appears shifted by the Doppler effect. In the lab frame all atoms then have effectively a
slightly different resonance frequency. The distribution of these individual frequencies reflects
the velocity distribution in the gas. The corresponding lineshape is typically Gaussian instead
of Lorentzian. At room temperature Doppler broadening usually dominates over the natu-
ral linewidth, however for very cold atomic vapors this is no longer the case. Nowadays we
20 II ATOM-LIGHT INTERACTION

can cool atomic vapors to K or even nK temperatures where Doppler broadening practically
disappears.
In liquids or solids the situation is even more complex due to the higher densities and
stronger interactions. For example, individual atoms may be subject to different environments
due to crystal stresses etc.

3. Induced dipole moment

The off-diagonal elements, or coherences, give us access to the dipole moment. We should
note that the steady state (ss) was only steady in a time dependent basis. In the original basis
(ss)
we have eg = eit eg , etc., and for the dipole moment we get
D E       
0 1 0 1 it
d = d = tr d = d (eg + ge ) = d ((ss)
eg e + c.c.) (II.37)
1 0 1 0
The dipole moment oscillates at the laser frequency, with a phase difference given by the com-
(ss) (ss) (ss)
plex argument of eg . Writing eg = |eg | ei we get
D E
d = 2d |(ss) (ss)
eg | cos(t ) = 2d |eg | cos[(t )] (II.38)

For the phase of the dipole we must be careful because 1 = dE0 /~ < 0 if E0 > 0. We then
(ss)
see that the phase is in the range 0 < arg(eg ) < , so that the dipole moment lags behind the
field. Lets look at the low saturation limit, and set = /2 (only spontaneous emission) in the
OBE, Eq. (II.31),
i dE0 /~
(1)
eg = + (s  1) (II.39)
2 /2 i
We can distinguish:
= 0 (on resonance): the dipole lags the field by a /2 phase delay (or = T /4 time
delay, one quarter period)
 /2 (far above resonance): the dipole has a phase lag of approximately (or
T /2), so it is in anti-phase with the field
 /2 (far below resonance): the dipole has almost no phase lag, 0 ( 0), and
is in phase with the field
The behavior of the phase lag near 0 has the characteristic resonant behavior, varying as
an arctan(2/).

4. Linear polarizability of the atom

The polarizability is the induced dipole moment per unit field applied. We define the linear
(ss)
complex polarizability by d eg = (1) (E0 /2), so
i d2 /~ d2 /~
(1) = = (II.40)
/2 i + i/2
If the polarizability is divided by 40 we get something with the units of volume. If we fur-
thermore use the relationship = 03 d2 /30 ~c3 , [see Eq. (GAF:6.119)], we obtain

(1) 33 /2
=i (II.41)
40 1 2i/

This effective volume is of order 3 = (/2)3 , quite large considering that 100 nm,
whereas the size of the electron wave function is typically 0.1 nm.
I Steady state, polarizability, susceptibility, etc. 21

Figure II.3. Linear susceptibility of a medium consisting of identical two level atoms. Shown are the real
i
and imaginary parts of the function 12i/ (1) . The real part (red) is proportional to n 1 with n the
refractive index. The imaginary part (blue) is proportional to the absorption coefficient and absorption
cross-section.

5. Linear susceptibility of the medium

While the polarizability is a property of a single atom, we can also define the response
of a medium containing N/V atoms per unit volume. Defining the linear susceptibility by
P(+) = 0 (1) E(+) , with P(+) the induced dipole moment per unit volume (positive frequency
component), we get (1) = ((1) /0 ) N/V ,

