UNIVERSITY OF CALGARY

Double Electromagnetically Induced Transparency

With Application to Nonlinear Optics at Low Light Levels

by

Andrew John MacRae

A THESIS
SUBMITTED TO THE FACULTY OF GRADUATE STUDIES
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF MASTER OF SCIENCE

DEPARTMENT OF PHYSICS AND ASTRONOMY

CALGARY, ALBERTA
August, 2008

c Andrew John MacRae 2008

UNIVERSITY OF CALGARY
FACULTY OF GRADUATE STUDIES
The undersigned certify that they have read, and recommend to the Faculty of Graduate
Studies for acceptance, a thesis entitled Double Electromagnetically Induced Transparency submitted by Andrew John MacRae in partial fulllment of the requirements
for the degree of MASTER OF SCIENCE.

Chairman, Dr. Alexander Lvovsky

Department of Physics and Astronomy

Chairman, Dr. Wolfgang Tittel

Department of Physics and Astronomy

Chairman, Dr. Karl-Peter Marzlin

Department of Physics, St. Francis
Xavier University

Chairman, Dr. Sergey Moiseev

Kazan Physical-Technical Institute,
Russian Academy of Sciences

Date
ii

Abstract
Electromagnetically Induced Transparency (EIT) has become an invaluable resource for
numerous applications in quantum information theory, fundamental physics, nonlinear
optics and precision metrology. This thesis is concerned with an extension to EIT: doubleEIT, in which two separate optical elds experience EIT due to a coherence between a
common ground state. Double EIT has appeared as the basis for a number of theoretical
proposals but has yet to be demonstrated and analyzed experimentally.
In this thesis, a double-EIT system is experimentally implemented and analyzed,
using a hot vapor of rubidium-87 atoms. The interplay between the two signal elds is
studied and properties are found to emerge which may prove useful for application. An
application of double EIT which in principle allows large nonlinear interactions between
light pulses as low as several photons per atomic cross section is discussed. Progress
towards a physical demonstration of this eect is reported and prospects for further work
in this direction are then described.

iii

Acknowledgments
One thing that Ive learned is that everything requires more time and eort than originally
planned, and the work described in this thesis is no exception. I can say without hesitation
that without the people I was surrounded with over the past two years, I would most
likely be burned out and bitter. Instead, I can say that life is good (and I can mean it.)
First and foremost, I owe much of this last statement to Sheralyn, who from the time
I signed on with the group, became my ance, and then became my beautiful wife. I
dedicate this you.
Secondly, I would like to thank my supervisor Alex. When I rst applied to do
graduate work here, my academic history was - to say the least, less than ideal. Despite
this, and for reasons I may never understand, he decided to give it a shot and supervise
me. Alex, thanks for lending your wisdom and teaching me how to do experimental
physics. I hope youre not lled with regret and despair for your decision!
Next, I need to point out that nearly all of the work described in the following thesis
was done together with my coworker Geo Campbell, who always brought new ideas to
the table. As well, Id like to thank you Keanu, for your acting, and for your inspiration.
I would be a real jerk if I tried to deny that I learned everything I know about
quantum optics from the quantum information technology group, here at the University
of Calgary. Therefore, I want to thank everyone who I worked with over the past few
years: Eden, J
urgen, Geo, Dmitry, An-ning, Dallas, Sun-Hyun, Chris, Peter, Conner,
Anna, (et al.)
Finally, to my family: Grandpa, Grandma, Mom, Dad, Alex, Catherine, and Heather,
Aunts, Uncles, and everyone at the HOH: thanks for taking me in and keeping me well
fed everytime I return to Victoria, and thanks for the unconditional support that only
family can give.
iv

Table of Contents
Approval Page . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . .
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2
Underlying Physical Principles . . . . . . . . . . . . . . . . . .
2.1 Electromagnetically Induced Transparency . . . . . . . . . . .
2.1.1 Basic Formalism . . . . . . . . . . . . . . . . . . . . .
2.1.2 Absorption and Dispersion Proles . . . . . . . . . . .
2.1.3 Slow Light . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.4 The Storage of Light . . . . . . . . . . . . . . . . . . .
2.1.5 Theory vs. Experiment . . . . . . . . . . . . . . . . . .
2.2 Rubidium-87: The Physical System . . . . . . . . . . . . . . .
2.2.1 Atomic Structure . . . . . . . . . . . . . . . . . . . . .
2.2.2 Interaction with a Magnetic Field: Zeeman Splitting .
2.2.3 The AC-Stark Shift . . . . . . . . . . . . . . . . . . . .
3
Experimental Methods . . . . . . . . . . . . . . . . . . . . . .
3.1 Stable Diode Lasers . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Laser Phase Lock Loop . . . . . . . . . . . . . . . . . . . . . .
3.3 Scanning and Pulsing the Light: Acousto-Optical Modulators
3.4 Housing the Atoms . . . . . . . . . . . . . . . . . . . . . . . .
3.5 Detecting the Light . . . . . . . . . . . . . . . . . . . . . . . .
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
Double Electromagnetically Induced Transparency . . . . . . .
4.1 Experimental Results . . . . . . . . . . . . . . . . . . . . . . .
4.1.1 Experimental Setup . . . . . . . . . . . . . . . . . . . .
4.1.2 CW DEIT . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.3 Simultaneous Slowdown of Co-Propagating Pulses . . .
4.1.4 Simultaneous Storage . . . . . . . . . . . . . . . . . . .
4.2 A Theoretical Model . . . . . . . . . . . . . . . . . . . . . . .
4.3 Further Prospects . . . . . . . . . . . . . . . . . . . . . . . . .
5
Giant Optical Nonlinearities With DEIT . . . . . . . . . . . .
5.1 EIT-Based Nonlinearities . . . . . . . . . . . . . . . . . . . . .
5.2 Physical Implementation in 87 Rb . . . . . . . . . . . . . . . .
5.3 Experimental Observations and Progress . . . . . . . . . . . .
5.4 Prospects for Experiment . . . . . . . . . . . . . . . . . . . . .
6
Conclusion and Outlook . . . . . . . . . . . . . . . . . . . . .
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

ii
iii
iv
v
vi
1
3
3
3
5
9
10
12
13
15
16
21
22
22
24
28
29
31
33
34
35
35
37
40
43
45
49
50
50
52
54
58
60
62

List of Figures
2.1
2.2
2.3
2.4

2.8

Atomic level scheme for EIT . . . . . . . . . . . . . . . . . . . . . . . . .

Absorption and dispersion proles associated with EIT. . . . . . . . . . .
Experimental plot displying the slow light eect . . . . . . . . . . . . . .
EIT absorption in the frequency domain - pulse distortion occurs as a
result of asymmetric absorption. . . . . . . . . . . . . . . . . . . . . . . .
Storage of light using EIT. . . . . . . . . . . . . . . . . . . . . . . . . . .
Experimental EIT dispersion and absorption proles. . . . . . . . . . . .
Overview of the atomic system used in the following experiments. Experiments were performed on the D1 transition at 794.979 nm. . . . . . . . .
Splitting of the degenerate Zeeman sublevels with applied eld. . . . . .

17
20

3.1
3.2
3.3
3.4
3.5
3.6

Diode lasers used in the experiments. . . . . .

Block diagram of the optical phase-locked loop
Beat frequency between locked lasers . . . . .
AOM driver block diagram. . . . . . . . . . .
Delay generator designed for the experiments .
Phase detection scheme for the experiment . .

23
25
27
29
30
32

2.5
2.6
2.7

. . .
used
. . .
. . .
. . .
. . .

. . . . . . . . . . .
in the experiments
. . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .

.
.
.
.
.
.

Atomic scheme used in experiment. Observation of simultaneous EIT

transparency dips. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Experimental setup used to realize double EIT. . . . . . . . . . . . . . .
4.3 Zeeman EIT for various transitions. . . . . . . . . . . . . . . . . . . . . .
4.4 Cross-talk between EIT signal elds. . . . . . . . . . . . . . . . . . . . .
4.5 Simultaneous slow light with no atomic preparation. . . . . . . . . . . . .
4.6 Pulse sequence used for atomic preparation. . . . . . . . . . . . . . . . .
4.7 Simultaneous group velocity of the pulses as a function of pump power.
Matched Pulses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.8 Simultaneous storage and retrieval of two signal pulses using a single pump
eld. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.9 Single EIT storage on the hyperne eld . . . . . . . . . . . . . . . . . .
4.10 The theoretical tripod scheme analyzed in the text. . . . . . . . . . . . .
4.11 Theoretical prediction and experimental observation of optical pumping
eect. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5
8
9
11
12
14

4.1

5.1
5.2
5.3
5.4

N-type atomic level scheme proposed by Schmidt and Imamoglu .

Atomic system used to explore large XPM. . . . . . . . . . . . . .
Splitting of EIT lines with applied magnetic eld . . . . . . . . .
Deviation of 2-photon resonance of the Stark eld EIT as a result
nonlinear Zeeman shift. . . . . . . . . . . . . . . . . . . . . . . . .

vi

. . . .
. . . .
. . . .
of the
. . . .

36
37
38
39
41
42
43
44
45
46
48
51
53
55
56

Chapter 1
Introduction
In recent years, advances in the theoretical understanding of quantum optics have lead to
the prospect of using quantum properties of light for the construction of quantum computers [1], the encryption of sensitive information [2], and ultra-sensitive measurements
of fundamental parameters [3]. In parallel with the theoretical progress, new experimental techniques and technologies have made the implementation of these ideas feasible.
Among these advances is the Electromagnetically Induced Transparency (EIT) eect [4].
EIT allows for the ultra-slow propagation of a light pulse with minimal loss [5], and the
temporary storage of classical [6] or non-classical light [7] for use as a quantum memory
[8]. As this progress has continued in classical and quantum optical technology, several
proposals have sought to further exploit the EIT eect to control the relative properties
of separate light pulses using multiple EIT systems [9, 10]. In particular, double EIT,
which involves the EIT eect on two separate signal elds, has been proposed to realize
large optical nonlinearities between single photons [11], all optical light buering [12],
and quantum logic gates for use in a quantum computer [13].
One of the benets of using light as a carrier of quantum information is its ability
to travel (as quickly as nature will allow) with minimal unwanted interaction with the
environment. However, given the robustness of light against cross-talk with outside
sources and its tendency to propagate, it is dicult to manipulate information carried by
this light solely with another light channel. In order to overcome this challenge, nonlinear
interactions requiring light levels as low as several photons per pulse have been proposed
in which an ensemble of atoms mediate the eect of one light pulse on another. In the
following chapters, a scheme is explored which shows promise towards the realization of
1

2
such an eect.
Chapter 2 outlines the theoretical concepts addressed in the remainder of the thesis.
The formalism of EIT is rst briey presented and its use in creating slow and stored
light is described. A description of the atomic system used to perform the experiments is
then given and the relevant features of the interaction of the atomic system with external
elds are outlined.
In chapter 3, the technical parameters of the experiment are described in detail. Since
the majority of the work of an experimental project lies in developing techniques and
technology to measure physical parameters, this chapter outlines the bulk of the eort
to obtain the results described in the remaining portion of the thesis.
The main results of the thesis are presented in chapter 4. The observation of double
EIT is rst reported and a full study of the interplay between the elds is given. The slow
light dynamics are explored and with the use of preparatory pulses, the group velocity
of the slowed light is controlled. The simultaneous storage of separate light pulses is
described. A simple model of a four-level double EIT scheme is analyzed and shown to
display the behavior present in the experiment.
Finally, chapter 5 discusses the possible use of this system to exhibit a previously
proposed [10] EIT-based nonlinear scheme. Progress towards this goal is reported and
further prospects are explored.

