Ref 5

REVIEWS OF MODERN PHYSICS, VOLUME 77, APRIL 2005
Electromagnetically induced transparency: Optics in coherent media

Michael Fleischhauer
Fachbereich Physik, Technische Universität Kaiserslautern, D-67663 Kaiserslautern,
Germany
Atac Imamoglu
Institute of Quantum Electronics, ETH-Hönggerberg, HPT G12, CH-8093 Zürich,
Switzerland
Jonathan P. Marangos
Quantum Optics and Laser Science Group, Blackett Laboratory, Imperial College, London
SW7 2BW, United Kingdom
共Published 12 July 2005兲
Coherent preparation by laser light of quantum states of atoms and molecules can lead to quantum
interference in the amplitudes of optical transitions. In this way the optical properties of a medium can
be dramatically modified, leading to electromagnetically induced transparency and related effects,
which have placed gas-phase systems at the center of recent advances in the development of media
with radically new optical properties. This article reviews these advances and the new possibilities they
offer for nonlinear optics and quantum information science. As a basis for the theory of
electromagnetically induced transparency the authors consider the atomic dynamics and the optical
response of the medium to a continuous-wave laser. They then discuss pulse propagation and the
adiabatic evolution of field-coupled states and show how coherently prepared media can be used to
improve frequency conversion in nonlinear optical mixing experiments. The extension of these
concepts to very weak optical fields in the few-photon limit is then examined. The review concludes
with a discussion of future prospects and potential new applications.
CONTENTS B. Cross-phase modulation using single-photon pulses

with matched group velocities 664
C. Few-photon four-wave mixing 665
I. Introduction 633
D. Few-photon cavity EIT 666
II. Physical Concepts Underlying Electromagnetically
VII. Summary and Perspectives 667
Induced Transparency 635
Acknowledgments 670
A. Interference between excitation pathways 635
References 670
B. Dark state of the three-level ⌳-type atom 637
III. Atomic Dynamics and Optical Response 638
A. Master equation and linear susceptibility 638 I. INTRODUCTION
B. Effective Hamiltonian and dressed-state picture 642
Advances in optics have frequently arisen through the
C. Transmission through a medium with EIT 643
development of new materials with optimized optical
D. Dark-state preparation and Raman adiabatic passage 644
properties. For instance, the introduction of new optical
E. EIT inside optical resonators 645
crystals in the 1970s and 1980s led to substantial in-
F. Enhancement of refractive index, magnetometry,
creases in nonlinear optical conversion efficiencies to the
and lasing without inversion 646
ultraviolet 共UV兲. Another example has been the devel-
IV. EIT and Pulse Propagation 647
opment of periodically poled crystals that permit qua-
A. Linear response: Slow and ultraslow light 647 siphase matching in otherwise poorly phase-matched
B. “Stopping of light” and quantum memories for nonlinear optical processes, which greatly enhance non-
photons 650 linear frequency-mixing efficiencies. As we shall see, co-
C. Nonlinear response: Adiabatic pulse propagation herent preparation is a new avenue that produces re-
and adiabatons 652 markable changes in the optical properties of a gas-
V. Enhanced Frequency Conversion 654 phase atomic or molecular medium.
A. Overview of atomic coherence-enhanced nonlinear The cause of the modified optical response of an
optics 654 atomic medium in this case is the laser-induced coher-
B. Nonlinear mixing and frequency up-conversion with ence of atomic states, which leads to quantum interfer-
electromagnetically induced transparency 657 ence between the excitation pathways that control the
C. Nonlinear optics with maximal coherence 659 optical response. We can in this way eliminate the ab-
D. Four-wave mixing in double-⌳ systems 660 sorption and refraction 共linear susceptibility兲 at the reso-
VI. EIT with Few Photons 661 nant frequency of a transition. This was termed electro-
A. Giant Kerr effect 661 magnetically induced transparency 共EIT兲 by Harris and
0034-6861/2005/77共2兲/633共41兲/$50.00 633 ©2005 The American Physical Society

634 Fleischhauer, Imamoglu, and Marangos: Electromagnetically induced transparency
co-workers 共Harris et al., 1990兲. The importance of EIT

stems from the fact that it gives rise to greatly enhanced
nonlinear susceptibility in the spectral region of induced
transparency of the medium and is associated with steep
dispersion. For some readable general accounts of ear-
lier work in the field, see, for example, Harris 共1997兲 or
Scully 共1992兲. Other more recent reviews on specific as-
pects of EIT and its applications can be found in the
articles of Lukin, Hemmer, and Scully 共2000兲; Matsko,
Kocharovskaya, et al. 共2001兲; Vitanov et al. 共2001兲, as
well as in the topical Colloquium by Lukin 共2003兲. It
should be emphasized that the modification of atomic
properties due to quantum interference has been studied
extensively for 25 years; see, for example, Arimondo
共1996兲. In particular, the phenomenon of coherent popu-
lation trapping 共CPT兲 observed by Alzetta et al. 共1976兲 is
closely related to EIT. In contrast to CPT which is a
“spectroscopic” phenomenon that involves only modifi-
cations to the material states in an optically thin sample, FIG. 1. Susceptibility as a function of the frequency ␻p of the
EIT is a phenomenon specific to optically thick media in applied field relative to the atomic resonance frequency ␻31,
for a radiatively broadened two-level system with radiative
which both the optical fields and the material states are
width ␥31 共dashed line兲 and an EIT system with resonant cou-
modified.
pling field 共solid line兲: top, imaginary part of ␹共1兲 characterizing
The optical properties of atomic and molecular gases absorption; bottom, real part of ␹共1兲 determining the refractive
are fundamentally tied to their intrinsic energy-level properties of the medium.
structure. The linear response of an atom to resonant
light is described by the first-order susceptibility ␹共1兲.
The imaginary part of this susceptibility Im关␹共1兲兴 deter- ting 共Autler and Townes, 1955兲, dark states, and EIT,
mines the dissipation of the field by the atomic gas 共ab- will be the subject of this review. These phenomena can
sorption兲, while the real part Re关␹共1兲兴 determines the re- be understood within the basis of either bare atomic
fractive index. The form of Im关␹共1兲兴 at a dipole-allowed states or new eigenstates, which diagonalize the com-
transition as a function of frequency is that of a Lorent- plete atom-field interaction Hamiltonian. In both cases
zian function with a width set by the damping. The re- we shall see that interference between alternative exci-
tation pathways between atomic states leads to modified
fractive index Re关␹共1兲兴 follows the familiar dispersion
optical response.
profile, with anomalous dispersion 共decrease in Re关␹共1兲兴
The linear and nonlinear susceptibilities of a ⌳-type
with field frequency兲 in the central part of the absorp-
three-level system driven by a coherent coupling field
tion profile within the linewidth. Figure 1 illustrates both
will be derived in Sec. III. Figure 1 shows the imaginary
the conventional form of ␹共1兲 and the modified form that and real parts of the linear susceptibility for the case of
results from EIT, as will be discussed shortly. a resonant coupling field as a function of the probe field
In the case of laser excitation where the magnitude of detuning from resonance. Figure 2 shows the corre-
the electric field can be very large we reach the situation sponding third-order nonlinear susceptibility. Inspection
where the interaction energy of the laser coupling di- of these frequency-dependent dressed susceptibilities re-
vided by ប exceeds the characteristic linewidth of the veals immediately several important features. One rec-
bare atom. In this case the evolution of the atom-field ognizes that Im关␹共1兲兴 undergoes destructive interference
system requires a description in terms of state-amplitude
or density-matrix equations. In such a description we
must retain the phase information associated with the
evolution of the atomic-state amplitudes, and it is in this
sense that we refer to atomic coherence and coherent
preparation. This is of course in contrast to the rate-
equation treatment of the state populations often appro-
priate when the damping is large or the coupling is
weak, for which the coherence of the states can be ig-
nored. For a full account of the coherent excitation of
atoms the reader is referred to Shore 共1990兲.
For a two-level system, the result of coherent evolu-
tion is characterized by oscillatory population transfer
共Rabi flopping兲. The generalization of this coherent situ- FIG. 2. Absolute value of nonlinear susceptibility for sum-
ation to driven three-level atoms leads to many new frequency generation 兩␹共3兲兩 as a function of ␻p, in arbitrary
phenomena, some of which, such as Autler-Townes split- units. The parameters are identical to those used in Fig. 1.
Rev. Mod. Phys., Vol. 77, No. 2, April 2005