N d2 1 N 63
(1) = i =i (II.42)
V 0 ~ /2 i V 1 2i/

Note that the susceptibility is dimensionless. The real and imaginary parts of (1) = 0 +i00 are
related to the refractive index n and absorption coefficient [m1 ] of the medium, respectively,
see Fig. II.3 and Sec. 2.4.4 in GAF. For |(1) |  1 we have n 1 + 0 /2 and k00 = 200 /.
For the absorption coefficient we find that the intensity of a light beam is attenuated by the
00
medium as I e z/c . If we compare this with the Lambert-Beer law, I e(N/V ) z , we
find the absorption cross-section,

32 1
= (II.43)
2 1 + (2/)2

Attenuation occurs as if the medium consists of a collection of black disks of area each. On
resonance the size of this disk for a single atom is 32 /2, again much larger than the size of
the atom. The atom is a very sensitive antenna for resonant light.

III. INTERACTION OF ATOMS WITH QUANTIZED FIELDS

A. Jaynes-Cummings model (JCM)

The Jaynes-Cummings model (see GK, Sec. 4.5) describes the interaction of a two-level sys-
tem with a single mode of the radiation field. The JCM functions as a real work horse of cavity
QED, but also for quantum optics in general describing the coupling of any two-level system
to any single mode of a quantum harmonic oscillator. The Hamiltonian is given by
1
~0 3 + ~ a a + ~ a+ + a

HJCM = (III.1)
2
Here the RWA has already been assumed, neglecting the following terms: ~ a + + a .

The ground-state energy of the field mode, ~/2, has also been ignored as it is no more than a
22 III INTERACTION OF ATOMS WITH QUANTIZED FIELDS

Figure III.1. Dressed state energy levels. The crossing dashed lines show the energies of the bare states
|e, ni and |g, n + 1i, in the absence of coupling. The coupling (interaction) lifts the degeneracy in the
crossing point, producing an avoided crossing with a minimum splitting given by the resonant Rabi
frequency. Near the avoided crossing the dressed eigenstates |n, i are superpositions of the bare states.
Far from the avoided crossing the dressed eigenstates asymptotically approach the bare states, with only
a small admixture of the other bare state, and an energy (light shift).

shift of the zero on the energy axis. The coupling strength ~ has been called the vacuum Rabi
splitting, although single-photon Rabi frequency is more appropriate.
Comparing HJCM to the semiclassical atom-light interaction, we notice that:

The field energy is now included in the Hamiltonian; HJCM works in a Hilbert space of
combined states of the two-level system and the quantized field mode. These combined
states are called dressed states.

The interaction term not only changes the state of the atom but also the state of the field,
by adding or removing a photon. In the semiclassical case the field was a c-number (not
an operator) and was imposed from the outside.

In order to see any interaction effects at small photon numbers, we must


somehow in-
crease the single-photon Rabi frequency . Remembering that 1/ V where V is the
quantization volume, this leads naturally towards optical microcavities. The boundary
conditions imposed by the mirrors define modes confined to small volumes, leading to
values of that are routinely pushed into the MHz regime or even higher. In this way, a
single photon can saturate an atomic transition.

B. Dressed states and light shifts

If the coupling is set to zero, the eigenstates of the Hamiltonian are called bare states: simple
product states of the form |ei |ni or |gi |ni, where |ni denotes the number of photons. The
bare states form a pair of ladders of energy levels, with energies n~ for the |gi |ni states, and
~0 + n~ for the |ei |ni states.
If the field is near resonance with the atomic transition, = 0 + , we have pairs of states,
|gi |n + 1i and |ei |ni, that are nearly degenerate, differing only by ~. Furthermore, we can
see from Eq. (III.1) that HJCM couples only states within the same pair, |gi |n + 1i and |ei |ni.
Within the subspace of such a pair we again have a two-level system and we can easily solve
A First order coherence 23

the eigenvalues and eigenstates. We can immediately write down the energy eigenvalues,
   