Chapter 2
Underlying Physical Principles
The purpose of this chapter is to give a theoretical description of the physical eects
relevant to the work presented in this thesis. A full description of EIT itself could
ll a large chapter in a textbook and still be lacking in completeness, however such a
description certainly does not belong here. The purpose of this chapter is therefore not
to reinvent the wheel, but rather to provide a picture of the eects specic to this
thesis. For a full description, refer to [14].
This chapter begins with a description of the cornerstone of this work: electromagnetically induced transparency, including a comparison of what is normally presented
in textbooks, with what is observed in a typical experiment. Next a description of the
atomic system used is given, including its interaction with electromagnetic elds.

2.1 Electromagnetically Induced Transparency

The primary physical principle underlying the research described in this thesis is the
electromagnetically induced transparency (EIT) eect. In EIT, the optical response of
an atomic medium is modied by a coherence induced via applied light elds. The
physical basis of this eect is quantum interference between excitation pathways.
2.1.1 Basic Formalism
The EIT eect can be understood in terms of the phenomenon of a dark state, rst used to
describe coherent population trapping [18]. To address the concept of a dark state, consider and atom in a lambda-type scheme which is coupled to light elds with Rabi frequencies p and s (Fig. 2.1). Assuming the dipole and rotating wave approximations[19],
3

4
the full Hamiltonian for this system in the absence of decoherence is given by

H=
2

2(1 2 )

p
.

(2.1)

22

where i is the detuning from state |i and the basis has been transformed to the corotating frame via the unitary transformation U = |e e| + eip t |1 1| + eic t |2 2|,
with p and c being the laser frequencies of pump and coupling elds respectively. For
the case of two-photon resonance (1 = 2 ) the eigenstates can be shown to be:

|+ = sin sin |1 + cos |e + cos sin |2

(2.2)

| = sin cos |1 sin |e + cos cos |2

(2.3)

|d = sin |1 cos |2

(2.4)

where = tan1 (s /p ), and = tan1



2p + 2s / /2 are known as the mixing

angles. Of particular interest is the dark state |d ; a coherent superposition of the

ground states which contains no |e term, and therefore does not couple to the excited
state. In other words, if the atoms are prepared in this state, they will not interact with
the incident light eld.
The idea behind EIT is to eciently drive the atomic system into this dark state using
a perturbative signal eld for s and a strong coupling eld for p . Under these conditions, the majority of the population is in the signal ground state (11 1), cos 1, and
sin 0 so that |d  |1, i.e. the system population is driven into a dark state from
which the weak signal eld couples. In this case, the signal eld experiences transparency.
The quantum mechanical basis for this transparency is a destructive interference of the
probability excitation pathways.

Figure 2.1: Atomic level scheme used to demonstrate CPT and EIT.
2.1.2 Absorption and Dispersion Proles
The atomic response to an optical eld is described by the susceptibility . For the
present discussion, we are interested in the linear susceptibility. However a key feature of
EIT is that the vanishing linear susceptibility which eliminates absorption is accompanied
by a signicant higher order term (3) [14] - a fact that will be exploited when discussing
the large Kerr eect in chapter 5.
In order to obtain an expression for , the master equation for the density operator
for a laser-driven atomic system is rst solved in the steady-state ( = 0.) The resultant
expression for the density matrix is then related to the susceptibility. In Lindblad form,
the master equation is given by:

=
[H, ] + Ld [] .
t
ih

(2.5)

The rst term describes the coherent evolution of the atomic system under the driving
eld, and the second term Ld describes relaxation terms including spontaneous decay
and dephasing. In the Lindblad formalism, each relaxation term may described by the
super-operator

Lij 2ij ij ij ij ij ij

(2.6)

where ij |i j| is the atomic projection operator. Ld is then given as a sum of these
individual superoperators operators Lij . Spontaneous emission from |e to |1 and |e to
|2 with rates 1 and 2 respectively, is included as
ground states |1 and |2 at rate 12 is described by
via

e
L .
2 ee

1
L
2 1e

12
L11 ,
2

2
L .
2 2e

Dephasing between

and dephasing on |e is added

The density matrix is then obtained by solving the master equation in the

steady-state:

0=

1
1
[H, ] + (1 L1e + 2 L2e + 12 L22 + e Lee ) .
ih
2

(2.7)

In order to relate the susceptibility to the density matrix obtained by solving equation
(2.7), rst note that for a linear dielectric such as a dilute gas, we can write the atomic
polarization as:

P = 0 E

(2.8)

where 0 is the permittivity of free space, and E is the electric eld.

Furthermore, the atomic polarization is dened as the mean dipole moment per unit
volume so that, in the density matrix formalism;

P=


N

d = Tr d
V

(2.9)

where N is the number of relevant atoms, V is the eective volume,  = N/V is the
is the dipole operator. For a given transition from state |i to |j,
atomic density, and d
ij = ij |i j| where ij is the electronic dipole element
the dipole operator is given by d
for the |i |j transition. Using this in equation (2.9) and dividing out the Rabi
frequency corresponding to the optical eld (
hp E) gives:

 |ij |2
ij .
0 h

(2.10)

For an optical medium such as an atomic vapor, the transmission is described by the
amplitude transfer function [14]:

T (, z) = eikz()/2

(2.11)

where k = /2 is the wave-number of the transmitted light. To see this, note that all
dilute gases may be treated as linear dielectrics [15], for which P = 0 E = D 0 E,
where D = E = 0 (1 + ) E is the electric displacement and is the permittivity of the
medium. Equating the two expressions for D yields 1+ = /0 /00 = c2 /v 2 = n2 ,
where it was noted that /0 1 and that the index of refraction n c/v. Since ||  1,

n = 1 + 1 + /2, and so einkz eikz/2 , neglecting the phase factor eikz . From
equation (2.11), it is apparent that the imaginary portion of gives rise to absorption,
whereas the real component leads to dispersion.
Since spontaneous emission from the excited state |e to the meta-stable groundstates |1 and |2 overwhelms other decoherence processes, additional coherence decay
terms on |e may be neglected and the total decoherence decay rates can be dened as:
e1 = 1 + 2 + e 1 + 2 , and e2 = 1 + 2 + e + 12 1 + 2 + 12 .
Using the above coherence decay terms, inserting the o-diagonal density matrix from
equation (2.7) solved to rst order, and identifying terms which rotate at exp[ip t] gives
the rather cumbersome equation:


2
2
2
2
|1e |2  4 |c | 4 412 + i 8 e1 + 212 |c | + 12 e1
=

2
0 h

|c | + (e1 + 2i)(e1 + 2i)

(2.12)

where the one photon detuning has been dened as 1 and the two-photon detuning
as 1 2 (see equation (2.1).) A detailed derivation may be found in [16].

Figure 2.2: Imaginary and real components of the susceptibility in equation (2.10) corresponding to the absorption and dispersion proles respectively. The steep dispersion
leads to slow light (see next section) while the vanishing absorption allows the lossless
transmission of this light. When the atoms are in this dark state superposition, the
probability amplitude of the pump exciting an atom from |1 is equal and opposite to
that of the signal exciting from state |2for the lossless transmission of this light.
Figure 2.2 displays the real and imaginary components of susceptibility corresponding
to dispersion and absorption proles respectively in the absence of decoherence. The xaxis is the two-photon detuning , which illustrates the fact that it is the two-photon
resonace condition which is crucial for EIT. Note that with no decoherence, perfect
transparency is always attained for = 0. The presence of decoherence lessens the eect
of transparency which can be partially compensated for by driving the atoms harder with
a stronger pump eld. Note as well that the FWHM of the EIT signal is proportional to
the strength of the pump eld |p | [20]. This provides a simple method for manipulating
bandwidth for the EIT system, which is important for the applications discussed in the
following subsections.

Figure 2.3: Slow light eect on the hyperne channel. The square trace show the original
pulse obtained in the absence of the atomic gas, the distorted pulse is measured after
placing the atoms in the beam path. The distortion is a result of the narrowness of the
EIT window in this conguration ( 100 kHz.) The delay of the pulse corresponds to a
group velocity of 33 km/s and a pulse compression of 104 (300 m 3 cm.)
2.1.3 Slow Light
The steep dispersion curve associated with EIT (Fig. 2.2) has a profound impact for
pulsed light, which by nature is spread out in frequency. Specically, since the group
velocity of a pulse of light is dened as

vg

d
= c [n() + (dn/d)]1 ,
dk

(2.13)

the steep dispersion curve leads to a large dn/d in the denominator, leading to ultra-slow
light pulses. Group velocities on the order of 10m/s [5] have been obtained, a fact which
is worth a second thought - one can ride a bike faster than a pulse of light progagates in
an EIT medium!
Figure 2.3 displays the slow light observed in our experiment in which a 1 s pulse is

10
delayed as is passes through a 10 cm cell, corresponding to a group velocity of 33 km/s.
A often quoted character of slow light with EIT is that whereas the group velocity (vg
dn/d)  c, the group velocity dispersion d2 n/d 2 is vanishing. This means that in
principle, there is no distortion of the waveform as it propagates through the EIT medium.
A quick glance at Fig. 2.3 shows that this is not the case in the above experiment. The
reason for this lies in a fundamental tradeo o an EIT system: one would like for the
pulse to t spatially in the cell, requiring a large group velocity reduction and a short
pulse. However a shorter pulse means wider spectral bandwidth that must t into the
EIT window. A wider EIT window in turn results in a less pronounced group velocity
reduction, and so a happy medium must be found. In the case of the pulse shown above,
the large group velocity reduction occurred at the expense of the absorption of the side
bands (see Fig. 2.4.)

2.1.4 The Storage of Light

As a pulse of light propagates through an EIT medium, it becomes compressed by a
factor of c/vg . During this propagation, the interaction of the light with the atomic
ensemble forms a spin wave which propagates at velocity vg . The light-spin subspace
form a quasi-particle known as a dark-state polariton [21]. If the pump eld is switched
o adiabatically, the polariton becomes a pure spin wave (between the two EIT ground
states), which maintains the coherent properties of the light pulse, but does not propagate. If the eld is switched back on adiabatically, the dark-state polariton accelerates
once again, and the original light pulse is recovered from the atoms. In this way, light
may be stored in atoms and storage times of over 1 ms have been reported in an atomic
vapour [6] and up to 1 second in a solid [17], as well as the storage of quantum light
[7, 22, 23]. It should be noted however, that the retrieval eciency of a stored light pulse
is limited due to decoherence of the atoms as well as sideband absorption as described in

11

Figure 2.4: Transmission through a narrow EIT medium results in pulse distortion.
Inset: Fourier transform of the square pulses used in the experiments is a sinc function.
Transmission through the Lorentzian EIT prole with FWHM much less than the the sinc
sideband leads to their attenuation. The retrieved pulse is no longer a top hat function
in the time domain.