Fleischhauer, Imamoglu, and Marangos: Electromagnetically induced transparency 635
in the region of resonance, i.e., the coherently driven Throughout this review we shall try to point out the
medium is transparent to the probe field. The fact that important applications that arise as a result of EIT and
transparency of the sample is attained at resonance is describe in outline some of the seminal experiments. We
not in itself of great importance, as the same degree of attempt to cover comprehensively the alteration of lin-
transparency can be obtained simply by tuning suffi- ear and nonlinear optical response due to electromag-
ciently away from resonance. What is important is that netically induced transparency and related phenomena.
in the same spectral region where there is a high degree We do not, however, cover the related topic of lasing
of transmission the nonlinear response ␹共3兲 displays con- without inversion 共Kocharovskaya and Khanin, 1988;
structive interference, i.e., its value at resonance is larger Gornyi et al., 1989; Harris, 1989; Scully et al., 1989兲 and
than expected from a sum of two split Lorentzian lines. the interested reader is advised to look elsewhere for
Furthermore, the dispersion variation in the vicinity of reviews on that subject 共Knight, 1990; Kocharovskaya,
the resonance differs markedly from the steep anoma- 1992; Mandel, 1993; Scully and Zubairy, 1997兲. In the
lous dispersion familiar at an undressed resonance 共see following section we shall introduce the underlying
Fig. 1兲. Instead there is a normal dispersion in a region physical concepts of EIT through the coupling of near-
of low absorption, the steepness of which is controlled resonant laser fields with the states of a three-level sys-
by the coupling-laser strength 共i.e., very steep for low tem. In Sec. III we discuss in detail the dynamics in
values of the drive laser coupling兲. Thus despite the three-level atoms coupled to the applied laser fields and
transparency the transmitted laser pulse can still experi- determine the optical response. In Sec. IV we treat the
ence strong dispersive and nonlinear effects. It is most topic of pulse propagation in EIT and review the devel-
significant that the refractive index passes through the opments that have culminated in the demonstration of
vacuum value and the dispersion is steep and linear ex- ultraslow group velocities down to a few meters per sec-
actly where absorption is small. This gives rise to effects ond and even the “stopping” and “storing” of a pulse of
such as ultraslow group velocities, longitudinal pulse light. The utility of coherent preparation in enhancing
compression, and storage of light. Furthermore, through the efficiency of frequency conversion processes is dis-
the destructive 共constructive兲 interference in Im␹共1兲共␹3兲 cussed in Sec. V. Then in Sec. VI we shall turn to the
and the elimination of the effect of resonance upon the treatment of EIT in the few-photon limit, where it is
refractive index, the conditions for efficient nonlinear necessary to apply a fully quantum treatment of the
mixing are met. fields. Finally, we draw conclusions and discuss future
Coherent preparation techniques are most effective— prospects in quantum optics and atom optics as a result
and have been most studied—in atomic and molecular of EIT.
samples that are in the gas phase. The reason for this is
that the coherent evolution of states will in general be
inhibited by dephasing of the complex-state amplitudes,
II. PHYSICAL CONCEPTS UNDERLYING
but the coherence dephasing rates for gases are rela- ELECTROMAGNETICALLY INDUCED TRANSPARENCY
tively small when compared to those in solid-state me-
dia. Nevertheless, several workers have recently made A. Interference between excitation pathways
progress in applying these techniques to a variety of
solid-state systems 共Faist et al., 1997; Ham, Hemmer, and To understand how quantum interference may modify
Shahiar, 1997; Schmidt et al., 1997; Zhao et al., 1997; Wei the optical properties of an atomic system it is informa-
and Manson, 1999; Serapiglia et al., 2000兲. Atomic gases tive to follow the historical approach and turn first to the
remain very important in optical applications for several subject of photoionization of multielectron atoms. Due
key reasons; for example, they are often transparent in to the existence of doubly excited states the photoion-
the vacuum UV and infrared, and there exist techniques ization spectrum of multielectron atoms can show a rich
to cool them to ultracold temperatures, thus eliminating structure of resonances. These resonances are broad-
inhomogeneous broadening. They can also tolerate high ened due to relatively rapid decay 共caused by the inter-
optical intensities. Gas-phase media are, for instance, action between the excited electrons of these states兲 to
necessary for frequency conversion beyond the high- degenerate continuum states, with lifetimes in the pico-
and low-frequency cutoff of solid-state materials. second to subpicosecond range. For the reason that they
The dramatic modifications of the optical properties decay naturally to the continuum, these states are called
that are gained through coherent preparation of the at- autoionizing states. Fano introduced to atomic physics
oms or molecules in a medium have ushered in a renais- the concept of the interference between the excitation
sance in activity in the area of gas-phase nonlinear op- channels to the continuum that exist for a single au-
tics. New opportunities also arise due to the extremely toionizing resonance coupled to a flat continuum 关Fano,
spectrally narrow features that can result from coherent 1961; Fig. 3共a兲兴. In the vicinity of the autoionizing reso-
preparation; in particular, this gives ultralow group ve- nance the final continuum state can be reached via ei-
locities and ultralarge nonlinearities. These are of poten- ther direct excitation 共channel 1兲 or through resonance
tial importance to new techniques in optical information with the configuration interaction, leading to the decay
storage and also in nonlinear optics at the few-photon that provides a second channel to the final continuum
level, both of which may be important to quantum infor- state. The interference between these channels can be
mation processing. constructive or destructive and leads to frequency-

FIG. 3. Fano interferences of excitation channels into a con-

tinuum: 共a兲 for a single autoionizing resonance; 共b兲 for two
autoionizing states. FIG. 4. Interference generated by coherent coupling: left, co-
herent coupling of a metastable state 兩2典 to an excited state 兩3典
by the dressing laser generates 共right兲 interference of excita-
dependent suppression or enhancement in the photoion- tion pathways through the doublet of dressed states 兩a±典
ization cross section. 共Autler-Townes doublet兲 provided the decay out of state 兩2典 is
Experiments by Madden and Codling 共1965兲 showed negligible compared to that of state 兩3典.
the resulting transparency windows in the autoionizing
spectrum of helium. Sometime later the autoionizing in-
terference structures in strontium were used by Arm- spaced lifetime-broadened resonances were equivalent
strong and Wynne to enhance sum-frequency mixing in to dressed states created by coupling a pair of well-
a frequency up-conversion experiment to generate light separated atomic bound levels with a resonant laser field
in the vacuum UV 共Armstrong and Wynne, 1974兲.1 In 共Fig. 4兲. They thus proposed that the energy-level struc-
this wave-mixing experiment the absorption was elimi- ture required for quantum interference could be engi-
nated in spectral regions where the nonlinear response neered by use of an external laser field. Harris et al.
remained large, hence an improved efficiency for fre- 共1990兲 then showed how this same situation could be
quency conversion was reported. extended to frequency conversion in a four-wave mixing
During the 1960s autoionizing spectra were much scheme among atomic bound states with the frequency
studied 共see, for example, Garton, 1966兲. Several authors conversion hugely enhanced. This is achieved through
addressed the issue of interaction between two or more the cancellation of linear susceptibility at resonance as
spectral series in the same frequency range, where the shown in Fig. 1, while the nonlinear susceptibility is en-
interference between closely spaced resonances needs to hanced through constructive interference. The latter pa-
be considered 共Fano and Cooper, 1965; Shore, 1967兲. per was the first appearance of the term electromagneti-
Shapiro provided the first explicit analysis of the case of cally induced transparency 共EIT兲, which was used to
interference between two or more resonances coupled describe this cancellation of the linear response by de-
to a single continuum 共Shapiro, 1970兲. In this case the structive interference in a laser-dressed medium.
interference is mainly between the two transition path- Boller et al. 共1991兲, in discussing the first experimental
ways from the ground state to the final state via each of observation of EIT in Sr vapor, pointed out that there
the two resonances 关Fig. 3共b兲兴. Naturally interference are two physically informative ways that we can view
will be significant only if the spacing between these reso- EIT. In the first we use the picture that arises from the
nances is comparable to or less than their widths. work of Imamoglu and Harris 共1989兲, in which the
Hahn, King, and Harris 共1990兲 showed how this situa- dressed states can be viewed as simply comprising two
tion could be used to enhance four-wave mixing. In ex- closely spaced resonances effectively decaying to the
periments in zinc vapor that showed significantly in- same continuum 共Boller et al., 1991; Zhang et al., 1995兲.
creased nonlinear mixing, one of the fields was tuned to If the probe field is tuned exactly to the zero-field reso-
a transition between an excited bound state and a pair of nance frequency, then the contributions to the linear sus-
closely spaced autoionizing states that had a frequency ceptibility due to the two resonances, which are equally
separation much less than their decay widths. spaced but with opposite signs of detuning, will be equal
The case of interference between two closely spaced and opposite and thus lead to the cancellation of the
lifetime-broadened resonances, decaying to the same response at this frequency due to a Fano-like interfer-
continuum, was further analyzed by Harris 共1989兲. He ence of the decay channels. An alternative and equiva-
pointed out that this will lead to lasing without inver- lent picture is to consider the bare rather than the
sion, since the interference between the two decay chan- dressed atomic states. In this view EIT can be seen as
nels eliminates absorption while leaving stimulated arising through different pathways between the bare
emission from the states unchanged. Although we shall states. The effect of the fields is to transfer a small but
not discuss the subject of lasing without inversion fur- finite amplitude into state 兩2典. The amplitude for 兩3典,
ther in this review, this work was an important step in which is assumed to be the only decaying state and thus
the story with which we are concerned. A breakthrough the only way to absorption, is thus driven by two
was then made in the work of Imamoglu and Harris routes—directly via the 兩1典-兩3典 pathway, or indirectly via
共1989兲 when it was realized that the pair of closely the 兩1典-兩3典-兩2典-兩3典 pathway 共or by higher-order variants兲.
Because the coupling field is much more intense than
the probe, this indirect pathway has a probability ampli-
1
See the improved ␹共3兲 fit of Armstrong and Beers, 1975. tude that is in fact of equal magnitude to the direct path-

The physics underlying the cancellation of absorption

in EIT is identical to that involved in the phenomena of
dark-state and coherent population trapping 共Lounis
and Cohen-Tannoudji, 1992兲. We shall therefore review
briefly the concept of dark states. Alzetta et al. 共1976兲
made the earliest observation of the phenomenon of co-
herent population trapping 共CPT兲 followed shortly by
theoretical studies of Whitley and Stroud 共1976兲. Ari-
FIG. 5. Generic system for EIT: lambda-type scheme with mondo and Orriols 共1976兲 and Gray et al. 共1978兲 ex-
probe field of frequency ␻p and coupling field of frequency ␻c. plained these observations using the notion of coherent
⌬1 = ␻31 − ␻p and ⌬2 = ␻32 − ␻c denote field detunings from population trapping in a dark eigenstate of a three-level
atomic resonances and ⌫ik radiative decay rates from state 兩i典 lambda medium 共see Fig. 5兲. In this process a pair of
to state 兩k典. near-resonant fields are coupled to the lambda system
and result in the Hamiltonian H = H0 + Hint, where the
way, but for resonant fields it is of opposite sign. Hamiltonian for the bare atom is H0 and that for the
interaction with the fields is Hint. The Hamiltonian H
has a new set of eigenstates when viewed in a proper
rotating frame 共see below兲, one of which has the form
B. Dark state of the three-level ⌳-type atom
兩a0典 = ␣兩1典 − ␤兩2典, which contains no amplitude of the bare
The use in laser spectroscopy of externally applied state 兩3典 and has amplitudes ␣ and ␤ proportional to the
electromagnetic fields to change the system Hamiltonian fields such that it is effectively decoupled from the light
of course predates the idea of using this in nonlinear fields. In the experiments of Alzetta, population was
optics or in lasing without inversion. We must mention pumped into this state via spontaneous decay from the
the enormous body of work treating the effects of static excited states and then remained there since the excita-
magnetic fields 共Zeeman effect兲 and static electric fields tion probability of this dark state is canceled via inter-
共Stark effect兲. The case of strong optical fields applied to ference. An early account on the effect of CPT on the
an atom began to be extensively studied following the propagation of laser fields was given by Kocharovskaya
invention of the laser in the early 1960s. Hänsch and and Khanin 共1986兲. A very informative review of the
Toschek 共1970兲 recognized the existence of these types applications of dark states and the coherent population
of interference processes for three-level atoms coupled trapping that accompanies them in spectroscopy has
to strong laser fields in computing the system suscepti- been provided by Arimondo 共1996兲.
bility from a density-matrix treatment of the response. We would now like to look a bit more closely at the
They identified terms in the off-diagonal density-matrix structure of the laser-dressed eigenstates of a three-level
elements indicative of the interference, although they atom illustrated in Fig. 5. This discussion is intended to
did not explicitly consider the optical and nonlinear op- provide a simple physical picture that establishes the
tical effects in a dense medium. connection between the key ideas of EIT and that of
Our interest is in the case of electromagnetic fields in maximal coherence.
the optical frequency range, applied in resonance to the Within the dipole approximation the atom-laser inter-
states of a three-level atom. We illustrate the three pos- action Hint = ␮ · E is often expressed in terms of the Rabi
sible coupling schemes in Figs. 5 and 6. For consistency coupling 共or Rabi frequency兲 ⍀ = ␮ · E0 / ប, with E0 being
states are labeled so that the 兩1典-兩2典 transition is always the amplitude of the electric field E, and ␮ the transition
the dipole-forbidden transition. Of these prototype electronic dipole moment. After introducing the
schemes we shall be most concerned with the lambda rotating-wave approximation, we can represent the
configuration in Fig. 5, since the ladder and vee configu- Hamiltonian of the three-level atom interacting with a
rations illustrated in Fig. 6 are of more limited utility for coupling laser with real Rabi frequency ⍀c and a probe
the applications that will be discussed later. laser with Rabi frequency ⍀p 共Fig. 5兲 in a rotating frame
as
冤冥
0 0 ⍀p
ប
Hint = − 0 − 2共⌬ 1 − ⌬ 2 兲 ⍀c . 共1兲
2
⍀p ⍀c − 2⌬1
Here ⌬1 = ␻31 − ␻p and ⌬2 = ␻32 − ␻c are the detunings of

the probe and coupling laser frequencies ␻p and ␻c from
the corresponding atomic transitions.
A succinct way of expressing the eigenstates of the
FIG. 6. Ladder 共left兲 and vee-type 共right兲 three-level schemes. interaction Hamiltonian 共1兲 is in terms of the “mixing
These do not show EIT in the strict sense because of the ab- angles” ␪ and ␾ that are dependent in a simple way
sence of a 共meta兲stable dark state. upon the Rabi couplings as well as the single-photon