1 1 1 1 p
E (n) = n + ~ ~n () = n + ~ ~ 2 + 42 (n + 1) (III.2)
2 2 2 2

This is shown graphically in Fig. III.1. The avoided crossing has a minimum splitting of
~n (0) = 2~ n + 1. Far from the avoided crossing the dressed eigenstates asymptotically
become equal to the bare states, but with an energy shift. These so-called light shifts, or AC
Stark shifts, are the basis foroptical dipole traps. The picture of
a shifted bare state is correct in
the limit ||  n (0) = 2 n + 1. We can then approximate 1 +  1 + /2, and obtain the
shift for the ground bare state,

2n (0) I
Eg (III.3)
4
For the ground state the light shift is negative for < 0 (red detuning), and positive for
> 0 (blue detuning) [12]. Thus, for < 0 the ground state atoms experience a force pulling
them into the laser beam towards higher laser intensity. If the laser intensity I is sufficiently
large, the atoms can be trapped in the intensity maximum. Conversely, for > 0 the ground
state atoms are expelled from the laser beam. Atom trapping can then be achieved by creating
a dark spot surrounded by light. Atom traps based on the light shift are called optical dipole
traps. Intuitively, one can see the potential as arising from the interaction of the laser-induced
atomic dipole with the laser electric field, d E. Far below resonance, < 0 the induced
dipole is in phase with the field, d E < 0, so that the dipoles are pulled into the field. Above
resonance the dipole is in antiphase and the opposite happens.
For the excited state the light shift has the opposite sign. This is usually not so useful for
trapping because excited states tend to have short lifetimes. However, the level shift can still
show up in spectroscopy, for example when driving a transition from the excited state to some
third level.
The light shift is not a strictly conservative potential. The finite admixture of excited state
causes some spontaneous emission, which heats the atoms in the trap. The associated photon
scattering rate is given by sc = s/2(1 + s) s/2 (see the steady state solution of the OBE).
In the regime discussed here we have s I/2 so that sc drops faster with detuning than
Eg . Given sufficient laser power, one can then increase the intensity and detuning propor-
tionally, keeping the light shift ( I/) constant, while suppressing the photon scattering rate
( I/2 1/). The trapping potential then becomes more and more conservative.

IV. COHERENCE FUNCTIONS AND INTERFEROMETRY

A. First order coherence

Suppose we send a light field E(t) into a Michelson interferometer consisting of a 50/50
beam splitter and two arms with a round trip length difference L = c . At the detector one
2
measures a signal i(t) E (+) (t) E (+) (t + ) . For a stationary light field,
D E hD Ei
i( ) E () (t)E (+) (t) Re E () (t)E (+) (t + ) (IV.1)

For small values of the fields at t and at t + are entirely correlated and the signal is fully
modulated from zero to its maximum, i( ) = imax (1cos )/2. The fringe visibility is then said
to be unity. In the limit of large , on the other hand, the fields are uncorrelated, the visibility
drops to zero, and i( ) imax /2.
One defines the normalized first order coherence function

()

(1) E (t)E (+) (t + ) a (0)a( )
g ( ) =
() = (IV.2)
E (t)E (+) (t) ha (0)a(0)i
24 IV COHERENCE FUNCTIONS AND INTERFEROMETRY

Note that the time dependence


of the field operators implies that we use the Heisenberg pic-
ture. This function g (1) ( ) drops from unity at = 0 to zero as  c . The characteristic
width c over which this happens is called the coherence time. We can think of it as the time
during which the memory of the phase of the field is preserved. The coherence time is also the
inverse of the spectral width. It is easy to check that the Fourier transform of the autocorrela-

2
tion function E () (t)E (+) (t + ) yields the power spectral density |E()| , where E() is the
Fourier transform of E(t). This relationship is also called the Wiener-Khintchine theorem.