12

Figure 2.5: Typical light storage run in the experiment. Dotted curve shows the switching
sequence of the control eld. Original pulse has amplitude 1 and width 2 s.
the previous section. As a result, storage eciencies are typically no higher than 25%.
Figure 2.5 shows a typical storage run obtained in the lab.

2.1.5 Theory vs. Experiment

In the literature, it is common to see the EIT dispersion curves plotted as in gure 2.2
which are for a single atom in the absence of decoherence. In practice however, several
forms of decoherence are always present. Furthermore, in a hot atomic vapor, the EIT
eect occurres over an ensemble of atoms which are Doppler broadened. Additionally,
the condition of one-photon detuning = 0 is rarely met precisely. All of these eects
conspire to give absorption and dispersion proles which dier markedly from that of a
decoherence-free single atom.
With the presence of decoherence, the ground state coherence which causes EIT is no
longer pure, and total transparency is never obtained at two-photon resonance. By using
a suciently strong pump however, the transparence at resonance can be improved, but

13
at the the expense of a less steep dispersion curve. If decoherence rates become too large,
the dark state can not be induced at all, and the EIT eect will not occur [14].
In a hot atomic vapor, the random motion of the atoms causes a distribution of
Doppler shifts, so that each atom has a velocity-dependant resonant frequency with
respect to the incoming light. Since the probe and pump beams in a typical EIT setup are
co-propagating, they see the same velocity distribution in the atoms and therefore see the
same relative (two-photon) detuning. Although the two photon resonance is independent
of the Doppler broadening, the one photon detuning is not. With
= 0, noticeable
asymmetry is present in the EIT linewidth. The absorption prole is a convolution
of the prole of a single atom with a Maxwell distribution having full width at half


maximum F W HM =

8ln2kT /mc2 0 , where 0 is the resonant wavelength of the

atomic transition, T is the temperature of the atomic vapour, k = 1.38 1023 J/K
is the Boltzmann constant, and m is the atomic mass of the medium. This has the
eect of signicantly narrowing the EIT linewidth since the FWHM of the EIT line now
depends on the inverse Doppler linewidth WD rather than the inverse natural linewidth
[20]. Since WD / 100, the EIT linewidth is considerably narrower in a Doppler
Broadened system. Figure 2.6 displays the absorption and dispersion proles present in
the laboratory.

2.2 Rubidium-87: The Physical System

The experiments described in this work involve the interaction of light elds with a
dilute gas of

87

Rb atoms. Rubidium is an alkali metal having a single valence electron

orbiting a lled shell. Such hydrogenic atoms have the favorable property that they are
aected by a (spin-independent) central potential V (r) and as a result, may be described
in the well known hydrogen atom formalism [24]. In addition to simplicity of analysis,

14

Figure 2.6: Phase and transmission of signal eld under EIT is monitored while scanning
over 2-photon resonance.

15
87

Rb has transitions in the neighborhood of 800nm, a wavelength accessible by many

commercially available diode lasers. The particular atomic transitions addressed in this
thesis are sublevels of the 5S 5P transition which are due to elds internal to the
atom. These levels experience additional energy shifts due to the presence of static
and time-varying electromagnetic elds. The following subsection describes the precise
atomic level addressed by the light elds used in the experiments.

2.2.1 Atomic Structure

Due to the coupling of the total electronic orbital momentum L with the electronic spin
S (so called L S coupling), the spectral line corresponding to the 5S 5P transition
exhibits ne structure and is split into several lines. The number of distinct energies
due to the ne structure splitting is determined by the standard addition of angular
momentum [25]. Under the L S coupling Hamiltonian, the observable J = L + S
is conserved. The number lines in the ne structure is determined by the number of
possible eigenvalues of J satisfying:

|L S| J L + S.

(2.14)

For the S-shell, L = 0 and since S = 1/2 for the electron, there is only one possible
value for J, namely J = 1/2. For the P-shell (L = 1), J = 1/2 or J = 3/2 are each valid
and so the 5S 5P transition is split into a ne structure doublet: the D1 transition
(52 S1/2 52 P1/2 ), and the D2 transition (52 S1/2 52 P3/2 ). Since these transitions are
separated by over 6 THz (the D1 transition occurs at 794.979 nm and the D2 at 780.241
nm) they are often treated separately. The remainder of this chapter will deal with the
D1 transition.
Just as the interaction of the orbital momentum with the electronic spin angular
momentum leads to ne-structure within the lines, the interaction of the nuclear spin I

16
with the electronic spin yields an additional energy shifts leading to hyperne structure.
Just as before, total angular momentum F = J + I is used as in equation (2.14), giving
|J I| F J + I. For Rubidium, I = 3/2 and since, for both shells on the D1
transition. J = 1/2, we have the possibility to use F = 1 or F = 2. Note that for the
D2 transition, we can also have J = 3/2, yielding F = 0, F = 1, F = 2, or F = 3 for
the P -shell. This is one advantage of using the D1 transition in an experiment; there are
simply less levels to worry about!
To get a value for the hyperne splitting, one uses rst order perturbation theory
with the hydrogenic Hamiltonian experiencing a perturbation in the form of a magnetic
eld due to the dipole moment of the nucleus. Neglecting the quadrapole and octopole
moments [26], the rst order correction is given by H (1)  = 52 Lj I J = 52 Lj /2F 2
I 2 J 2 , where the substitution F 2 = |I + J|2 = I 2 + J 2 + 2I J was used. Noting that
I = 3/2 and J = 1/2, the splitting is:

hf =

52 Lj
4F (F + 1) 18
[F (F + 1) I(I + 1) J(J + 1)] =
52 Lj
2
4

(2.15)

where 52 Lj is the magnetic dipole constant for the particular state. Numerically,
52 S1/2 /h = 3.417GHz and 52 P1/2 /h = 408.3MHz [26]. Plugging in these values for each
allowable F , we nd that the S-shell of the D1 transition is split into two lines separated
by 6.834GHz and the P -shell is split into two lines 816.7MHz apart. The structure
described above is summarized in gure 2.7.

2.2.2 Interaction with a Magnetic Field: Zeeman Splitting

The previous subsection described the 5S 5P transition which contains ne structure
sub-levels, each of which, in turn, contain hyperne sublevels. To go one step further
down the rabbit hole[27], each of the hyperne levels F contains 2F + 1 magnetic

17

Figure 2.7: Overview of the atomic system used in the following experiments. Experiments were performed on the D1 transition at 794.979 nm.
sublevels labeled by their magnetic number mF . The degeneracy of the ne and hyperne sub-levels is lifted by magnetic elds dynamically generated within the atom. The
magnetic sublevels become non-degenerate in the presence of an external magnetic eld
B. In the experiments described in this thesis, the atoms are placed in a uniform, time
independent magnetic eld aligned with the beam path which is chosen to be the axis
of quantization: B(t) = B z. The eective Hamiltonian describing the interaction of the
atoms with this magnetic eld is:

HB =

B
B B
(gS S + gL L + gI I) B =
(gS Sz + gL Lz + gI Iz )
h

(2.16)

where B = h(1.3996)MHz/G is the Bohr magneton, and gS ,gJ , and gI are the spin,
orbital, and nuclear g-factors respectively. If the splitting due to the magnetic eld is
small due to the ne structure splitting ( 7 THz here), then J is a good quantum
number and we can write

18

HB =

B B
(gJ Jz + gI Iz )
h

(2.17)

where gJ is the Lande g-factor (gJ (52 S1/2 ) = 2.0023 2 and gJ (52 P1/2 ) = 0.6666 2/3.)
gI 0.0009 is the nuclear g-factor and may be neglected, as it accounts for a correction
on the order of 103 . The calculation of Zeeman splitting in the weak or strong eld
regimes may be treated simply using rst-order perturbation theory. Naturally, the
general calculation of energy splitting requires more eort than in the perturbative cases,
however it is this intermediate eld regime which is required for the experiments described
in chapter 5 and therefore, the calculation must be carried out.
The approach will be to diagonalize the Hamiltonian including the hyperne and
Zeeman interaction in order to determine the energy splitting. The full Hamiltonian is:

Hhf s + HB = 52 Lj I J +

gJ B B
gJ B B
Jz = 52 Lj (Ix Jx + Iy Jy + Iz Jz ) +
Jz (2.18)
h

where the additional oset due to the nuclear spin has been neglected, since gI /gJ  1.
The product of the J and I operators is understood to be the tensor product, so that
Jz 14 Jz , where 14 is the 4-dimensional identity matrix. Using the Pauli spin matrices
for J and their spin 3/2 analogs for I, and working in the (Jz , mJ , Iz , mI ) basis, we have
and = B gJ for brevity):
(with 52 Lj h

19

H=

3
4

B
2

3
+ 2 B
4

3
2

3
2

+ 2 B

4 2 B

4 + 2 B

2 B

3
2

3
2

3
+ 2 B
4
0

0
3
4

(2.19)
Since this matrix is block diagonal, its eigenvalues are straightforward:

B gJ
352 Lj h

B
4
2


1
= 52 LJ h
2 (252 Lj h
)2 + (B gJ B)2
4


1
= 52 Lj h
2 (252 Lj h
)2 2B 52 Lj h
gJ B + (B gJ B)2
4



1
= 52 Lj h
2 (252 Lj h
)2 2B 52 Lj h
gJ B + (B gJ B)2 .
4

E1,2 =
E3,4
E5,6
E7,8

Figure 2.8 displays the energy lines as a function of applied eld. Note that for large
elds, for which B 

52 L

B gJ
h

(i.e.  B), each eigenvalue in equation (2.20) reduces to

E (B gJ /2)B, or

Estrong

B gJ
BmJ
2

(2.20)

for mJ = 1/2. Thus for strong elds, the hyperne lines are split into two separate
lines corresponding to the value of mJ (see gure 2.8.) This result can also be obtained
by using rst order pertubation theory.

2 B

20

Figure 2.8: Splitting of the degenerate Zeeman sublevels with applied eld. Note that in
the weak eld approximation, the splitting is linear in mf and in the strong eld regime,
linear in mj .)
On the other hand, if B 

52 L

B gJ
h

whose shift is linear in B. Letting

, the two hyperne lines are split into eight lines,

3
4

E(F =2) , and 5

E(F =1) , setting gF = gJ /4
4

[26], the energy eigenvalues from 2.20 become:

Eweak =

EF =1 B gF mF B

for F = 1, mF {1, 0, 1}

EF =2 + B gF mF B

for F = 2, mF {2, 1, 0, 1, 2}

which is identical to the result obtained from rst order perturbation theory [26]. This
weak eld splitting is shown in gure 2.8 inset.