共⌬1 = ⌬兲 and two-photon 共␦ = ⌬1 − ⌬2兲 detunings 共see Fig. first demonstrated by Bergmann and co-workers 共Kuk-
5兲. For two-photon resonance 共⌬2 = ⌬1, or ␦ = 0兲 the mix- linski et al., 1989; Gaubatz et al., 1990兲, who termed this
ing angles are given by stimulated Raman adiabatic passage 共STIRAP兲. With the
system starting with all the amplitude in the ground state
⍀p 兩1典 the laser field at ␻c is first applied. This is similar to
tan ␪ = , 共2兲
⍀c the EIT situation since ⍀p Ⰶ ⍀c and the dark state in-
deed corresponds exactly to 兩1典 with negligible popula-
冑⍀p2 + ⍀2c tion in either 兩2典 or 兩3典. But now ⍀p is gradually 共i.e.,
tan 2␾ = . 共3兲 adiabatically兲 increased and at the same time ⍀c is
⌬
gradually decreased so that eventually ⍀c Ⰶ ⍀p and the
The eigenstates can then be written in terms of the bare dark state becomes 兩a0典 = −兩2典. This population transfer is
atom states: attained via the “maximally coherent” dark state when
⍀p = ⍀c, i.e., ␪ = ␲ / 4 in which the state takes the form
兩a+典 = sin ␪ sin ␾兩1典 + cos ␾兩3典 + cos ␪ sin ␾兩2典, 共4兲兩a0典 = 共1 / 冑2兲共兩1典 − 兩2典兲. A full description of this process
and the conditions for adiabaticity are given in the next
兩a0典 = cos ␪兩1典 − sin ␪兩2典, 共5兲 section.
兩a−典 = sin ␪ cos ␾兩1典 − sin ␾兩3典 + cos ␪ cos ␾兩2典. 共6兲
III. ATOMIC DYNAMICS AND OPTICAL RESPONSE
The reader’s attention is drawn to the following features:
While the state 兩a0典 remains at zero energy, the pair of The essential features of EIT and many of its applica-
states 兩a+典 and 兩a−典 are shifted up and down by an tions can be quantitatively described using a semiclassi-
amount ប␻±, cal analysis. This will be the focus of this section, where
we shall assume continuous-wave 共cw兲 classical fields in-
ប
ប␻± = 共⌬ ± 冑⌬2 + ⍀p2 + ⍀2c 兲. 共7兲 teracting with a single atom that can be modeled as a
2 lambda system 共Fig. 5兲. For the derivation of linear and
nonlinear susceptibilities, we concentrate on the steady-
The states 兩a±典 retain a component of all of the bare state solution of the atomic master equation. While a
atomic states, but in contrast state 兩a0典 has no contribu- master equation analysis is general and can be used for
tion from 兩3典 and is therefore the dark state, since if the the treatment of nonperturbative field effects, a single-
atom is formed in this state there is no possibility of atom wave-function approach is simpler and more illus-
excitation to 兩3典 and subsequent spontaneous emission. trative; we shall therefore use the latter to discuss the
It should be noted that the dark state will always be dressed-state interpretation of EIT. As many of the cen-
one of the possible states of the dressed system, but that tral features of EIT are related to optically thick media,
the details of the evolution of the fields can determine we shall discuss the light transmission of such a medium.
whether the atom is in this state 兩a0典, in 兩a±典, or an ad- The process of establishing EIT will then be examined
mixture. Evolution into the dark state via optical pump- and the connection to Raman adiabatic passage will be
ing 共through spontaneous decay from 兩3典兲 is one way to established. Finally, we shall discuss EIT inside optical
trap population in this state that is well known in laser resonators and briefly review some applications of EIT
spectroscopy and laser-atom manipulation. EIT offers that will not be expanded upon in the following sections
an alternative, adiabatic, and much more rapid route to of the review.
evolve into this state.
To see the origin of EIT using the dressed-state pic-
ture above, consider the case of a weak probe, that is, A. Master equation and linear susceptibility
⍀p Ⰶ ⍀c. In this case sin ␪ → 0 and cos ␪ → 1, which re-
sults in the dressed eigenstates shown in Fig. 4. The We consider an ensemble of identical atoms whose
ground state becomes identical to the dark state 兩a0典 dynamics can be described by taking into account only
= 兩1典 from which excitation cannot occur. Furthermore, three of its eigenstates. In the absence of electromag-
when the probe is on resonance 共⌬ = 0兲 then tan ␾ → 1 netic fields, all atoms are assumed to be in the lowest-
共i.e., ␾ = ␲ / 2兲, and then 兩a+典 = 共1 / 冑2兲共兩2典 + 兩3典兲 and 兩a−典 energy state 兩1典共Fig. 5兲. State 兩2典 has the same parity as
= 共1 / 冑2兲共兩2典 − 兩3典兲; these are the usual dressed states rel- 兩1典 and is assumed to have a very long coherence time.
evant to EIT in the limit of a strong-coupling field and a The highest-energy state 兩3典 is of opposite parity and has
weak probe. Conversely the picture of dark states show a nonzero electric dipole coupling to both 兩1典 and 兩2典. A
us that the EIT interference will survive in a lambda 共near兲 resonant nonperturbative electromagnetic field of
system in the limit of a pair of strong fields and so is frequency ␻c, termed the coupling field, is applied on the
relevant to the situation of maximal coherence. 兩2典-兩3典 transition. A probe field of frequency ␻p is applied
To ensure that a dark state is formed, an adiabatic on the 兩1典-兩3典 transition. EIT is primarily concerned with
evolution of the field is often adopted 共Oreg et al., 1984兲. the modification of the linear and nonlinear optical
The physically most interesting example of this involves properties of this—typically perturbative—probe field.
a so-called counterintuitive pulse sequence. This was We emphasize that most atomic systems have a subspace

of their state space that can mimic this simplified picture dependent exponentials arise from the conversion to the
when driven by near-resonant polarized electromagnetic Schrödinger picture. Assuming ␮13 = ␮13ẑ and ␮23 = ␮23ẑ,
fields. we let = Natom / V and obtain Pz共t兲 = P共t兲 as
The time-dependent interaction Hamiltonian in the
interaction picture that describes the atom-laser cou- P共t兲 = 关␮13␳31e−i␻31t + ␮23␳32e−i␻32t + c.c.兴. 共11兲
pling is We now focus on the perturbative regime in the probe
ប field and evaluate the off-diagonal density-matrix ele-
Hint = − 关⍀p共t兲␴ˆ 31ei⌬1t + ⍀c共t兲␴ˆ 32ei⌬2t + H.c.兴, 共8兲 ments ␳31共t兲, ␳32共t兲, and ␳12共t兲 to obtain P共t兲, or, equiva-
2
lently, the linear susceptibility ␹共1兲共−␻p , ␻p兲. Taking ␳11
where ⍀c共t兲 and ⍀p共t兲 denote the Rabi frequency asso- ⯝ 1 and using a rotating frame to eliminate fast expo-
ciated with the coupling and probe fields. ␴ˆ ij = 兩i典具j兩 is the nential time dependences, we find
atomic projection operator 共i , j = 1,2,3兲. This semiclassi-
i⍀cei⌬1t
cal Hamiltonian can be obtained from the fully quan- ␳32 = ␳12 ,
tized interaction-picture interaction Hamiltonian by re- 共␥32 + i2⌬2兲
placing annihilation and creation operators by c
numbers. i⍀cei⌬2t
␳12 = − ␳13 ,
The dynamics of laser-driven atomic systems are gov- ␥21 + i2共⌬2 − ⌬1兲
erned by the master equation for the atomic density op-
erator: i⍀pei⌬1t i⍀cei⌬2t
␳31 = + ␳21 . 共12兲
d␳ 1 ⌫31 共␥31 + i2⌬1兲共␥31 + i2⌬1兲
= 关Hint, ␳兴 + 关2␴ˆ 13␳␴ˆ 31 − ␴ˆ 33␳ − ␳␴ˆ 33兴
dt iប 2 As in Sec. I we define the single-photon detuning as ⌬
⌫32 = ⌬1 = ␻31 − ␻p and two-photon detuning as ␦ = ⌬1 − ⌬2
+ 关2␴ˆ 23␳␴ˆ 32 − ␴ˆ 33␳ − ␳␴ˆ 33兴 = ␻21 − 共␻p − ␻ − c兲. Keeping track of the terms that oscil-
2
late with exp关−i␻pt兴, we obtain
␥2deph
+ 关2␴ˆ 22␳␴ˆ 22 − ␴ˆ 22␳ − ␳␴ˆ 22兴兩␮13兩2
2 ␹共1兲共− ␻p, ␻p兲 =
⑀ 0ប
冋
␥3deph
+ 关2␴ˆ 33␳␴ˆ 33 − ␴ˆ 33␳ − ␳␴ˆ 33兴, 共9兲 4␦共兩⍀c兩2 − 4␦⌬兲 − 4⌬␥21 2
2 ⫻
兩兩⍀c兩2 + 共␥31 + i2⌬兲共␥21 + i2␦兲兩2
册
where the second and third terms on the right-hand side
describe spontaneous emission from state 兩3典 to states 兩1典 8␦2␥31 + 2␥21共兩⍀c兩2 + ␥21␥31兲
+i .
and 兩2典, with rates ⌫31 and ⌫32, respectively. We assume 兩兩⍀c兩2 + 共␥31 + i2⌬兲共␥21 + i2␦兲兩2
here that the thermal occupancy of the relevant radia- 共13兲
tion field modes is completely negligible. For optical fre-
quencies this approximation is easily justified. A de- The linear susceptibility given in Eq. 共13兲 contains
tailed derivation of this master equation is given by many of the important features of EIT. First, it must be
Cohen-Tannoudji et al. 共1992兲. We have also introduced mentioned that Eq. 共13兲 predicts Autler-Townes splitting
energy-conserving dephasing processes with rates ␥3deph of an atomic resonance due to the presence of a nonper-
and ␥2deph; the latter determines one of the fundamental turbative field. There is more, however, in this expres-
time scales for practical EIT systems, as we shall see sion than the modification of absorption due to the ap-
shortly. For convenience, we define the total spontane- pearance of dressed atomic states: in particular, for two-
ous emission rate out of state 兩3典 as ⌫3 = ⌫31 + ⌫32. The photon Raman resonance 共␦ = 0兲, both real and
coherence decay rates are defined as ␥31 = ⌫3 + ␥3deph, imaginary parts of the linear susceptibility vanish in the
␥32 = ⌫3 + ␥3deph + ␥2deph, and ␥21 = ␥2deph. ideal limit of ␥21 = 0. As depicted in Fig. 7, this result is
The polarization generated in the atomic medium by independent of the strength of the coherent coupling
the applied fields is of primary interest, since it acts as a field. It should be noted that changing the Rabi fre-
source term in Maxwell’s equations and determines the quency of the coupling laser only changes the spectral
electromagnetic field dynamics. The expectation value profile of absorption, the integral of Im关␹兴 over ⌬ is con-
of the atomic polarization is served as ⍀c is varied. In the limit 兩⍀c兩 ⬎ ␥31, the absorp-
tion profile carries the signatures of an Autler-Townes
ជ 共t兲 = − 兺具er 典/V
P doublet. Important features in absorption, such as van-
i
i ishing loss at ␦ = 0 and enhanced absorption on the low-
Natom and high-energy sides of the doublet, become evident on
= 关␮13␳31e−i␻31t + ␮23␳32e−i␻32t + c.c.兴. 共10兲 a closer inspection. For 兩⍀c兩 Ⰶ ␥31, we obtain a sharp
V
transmission window with a linewidth much narrower
To obtain the expression in the second line, we assumed than ␥31. In this latter case, it becomes apparent that the
that all Natom atoms contained in the volume V couple modifications in the linear susceptibility call for an ex-
identically to the electromagnetic fields. The time- planation based on quantum interference phenomena