B. Second order coherence

For many interesting effects in quantum optics we must look at higher order coherence func-
tions. In particular second order, or intensity-intensity correlation functions g (2) ( ) can display
many peculiar features that often cannot be understood classically. Examples are the violation
of Bells inequality, the Hong-Ou-Mandel effect and the Hanbury Brown-Twiss effect (although
the HBT effect can in fact be understood classically).
Roughly speaking, the first-order coherence function describes a single-photon property,
whereas second-order coherence functions describe two-photon properties, or correlations be-
tween pairs of photons. Second order coherence functions are usually measured by looking at
coincidence counts on photodetectors, essentially measuring conditional probabilities: if one
detector registers a photon, does that influence the probability that a second detector also reg-
isters one within some time or separation x? Rather than just measuring the intensity of
a source, one looks at its fluctuations, or more precisely one compares the fluctuations at two
different detectors.
For a source emitting in a single mode of the field we can define


a ( )a (0)a(0)a( )

(2) hI(t + )I(t)i
g ( ) = = 2
hI(t)i hI(t)i ha (0)a(0)i
The second order coherence gives the conditional probability (per unit time) that, given that a
photon has been detected at time t, another photon is detected at time t + .
For chaotic (thermal) light the second order coherence function is fully determined by the
first order coherence,
2
g (2) ( ) = 1 + g (1) ( ) (IV.3)

and in particular g (2) (0) = 2. This is called photon bunching: immediately after detecting a
photon the probability of detecting another one is increased by a factor two.
For coherent states such as emitted by a laser, Eq. (IV.3) does not hold, but instead g (2) ( ) =
1 for all . This means that the probability to detect a photon is independent of earlier photon
detections. The number of photons counted in a certain time interval then follows the Poisso-
nian statistical distribution.
For number states we have g (2) (0) < 1, a condition that is called photon anti-bunching,
which is classically impossible. An important experimental goal is to develop sources that
emit a single photon on demand. Such sources are benchmarked by showing that g (2) (0) =
0, indicating that you dont find more than one photon at one time. This also identifies a
single photon source as nonclassical. Single photon sources are a crucial ingredient in quantum
cryptography systems[13].
For all cases, the correlations disappear for long times, g (2) ( ) 1 as  c .
If we measure coincidences between two detectors 1 and 2, a generalized second order co-
herence function is accessible,
D E
hI (r , t + )I (r , t)i a1 ( )a2 (0)a2 (0)a1 ( )
(2) 1 1 2 2
g12 (r1 , r2 , ) = =D ED E
hI1 (r1 , t)i hI2 (r2 , t)i a (0)a (0) a (0)a (0)
1 1 2 2

This cross correlation plays a central role for example in detecting two-photon entanglement,
as in Einstein-Podolsky-Rosen (EPR) type experiments showing violation of Bells inequality.
C Quantifying entanglement 25

V. ENTANGLEMENT

A. Entanglement

We consider a quantum system that can be divided into two subsystems, for example two
particles or two photons. A simple definition for an entangled state is then a state that cannot
be factorized. For example, for two spin-1/2 particles, the following four states, called the Bell
states, are entangled:

= 1 (|i |i |i |i )

A B A B
2
= 1 (|i |i |i |i )

A B A B
2
None of the Bell states can be factorized, |i 6= (a |iA + a |iA ) (b |iB + b |iB ). Stated
this way, it seems that entanglement is a yes/no property. Actually, this is not the case; entan-
glement can be quantified and can vary between zero and a maximum value, see below.
Experimentally, the benchmark for entanglement is a violation of Bells inequality. If two
subsystems are subjected to measurement in spacelike separate locations, the measurement
outcomes are correlated. Bells inequalities set a maximum to the level of correlations that
are consistent with a classical, local realistic viewpoint. In quantum mechanics this limit can
be broken, yielding stronger correlations, leading to a violation of Bells inequality. The most
famous example is the experiment by Aspect and coworkers in 1981, using pairs of photons
that were entangled in their polarizations.