21
2.2.3 The AC-Stark Shift
The nal relevant physical eect which will be discussed in this chapter is the AC Stark
shift. Consider a two level atom coupled o resonantly to a light eld. The bare-state
Hamiltonian gives energy eigenvalues corresponding to the energy dierence between the
atomic levels. Turning on an o-resonant coupling eld between the levels introduces
transition terms to the Hamiltonian. Since the introduction of this eld has modied the
Hamiltonian, it is reasonable to expect that the energy eigenvalues have changed, in turn
changing the spacing between the two levels - an eect known as the AC-Stark shift.
A full analysis of this eect is given in [28], but a simple expression for the magnitude
of the energy shift E can be arrived at by considering a far o-resonant eld so that the
transition probability is small, and using second-order perturbation theory. Note that
rst order perturbation theory gives no correction to the energy level [29]. Recall that
the second order correction to the energy of a system with Hamiltonian experiencing a
perturbation H  is given by:

E (2) =


i=j

|i| H  |j|2
Ei Ej

(2.21)

where Ek terms correspond to the unperturbed eigen-energies. In the case of o resonant

coupling, the perturbative Hamiltonian is given by h
12 /2 |1 2| + c.c.. Denoting E2
, and inserting into (2.21) gives the AC-Stark shift:
E1 h

EStark

h
||2
.
=

(2.22)

Chapter 3
Experimental Methods
In this chapter, the primary tools required to perform the experiments are described.
Since a large part of the work performed during the past 18 months involved the design,
construction, and optimization of the techniques and devices necessary to accurately
measure the physical eects of interest, this thesis would not be complete without a
summary of the main aspects of the experimental implementation.

3.1 Stable Diode Lasers

A quantum optics experiment such as described in this thesis would not be possible
without a high quality light source. It should provide a single mode collimated beam, be
tunable over a wide range of frequencies, while at the same time have a narrow range of
uctuations from its central frequency. At the same time, since funding is unfortunately
never an inexhaustible resource, it should not be prohibitively expensive. In order to
meet each of these requirements, external-cavity diode lasers (ECDL) based on a design
described in [31] were constructed for use in this experiment.
The basic components of the ECDL are displayed in gure 3.1. The laser diode
is formed by a PN-junction semiconductor which emits light over its relatively wide
( 1 THz) gain prole whose center frequency at a given temperature depends on the
injection current. The surfaces of the diode form a Fabry-Perot cavity, which has a
comb of resonant frequencies giving a number of spectral modes, each having short term
linewidth on the order of tens of MHz. The purpose of the external cavity is to select a
single mode and further reduce the bandwidth by two orders of magnitude. The external

22

23

Figure 3.1: (a) Sketch of the laser diode system used in this experiment: External cavity
is formed between diode and rst order reection from grating. Output frequency is set
by holding temperature constant and modulating both the external cavity length and
the injection current.(b) Photograph of one of the ECDLs in the lab.
cavity consists of a reective, blazed diraction grating: the rst order couples back into
the laser diode forming a cavity, and the zeroth order is reected to an out-coupling
mirror. The light which leaves the laser is at this point a stable and narrow-band source
suitable for use in an experiment.
In order to tune the output frequency of the laser, one may change either the temperature of the diode, the injection current, or the length of the external cavity. Since a
change in temperature is a considerably slow process, the temperature which corresponds
to the neighborhood of the desired output frequency is determined ahead of time and
then held constant by active feedback using a Peltier thermoelectric cooler (TEC) and
a temperature sensor for continuous measurement. For ner corrections, the length of
the external cavity is adjusted by a piezoelectric transducer which is attached to the
grating mount. Using this technique, the output frequencies can be scanned over a range
of several GHz at a bandwidth of several kHz. If this scan is made to be too large, the
laser may hop modes and will have to be re-calibrated to obtain single mode operation.
Finally, for higher bandwidth corrections, the injection current to laser diode may be

24
modulated, allowing for a correction bandwidth of up to 1MHz, over a limited frequency
range.
Putting the system together is a simple task but the initial alignment can be a humbling experience. However, once this stage is complete, high quality operation is maintained with little further work. Self made diode lasers have a tendency to take on a
personality of their own and require a good deal of getting used to; so much so that our
lasers have acquired their own names: Hubert and Alvar.

3.2 Laser Phase Lock Loop

When a number of light elds coherently interact with an atomic ensemble, it is imperative that these light sources are phase locked with respect to each other. If each of
these elds originate from the same coherent source, then this condition is automatically
satised. However, if the required elds dier too greatly in frequency, they need to come
from a separate source. Even if each source is itself coherent, the relative phase between
the two elds may uctuate so rapidly as to render their coherent interaction impossible.
This problem may be circumvented by locking the two sources together in phase by using
a feedback loop. This section describes the Laser Phase-Lock Loop (PLL) used to lock
the ECDLs in this experiment. The PLL constructed in this experiment was based on
the design by J. Appel [16], an earlier student in the research group. Full details will be
described in a publication currently under preparation.
A slave laser operating at frequency 1 is locked to a master laser at 2 when their
relative phase changes at xed rate, specically, the beat frequency 2 1 . This is
especially important for experiments in EIT, where it is the relative (2-photon) detuning
which is the crucial parameter (see chapter 2.) In order to monitor this beat frequency,
the two signals are interfered on a beam splitter and observed with a photodiode. Letting

25

Figure 3.2: Block diagram for the operation of the Optical PLL. The optical phase is
measured as the AC-component of the interference between the lasers. A stable frequency
source provides the reference. The two signals are divided to a common frequency and
passed into an Analog Devices ADF 4107 PLL chip producing an error signal. The FET
and external cavity feedback signals are then separately outcoupled
E1 (t) = E1 ei1 t and E2 (t) = E2 ei2 t be the light elds in question, the signal observed at
the photodiode is proportional to |E1 (t) + E2 (t)|2 , or

I = I0 + I1 cos (t + )

(3.1)

where I0 = |E1 |2 + |E2 |2 and I1 = 2|E1 E2 |, and = arg[E1 E22 ]. In other words, is
measured directly as the AC component of photo-detector output. Since the hyperne
levels in

87

Rb are separated by 6.8GHz, a photodetector with at least 7GHz bandwidth

is required. Originally this was accomplished by using a New Focus 1577A diode, but
by placing an additional lter at the input of the PLL circuit, the Hamamatsu G-4176
proved to provide the same results with more than an order of magnitude reduction in
price.
The basic operation of the laser PLL is sketched in gure 3.2. The beat signal and
a stable reference are sent to a phase-frequency detector (PFD). The PFD operates as
follows: the frequency of the input signal is divided down by an integer factor N, while

26
the reference is divided by an integer R. N and R are chosen so that /N = ref /R.
In the case of the following experiments, the laser frequencies are separated in frequency
by the hyperne ground-state splitting of 6.831GHz. Since it is preferable to work with
much lower frequencies (say, several tens of MHz), N is chosen to be 96, so that the input
frequency to the PFD is 71.156 MHz. A stable frequency reference of 213.469 MHz is
then divided down by R = 3 before being fed to the PFD and compared to the input.
The setting (N, R) = (96, 3) is not unique ((384, 1) with reference frequency 17.35 MHz
was also used), but slightly better performance may be obtained by experimenting with
dierent values. The frequency divided signals are then locked in phase by comparing
them at a charge pump with a small but nite dead time: if the divided reference signal
switches high before the divided optical signal, a positive voltage results. Similarly, if the
divided optical signal switches before the reference, a negative voltage is produced. After
a brief programable dead time, the charge pump is reset and begins listening again. This
forms the error signal for the feedback loop: if the signals begin to change too quickly
or slowly, a correcting signal is applied to the slave ECDL. The error signal is split up
into low and high frequency components which are sent to the ECDL as piezoelectric
and injection current modulation respectively. When the relative optical phase occurs at
Nref /R, the error signal is zero, and the lasers are phase locked. Using this technique,
the relative phase uctuations between the two lasers can be reduced to below a few
tens of Hz, with this value being due to the limited resolution of the spectrum analyzer.
Without such a phase lock, EIT using two separate sources would not be possible. Figure
3.3 displays the beat frequency spectrum between the two lasers with the PLL applied.

27

Figure 3.3: With the use of an OPLL, the beat frequency between the optical beams
is xed to less than one part in 106 . Here the lasers each at frequencies of 377THz,
are set at a xed dierence of f0 = 6.75470GHz. The phase noise is eectively pushed
away from the carrier band. Note that the logarithmic scale used since on a linear scale,
the sideband noise is unidentiable as compared to the carrier band. Settings used are
resolution bandwidth = video bandwidth = 10 kHz

28

3.3 Scanning and Pulsing the Light: Acousto-Optical Modulators

While the injection current and piezoelectric-controlled grating allow the frequency of
the laser to be roughly set, they do not suce to ne tune and scan the frequency on
the kHz level. Furthermore one often requires short ( 1s) pulses with rise time of
100ns or less, which is not possible by using the laser controller alone (the transient
behavior present when switching on a diode laser is on the order of several seconds.) The
rst problem may be circumvented by tuning the reference frequency of the phase lock,
but this technique can not produce the required switching of the amplitude to create
pulses. In order to achieve the required frequency and amplitude modulation one can
use an acousto-optical modulator (AOM). An AOM is a crystal with a piezo-electric
modulator attached to one side. Applying a periodic signal to the piezo with frequency
sends traveling waves across the crystal creating a moving, periodic phase-grating.
Light at frequency 0 incident on the crystal is then Bragg-scattered via a photon-phonon
interaction creating spatially separated sidebands at frequencies = 0 . Typically,
only one of the spatially separated sidebands is selected with an iris and then out-coupled
to the experiment. By modulating the frequency or amplitude of the signal passed to
the AOM piezo, amplitude frequency or amplitude modulation is performed on the light.
Using this technique, steady frequency scans of anywhere from 500Hz to 100MHz are
possible, depending on the bandwidth of the particular AOM. By sending a pulsed signal
at the desired frequency to the AOM, a pulse of light will emerge from the device which
has a duration limited only by the time required for the compressional wave to traverse
the crystal: min = Lcrystal /vsound 100ns.
In order to create the signal to be sent to the AOM, a switchable high-frequency
driver was designed to operate the 80MHz and 200MHz center frequency AOMs used
in this experiment. The block diagram of this driver is shown in gure 3.4. A 10MHz

29

Figure 3.4: Block diagram of the electronic driver used for the AOMs in the lab. Since
stable 10MHz sources are abundant, but higher frequency are modulations are required
for the AOMs, this driver multiplies the 10MHz input by a programable integer factor,
the signal is then amplied to 30dBm and fed directly into the AOM. A high speed switch
allows the signal to be turned on for a brief period, allowing the creation of pulses.
stable reference input is multiplied by a programmable integer and send to a fast (4.5ns
rise time) switch. This allows a stable input to always be present while using a dierent
source to create the switching signal.
In order to precisely control the arrival time of the light pulses in the experiments,
the switching signals to the driver must be independently controllable. To this end, a
triggerable delay generator was constructed which when triggered, produces an output
pulse of programmable width after a programmable delay into both of its output channels.
This delay generator is summarized in gure 3.5.

3.4 Housing the Atoms

The experiments described in this thesis were performed in Rb vapor at a temperature
ranging from 35 C to 70 C contained in a 10cm long, 2.5cm diameter glass tube. The
primary advantage of using an atomic vapor over a collection of cold/trapped atoms is
the minimal amount of preparation needed to perform experiments on the atoms. This
simplicity comes at the cost of higher decoherence rates and Doppler broadened atomic
transitions as compared to a cold, trapped atom system such as a magneto-optical trap.