FIG. 7. EIT absorption spectrum for different values of cou-

pling field and ␥21 = 0: 共a兲 ⍀c = 0.3␥31; 共b兲 ⍀c = 2␥31.
rather than a simple line splitting. Before proceeding

with this discussion, we highlight some of the other im-
portant properties that emerge in Eq. 共13兲.
First and foremost it must be noted that the possibility
of eliminating absorption in an otherwise optically thick
medium has been demonstrated in several laboratories
in systems ranging from hot atomic vapor cells to mag-
netically trapped Bose-condensed atoms. Figure 8 shows
the absorption profile obtained in the first experiment to
demonstrate the effect, carried out by Boller and co- FIG. 8. First experimental demonstration of EIT in Sr vapor
workers 共Boller et al., 1991兲. The application of a by Boller et al., 1991: top, transmission through cell without
coupling field; bottom, with coupling field on. From Boller et
coupling field opens up a transparency window in an
al., 1991.
otherwise completely opaque atomic cell. This experi-
ment was carried out using an autonionizing state 兩3典
of strontium with ␥31 ⬃ 2 ⫻ 1011 s−1. The decay rate of the cient, implying that ultraslow group velocities for light
lower state coherence ␥21 ⬃ 4 ⫻ 109 s−1 is determined pulses can be obtained in transparent media, which will
by the collisional broadening for an atomic density of be discussed extensively in the following section.
⬃ 5 ⫻ 1015 cm−3. Typically, observation of coherent phenomena in
The transparency obtained at two-photon resonance atomic gases is hindered by dephasing due to collisions
is independent of the detuning of the probe field from and laser fluctuations as well as inhomogeneous broad-
the bare 兩1典-兩3典 transition 共⌬兲. As ⌬ increases, however, ening due to the Doppler effect. While recent EIT ex-
the distance between the frequencies where one obtains periments have been carried out using ultracold atomic
transparency and maximum absorption becomes smaller, gases driven by highly coherent laser fields, where Eq.
thereby limiting the width of the transparency window. 共9兲 provides a perfect description of the atomic dynam-
At the same time, the transmission profile and the asso- ics, it is important to analyze the robustness of EIT
ciated dispersion become highly asymmetric for ⌬ ⬎ ␥31. against these nonideal effects.
Figure 9 shows contour plots of Im关␹共1兲兴 as a function of In actual atomic systems, the dephasing rate of the
the detunings ⌬1 and ⌬2 of the two fields as well as the forbidden 兩1典-兩2典 transition is nonzero due to atomic col-
single-photon 共⌬兲 and two-photon 共␦兲 photon detunings. lisions. All the important features of EIT remain observ-
It is evident that for large detuning ⌬2 of the coupling able even when ␥21 ⫽ 0, provided that the coupling field
field the absorption spectrum is essentially that of a two- Rabi frequency satisfies
level system with an additional narrow Raman peak 兩⍀c兩2 Ⰷ ␥31␥21 . 共14兲
close to the two-photon resonance. Exactly at the two-
photon resonance point ⌬1 = ⌬2 the absorption vanishes Figure 10 shows the imaginary part of ␹共1兲共−␻p , ␻p兲 for
independent of the single-photon detuning. ␥21 = 0 关Fig. 10共a兲兴, ␥21 = 0.1␥31 关Fig. 10共b兲兴, and ␥21
In the limit ⌬2 = 0 and for ⍀c Ⰶ ␥31, the real part of the = 10␥31 关Fig. 10共c兲兴. In all plots, we take ⍀c = 0.5␥31. We
susceptibility seen by a probe field varies rapidly at reso- first note from Fig. 10共b兲 that the absorption profile ap-
nance 共⌬ ⬃ 0兲. In contrast to the well-known 共single兲 pears virtually unchanged with respect to the ideal case,
atomic resonances, the enhanced dispersion in the EIT provided that the inequality 共14兲 is satisfied. The trans-
system is associated with a vanishing absorption coeffi- parency is no longer perfect, however, and the residual

FIG. 9. Contour plot of imagi-

nary part of susceptibility
Im关␹共1兲兴: left, as a function of
detunings ⌬1 and ⌬2; right, as a
function of single-photon de-
tuning ⌬ and two-photon de-
tuning ␦ in units of ␥31. White
areas correspond to low absorp-
tion, dark to large absorption.
The insensitivity of the induced
transparency at ␦ = 0 on the
single-photon detuning is ap-
parent.
absorption due to ␥21 provides a fundamental limit for Amplitude or phase fluctuations of the nonperturba-
many of the EIT applications. For the ␥21 Ⰷ ␥31 case de- tive coupling laser can have a detrimental effect on the
picted in Fig. 10共c兲, the absorption minimum is absent: in observability of EIT. Instead of trying to present a gen-
this limit a constructive interference enhances the ab- eral analysis of the effect of coupling laser linewidth on
sorption coefficient in between the two peaks. linear susceptibility, we discuss several special cases that
The linear susceptibility given in Eq. 共13兲 depends are of practical importance. When the laser linewidth is
only on the total coherence decay rates, and not on the due to phase fluctuations that can be modeled using the
population decay rates of the atomic states. As a conse- Wiener-Levy phase-diffusion model, it can be shown
quence, the strong quantum interference effects de- that the resulting coupling laser linewidth directly con-
picted in Figs. 3共a兲–3共c兲 are observable even in systems tributes to the coherence decay rates ␥21 and ␥32 共Ima-
where collisional broadening dominates over lifetime moglu, 1991兲. In contrast, for a coupling field with large
broadening. This implies that EIT effects can be ob- amplitude fluctuations, the absorption profile could be
served in dense atomic gases, or even solids, provided smeared out. In applications where a nonperturbative
that there is a metastable transition with a relatively probe field is used, fluctuations in both laser fields are
long dephasing rate that satisfies ␥21 Ⰶ ␥31. EIT in such a important. If, however, the two lasers are obtained from
collisionally broadened atomic gas was first reported in the same fluctuating laser source using electro-optic or
1991 共Field et al., 1991兲. acousto-optic modulation, then EIT is preserved to first
order.
Doppler broadening is ubiquitous in hot atomic gases.
If the lambda system is based on two hyperfine-split
metastable states 兩1典 and 兩2典, then the Doppler broaden-
ing has no adverse effect on EIT, provided that one uses
copropagating probe and coupling fields. In other cases,
the susceptibility given in Eq. 共13兲 needs to be integrated
over a Gaussian density of states corresponding to the
Gaussian velocity distribution of the atoms. Qualita-
tively, the presence of two-photon Doppler broadening
with width ⌬␯Dopp ⬎ ⍀c will wash out the level splitting
and the interference. EIT can be recovered by increas-
ing the coupling field intensity so as to satisfy ⍀c
⬎ ⌬␯Dopp; in this limit we obtain
⌬␯Dopp
2
␥31
Im关␹共1兲共− ␻p, ␻p兲兴␻p=␻31 ⬀ . 共15兲
⍀4c
This strong dependence on reciprocal ⍀c is a direct con-

sequence of robust quantum interference. We emphasize
that in this large coupling field limit, we can consider ⍀c
as the effective detuning from dressed-state resonances.
As a result of quantum interference, this effective detun-
ing can be used to suppress absorption much more
FIG. 10. Absorption spectrum for nonvanishing decay of 兩2典-兩1典 strongly than in noninterfering systems, where the ab-
transition in arbitrary units: 共a兲 ␥31 = 1, ␥21 = 0; 共b兲 ␥31 = 1, ␥21 sorption coefficient is only proportional to the inverse
= 0.1; 共c兲 ␥31 = 1, ␥21 = 10. In all cases ⍀c = 0.5. detuning squared.