B. Reduced density matrix

If two subsystems are entangled and we wish to have a description for one subsystem alone,
the result will be a statistical mixture, to be described by a density matrix. Therefore, if we
consider subsystem B as an environment for subsystem A, entanglement with the environ-
ment leads to decoherence after discarding the environment. Formally, this is obtained by a
procedure of taking a partial trace.
Suppose for example that A and B are two spin-1/2 particles in a joint quantum state | i.
The density matrix of this pure state is = | i h |. A reduced density matrix for subsystem
A is obtained by tracing over the basis states of subsystem B,
A = TrB () = B B + B B




1
= (|A i hA | + |A i hA |)
2
The reduced density matrix A is identical to a completely incoherent mixture of the two states:
the state of A is |A i or |A i with equal probability. It is not a coherent superposition, it is a
statistical mixture. In matrix notation,
 
1/2 0
A =
0 1/2

The pure state thus reduces to an entirely mixed state A by ignoring the entanglement with
the environment subsystem B. We say that the interaction with the environment has led to
decoherence, as shown by the zero off-diagonal elements in A .

C. Quantifying entanglement

The entanglement in the pure state can be quantified by taking the von Neumann entropy
for the reduced A . The von Neumann entropy for a density matrix is given by
S() = Tr( log )
26 V ENTANGLEMENT

The entropy of entanglement is a measure of the entanglement of the two subsystems. The en-
tropy of entanglement of is the von Neumann entropy of the reduced density matrix:

E() = S(A ) = S(B ) = 1

In the context of quantum information we usually take the logarithm with base 2, so that the
maximum entropy of a qubit is 1. In this sense the four Bell states are maximally entangled: their
entropy of entanglement is indeed 1.

Unfortunately, calculating the von Neumann entropy for a given is easier said Pthan done.
We can easily do it if we
P can diagonalize , so if we can find a basis such that = P (i) |ii hi|.
In that case S() = i P (i) log P (i).

The von Neumann entropy can be seen as the analog of entropy in classical statistical me-
chanics. For a system with microstates labeled by i and a probability
P distribution over those
microstates P (i) the entropy is usually defined as S = kB P (i) ln P (i). If the number of
microstates is W and all microstates are equally probable, P (i) = 1/W , we recover the famous
formula

S = kB ln W

Final remark: the entropy of entanglement gives us a quantitative measure for entanglement
in a bipartite system. The more general case of quantifying entanglement between more than
two subsystems is still an open problem.
2 Circular polarization 27

Appendix A: Appendix

1. Real frequencies vs. angular frequencies

Its important to always know if a frequency denotes a real frequency (cycles per second,
unit: Hz) or an angular frequency (radians per second, rad/s). Since one cycle is 2 radians,
the numerical value of the angular frequency is larger by a factor 2. To minimize confusion
we make it a habit to denote a real frequency by the symbol f or , whereas we denote an
angular frequency as or . We mostly prefer angular frequencies because the formulas look
cleaner: cos t vs. cos 2f t. If an angular frequency must be given a value, we do it as follows:
/2 = 1 MHz, or: = 2 1 MHz, or = 2 106 s-1 . For real frequencies we simply state
= 1 MHz. If we say a signal has a frequency of 1 MHz, we usually mean its real frequency f .

2. Circular polarization

Consider a plane wave propagating in the +z direction, with wave vector k = kez . Circular
polarization vectors can be defined as follows:

   
1 1 1 1 1 1
sL = (ex + i ey ) = and sR = (ex i ey ) = (A.1)
2 2 i 2 2 i

(note that different conventions exist where these vectors are multiplied by an overall phase
factor)
To see why this is represents circular polarization, write the (real-valued) electric field as:

    
1 E0 1 1
E(r, t) = E0 sL ei(kzt+) + c.c. = ei(kzt+) + ei(kzt+) (A.2)
2 2 2 i i
 
E0 cos (kz t + )
= (A.3)
2 sin (kz t + )