30

Figure 3.5: Using two pairs of monostable multivibrators, one triggered on the rising
edge and the other on the falling, a signal input pulse triggers two pulses of individually
programable width and delay. On the right, the signal at dierent points in the circuit
is displayed.

31
The vapor cell is housed in a large steel oven with a heater and temperature controller
keeping the system at a xed temperature. In order to avoid random Zeeman shifts of the
atoms energy levels due to external elds, several layers of metal shielding surround
the atomic cell in the oven [30]. Two large solenoids surround the oven as well allowing
for the application of a reasonably stable, uniform magnetic eld along the axis of laser
propagation.

3.5 Detecting the Light

It is safe to say that every experimental result presented in this thesis relied on the detection of light; a few words should then be said about the method used to accomplish
this task. Since macroscopic amounts of light were used (1 W of 795 nm light corresponds to a photon ux of 1012 s1 ), simple Si photodetectors could be used in place
of single photon detectors or photo-multiplier tubes. In such a detector, the photovoltaic
eect transforms the energy from an incident light beam into an electrical signal. This is
accomplished through a small energy gap between the valence and conduction bands of
the detector: when the incident light has enough energy to transfer an electron from the
valence to conduction band, the accumulated charge results in a ow of current. Since
there is always electronic noise present in an electronic circuit, there is a minimum light
level detectable by a photo-detector, below which, the current due to the incident light
is indecipherable from the current due to noise.
There are two ways of looking at a signal coming into a detector: one can plot the
amplitude of the signal as a function of time (time-domain detection), or one can plot
the amplitude of a given frequency range (frequency domain detection.) Figure 2.3 is
an example of a time domain measurement and Fig. 3.3 shows a frequency domain
measurement. Since the typical frequency of an optical eld is 1014 Hz which is orders

32

Figure 3.6: The phase detection system used in the experiment. The same frequency
source which drives the AOM is compared to the optical beat frequency between the modulated and unmodulated light signals. Each signal is compared at a AD8302 RF/Phase
detector and the dc-output which corresponds to the relative phase between the signals
is observed on a scope with a resolution of 17.5 mrad/mV.
of magnitude faster than possible with electronics, the spectrum of a light signal cannot
be measured directly. To circumvent this problem, one can observe the beat frequency
spectrum of the optical signal of interest mixed with a local oscillator; a technique known
as heterodyne detection.
In addition to the amplitude of an optical signal, the phase is also of interest. For
example, one may be interested in the phase accumulated while passing through a nonlinear medium as compared to the case where this medium is not present. In chapter
ve, a experiment which requires the measurement of the phase of short (1 s) pulse
of light is described. A signicant challenge to overcome was the determination of the
optical phase given only a brief pulse. In order to overcome this obstacle, a fast electronic
phase detector was designed and constructed capable of detecting the phase dierence
between two signals oscillating at up to several GHz. The schematic for the phase detection system developed is shown in gure 3.6. The heart of the system is a Analog
Devices AD8302 phase comparator chip which provides an output voltage proportional

33
to the relative phase of its two inputs, with phase accuracy of 1 /mV. Since the rise
time for a full swing of the output is 400 ns, it is fast enough to accurately detect the
phase of pulses with duration on the order of 1 s. The 200 MHz phase reference is sent
to the phase detector as well as an AOM driving the signal of interest. The signal is
then mixed on a fast photodetector with a local oscillator taken from the signal before
passage through the AOM. The resultant 200 MHz beat signal is then phase stable with
the reference which provides the input for the phase detector. The output channel then
gives a direct measure of the optical phase, modulo 2.

3.6 Summary
In this chapter, the primary components of the experiments in the next two chapters were
described in detail. Specically, two external cavity diode lasers were constructed. A laser
PLL was constructed and congured to lock these two lasers together at 6.831GHz, with
sub-Hz precision. AOM drivers operating at 80MHz and 200MHz were designed and
constructed in order to scan and switch the light signals independently in order to netune the arrival of simultaneously created pulses, a delay generator was designed. Finally
a phase detector capable of detecting the optical phase of pulses as short as 1 s was
designed, manufactured, and put into operation.

Chapter 4
Double Electromagnetically Induced Transparency
At the beginning of this project, when the idea of demonstrating a EIT based nonlinearity
for pulsed light was rst discussed, perhaps the biggest obstacle was the creation of a
double-EIT system - a task which had yet to be experimentally realized. Specically, it
was not known whether the coherence created in one lambda-system would detrimentally
aect that of the other. Furthermore, if we were able to obtain the simultaneous slowdown
of the co-propagating pulses, would we be able to manipulate the system in order to adjust
the group velocity of these pulses? Finally, could the pulses be matched so that they
may be of use in the future experiment which aims to create a large EIT-based nonlinear
optical eect? The rst task was therefore to attempt an experimental realization of a
double-EIT scheme and use it to produce matched pulses. This endeavor turned out to
be not only challenging, but interesting in its own right.
As further motivation for studying double EIT, note that while EIT itself has become
a cornerstone of many methods for controlling optical elds, several theoretical proposals
have suggested double EIT (DEIT) as a vehicle for extending the utility of EIT schemes
by creating transparency conditions for two signal elds simultaneously [10, 13, 32]. This
provides the possibility for coherent control [9, 19, 33] and nonlinear interaction between
weak optical elds [10, 32]. DEIT allows propagation of the two signal elds with minimal
loss, and increases the interaction time between pulses due to group velocity reduction,
making it a promising candidate for numerous applications in quantum computation and
communication.
In this chapter, a double EIT system is thoroughly investigated including the interdependency of the two signal elds in both continuous-wave (CW) and pulsed cases. In
34

35
the CW case the eect of optical pumping due to the second signal eld is investigated.
It is shown that the delay induced in the pulses can be adjusted and matched through
preparation of the atomic states. In addition to its use for quantum computational gates,
this technique could nd application in quantum communication protocols that use frequency multiplexed information channels, such as simultaneous quantum memory, simple
qubit operations and correction of time delays [12]. Finally, a theoretical model exploring
the double EIT system used for the experiments is investigated, and many of the key
experimental observations are reproduced. Many details of this chapter may be found in
[34].

4.1 Experimental Results

The double EIT system explored in this work is displayed in gure 4.1, operating on the
D1 transition of

87

Rb (see chapter 2.) The pump eld is right-circularly polarized, and

couples the F = 2 F  = 2 transition. The rst signal eld is of left circular polarization
and couples the same transition, except addresses a dierent Zeeman sublevel. For this
reason, we name this eld the Zeeman field. Similarly, the hyperfine field is of right
circular polarization and couples out of the second hyperne ground level F = 1
F  = 2. These particular transitions were chosen to form the tripod scheme due to their
applicability to a nonlinear optics scheme (see Chapter 5.) It should be noted that not
all combinations of atomic transitions give the same quality EIT. Figure 4.3 shows the
Zeeman EIT for various transitions.

4.1.1 Experimental Setup

The experimental setup is displayed in Fig. 4.2. The experiments were performed in a
10-cm atomic cell containing the 87 Rb gas and 10 torr Ne buer gas maintained at 45 C.
The elds were produced by two self-made external cavity diode lasers. As mentioned

36

Figure 4.1: Left: the double EIT scheme is of tripod conguration formed from the
Zeeman sublevels of the hyperne levels of the 87 Rb D1 transition. The + -polarized
pump couples F = 2 F  = 2, the + hyperne signal eld couples F = 1 F  = 2,
and the Zeeman signal eld is -polarized and couples F=2 F=2. Right: Double EIT
transparency peaks were observed by scanning the pump-eld detuning, simultaneously
scanning 2-photon detuning for each signal eld. Note: signal elds are separated in
frequency here by 6.863 GHz
in chapter 3, in order that the elds act coherently on a single set of atoms, the lasers
were phase locked at 6.834 GHz. One of the lasers provided the pump and Zeeman elds,
while the other generated the hyperne eld and the local oscillator for its detection.
The pump and each of the signal elds passed through AOMs allowing them to be
scanned in frequency, or switched independently. The beams were then spatially mode
matched and passed through a quarter-wave plate to convert from linear to circular
polarization. Typical laser beam power in these experiments amounted to 2 mW 6
mW for the pump, and 1 W 150 W for the Zeeman and hyperne elds. The beam
diameters in the cell were 750 m. After passage through the atomic medium, another
quarter-wave plate and a polarizing beam splitter (PBS) separated the elds into two
paths for detection; one with the Zeeman eld, and the other containing the hyperne
and pump elds.
The Zeeman eld was measured directly with a photodiode. However, the detection

37

Figure 4.2: Experimental setup used to realize double EIT.

of the hyperne eld proved to be more challenging - since the weak hyperne eld was
in the same spatial and polarization mode as the strong pump, it could not be spatially
separated and had be measured in the presence of the much stronger pump. To this
end heterodyne detection was employed, with a local oscillator diering in frequency by
200 MHz, taken from before the hyperne beam passes through the AOM. The resultant
beat frequency was observed with a spectrum analyzer in a time-resolved, zero frequency
span setting with a resolution of 200 ns.

4.1.2 CW DEIT
The rst step in exploring the double EIT system was to look for the simultaneous
appearance of the EIT transparency dip on each signal eld. This was accomplished by
scanning the frequency of the pump eld, eectively scanning the two-photon detuning
of the Zeeman and hyperne EIT systems simultaneously. Simultaneous transparency
peaks occurred as 2-photon resonance was met between each signal eld and the pump
as seen in gure 4.1b. It was noted that the transparency at resonance was signicantly

38

1
F=1>F=1
F=2>F=1
F=2>F=2

0.9
0.8

Transmission

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1000 800

600

400

200

200

400

600

800

1000

2Photon Detuning [kHz]

Figure 4.3: EIT in the Zeeman conguration for various transitions. Note the transition
F = 1 F  = 2 is not displayed here since the transparency was negligibly shallow.
greater for the hyperne EIT than for the Zeeman. This can be explained by the fact that
the Zeeman eld couples to an excited level which is not part of a Lambda system (F = 2,
mF = 2, see gure 4.1), leading to absorption for any atoms in the mF = 1 ground
state, whereas the hyperne eld does not. Additionally the decoherence rate between
the Zeeman ground state and pump is in general not equal to that of the hyperne ground
state and pump and can in principle be much higher.
Once double EIT was observed, this rst question to explore was the eect of one EIT
system on the other: could the tripod system be treated as two independent EIT systems
or is there some cross-talk between the two signals? To answer this, a single EIT line was
observed with the other eld completely blocked. The second eld, set to a relatively low

39

Figure 4.4: The eect of one signal eld on the other in the DEIT system. In (a),
the absorption spectrum of the hyperne eld, xed at 50 W, is measured for various
powers of the Zeeman EIT signal. In (b), the Zeeman signal is of constant intensity and
the hyperne power is varied.
power was then unblocked and its power was increased while monitoring the EIT prole
of the rst eld. The results of this experiment are displayed in gure 4.4. The hyperne
eld was set to 50W and the prole was observed for various Zeeman eld powers. It
became immediately apparent that even for weak Zeeman elds (< 10W) there was a
signicant eect on the hyperne EIT. Specically, the EIT contrast (dened as the ratio
of absorption at 2-photon resonance and the absorption outside the transparency dip)
was considerably enhanced. Increasing the Zeeman signal strength furthered the contrast
enhancement without noticeably aecting the transparency at two-photon resonance.
Finally when the Zeeman eld became much stronger than the hyperne eld (> 150W)
the two-photon resonant transparency began to decrease, due to additional decoherence
from the stronger Zeeman eld, acting as an additional pump eld [20]. The experiment
was then repeated for the Zeeman eld, with the hyperne eld slowly increasing in
intensity and a similar eect was noted (Fig. 4.4.)
The above eect can be explained by optical pumping from one ground state to the
other. From the atomic level scheme (Fig. 4.1), we see that when the hyperne eld is

40
absent, the atomic population will collect in the F = 1 ground state. Turning on the
hyperne eld will now serve to repopulate the F = 2 ground state, specically the m = 2
sublevel, increasing the eective atomic density of atoms experiencing EIT. Similarly,
with only the hyperne eld present, the |F = 2, m = 2 ground state will accumulate,
and turning on the Zeeman eld will have the eect of increasing the number of atoms
in the F = 1 ground state. Thus the relative contrasts can be adjusted by selecting
appropriate powers for each eld. The ability to manipulate the EIT lineshape proved
extremely useful when dealing with slow light as described in the next subsection.