B. Effective Hamiltonian and dressed-state picture Since the population in state 兩3典 is zero, there is no spon-
taneous emission or light scattering and hence no ab-
It is well known in quantum optics that for a given sorption. In the limit of a perturbative probe field where
master equation in the so-called Lindblad form, one can 兩a0典 ⬃ 兩1典, we have an alternative way of understanding
obtain an equivalent stochastic wave-function descrip- the vanishing amplitude in 兩3典: the atom has two ways of
tion of the dynamics of the system based on the evolu- reaching the dissipative state 兩3典—either directly from
tion via a non-Hermitian effective Hamiltonian and state 兩1典 or via the path 兩1典-兩3典-兩2典-兩3典. The latter has com-
quantum-jump processes 共Dalibard et al., 1992; Gardiner parable amplitude to the former, since ⍀c is nonpertur-
et al., 1992; Carmichael, 1993兲. Further simplification of bative. As a result, the amplitudes for these paths can
the description of the dynamics is obtained when the exhibit strong quantum interference.
probability of a quantum-jump process is negligibly We can alternatively view the atomic system in a dif-
small: in this limit, the nonunitary wave-function evolu- ferent basis—one that would have yielded a diagonal
tion induced by the effective Hamiltonian captures most Hamiltonian matrix had the dissipation and the probe
of the essential physics and can be used to calculate lin- coupling been vanishingly small. This dressed-state basis
ear and nonlinear susceptibilities. This simplification ap- is obtained by applying the unitary transformation on
plies for an atomic system in which the ground state the 共bare-basis兲 state vector 兩⌿共t兲典:
共which has near-unity occupancy兲 is coupled to excited
states via weak electromagnetic fields. In the case of 兩⌿d共t兲典 = U兩⌿共t兲典, 共18兲
EIT, this is exactly the case when atoms in state 兩1典 are where
driven by a weak probe field, independent of the
冤冥
strength of the coupling field. 1 0 0
Our starting point is the Hamiltonian of Eq. 共8兲. To U = 0 cos ␾ sin ␾ . 共19兲
incorporate the effect of radiative decay out of state 兩3典,
0 − sin ␾ cos ␾
we introduce an anti-Hermitian term, −iប⌫3␴33 / 2, which
reproduces the correct decay terms in the master equa- Here ␾ is defined by tan共2␾兲 = ⍀c / ⌬2. Application of this
tion. Conversely, the physics that this 共combined兲 non- transformation to Heff yields
冤冥
Hermitian Hamiltonian fails to capture is the repopula-
0 ⍀p sin ␾ ⍀p cos ␾
tion of atomic states due to spontaneous emission down
to states 兩1典 and 兩2典. Since we cannot describe pure i
ប ⍀p sin ␾ − 2⌬− + i⌫− − ⌫+−
dephasing processes using a wave-function model, we H̃eff = − 2 ,
can use the population decay of state 兩2典共at rate ⌫2兲 to 2
i
mimic dephasing. We therefore introduce an additional ⍀p cos ␾ − ⌫+− − 2⌬+ + i⌫+
anti-Hermitian decay term to obtain an effective Hamil- 2
tonian in the rotating frame 共20兲
ប
冉
i
Heff = − 关⍀p␴ˆ 31 + ⍀c␴ˆ 32 + H.c.兴 + ប ⌬ − ⌫3 ␴ˆ 33
2 2
冊 where ⌬− and ⌬+ denote the detunings of the probe field
from the dressed resonances, ⌫− = 共⌫2 cos2 ␾ + ⌫3 sin2 ␾兲,
⌫+ = 共⌫2 sin2 ␾ + ⌫3 cos2 ␾兲, and ⌫+− = 共⌫2 − ⌫3兲sin 2␾. We
冉 i
冊
+ ប ␦ − ⌫2 ␴ˆ 22 .
2
共16兲
can calculate the absorption rate of the probe photons
by determining the eigenvalues of H̃eff in the limit of
We reemphasize that for ⌫2 = 0 and ␦ = 0, one of the perturbative ⍀p. The imaginary part of the eigenvalue
eigenstates of Heff is 关Im共−␭1兲兴 that corresponds to 兩1典 as ⍀p → 0 for ⌫2 = 0 is
− ⍀p2 ⌫3共⌬+ sin2 ␾ + ⌬− cos2 ␾兲2
1 Im共␭1兲 =
兩a0典 = 关⍀c兩1典 − ⍀p兩2典兴, 共17兲 4⌬−2⌬+2 + 共⌬+⌫3 sin2 ␾ + ⌬−⌫3 cos2 ␾兲2
⍀2c + ⍀p2
⬀ Im关␹共1兲共− ␻p, ␻p兲兴. 共21兲
with an eigenenergy ប␻0 = 0 independent of the values of
the Rabi frequencies and the single-photon detuning. While this expression contains the same information as
We here have assumed real-valued Rabi frequencies, Eq. 共13兲, appearance of the square of the sum of two
which are always possible as long as propagation effects amplitudes in the numerator of Eq. 共21兲 makes the role
are not considered. In the time-dependent response, we of quantum interference clearer.
shall see that the emergence of EIT can be understood H̃eff given in Eq. 共20兲 has exactly the same form as the
as a result of the system’s moving adiabatically from one describing Fano interference of two autoionizing
bare state 兩1典 in the absence of fields to 兩a0典 in the pres- states 共Fano, 1961; Harris, 1989兲. It is the presence
ence of the fields. of imaginary off-diagonal terms ␬23 = ␬32 = 0.5共⌫3
The structure of this dark eigenstate provides us the − ⌫2兲sin 2␾ that gives rise to quantum interference in ab-
simplest picture with which we can understand coherent sorption. When ⌫2 = 0, we find that ␬23 = 冑⌫2d⌫3d, where
population trapping and EIT. The atomic system under ⌫2d = ⌫2 cos2 ␾ + ⌫3 sin2 ␾ and ⌫3d = ⌫2 sin2 ␾ + ⌫3 cos2 ␾.
the application of the two laser fields satisfying ␦ = 0 In this case, we find that the interference in absorption is
moves into 兩a0典, which has no contribution from state 兩3典. destructive when the probe field is tuned in between the

two dressed-state resonances, with a perfect cancellation sfg

ing Heff and has two contributions: the first contribution
for ␦ = 0. When ⌫3 = 0, we have ␬23 = −冑⌫2d⌫3d; in this proportional to ⍀p simply gives us the same linear sus-
case, the quantum interference is constructive for probe ceptibility we calculated in the previous subsection using
fields tuned in between the two dressed states and de- the master-equation approach. The second contribution
structive otherwise. Even though absorption is sup- proportional to ⍀eff⍀c ⬀ ⍀a⍀b⍀c gives us the nonlinear
pressed for 兩␦兩 Ⰷ ⍀c, we never obtain complete transpar- susceptibility
ency or a steep dispersion in this 共⌫3 = 0兲 case. Finally, for
␹共3兲共− ␻p, ␻a, ␻b, ␻c兲
⌫2 = ⌫3, the imaginary off-diagonal terms vanish; in this
limit, there are no interference effects. These results 2␮23␮31 1
were already depicted in Fig. 10. =
3⑀0ប ⍀c + 共␥31 + i2⌬兲共␥21 + i2␦兲
3 2
冋册
The complete elimination of absorption for ␦ = 0 in the
limit ⌫2 = 0 can therefore be understood by invoking the 1 1
⫻ 兺 ␮1i␮i2 + . 共25兲
arguments used by Fano 共1961兲. Let us assume that the i ␻i1 − ␻a ␻i1 − ␻b
decay of state 兩3典 is due to spontaneous emission into
state 兩f典. For an atom initially in state 兩1典, a probe photon This third-order nonlinear susceptibility that describes
absorption event results in the generation of a 共Raman-兲 sum-frequency generation highlights one of the most im-
portant properties of EIT, namely, that the nonlinear
scattered photon and leaves the atom in state 兩f典. To
susceptibilities need not vanish when the linear suscep-
reach this final state, the atom can virtually excite either
tibility vanishes due to destructive quantum inter-
of the two dressed states; since excitations are virtual,
the probability amplitudes corresponding to each have ference. In fact, for ⍀2c ⬍ ⌫23 / 2, we notice that
to be added, leading to quantum interference. State 兩f典 ␹共3兲共−␻p , ␻a , ␻b , ␻c兲 has a maximum at ␦ = 0 where
could either be a fourth atomic state not coherently ␹共1兲共−␻p , ␻p兲 = 0.
coupled to the others or it could be the state 兩1典 itself. In It is perhaps illustrative to compare ␹共3兲 obtained for
the latter case, the absorption process is Rayleigh scat- the EIT system with the usual third-order nonlinear
tering of probe photons. susceptibility for nonresonant sum-frequency genera-
Having seen the simplicity of the analysis based on tion. As expected, the expressions in the two limits
effective non-Hermitian Hamiltonians, we next address have the same form, with the exception of detunings:
the question of nonlinear susceptibilities. Since the latter the two- and three-photon detunings in the case of
are by definition calculated in the limit of near-unity nonresonant susceptibility 共1 / ⌬␻2⌬␻3兲 are replaced by
ground-state occupancy, Heff of Eq. 共16兲 provides a natu- 1 / 共⍀2c + ␥21␥31兲 in the case of EIT with ␦ = ⌬ = 0.
ral starting point. In this section, we shall focus on sum-
frequency generation to highlight the different role C. Transmission through a medium with EIT
quantum interference effects play in nonlinear suscepti-
bilities as compared to linear ones. To this end, we in- Many of the interesting applications of EIT stem from
troduce an additional driving term ⍀eff that couples the spectroscopic properties of dark resonances, i.e., nar-
states 兩1典 and 兩2典 coherently. Since we have assumed this row transmission windows, in otherwise opaque media.
transition to be dipole forbidden, we envision that this It is therefore worthwhile to consider the specifics of the
effective coupling is mediated by two laser fields 共at fre- EIT transmission profile in more detail. In an optically
quencies ␻a and ␻b兲 and a set of intermediate 共nonreso- thick medium it is not the susceptibility ␹共1兲共−␻p , ␻p兲 it-
nant兲 states 兩i典: self that governs the spectroscopic properties but the
⍀eff = 兺
i
⍀ a⍀ b
2
冋 1
+
1
␻i1 − ␻a ␻i1 − ␻b
. 册共22兲
collective response of the entire medium, characterized
by the amplitude transfer function
T共␻p,z兲 = exp兵ikz␹共1兲共− ␻p, ␻p兲/2其, 共26兲
Here ⍀a = ␮1iEa / ប and ⍀b = ␮i2Eb / ប.
We can now rewrite the effective Hamiltonian for with k = 2␲ / ␭ the resonant wave number and z the me-
sum-frequency generation as dium length. In contrast to absorbing resonances, the
width of the spectral window in which an EIT medium
ប appears transparent decreases with the product of nor-
sfg
Heff = − 关⍀eff␴ˆ 21 + ⍀p␴ˆ 31 + ⍀c␴ˆ 32 + H.c.兴
2 malized density and medium length. Substituting the
冉冊冉冊
susceptibility of an ideal EIT medium with resonant
i i control field into Eq. 共26兲, one finds that the transmittiv-
+ ប ⌬ − ⌫3 ␴ˆ 33 + ប ␦ − ⌫2 ␴ˆ 22 . 共23兲
2 2 ity near resonance ␦ = 0 is a Gaussian with width
To obtain the nonlinear susceptibility, we assume both ⍀2c 1
⍀eff and ⍀p to be perturbative. The generated polariza- ⌬␻trans = 共␴z Ⰷ 1兲. 共27兲
冑⌫31␥31 冑␴z
tion at ␻p is given by
Here ␴ = 3␭2 / 2␲ is the absorption cross section of an
P共t兲 = ␮13a3共t兲exp共− i␻pt兲 + c.c., 共24兲
atom and is the atom number density. One recognizes
where a3共t兲 is the probability amplitude for finding a a power broadening of the resonance with increasing
given atom in state 兩3典. This amplitude is calculated us- Rabi frequency ⍀c or intensity of the coupling field,