This shows that the Ex and Ey components oscillate with a /2 phase difference. If we look at a
fixed position (e.g. set kz + = 0), the vector (Ex , Ey ) (cos t, sin t) rotates anti-clockwise in
the xy plane. Thus, an observer looking into the direction from where the sL -polarized light is
coming sees an anti-clockwise rotation. This is called left-handed polarization, in the convention
of Born and Wolf. If we plot the E vector at fixed time (snapshot), we get a left-handed spiral.
Similarly, sR is called right-handed polarization.
Generalization: if the plane wave propagates in an arbitrary direction, define the unit vector
k = k/k, and choose two linear, real-valued, orthonormal polarizations s1 , s2 that together
with k form a right-handed triplet, s1 s2 = k, etc. The circular polarization vectors are then:

1 1
sL = (s1 + i s2 ) and sR = (s1 i s2 ) (A.4)
2 2

a. Spin and helicity

For any polarization vector s we can express the sense of rotation (at fixed r) in a real-valued
spin vector, as follows

= i s s (A.5)

For the linear polarizations s1 , s2 this obviously evaluates to zero. For the circular polarizations
sL , sR on the other hand this vector is equal to k and k, respectively. In other words, k = 1
28 A APPENDIX

for sL and k = 1 for sR . Thus left-handed polarization corresponds to positive helicity, and
vice versa.
When coupling to atomic transitions we usually choose a quantization axis. Now what mat-
ters is the component of along the quantization axis. Suppose that the quantization axis is
+z. For sL light propagating in the +z direction, i.e. k = z, we get a spin = z. This we
normally call + polarization, because it couples to m = 1 electric dipole transitions. If light
with the same handedness propagates in the opposite direction, z, we get a spin = z, so
now we must call it polarization. Thus, left handed polarization (positive helicity) can be
either + or , depending on the direction of propagation.
Reflection from a mirror at normal incidence flips the handedness, but preserves the spin,
so sL becomes sR but + stays + . The common description of MOT beams in terms of coun-
terpropagating + / beam pairs is confusing because it assumes a fixed quantization axis,
whilst in the center of the MOT the magnetic field direction is flipped. A more consistent defi-
nition would be: left-handed light sL for directions where the MOT field points outward, and
vice versa. If the MOT field points outward in the z direction and inward in the x, y directions,
the z beam pair must be sL polarized, the x and y beam pairs must be sR polarized.

3. Operators as matrices

If M is a linear operator in Hilbert space, it is fully determined by all complex numbers


Mij = hi|M |ji, where |ii, |ji are basis vectors from an orthonormal basis. P The quantities
hi|M
P |ji are called
P 0 matrix elements. It is easy to see why if we write |i = i i |ii and M |i =
i i M |ii = i i |ii.PTaking the Hermite product with hj| and using hj|ii = ji (orthonor-
mality) we find j0 = i Mji i . This is exactly a matrix-vector product where the i are the
elements of a (column) vector and Mji is the matrix element at row j and column i. Using the
matrix notation a hermitian matrix is characterized by Hij = (H )ij = Hji
, i.e. the matrix is
equal to P the complex conjugate of its own transpose. Note that an operator can also be written
as M = i,j Mij |ii hj| (double summation); |ii hj| is just a matrix with a 1 on position i, j and
0 everywhere else.

4. Projection operators

For any normalized state |iiP we define its projection operator, or projector, as Pi = |ii hi|.
When acted on a state |i = an |ni the projector picks out the component along |ii. If |ii is
one of the basis vectors out of an orthonormal basis {|ni}, we have
X
Pi |i = an |ii hi|ni = ai |ii (A.6)
n

From the definition one sees immediately that a projection operator is idempotent: Pi2 = Pi .
Furthermore, if {|ni} is an orthonormal basis, then
X X
Pn = |ni hn| = I (A.7)
n n

is the identity operator. Equation (A.7) above is also called closure relation.
For example, if P0 acts on a qubit state vector:

P0 (a0 |0i + a1 |1i) = a0 |0i


     
1 1  1 0
or, using matrix notation, with |0i = and P0 = 1 0 = :
0 0 0 0
    
1 0 a0 a0
P0 (a0 |0i + a1 |1i) = =
0 0 a1 0
6 Interaction picture 29

5. Heisenberg picture

In quantum optics we are frequently interested in calculating time-dependent expectation


values of electromagnetic field variables. In many cases these are more easily calculated in the
Heisenberg picture, which is the opposite of the Schrdinger picture. In the latter the operators
are time independent and only the wavefunction changes in time. In the Heisenberg picture it
is the opposite: the wavefunctions are independent of time, but the operators are not. Denoting
quantities in the two pictures by subscripts H and S, we have
H = eiHt/~ S (t) = S (0) (A.8)
where H and H are time independent (direct check for H by differentiation).
Operators are related as follows,
AH (t) = eiHt/~ AS eiHt/~ (A.9)
Note that any operator that commutes with H remains unchanged. This is true in particular
for H itself, HH = HS . By differentiation one finds directly
dAH (t) ih i
= AH (t), H (A.10)
dt ~
As an example, taking H = ~ aS aS = ~ aH (t) aH (t) we find
daH (t)
= i aH (t) so that aH (t) = eit aH (0) = eit aS (A.11)
dt

6. Interaction picture

Lets consider the situation that the Hamiltonian consists of a constant part, H0 plus an in-
teraction VS (t) driving transitions between eigenstates of H0 . We are not interested in seeing
the time evolution due to H0 since it just consists of rapidly rotating phase factors:

eiH0 t/~
X X
|(0)i = n |ni n eiEn t/~ |ni
We can transform these phase factors away with a unitary basis transformation:
|I (t)i = eiH0 t/~ |S (t)i |S (t)i = eiH0 t/~ |I (t)i (A.12)
We have labeled the untransformed state vector by S (for Schrdinger picture) and the trans-
formed state vector by I (for interaction picture). We now take the time derivative:
d i d
|I (t)i = H0 |I (t)i + eiH0 t/~ |S (t)i (A.13)
dt ~ dt
i i
= H0 |I (t)i eiH0 t/~ (H0 + VS (t)) eiH0 t/~ |I (t)i (A.14)
~ ~
i
= eiH0 t/~ VS (t) eiH0 t/~ |I (t)i (A.15)
~
We then see that |I (t)i obeys a Schrdinger equation with a transformed Hamiltonian:
HI (t) = VI (t) = eiH0 t/~ VS (t) eiH0 t/~ (A.16)
Thus, in this transformed basis the only remaining time evolution is due to the interaction term.
The basis in which this happens is called the interaction picture. The matrix elements of VI (t)
are related to those of VS (t) as
hm |VI (t)| ni = ei(En Em )t/~ hm |VS (t)| ni (A.17)
One should keep in mind that all other operators in the interaction picture must also be trans-
formed, for example for an expectation value:
D E D E D E
S (t) AS S (t) = I (t) eiH0 t/~ AS eiH0 t/~ I (t) I (t) AI (t) I (t) (A.18)

30 A APPENDIX

7. Special (square) matrices/operators

All of the below are square matrices:

Unitary: U U = U U = I
unimodular eigenvalues (ei ).
a unitary matrix is normal.

Normal: N N = N N .
spectral decomposition: an operator is normal if and only if it is diagonalizable.
a normal  matrix is Hermitian if and only if it has real eigenvalues.
Since a, a 6= 0, the creation and annihilation operators are not normal.


Hermitian: Subclass of normal operators; means H = H


real eigenvalues
exp(iH) is unitary.
Hermitian operators are normal.

Positive: Subclass of Hermitian operators; means hv| P |vi 0 for any |vi. (if strictly > 0, then
positive definite)
non-negative (real) eigenvalues
positive operators are Hermitian.
Examples of positive operators are kinetic energy p2 , photon number n = a a, and the
density matrix .