4.1.3 Simultaneous Slowdown of Co-Propagating Pulses

Once DEIT had been observed in continuous wave (CW) regime, the next step was to
observe the slow light eect simultaneously on two pulses. The majority of the proposals
involving DEIT mentioned at the beginning of this chapter involve co-propagating pulses,
each traveling with identical group velocity. By sending a 1 s pulse from the Zeeman
and hyperne elds simultaneously through the EIT medium, each experienced a reduced
group velocity. However a signicant problem was that while the slow light eect was
present for each eld, there was a severe group velocity mismatch between the pulses (Fig.
4.5.) Furthermore, without the ability to control the relative group velocity between the
pulses, the simultaneous slowdown of two separate pulses observed in this experiment
would be of very limited applicability to the aforementioned proposals.
This initial limitation can be overcome by noting that the group velocity of a light
pulse is ultimately determined by the EIT dispersion prole - that is, the real component
of the linear susceptibility. In the previous section, it was found that by having a second
eld present in the EIT system, the lineshape could be manipulated by changing the
strength of that eld.
To exploit the change in the dispersion and therefore the slow-light dynamics, a

41

1
Zeeman Pulse: No Atoms
Zeeman Slowed Pulse
Hyperfine Pulse: No Atoms
Hyperfine Slowed Pulse

Normalized Signal at Detector [au]

0.9
0.8
0.7
0.6
0.5
0.4
v = 298 km/s
g

v = 33.0 km/s
g

0.3
0.2
0.1
0

10

11

12

13
14
Time [s]

15

16

17

18

Figure 4.5: Without atomic preparation pulses, the group velocities of the simultaneous
slowed pulses are severely mismatched.

42

Figure 4.6: Pulse sequence used for atomic preparation. A preparatory pulse in the
hyperne channel is applied to optically pump atoms into the Zeeman ground state.
After a brief dead-time, set to be much less than the relaxation time of the system,
simultaneous pulses are sent through the atoms.
preparatory pulse was sent in the hyperne chanel and then switched o momentarily
before sending the simultaneous pulses through the atoms. By adjusting the strength
and duration of the preparatory pulse, the amount of optical pumping from one ground
state to another was increased or decreased. Figure 4.6 illustrates the pulse sequence
described above.
Since the group velocity of a pulse depends on the atomic density in the relevant
transition [5], optical pumping out of the hyperne ground state into the Zeeman ground
state has the eect of increasing the group velocity of the hyperne pulse while reducing
that of the Zeeman. Noting that the group velocity reduction was initially less dramatic
for the Zeeman pulse (Fig. 4.5,) the hyperne channel preparatory pulse had the eect
of correcting for this initial discrepancy. Figure 4.7(a) shows the results obtained by
using a 500 s hyperne initial pulse with varying power followed by simultaneous 2 s
pulses in each signal eld. In the neighborhood of Pprep = 55 W, the group velocities
of the pulses are found to be matched at 155 km/s. The solid lines are produced from
a numerical simulation of a tripod DEIT system (see next section.) Figure 4.7(b) shows
the signals at the detector for the 55 W data point. Note that, as mentioned before, the

43

Figure 4.7: (a) The group velocity of each pulse is shown as a function of preparatory
pulse power. Note that the curves intersect, at which point the pulses are matched. (b)
Traces for the Pprep = 55 W data point.
transparency at two photon resonance for the Zeeman EIT is considerably less than that
of the hyperne EIT. Consequently, a transmitted slow Zeeman pulse suers considerably
more absorption than the hyperne pulse. For this reason, the Zeeman pulse is scaled
by a factor of 10 in the gure.
In addition to using preparatory pulses in the hyperne eld, pulses were applied
in the Zeeman channel and was found to further enhance the slow light eect on the
hyperne pulse by nearly a factor of 10. Although not of interest for creating matched
light pulses, the Zeeman eld could be used to set a programmable delay in an optical
pulse. This itself could nd application as an optical light buer acting on a single
channel.

4.1.4 Simultaneous Storage

After exploring the simultaneous slowdown of pulses with matched group velocities, then
next step was to switch o the control eld during the propagation of these pulses in
order to store the pulses simultaneously and retrieve the pulses a time later by switching
the control eld back on. Figure 4.8 displays the results of a typical storage run with

44

Figure 4.8: Simultaneous storage and retrieval of two signal pulses using a single pump
eld.
a storage time of 10 s. Storage times of 100 s were routinely observed, but with
decreasing retrieval eciency.
One of the main limitations of simultaneous storage is that the pulse compression
factor vg /c may not be small enough to have the pulse t spatially in the cell. A 2 s
pulse in free space has spatial length ct = 600 m. For our 10 cm cell, this requires
vg /c = 2 104 corresponding to vg = 20 km/s, which is an order of magnitude lower
than obtained in the experimental setup. When a pulse does not t entirely into the cell,
only the portion of the pulse which is within the cell at the instant that the pump eld
is switched o is stored, the rest of which escapes. Thus aside from the loss of eciency
due to decoherence in the atoms, another practical limitation is the pulse compression.
It should be noted that in the presence of the preparatory eld, the spatial extent of the
hyperne pulse could be reduced to t within the cell. In this case, the retrieval eciency
was only limited by absorption and atomic decoherence and reached 51% (Fig. 4.9).

45

1
Stored Pulse
Slowed Pulse
Pulse Without Atoms

Normailized Amplitude at detector

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

10

15

20

25

Time [s]

Figure 4.9: Single EIT storage on the hyperne eld: only light which is in the atomic
medium when the pump eld is switched o can be stored. For comparison, the dotted
curve show the slow light produced when the the pump eld is left on.

4.2 A Theoretical Model

In order to gain understanding of our observations, a theoretical model of a tripod atomic
system was evaluated for which the susceptibility was numerically determined. Although
the double EIT system studied in the previous section consists of 16 levels, forming
several tripod EIT systems, these tripod systems are degenerate and the properties of
the system as a whole can be eectively modelled by examining a single tripod system.
The model consists of a common excited state |e and three ground states |z, |h, |p,
coupled by two weak signal elds z , h , and a pump p respectively. The decoherence
mechanisms included dephasing (decay of the o-diagonal matrix elements) as a result of
atomic collisions ij acting within each pair of ground states, and population exchange
G describing atoms which drift in and out of the beam, acting on levels |h and |z (Fig.
4.10).
In the rotating wave approximation, the light-atom Hamiltonian is given by:

46

Figure 4.10: The theoretical tripod scheme analyzed in the text.

h p z

(4.1)

where i is the detuning of the light eld with respect to the |i |e atomic transition.
The above mentioned decoherence terms may be modeled by introducing the relaxation
matrix (see chapter 2):

ee (h + p + z )

12 eh h

12 ep p

12 ez z

12 he h

ee h + G (zz hh )

1

2 ph ph

Ghz 12 hz hz

12 pe p

12 hp ph

ee p

12 pz pz

12 ze z

Gzh 12 zh hz

12 zp pz ee G (hh zz ) + ee z

(4.2)
By solving the Liouville equation using eq. (4.1) and the above decoherence matrix, the
steady-state density matrix was obtained (see equation 2.5.)
The main purpose of the theoretical analysis of this system was to investigate the

47
eect of the preparatory pulse on the slow-light dynamics of the system. A nontrivial
assumption in this model is that the steady-state approach is valid. Recall that in the
experiment the preparatory pulse is relatively long (500 s), which is switched o for
of f 5 s before sending the twin pulses. In order to verify the validity of the steady
state approximation the relaxation time of the system was measured. Specically, with
the preparatory pulse on, the other eld sees a higher atomic density than it does when
the eld is o. For a nite time after the preparatory eld is switched o, the atomic
system remains unchanged before it begins to decay to the no preparation eld state.
This time was measured to be on the order of relax 100 s. Since relax  of f , the
steady-state approach is justied.
The group velocities of the signal pulses were calculated using the linear response
theory with the initial state given by the steady state density matrix. The susceptibility
of the atomic gas for the two signal elds was evaluated numerically for varying strengths
of the preparation eld. The parameters of the system were set to t the group velocity
behavior to that observed experimentally. A good t was obtained with a Doppler width
of 500 MHz, z = h = p = 6 MHz, hz = hp 5 kHz. zp 40 kHz, and G = 50 Hz.
The results of this analysis using the above parameters is displayed as the solid line in
gure 4.7.
In the CW picture, our theoretical model has also been able to qualitatively reproduce
the enhancement of the transparency contrast for one signal eld when increasing the
strength of the other. Figure 4.11 shows the experimental and theoretical plots of the
CW EIT linewidth with increasing eld strength on the opposite EIT signal.
It was found that a crucial parameter in obtaining a good t was the decoherence
between the two signal ground states |h and |z. As hz was increased, the optical
pumping eect became much more signicant. With small hz , enhanced transparency
on 2-photon resonance in the CW regime emerged due to the presence of an additional

48

Figure 4.11: Comparison of the optical pumping eect on the EIT contrast in the experiment to that predicted by the theoretical model
dark state formed by energy levels |h and |z. Experimentally this enhancement was
veried by manipulating the detuning of the hyperne pulse and observing an additional
small transparency dip, corresponding to this dark state.
Also interesting is the role of the population exchange rate G. As noted in [20],
the fraction of this mechanism in ground state decoherence is small compared to pure
dephasing ij . This nding is conrmed by the present experiment, yet the population
exchange mechanism cannot be completely neglected. If we set G = 0, the populations
of states |h and |z would depend only on the ratio between h and z , but not on their
absolute magnitude. The nonlinear eect of the signal elds on each other could then be
observed at arbitrarily low eld strength, which is not in agreement with experimental
observation. This discrepancy can be addressed by setting a small, but nonzero value of
G, which governs the ground state populations at low signal intensities. Intuitively, the
non-zero population exchange serves to continuously repopulate a ground state which
may otherwise have zero population, in which case there would be no optical pumping.
This allows for a more signicant optical pumping eect dependant on the strength of
the signal eld which couples out of this level, and at negligibly small eld strengths, the
optical pumping eect is overwhelmed by the population exchange.