acteristic of EIT, which distinguishes it from coherent

population trapping, is the preparation which in EIT
happens automatically through Raman adiabatic pas-
sage. This process naturally becomes important if an op-
tically thick medium is considered.
The formation of a dark state can happen either via
optical pumping, i.e., by an incoherent process, or via a
coherent preparation scheme. The first mechanism is ob-
viously required if the atomic ensemble is initially in a
mixed state. Optical pumping may not, however, be well
suited to achieve electromagnetically induced transpar-
ency in optically thick media. Here radiation trapping
can lead to a dramatic slowdown of the pumping pro-
cess. If the dark state has a finite lifetime, the resulting
effective pump rate may become even too slow to
achieve transparency at all 共Fleischhauer, 1999兲. Further-
more, the preparation by optical pumping is always as-
sociated with a nonrecoverable loss of photons from the
probe field, which could be fatal if the probe field were a
few-photon pulse.
The problems associated with radiation trapping and
FIG. 11. Transmission spectrum through an EIT medium: top, preparation losses can be avoided if the atoms are ini-
dependence on density length products ␴z; bottom, illustra-
tially in a pure state, e.g., state 兩1典. If 兩1典 is a nondegen-
tion of dispersive width.
erate ground state with a sufficient energy gap to state
兩2典, this is in general fulfilled. Under certain conditions
which can, however, be partially compensated for by in- the dark state is in this case prepared by a coherent
creasing the density-length product ␴z. This important mechanism that does not involve spontaneous emission
property is illustrated in Fig. 11, where a finite decay at all. The process that achieves this is called stimulated
rate ␥21 of the coherence between the two lower levels Raman adiabatic passage 共STIRAP兲 and was discovered
was taken into account as well. and developed by Bergmann et al. 共see also Kuklinski et
The perfect linear dispersion near the EIT resonance, al., 1989 and Gaubatz et al., 1990兲 after some earlier
expressed by the linear dependence of the real part of related but independent discoveries of Eberly and co-
␹ on ␦, can be used to detect shifts of the EIT resonance workers 共Oreg et al., 1984兲. In the STIRAP process rel-
by phase-sensitive detection. Let us define ⌬␻dis as evant here, the atoms are returned to the initial state
the frequency variation associated with an accumulated after the interaction with the pulsed probe field, thus
phase shift of 2␲ for the propagation through the me- conserving the number of photons in the pulse. The
dium. This is given by STIRAP technique is now a widely used method in
atomic and molecular physics for the preparation of spe-
⍀2c 1 cific quantum states and for controlling chemical reac-
⌬␻disp = 2␲ 共␴z Ⰷ 1兲, 共28兲
⌫31 ␴z tions. It also has a large range of applications in quan-
tum optics and matter-wave interferometry. For recent
much smaller than ⌬␻trans. This effect, which is also il- reviews on this subject see Bergmann et al. 共1998兲 and
lustrated in Fig. 11, plays an important role for applica- Vitanov et al. 共2001兲.
tions of EIT in optical detection of magnetic fields In STIRAP, the two coherent fields coupling the states
共Scully and Fleischhauer, 1992; Fleischhauer and Scully, of the three-level lambda system are considered to be
1994兲. time dependent and in two-photon resonance. In the ba-
The narrowing of the transparency and dispersive sis of the bare atomic states 兵兩1典,兩2典,兩3典其 the correspond-
width of an EIT medium with increasing density was ing time-dependent interaction Hamiltonian in the
experimentally observed by Lukin et al. 共1997兲 in a for- rotating-wave approximation and in a proper rotating
ward four-wave mixing experiment. frame is
冤冥
0 0 ⍀p共t兲
D. Dark-state preparation and Raman adiabatic passage ប
H共t兲 = − 0 0 ⍀c共t兲 . 共29兲
2
In the preceding sections we have discussed EIT un- ⍀p共t兲 ⍀c共t兲 − 2⌬
der steady-state conditions. We have seen that atoms in
a particular superposition state, the dark state 兩a0典, can We assume that the relative phase between the two
be decoupled from the interaction. If there is two- fields is arbitrary but constant and thus both ⍀p and ⍀c
photon resonance between atoms and fields, this state is can be taken real. This also implies that the two fields
stationary. Properly prepared atoms remain in this state are assumed to be transform-limited pulses, or more
and render the medium transparent. An important char- precisely that the beat note between the two is trans-

form limited. The essence of STIRAP is that a proper

adiabatic change of ⍀p共t兲 and ⍀c共t兲 allows a complete
transfer of population from the initial state 兩1典 to the
冏冓冏 d
dt
冏冔冏
a0共t兲 a±共t兲 Ⰶ 兩␻±共t兲兩, 共32兲
target state 兩2典 or vice versa without populating the in- which for the case of single-photon resonance ⌬ = 0 sim-
termediate excited state 兩3典 and thus without spontane- plifies to
ous emission.
The underlying mechanism can most easily be under- ⍀共t兲 ⬅ 冑⍀p2 共t兲 + ⍀2c 共t兲 Ⰷ 兩␪˙ 共t兲兩. 共33兲
stood if the time-dependent interaction Hamiltonian in
the rotating-wave approximation is written in the basis In lowest order of the nonadiabatic coupling ␪˙ and in the
of its instantaneous eigenstates, given in Eqs. 共4兲–共6兲, limit 兩␪˙ 兩 Ⰶ ⌫ , characteristic for long pulses, the excited-
3
with the corresponding eigenvalues given in Eq. 共7兲. The state population is given by 兩␪˙ 兩2 / ⍀2共t兲, such that spon-
instantaneous eigenstate taneous emission leads to an instantaneous loss rate
共Fleischhauer and Manka, 1996; Vitanov and Stenholm,
⍀p共t兲 1997兲
兩a0共t兲典 = cos ␪共t兲兩1典 − sin ␪共t兲兩2典, tan ␪ = 共30兲
⍀c共t兲
兩␪˙ 共t兲兩2
is the adiabatic dark state, which has no overlap with the ⌫eff = ⌫3 . 共34兲
⍀2共t兲
excited state 兩3典 and thus does not lead to spontaneous
emission. If the pump and coupling pulses are not transform
An important property of the adiabatic dark state is limited in such a way that the beat-note phase changes
that, depending on the value of the mixing angle ␪共t兲, it in time 共Dalton and Knight, 1982兲, there will be addi-
can coincide with either one of the bare states 兩1典 and 兩2典 tional nonadiabatic losses due to this effect.
as well as any intermediate superposition with fixed rela- If all atoms are initially in state 兩1典, the coupling field
tive phase. If ␪ = 0, which corresponds to a coupling field 共⍀c兲 is applied before the probe field 共⍀p兲, and if the
much stronger than the pump field, 兩a0典 = 兩1典. Likewise, if characteristic time change is slow on a scale set by the
␪ = ␲ / 2 we have 兩a0典 = −兩2典. Thus if all atoms are initially rms Rabi frequency, the fields prepare the dark state of
共i.e., for t = −⬁兲 prepared in state 兩1典 and 兩⍀c共−⬁兲兩 EIT by stimulated Raman adiabatic passage. If the two
Ⰷ 兩⍀p共−⬁兲兩, i.e., if the coupling pulse is applied first, the fields are constant, the dark state is also self-prepared by
the lasers, but this takes a finite amount of preparation
system is in the dark state at t = −⬁. If the mixing angle
energy from the probe field and transfers it into the me-
␪共t兲 is now rotated from 0 to ␲ / 2, the dark state changes
dium and the coupling field. This preparation energy
from 兩1典 to −兩2典. If this rotation is sufficiently slow, the
was first determined by Harris and Luo 共1995兲. It is
adiabatic theorem guarantees that the state vector of the
given by the number of atoms in the probe path N and
system follows this evolution. In this way population can
the stationary values of the Rabi frequencies,
be transferred with 100% efficiency and without
spontaneous-emission losses. The only requirements be- 兩⍀p兩2
sides the correct initial conditions are that the time rate Eprep = ប␻pN . 共35兲
兩⍀c兩2
of change of the fields be sufficiently slow, i.e., adiabatic,
and that the relative phase between the fields be at least After the preparation energy is transferred to the atoms
approximately constant. and the coupling field, the medium remains transparent.
To see how slow the change of the fields has to be to On the other hand, if the probe field is also switched off
stay within the adiabatic approximation, we transform before the coupling field, the dark state eventually re-
the Hamiltonian matrix 共29兲 for the state amplitudes in turns to the initial state 兩1典 and the preparation energy is
the bare basis into the basis of instantaneous eigenstates. transferred back from the medium and the coupling field
Since the transformation matrix R is explicitly time de- to the probe field. In this case there is no net loss of
pendent for the case of time-dependent fields, we find energy from the probe field but only a temporary trans-
H̃ = R−1HR − R−1Ṙ, or, explicitly, fer. As we shall see later in Sec. IV, this temporary trans-
fer of energy has some interesting consequences for the
冤冥
i˙ propagation of the fields.
␻+共t兲 ␪ 0
2
i i E. EIT inside optical resonators
H̃ = − ប − ␪˙ 0 − ␪˙ , 共31兲
2 2
The large linear dispersion associated with EIT in an
i˙
0 ␪ ␻−共t兲 optically thick medium can substantially affect the prop-
2 erties of a resonator system, especially since the medium
dispersion can easily exceed that of an empty cavity. It is
where an overdot means a time derivative. Adiabatic therefore natural to investigate the possibility of having
evolution occurs if the off-diagonal coupling, propor- an EIT medium inside a cavity, such that one of the
tional to ␪˙ , is sufficiently small compared to the eigen- high-finesse modes replaces the probe field and couples
values 兩␻±兩共7兲: the ground state 兩1典 to state 兩3典.