Figure A.1. The identity operator is unitary as well as positive (and Hermitian and normal). The Pauli
operators are Hermitian as well as unitary. The photon creation and annihilation operators are not even
normal.
Index

B
Bloch sphere, 7
Bloch vector, 6

C
Closure relation, 28
coherence time, 24

D
Decoherence, 25
Density matrix, 15
Doppler broadening, 19
dressed states, 22

E
Entangled state, 15, 16

F
first order coherence, 23

G
Group, 5

H
Heisenberg picture, 29

I
Identity operator, 28
Interaction picture, 10, 29

J
Jones matrix, 5
Jones vector, 5

L
Lie group, 5
light shift, 23
Lorentzian, 19

M
Matrix elements, 28

N
natural linewidth, 19
Normal ordering, 9

O
optical dipole trap, 23

P
Pauli matrices, 5
Poincar sphere, 6
Polarization, 5
Poynting vector, 4
Projection operator, 28
Projector, 16, 28
Pure state, 16

31
32 INDEX

Q
quantum Zeno effect, 11

R
Rabi frequency, 14
Rabi frequency, generalized, 15
rotating frame, 18
Rotating Wave Approximation, 11, 13

S
saturation, 19
saturation intensity, 19
saturation parameter, 19
Second order coherence, 24
Statistical mixture, 16
Stokes parameters, 6
SU(2), 5

T
Transition dipole, 13

V
visibility, 23

W
wave packets, 10
Wiener-Khintchine theorem, 24

[1] R. Loudon, The quantum theory of light (Oxford Univ. Press, 3rd ed., 2000); M. Fox, Quantum
Optics (Oxford Univ. Press, 2006); L. Mandel and E. Wolf, Optical Coherence and Quantum Optics
(Cambridge Univ. Press, 1995); M.O. Scully and M.S. Zubairy, Quantum Optics (Cambridge Univ.
Press, 1997), Grynberg, Aspect, and Fabre (GAF), Introduction to Quantum Optics: From the Semi-
classical Approach to Quantized Light (Cambridge Univ. Press, 2010).
[2] C. K. Hong, Z. Y. Ou, and L. Mandel, Measurement of subpicosecond time intervals between two photons
by interference, Phys. Rev. Lett. 59, 20442046 (1987).
[3] R. Lopes et al., Atomic Hong-Ou-Mandel experiment, Nature 520, 66 (2015); R. Islam, et al., Measuring
entanglement entropy through the interference of quantum many-body twins, Nature 528, 77 (2015).
[4] Careful: in some textbooks the intensity is defined as the square of the amplitude, without 0 c; the
quantity in Eq. (I.4) is then called the irradiance.
[5] An important example of a component that is not lossless, does not preserve the norm of the Jones
vector, is a polarizer. The Jones matrix of a polarizer is not unitary, so it is not an element of SU(2). It
is an element of a larger group called SL(2,C).
[6] For example Grynberg, Aspect & Fabre include an extra factor 1/2.
[7] Similarly, the Stokes parameters are conventionally defined in a different basis: (S1 , S2 , S3 )
(s3 , s1 , s2 ).
[8] J.-L. Basdevant, J. Dalibard, Quantum Mechanics, Springer (2002).
[9] Wayne M. Itano, D. J. Heinzen, J. J. Bollinger, D. J. Wineland, Quantum Zeno effect, Phys. Rev. A, 41,
2295 (1990).
[10] Note that GK use the symbol V for the Rabi frequency. This is rather unusual, so we do not adopt
this notation.
[11] For example, focusing a 1 W laser to a 10 m spot size, and taking ea0 as a typical dipole matrix
element, 21 /402 109 .
[12] Careful: Our definition of has an opposite sign compared to GK who define red detuning as
positive (GK define = 0 ).
[13] See e.g. GAF, Complement 5E.

Das könnte Ihnen auch gefallen