49

4.3 Further Prospects

The ability to manipulate the relative group velocities of light pulses could prove promising as an all-optical light buer for use in classical optical signal processing. Since the
pulses are at separate frequencies, a frequency multiplexing scheme could be implemented
in which light pulses are separated, independently addressed, and rejoined, based on their
frequency.
The ability to simultaneous store light pulses at separate frequencies shows that double EIT can be used as a broadband memory - acting independently on two pulses of
light. There are several limitations to this scheme, namely that the memory would only
work for discrete frequencies and the second pulse could not enter the atoms while a pulse
is being stored (since no pump eld would lead quickly to absorption.) Nevertheless, the
ability to independently store and retrieve light pulses at separate frequencies could nd
application in quantum memory.

Chapter 5
Giant Optical Nonlinearities With DEIT
One of the key features of EIT is that while vanishing linear absorption and unity index
of refraction are present at two-photon resonance due to destructive interference of excitation pathways, the third order susceptibility is at a maximum as a result of constructive
interference [26]. This allows for the creation of large optical nonlinearities such as crossphase modulation (XPM). In this chapter, progress is reported towards implementing a
large XPM scheme using double EIT.

5.1 EIT-Based Nonlinearities

The key attribute of an EIT system allowing for large cross-Kerr interaction is the steep
dispersion about two-photon resonance - the same mechanism responsible for slow light
(Fig. 2.2.) Since the optical phase shift introduced by an atomic medium is determined
by the real component of the susceptibility or equivalently, the index of refraction, a slight
change in the two photon detuning will correspond to a large shift in phase. Specically,
the optical phase shift induced by a medium of length L on a transmitted pulse is given
by:

(n) = L n().
c

(5.1)

Suppose then that the presence of an additional eld changes the two photon resonance
condition of the EIT system. If nf and ni are the indices of refraction with and without
the additional eld causing the shift in two-photon resonance, the XPM phase shift is
dened as XP M = (nf ) (ni ). Thus, an EIT medium has the capability to induce a

50

51

Figure 5.1: N-type atomic level scheme proposed by Schmidt and Imamoglu. O-resonant
coupling to |s creates a Stark shift leading to a large shift in the phase of the signal.
The inset shows a typical non-EIT XPM scheme for which the coecient is orders of
magnitude less [36].
large XP M between two elds, as long as one eld can aect the detuning of the other.
The rst proposal to exploit this nonlinear eect was published in 1996 by Schmidt
and Imamoglu [36]. The idea is to start with a standard EIT system, but to couple an
additional eld, o resonantly from the EIT signal ground state to an additional state
|s, forming an N-type scheme (Fig 5.1.) The |g |e |p lambda system produces the
standard steep dispersion associated with EIT, but the now ground state |p experiences
an AC-Stark shift due to an additional eld which we call the Stark eld, which couples
|p and |s o-resonantly. This shift in energy in turn shifts the 2-photon resonance of
the EIT system, creating a large phase shift. Experimental evidence of this scheme was
noted in [5] and fully analyzed in [37].
Compared to a conventional cross-phase modulation scheme involving three atomic
levels not experiencing EIT (see gure 5.1 inset,) the N-type scheme produces a nonlinear
coecient which is many orders of magnitude greater. Specically, the ratio of Re[(3) ]
nonlinearities from the EIT to non-EIT case is 3.3 109 [36]. The magnitude of this

52
nonlinearity allows for the possibility of nonlinear interactions at the single photon level.
The ability to create a conditional -phase shift between photons would be of tremendous
importance to the quantum information community since it would allow for the creation
of a quantum controlled-phase gate [13]; a universal quantum phase gate [1].
A major limitation of this scheme is that the generated nonlinear phase shift is proportional to the interaction length Lint , of the signal and Stark elds. For short pulses,
this becomes a severe restriction on the maximum possible phase shift, due to the group
velocity mismatch between the two pulses. The Stark pulse travels at the speed of light c,
whereas the signal pulse experiences the pronounced group velocity reduction associated
with EIT and travels at vs  c. As a result, the Stark pulse quickly escapes the slowed
signal. Quantitatively, the interaction length of the signal and Stark eld propagating
through a medium of length L is

Lint = min[L, Lh ]

(5.2)

where Lh = vs t is known as the Hau length and is dened as the length of propagation
which would correspond to a relative delay between the Stark and signal elds equal to
their temporal length t [11]. For 1 s pulses, with vs = 30 km/s, this corresponds to
Lint = 3 m, which gives a maximum relative phase shift of 0.1 rad, under ideal EIT
conditions [14]. Thus, the above scheme is rather limited in terms of its use as a quantum
phase gate, since the phase shift obtained is much less than .

5.2 Physical Implementation in

87

Rb

The above diculty may be overcome if both the signal and Stark elds have matched
group velocities. In this case, Lh and Lint is limited only by the cell length.
Therefore it would be benecial to have a system in which both elds experienced EIT,

53

Figure 5.2: (a) The large XPM scheme proposed to implement experimentally. The
|p |z |h provides the double EIT tripod scheme gives the slow light eect to both
signal elds. |p also couples o-resonantly to |s creating large cross phase modulation.
It is imperative that the strength of the magnetic eld is such that the Zeeman shift is
in the nonlinear, intermediate eld regime in order to break the |X |s |p EIT scheme.
(b) Illustrates the need for a magnetic eld.
such as the double EIT system described in the previous section (shown in Fig. 5.2(b)
for convenience.) In this case the hyperne eld acts as the signal and the Zeeman eld
acts as the Stark eld. Under these conditions the signal and Stark elds both experience
reduced, matched group velocities as well as forming an N-type Scheme. However, a
second glance reveals that this system is not suitable for a large XPM: an additional dark
state is formed between levels |x |s |p, rendering the atomic system transparent to the
Stark eld, and destroying the nonlinear eect.
A 2006 paper by Wang et al. proposed to overcome this obstacle by applying a strong
magnetic eld along the propagation axis of the laser, creating a Zeeman shift on the
magnetic sublevels [10]. The modied EIT scheme is shown in gure 5.2(a). In this case
the |p |h |z |e tripod system is on resonance while the the Zeeman eld couples to
|s o-resonantly. Care must be taken to choose a suitable magnetic eld: recall from

54
chapter 2 that for small eld strengths, the Zeeman shift is linear in mF . In this case, an
additional dark state is still formed from the coupling of the pump from |X to |s since,
although one-photon detuning is present, the two-photon detuning is identical to that of
the |p |e |z EIT scheme. For a suciently strong magnetic eld however, the Zeeman
shift is nonlinear in mF and two photon resonance can be maintained on |p |e |z while
having sucient two-photon resonance on |x |s |p to be outside of the EIT linewidth.
In this case, simultaneous matched light pulses can be created having the large XPM
described in the previous section.

5.3 Experimental Observations and Progress

In order to determine the magnetic eld strength required for the experiment, the splitting
of the EIT lines as a function of applied magnetic eld was measured. The magnetic eld
was produced by sending current through a tightly wound precision solenoid which rested
in the oven which held the cell. In general, when a magnetic eld is applied to the atoms,
the once degenerate lines are separated in frequency and two photon resonance is met
for each of these lines at dierent frequencies as described in chapter 2. Figure 5.3 shows
the hyperne EIT with and without a weak magnetic eld applied along the propagation
axis. Three transparency dips appear, corresponding to the mF = 1, mF = 0, and
mF = 1 ground states. At rst, the central mF = 0 dip does not deviate in frequency,
since to rst order there is no frequency shift. Note as well that the transparency at each
line is signicantly reduced as a result of the splitting. This can be explained by the
fact that while atoms in the particular ground-state which is on two-photon resonance
will experience EIT, there is also coupling to other atomic levels which have two-photon
detuning greater than the EIT bandwidth and therefore are not transparent to the signal
eld.

55

Figure 5.3: The splitting of the atomic lines as a result of an applied magnetic eld.
Since the two photon resonance condition is met for each EIT system separately, several
transparency dips emerge. Shown here is the hyperne EIT, which splits into three
separate EIT subsystems.
To determine the magnetic eld required to cause two-photon detuning on |X |s |p
greater than the EIT bandwidth, the technique for calculating the Zeeman splitting
for arbitrary elds from chapter two may be applied. From gure 5.2, the two photon
detuning in the |X |s |p EIT scheme is given by 2 = sx sp , where ij = (Ei
h ij . Also, |p |e |z EIT is resonant, h
sx = Ee Ep and h
sp = Ee Ez . Thus,
Ej )/
sx ) ((Es Ep ) h
sp ), or
(
h2 = (Es Ex ) h

2 = (Ex + Ez 2Ep )/h.

(5.3)

By observing the shift in the frequency at which the |X |s |p system becomes resonant, this shift may be observed experimentally. Figure 5.4 displays the experimental
observations of this shift as well as the theoretical curves obtained by the model obtained
by inserting eq. (5.3) into the model discussed in chapter two. It should be noted that
the accuracy at which the splitting can be measured by observing the magnetic eld lines
provides a precise measure of the magnetic eld obtained for a given value of current.

56

Figure 5.4: The 2 photon resonance of the |X |s |p EIT scheme while the |p |e |z
EIT scheme is resonant is plotted as a function of applied magnetic eld. The shaded
region is the oor, below which the Stark eld experiences EIT and the nonlinear eect
is destroyed.
This is the basis for additional proposals to use EIT for magnetometry [38].
The experimental setup used to observe the EIT-based nonlinearity is similar to that
of the double EIT setup, except a non-zero magnetic eld of strength 160 G corresponding
to 2 = 13.8 MHz is applied to the atoms. At this eld strength the one photon detuning
of the Stark eld is 144.2 MHz which for a Stark eld Rabi frequency of 500 kHz, gives
a Stark shift of 1.7 kHz.
One of the primary challenges in implementing this scheme experimentally is the
creation of a suciently stable and homogeneous magnetic eld. Fluctuations in current
through the solenoid will translate to uctuations in magnetic eld and therefore to
uctuations in the resonant frequency of the atoms - eectively smearing out the EIT

57
prole. As a minimum condition, it is required that the uctuations of the atomic
level due to the magnetic eld be less than the EIT linewidth. Since the Zeeman shift
is approximately 0.9MHz/G in the vicinity of 144.2MHz, and since the EIT line in this
experiment is on the order of a few hundred kHz, stability on the order of 0.1% is required.
Given that the solenoid used corresponded to 54.4G/A, the current uctuations must be
less than 3mA on a 3A signal. On top of the uctuations due to current instability,
external elds will also play a role in creating uctuations in the shift in the atomic
level. Special care must then be taken in order to suciently shield the atoms from
these external elds while at the same time, allowing the strong eld through the light
propagation axis.
As well as the above temporal inhomogeneities, spatial inhomogeneities will also tend
to destroy the EIT. If the magnetic eld strength or direction varies as a function of
position along the cell, atoms at dierent points along the cell will see light of dierent
polarization and dierent detunings. In order to minimize this eect, a solenoid which
is considerably longer than the atomic cell should be used with the cell placed in the
center.
As seen in gure 5.3, the EIT under split lines suers additional absorption due to
the coupling of the signal eld to additional levels which are not in two photon resonance
with respect to the pump. Since the atomic system used for this experiment is a Doppler
broadened cloud of

87

Rb, which at 40 C has Doppler FWHM WD 550 MHz, the

splitting is not sucient to prevent signicant coupling to these additional levels, and as
a result, the EIT contrast suers signicantly.
Another complication that the introduction of a large magnetic eld brings is that,
due to the splitting of the atomic lines, the degeneracy in the ground states is lifted.
Consequentially there is less atomic population in the single EIT ground-state addressed
as compared to the degenerate case, in which there are several EIT ground states each

58
experiencing the EIT eect simultaneously. This results in a broader EIT line and a
signicantly less steep dispersion curve at two-photon resonance. As a consequence,
group velocity reduction eect is signicantly hindered.