If an empty-cavity resonance ␻c0 is sufficiently close

to the frequency ␻0 of electromagnetically induced
transparency, the presence of the medium will lead to a
very strong pulling of the resonance frequency ␻c of the
combined atom-cavity system towards this value 共Lukin,
Fleischhauer, et al., 1998兲. At the same time the reso-
nance width is substantially reduced 共Lukin, Fleis-
chhauer, et al., 1998兲. To see this let us consider the re-
sponse function of a resonator containing an EIT
medium, i.e., the ratio of circulating to input intensity at
frequency ␻. For simplicity we assume a unidirectional
ring resonator of round-trip length L with a single in-out
coupling mirror. Then
Icirc共␻兲 t2
S共␻兲 ⬅ = , 共36兲
Iin共␻兲 1 + r2␬2 − 2r␬ cos关⌽共␻兲兴
where t and r are the amplitude transmittivity and re-
flectivity of the in-out coupling mirror. In the case of a
lossless mirror, r2 + t2 = 1. ⌽共␻兲 = ␻ / c共L + l Re关␹共1兲兴兲 is the FIG. 12. Experimental demonstration of linewidth narrowing:
phase shift and ␬ = exp兵−␻l Im关␹共1兲兴 / c其 the medium ab- 共a兲 transmission spectrum of empty cavity and with EIT me-
sorption per round trip L. l is the length of the EIT dium; 共b兲 magnification of transmission spectrum with EIT.
medium. Close to two-photon resonance, Re关␹共1兲兴 From Müller et al., 1997.
⬇ 共dn / d␻兲共␻ − ␻0兲. This yields for the resonance of the
cavity+ medium system
laser fields to control the refractive index has created
1 ␰ dn l much attention. A suggestion for controlling phase
␻c = ␻c0 + ␻ 0, where ␰ = ␻0 . 共37兲
1+␰ 1+␰ d␻ L matching in nonlinear mixing through the action of an
additional field was made by Tewari and Agarwal 共1986兲,
One recognizes a strong frequency pulling towards the but it is since the advent of EIT that these ideas have
atomic EIT resonance since the dispersion ␻0dn / d␻ can been extensively studied. Harris treated the refractive
be rather large. Likewise one finds from Eq. 共36兲 that properties of an EIT medium theoretically 共Harris et al.,
the width of the cavity resonance is changed according 1992兲. In pulsed-laser experiments the vanishing of the
to susceptibility at resonance leads to the elimination of
⌬␻c 1 − r␬ 1 distortion for pulses propagating through a very opti-
= . 共38兲 cally thick medium at resonance 共Jain et al., 1995兲. Re-
⌬␻c0 冑␬共1 − r兲 1 + ␰
lated suggestions by Harris for the use of strong off-
The first term accounts for a small enhancement of the resonant fields to control the refractive index 共Harris,
cavity width due to the additional losses induced by the 1994b兲 have been important in the implementation of
medium. Much more important, however, is the second maximal coherence frequency-mixing schemes with off-
term, which describes a substantial reduction of the cav- resonant lasers.
ity linewidth due to the linear dispersion of the medium. Extensive theoretical work was carried out by Scully
Since the EIT resonance depends on the atomic sys- and co-workers 共Scully, 1991; Fleischhauer et al., 1992a,
tem as well as on the coupling laser, frequency pulling 1992b; Scully and Zhu, 1992; Rathe et al., 1993兲 and oth-
and line narrowing are relative to the coupling laser. ers 共Wilson-Gordon and Friedman, 1992兲 on the use of
These effects have been observed experimentally by resonant cw fields for modification of the refractive in-
Müller et al. 共1997兲. Figure 12 shows the transmission dex. The experiments of several workers have exten-
spectrum of an empty cavity and a cavity with an EIT sively studied refractive index modifications in this cw
medium. Applications for difference-frequency locking limit 共Moseley et al., 1995; Xiao et al., 1995兲.
and stabilization have been suggested and the reduc- Working in Doppler-free conditions with the narrow
tion of the beat-note linewidth in a laser system bandwidths attainable through use of cw lasers it is pos-
below the Shawlow-Townes limit predicted 共Lukin, sible to observe very steep EIT-induced dispersion pro-
Fleischhauer, et al., 1998兲. files. This has led to the suggestion that very sensitive
interferometers and magnetometers be used, since tiny
variations in frequency couple to large changes in refrac-
F. Enhancement of refractive index, magnetometry, and tive index 共Scully and Fleischhauer, 1992; Fleischhauer
lasing without inversion and Scully, 1994; Budker et al., 1998, 1999; Nagel et al.,
1998; Sautenkov et al., 2000; Stahler et al., 2001; Af-
We now undertake a brief survey of some applications folderbach et al., 2002兲. In the magnetometer schemes
of EIT and related phenomena which will not be dis- this greatly increases the sensitivity of the measurement
cussed in detail in the following sections. The use of of the Zeeman shifts induced by a magnetic field.

As mentioned earlier, lasing without inversion is

closely related to the interference important in EIT.
There have by now been a number of experimental veri-
vph ⬅ 冏冏 ␻p
kp ␦=0
=
c
n
= c. 共41兲
fications of this concept with the actual realization of An important property of the EIT system is that the
laser action without population inversion in the visible second-order term in Eq. 共39兲 vanishes exactly if there is
spectral range including several reports of inversionless also single-photon resonance ⌬ = 0 of the probe field. As
amplification 共Gao et al., 1992; Fry et al., 1993; Nottle- a consequence there is no group velocity dispersion, i.e.,
man et al., 1993兲 as well as actual demonstrations of laser no wave-packet spreading. Using Eq. 共39兲 yields at two-
oscillation without population inversion 共Zibrov et al., photon resonance
1995; Padmabandu et al., 1996兲. An application of lasing
without inversion to short-wavelength lasing is yet to be c ⌫31
vgr = with ngr = ␴c , 共42兲
shown. 1 + ngr ⍀2c
where ␩k = ␴ was used. The reduced group velocity
gives rise to a group delay in a medium of length L:
冉冊
IV. EIT AND PULSE PROPAGATION
1 1 Lngr ⌫31
So far we have considered EIT only from the point of ␶d = L − = = ␴L 2 . 共43兲
vgr c c ⍀c
view of the atomic system and its linear response to sta-
tionary monochromatic fields, but have not paid atten- It should be noted that under nonideal conditions, i.e., if
tion to propagation effects of pulses. Although the sus- the dark resonance is not perfectly stable, for example
ceptibility discussed in the last section already contains due to collisional dephasing of the 兩1典-兩2典 coherence or
all necessary information, it is worthwhile to devote a fast phase fluctuations in the beat note between coupling
separate section to this issue, the more so as the almost and probe fields, the denominator in the expression for
perfect transparency at certain frequencies, characteris- ngr needs to be replaced by ⍀2c + ␥31␥21, where ␥31 and ␥21
tic of EIT, allows pulse propagation in otherwise opti- are, as defined in Sec. III, the transversal decay rates of
cally thick media. Here the action of the medium on the the probe transition and the 兩1典-兩2典 coherence. In this
light pulses is—apart from the single frequency for case there is a lower limit to vgr for a fixed density .
which the medium is transparent—quite substantial, and Due to the vanishing imaginary part of the suscepti-
a number of interesting propagation phenomena are en- bility, i.e., perfect transparency, at ␦ = 0, relatively high
countered. These effects often have a simple physical atom densities and low intensities of the coupling field
origin, but their size and their special properties make Ic ⬃ ⍀2c can be used. Thus the group index ngr can be
them very important for a variety of applications. rather large compared to unity, and extremely small
group velocities are possible. In the first experiments by
Harris and co-workers in lead vapor a group velocity of
A. Linear response: Slow and ultraslow light vgr / c ⬇ 165 was observed 共Kasapi et al., 1995兲. In later
experiments by Meschede et al. a dispersive slope corre-
Let us first consider the properties associated with the sponding to a value of vgr / c ⬇ 2000 was found 共Schmidt
linear response of an EIT medium to the probe field Ep. et al., 1996兲. Other earlier experiments in which directly
We have seen in Sec. III, Eq. 共13兲, that the most charac- or indirectly slow group velocities were measured were
teristic feature of the real part of the susceptibility spec- performed by Xiao et al. 共1995兲 and by Lukin et al.
trum is a linear dependence on the frequency close to 共1997兲. The slowdown of light by EIT attracted a lot of
the two-photon resonance ␦ = 0. For a negligible decay of attention when Hau and collaborators observed the
the 兩1典-兩2典 coherence one finds spectacular reduction of the group velocity to 17 m / s in
a Bose condensate of Na atoms, corresponding to a
2⌫31␦ pulse slowdown of seven orders of magnitude 共Hau et
Re关␹共1兲兴 = ␩ + O共␦2兲, 共39兲
⍀2c al., 1999兲. Similarly small values were later obtained in a
buffer-gas cell of hot Rb atoms by Kash et al. 共1999兲 and
where ␩ = 共3 / 4␲2兲␭3 is the normalized density and ␭ is by Budker et al. 共1999兲. More recently a substantial slow-
the transition wavelength in vacuum. Since the linear down of the group velocity was also observed in the
dispersion dn / d␻p of the refractive index n = 冑1 + Re关␹兴 solid state by Turukhin et al. 共2001兲.
is positive, EIT is associated with a reduction of the The lossless slowdown of a light pulse in a medium is
group velocity according to associated with a number of important effects. When a
冏冏
pulse enters such a medium, it becomes spatially com-
d␻p c pressed in the propagation direction by the ratio of
vgr ⬅ = , 共40兲
dkp ␦=0 n + ␻p共dn/d␻p兲 group velocity to the speed of light outside the medium
共Harris and Hau, 1999兲. This compression emerges be-
which was first pointed out by Harris et al. 共1992兲. At the cause when the pulse enters the sample its front end
same time the index of refraction of an ideal three-level propagates much more slowly than its back end. At the
medium is unity and thus the phase velocity of the probe same time, however, the electrical-field strength remains
field is just the vacuum speed of light: the same. The reverse happens when the pulse leaves