5.4 Prospects for Experiment

Although the double EIT system has been realized capable of demonstrating the large
cross-phase modulation described in [10] and a phase detection scheme has been put
in place, a nonlinear phase shift has yet to be observed. The primary factor which
hinders the magnitude of the nonlinear phase shift is the shallowness and poor contrast
of the EIT which results when the magnetic eld is present. This may be due to the
inhomogeneous broadening occurring in a hot atomic gas; an eect which has yet to
be analyzed theoretically. In order to overcome this setback, the detrimental eects
hindering the EIT mentioned in the previous section need to be compensated for, and
are discussed below.
In order to minimize the eect of temporal uctuations in the magnetic eld, a current regulating supply has been manufactured. Depending on the magnitude of the
uctuations, active compensation may be introduced by constructing a feedback system
consisting of an integrated hall probe controlling current modulation.
Although an ideal solenoid creates a uniform homogenous magnetic eld throughout,
there are in practice noticeable fringing eects present at the boundaries. The use of
a longer solenoid, or equivalently, a shorter cell will reduce the spatial inhomogeneity
throughout the cell, at the expense of less optical depth for a given temperature. In order
to overcome the contribution from outside magnetic elds, more -metal shielding can
be applied surrounding the wall of the cell, but not the faces. This has the disadvantage
however that the shielding itself will become signicantly magnetized over time. However,

59
if this shielding is demagnetized regularly by applying a strong AC-current through the
shielding which is adiabatically switched o, long-term magnetization eects will not
come into play.
The increased absorption in the experiment is due to the ground states which are
not part of an EIT dark state. By optically pumping out of these states via preparatory
pulses as in chapter 4, this eect may be minimized. However, some absorption is always
to be expected in a Doppler broadened system. Ideally, an initial pulse sequence would
prepare the atoms in a mixture of the signal and Stark ground states for which the light
elds would then not couple transitions which do not experience EIT. Such a sequence
is described in [10] in which the atoms are originally pumped into the common double
EIT excited state |e, followed by a Raman transition between the hyperne and Zeeman
ground-states.

Chapter 6
Conclusion and Outlook
In this thesis, a double EIT system was presented and fully studied. This system was used
to create matched slow pulses of light as well as to simultaneously store these pulses. In
order to accomplish this task, several technical challenges were required to be met. First,
a stable laser diode system was constructed capable of tunable single mode operation
over a wide frequency range. An optical phase-lock loop, capable of sub-Hz precision was
constructed to lock the diode lasers developed for use in this experiment. In order to
pulse and scan the frequency of the light, switchable drivers were made in order to drive
an 80-200 MHz AOM from a stable 10 MHz source. The timing of the optical pulses was
taken care of by designing a versatile delay generator used in conjunction with the AOM
drivers.
An EIT-based nonlinear scheme rst proposed in [10] was discussed and progress
towards its implementation was made. To this end, a phase detection scheme was developed which is capable of measuring the phase dierence between optical pulses as short
as 1 s. The oven containing the

87

Rb cell used in chapter 4 for double EIT was modi-

ed to accommodate a large (160G) magnetic eld and EIT was observed in the case of
nondegenerate Zeeman sublevels.
Although progress is being made towards nonlinear optics capable of creating a conditional phase shift of radians between single photons, the reported progress is a far cry
from this ideal case. It is doubtful that the experiment as described would ever provide
single photon level nonlinearities. As a short-term goal however, it should certainly be
possible to create a large conditional phase shift between two weak, 1 s pulses as a
proof of principle demonstration. In order to accomplish this, the experiment could be
60

61
moved to a trapped cold atom setup in order to minimize ground state decoherence and
additional absorption due to Doppler broadening. Such a cold atom trap is currently
being constructed as a separate project in the lab and it is planned to integrate the
current experiment with this atom trap. By suitable atomic preparation, the EIT eect
may be improved to minimize the absorption of the weak pulses, allowing for a weaker
pump eld, leading to a narrower dispersion curve, and a nonlinear phase shift of higher
magnitude.

Bibliography
[1] M. Nielsen and I. Chuang, Quantum Computation and Quantum Information,
Cambridge University Press (2000).
[2] D. Mayers, Unconditional Security in Quantum Cryptography, Journal of the ACM
48, 351406 (2001).
[3] V. Giovannetti, S. Lloyd, and L. Maccone, Quantum Metrology, Phys. Rev. Lett.
96, 010401 (2006).
[4] K.-J. Boller and A. Imamoglu and S.E. Harris, Observation of electromagnetically
induced transparency, Phys. Rev. Lett. 66, 2593 (1991).
[5] L. Hau, S. Harris, Z. Dutton and C. Behroozi, Light speed reduction to 17 metres
per second in an ultracold atomic gas, Nature 397, 594598 (1999).
[6] C. Liu, Z. Dutton, C. Behroozi and L. Hau, Observation of coherent optical information storage in an atomic medium using halted light pulses, Nature 409, 490493
(2000).
[7] J. Appel and E. Figueroa and D. Korystov and M. Lobino and A. I. Lvovsky, Quantum Memory for Squeezed Light, Phys. Rev. Lett. 100, 9 (2008).
[8] M. Fleischauer and M. Lukin, Quantum memory for photons: Dark-state polaritons, Phys. Rev. A 65, 022314 (2000).
[9] F. Vewinger, J. Appel, E. Figueroa and A. I. Lvovsky, Adiabatic frequency conversion of optical information in atomic vapor, Opt. Lett. 32, 27712773 (2007).
[10] K.-P. Marzlin, Z.-B. Wang and B. Sanders, Large Cross-Phase Modulation between
Slow Copropagating Weak Pulses in

87

Rb, Phys. Rev. Lett. 97, 063901 (2006).

62

63
[11] S. E. Harris and L. Hau, Nonlinear Optics at Low Light Levels, Phys. Rev. Lett.
82, 4611 4614, (1999).
[12] J. Khurgin, Optical buers based on slow light in electromagnetically induced transparent media and coupled resonator structures: comparative analysis, J. Opt. Soc.
Am. B, 22, 10621073 (2005).
[13] C. Ottaviani, P. Tombesi, M. Artoni, F. Cataliotti, S. Rebic, D. Vitali, and R. Corbalan, Polarization phase gate with a tripod atomic system, Phys. Rev. A. 70,
032317 (2004).
[14] M. Fleishhauer, A. Imamoglu and J.P. Marangos, Electromagnetically induced
transparency: Optics in coherent media, Rev. Mod. Phys. 77, 633 (2005).
[15] G. Pollack and D. Stump, Electromagnetism (Addison-Wesley, 2002).
[16] J. Appel, A Light-Atom Quantum Interface based on Electromagnetically Induced
Transparency (PhD. Thesis, 2007).
[17] J. J. Longdell, E. Fraval, M. J. Sellars, and N. B. Manson, Stopped Light with
Storage Times Greater than One Second Using Electromagnetically Induced Transparency in a Solid Phys. Rev. Lett. 95, 063601 (2005).
[18] G. Alzetta, A. Gozzini, L. Moi and G. Orriols, An experimental method for the
observation of r.f. transitions and laser beat resonances in oriented Na vapour, Nuovo
Cimento Soc. Ital. Fis. B. 36, 5 (1976)
[19] M. O. Scully and M. S. Zubairy, Quantum optics (Cambridge University Press,
1997).
[20] E. Figueroa, F. Vewinger, J. Appel, and A.I. Lvovsky, Decoherence of electromagnetically induced transparency in atomic vapor, Opt. Lett. 31, 26252627 (2006).

64
[21] M. Fleischauer and M. Lukin, Dark-State Polaritons in Electromagnetically Induced Transparency, Phys. Rev. Lett. 84, 5094 (2000).
[22] M. Eisaman, A. Andre, F. Massou, M. Fleischhauer, A. Zibrov and M. Lukin, Electromagnetically induced transparency with tunable single-photon pulses, Nature
438, 837841, (2005).
[23] T. Chaneli`ere, D. Matsukevich, S. Jenkins, S. Lan, A. Kennedy and A. Kuzmich,
Storage and retrieval of single photons transmitted between remote quantum memories, Nature 438, 833836, (2005).
[24] J.J. Sakurai, Modern Quantum Mechanics (Addison-Wesley, 1994).
[25] D.J. Griths, Introduction to Quantum Mechanics (Prentice-Hall, 1995).
[26] D. Steck, Alkali D Line Data, http://steck.us/alkalidata.
[27] L. Carroll, Alices Adventures in Wonderland (MacMillan, 1865).
[28] C. Gerry and P. Knight, Introductoroy Quantum Optics (Cambridge University
Press, 2006).
[29] N. Delone and V. Krainov, Multiphoton Processes in Atoms (Springer, 1994).
[30] E. Figueroa, Electromagnetically Induced Transparency and Squeezed Light (PhD.
Thesis, 2008).
[31] L. Ricci, M. Weidem
uller, T. Esslinger, A. Hemmerich, C. Zimmermann, V. Vuletic,
W. Konig and T. W. Hansch, A compact grating-stabilized diode laser system for
atomic physics, Optics Communications 117, 541549 (1995).

65
[32] X. Cao, C. Zang, C. Xie, H. Wang, S. Li and X. Yang, Large Cross-phase Modulation Based on Double EIT in a Four-level Tripod Atomic System, arXiv:0802.2136v1
[quant-ph] (2008).
[33] A. Imamoglu and M. D. Lukin, Controlling photons using electromagnetically induced transparency, Nature 413, 273276 (2001).
[34] A. MacRae, G. Campbell and A. I. Lvovsky, Matched Slow Pulses Using Double
Electromagnetically Induced Transparency, arXiv:0806.2656v [quant-ph] (2008)
[35] G. B. Arfken and H. J. Weber, Mathematical Methods for Physicists (Harcourt
Academic Press, 2001).
[36] H. Schmidt and V. Imamoglu, Giant Kerr nonlinearities obtained by electromagnetically induced transparency, Opt. Lett. 23, 1936 (1996).
[37] H. Kang and Y. Zifu, Observation of Large Kerr Nonlinearity at Low Light Intensities, Phys. Rev. Lett. 91, 9 (2003).
[38] D. Petrosyan and Yuri P. Malakyan, Magneto-optical rotation and cross-phase
modulation via coherently driven four-level atomsin a tripod conguration, Phys.
Rev. A 70, 023822 (2004).