the sample. In the case of the experiment of Hau et al. In the limit of linear response, i.e., in first order of
共1999兲 the spatial compression was from a kilometer to a perturbation in Ep, and under conditions of EIT, i.e., for
submillimeter scale! Since the refractive index is unity at negligible absorption, the set of one-dimensional
two-photon resonance, reflection from the medium Maxwell-Bloch equations can be solved 共Fleischhauer
boundary is usually negligible as long as the pulse spec- and Lukin, 2000, 2001兲. One finds that only the dark
trum is not too large 共Kozlov et al., 2002兲. polariton field ⌿ is excited, i.e., ⌽ ⬅ 0, and thus
Although in the absence of losses the time-integrated
photon flux through any plane inside the medium is con- Ep共z,t兲 = cos ␽⌿共z,t兲, 共47兲
stant, the total number of probe photons inside the me-
dium is reduced by a factor vgr / c due to spatial compres- e−i⌬kz
␳21共z,t兲 = − sin ␽⌿共z,t兲 . 共48兲
sion. Thus photons or electromagnetic energy must be 冑
temporarily stored in the combined system of atoms and
Furthermore, under conditions of single-photon reso-
coupling field. It should be noted that the notion of a
nance, ⌿ obeys the simple shortened wave equation
group velocity of light is still used even for vgr Ⰶ c, where
only a tiny fraction of the original pulse energy remains
electromagnetic.
It is instructive to consider slow-light propagation
冋 ⳵
⳵t
+ c cos2 ␽ 册
⳵
⳵z
⌿共z,t兲 = 0, 共49兲
from the point of view of the atoms. From this perspec- which describes a form-stable propagation with velocity
tive the physical mechanism for the temporary transfer
vgr = c cos2 ␽ . 共50兲
of excitations to and from the medium can be under-
stood as stimulated Raman adiabatic return. Before the The slowdown of the group velocity of light in an EIT
probe pulse interacts with three-level atoms, a cw cou- medium can now be given a very simple interpretation:
pling field puts all atoms into state 兩1典 by optical pump- EIT corresponds to the lossless and form-stable propa-
ing. In this limit state 兩1典 is identical to the dark state. gation of dark-state polaritons. These quasiparticles are
When the front end of the probe pulse arrives at an a coherent mixture of electromagnetic and atomic spin
atom, the dark state makes a small rotation from state excitations, the latter referring to an excitation of the
兩1典 to a superposition between 兩1典 and 兩2典. In this process 兩1典-兩2典 coherence. The admixture of the components de-
energy is taken out of the probe pulse and transferred scribed by the mixing angle ␽ depends on the strength of
into the atoms and the coupling field. When the probe the coupling field as well as the density of atoms and
pulse reaches its maximum, the rotation of the dark determines the propagation velocity. In the limit ␽ → 0,
state stops and is reversed. Thus energy is returned to corresponding to a strong coupling field, the dark-state
the probe pulse at its back end. The excursion of the polariton is almost entirely electromagnetic in nature,
dark state away from state 兩1典 and hence the character- and the propagation velocity is close to the vacuum
istic time of the adiabatic return process depends on the speed of light c. In the opposite limit, ␽ → ␲ / 2, the dark-
strength of the coupling field. The weaker the coupling state polariton has the character of a spin excitation and
field, the larger the excursion and thus the larger the its propagation velocity is close to zero. The concept of
pulse delay. To put this picture into a mathematical for- dark- and bright-state polaritons can easily be extended
malism one can employ a quasiparticle picture first in- to a quantized description of the probe field 共Fleisch-
troduced by Mazets and Matisov 共1996兲 and indepen- hauer and Lukin, 2000, 2001兲 as well as quantized matter
dently by Fleischhauer and Lukin 共2000兲, in which this fields 共Juzeliunas and Carmichael, 2002兲, in which case
concept was first applied to pulse propagation. One de- the polaritons obey approximately Bose commutation
fines dark 共⌿兲 and bright 共⌽兲 polariton fields according relations. This extension is of relevance for applications
to in quantum information processing and nonlinear quan-
tum optics, to be discussed later on.
⌿共z,t兲 = cos ␽Ep共z,t兲 − sin ␽冑␳21共z,t兲ei⌬kz , 共44兲 So far the atoms have been assumed to be at rest.
However, EIT in moving media shows a couple of other
⌽共z,t兲 = sin ␽Ep共z,t兲 + cos ␽冑␳21共z,t兲ei⌬kz , 共45兲 interesting features. First of all, if all atoms move with
the same velocity v 共兩v兩 Ⰶ c兲 relative to the propagation
with the mixing angle determined by the group index direction of the light pulse, Galilean transformation
rules predict that the group velocity 关Eq. 共50兲兴 will be
␴c⌫31 modified according to
tan2 ␽ = = ngr . 共46兲
⍀2c vgr = c cos2 ␽ + v sin2 ␽ . 共51兲
Ep is the normalized, slowly varying probe field strength, This expression can also be obtained from Eq. 共40兲 if
Ep = Ep冑ប␻ / 2⑀0, with ␻ being the corresponding carrier one takes into account that as a result of the Doppler
储
frequency. ⌬k = kc − kp, where kp is the wave number of effect the susceptibility or the index of refraction be-
储
the probe field and kc is the projection of the coupling- comes k dependent 共spatial dispersion兲. Light dragging
field wave vector along the z axis. ␳21 is the single-atom according to Eq. 共51兲 was recently observed by Strekalov
off-diagonal density-matrix element between the two et al. 共2004兲. It is interesting to note that vgr can now be
lower states. zero or even negative for ␽ ⫽ ␲ / 2 if the atoms move in

the opposite direction to the propagation of light. In ity of the control field was used by Dutton et al. 共2001兲 to
fact, since extremely large values of the group index can create strongly localized shock waves in a Bose-Einstein
be achieved in vapor cells, the thermal motion of atoms condensate and to study their dynamics.
even at room temperatures can have a significant effect Finally, interesting effects have been predicted by
on the effective group velocity. Making use of this effect, Leonhardt and co-workers when the propagation of ul-
Kocharovskaya et al. suggested “freezing” a light pulse traslow light is considered in nonuniformly moving me-
in a Doppler-broadened medium, where a specific veloc- dia. Using an effective classical field theory for slow
ity class is selectively excited 共Kocharovskaya et al., light, in which losses are neglected, they predicted an
2001兲. When the group velocity is decreased, the spectral Aharonov-Bohm-type phase shift in a rotating EIT me-
bandwidth of linear dispersion and transparency de- dium with a vortex flow 共Leonhardt and Piwnicki, 1999兲.
creases. In the limit of zero group velocity, the disper- Leonhardt and Piwnicki also pointed out an interesting
sion is infinite, which can clearly be supported only over analogy between slow light in a medium with a vortex
a vanishingly narrow frequency band. Thus in practice flow and general relativistic equations for a black hole
absorption and reflection will prevent the full stopping 共Leonhardt and Piwnicki, 2000a兲. This may allow for
of a light pulse with this method, unless a time- laboratory studies of some general relativistic effects
dependent coupling field is used, which leads to a dy- when adding an inward flow to the vortex 共Leonhardt
namical narrowing of the pulse spectrum 共see below兲 or and Piwnicki, 2000b; Visser, 2000兲. The practical obser-
acceleration of the atoms. If the flow of atoms is not vation of an optical analogy to the event horizon of a
parallel to the propagation of light, excitation carried by black hole will, however, most likely be prevented by
the atoms can be transported in a perpendicular direc- absorption and reflection effects.
tion 共Juzeliunas et al., 2003兲, as can be observed in vapor To assess the potential of slow light it is important to
cells where the atoms undergo diffusive motion 共Zibrov consider its limitations. A convenient figure of merit for
et al., 2002兲. this is not the achievable group velocity itself, but the
The slowing down of light has a number of important ratio of achievable delay time ␶d of a pulse in an EIT
applications. A reduction of the group velocity of pho- medium to its pulse length ␶p. One upper limit for the
tons leads to an enhanced interaction time in a nonlinear delay time is given by probe absorption due to the finite
medium, which is important in enhancing the efficiency lifetime of the dark resonance. Furthermore, for a
of nonlinear processes 共Harris and Hau, 1999; Lukin, pulsed probe field, i.e., for a probe field with a finite
Yelin, and Fleischhauer, 2000; Lukin and Imamoglu, spectral width, the absorption of the nonresonant fre-
2001兲. It should be noted in this context that, although quency components is nonzero even under ideal condi-
the number of photons in the medium is reduced by the tions of an infinitely long-lived dark state. Thus, in order
ratio of the group velocity to the vacuum speed of light, for the absorption to be negligible, first the decay time
the electric field is continuous at the boundaries and of the 兩1典-兩2典 transition must be very small and second
thus nonlinear polarization is not reduced as a result of the spectral width of the pulse ⌬␻p has to be much
compression. These applications will be discussed in smaller than the transparency width ⌬␻trans:
more detail in Secs. V and VI. Making use of the sub-
stantial pulse deformation at the boundary of an EIT ⍀2c 1
⌬␻p Ⰶ ⌬␻trans = 共53兲
medium is also of interest for the storage of information 冑⌫31␥31 冑␴L .
contained in long pulses, which in this way can be com-
pressed to a very small spatial volume. It is instructive to express ⌬␻trans in this condition in
As noted by Harris, pulse compression is associated terms of the group delay ␶d = ngrL / c:
冑
with a substantial enhancement of the spatial field gra-
⌫31
⌬␻trans = 冑␴L
dient in the propagation direction 共Harris, 2000兲, and 1
. 共54兲
thus ponderomotive dipole forces on atoms, resulting ␶d ␥31
from off-resonant couplings to the light pulse, are sub-
stantially amplified. If I0共z兲 denotes the probe intensity This discussion shows that it is not possible to bring a
at the input of the medium, one finds a force enhanced pulse to a complete stop by using EIT with a stationary
by the ratio c / vgr: coupling field, since in this case the transparency width
would vanish, leading to a complete absorption of the
d c d pulse. The same argument holds if one tries to stop an
Fdip ⬃ − I共z兲 = − I0共z兲. 共52兲 ultraslow pulse by means of atomic motion antiparallel
dz vgr dz
to the flow of light. Making use of ⌬␻p␶p 艌 1 one finds an
This has important potential applications in atom and upper limit for the ratio of group delay to pulse length
molecular optics as well as laser cooling. 共Harris and Hau, 1999兲
The light-matter coupling associated with EIT can
also be used for the preparation and detection of coher- ␶d 冑
ent matter-wave phenomena in ultracold quantum gases. Ⰶ ␴L. 共55兲
␶p
For example, EIT has been suggested as a probe for the
diffusion of the relative phase in a two-component Bose- The product ␴L is the opacity of the medium in the
Einstein condensate 共Ruostekoski and Walls, 1999a, absence of EIT, i.e., exp共−␴L兲 is the transmission co-
1999b兲. Stopping of a light pulse at a spatial discontinu- efficient at bare atomic resonance. Thus a noticeable

FIG. 13. Numerical simula-

tion of dark-state polariton
propagation with envelope
exp关−共z / 10兲2兴. The mixing
angle is rotated from 0 to ␲ / 2
and back according to cot ␽共t兲
= 100 „ 1 − 0.5 tanh 关 0.1 共 t − 15兲兴
+ 0.5 tanh 关 0.1 共 t − 125兲兴 …: top,
polariton amplitude; bottom
left, electric field amplitude;
bottom right, atomic spin co-
herence; all in arbitrary units.
Axes are in arbitrary units with
c = 1.
time delay of a pulse by EIT requires an optically thick vgr共t兲

medium with ␴L ⬎ 1. ⌬␻p共t兲 = ⌬␻p共0兲 , 共56兲
vgr共0兲
and the spectrum of the probe pulse stays within the
B. “Stopping of light” and quantum memories for photons transparency window at all times, provided it fulfills this
As discussed in the preceding subsection it is not pos- condition initially. The propagation equation for the
sible to bring a light pulse to a complete stop with sta- dark-state polariton is then still given by Eq. 共49兲, with
tionary EIT. To achieve this goal the group velocity has ␽ → ␽共t兲. Adiabatically rotating ␽共t兲 from 0 to ␲ / 2 leads
to be changed in time, as was shown by Fleischhauer and to a deceleration of the polariton to a full stop. At the
Lukin 共2000兲. It should be mentioned that the possibility same time its character changes from that of an electro-
of transferring spatial excitation distributions of atomic magnetic field to that of a pure spin excitation. Most
ensembles to light pulses was pointed out before by importantly, provided the rotation is adiabatic, all prop-
Csesznegi and Grobe 共1997兲. In the following the “stop- erties of the original light pulse are coherently trans-
ping” and “reacceleration” of a light pulse and its poten- ferred to the atomic spin system in this process modulo
tial applications will be discussed in more detail. At this an overall phase determined by that of the coupling
point a word of caution is needed, however. The expres- field.
sion “stopping of light” should not be taken literally. As Conditions of adiabaticity have been analyzed by
mentioned before, the reduction of the propagation ve- Matsko, Rostovtsev, Kocharovskaya, et al. 共2001兲 and by
locity of light in a lossless, passive medium is always Fleischhauer and Lukin 共2001兲. The resulting limitations
associated with a temporary transfer of its energy to the on the rate of change of the coupling field are rather
medium. In the extreme limit of zero velocity relative to weak. Furthermore, if the coupling field and thus the
the stationary medium no electromagnetic excitation is group velocity is already very small, even an instanta-
left at all. Nevertheless, the notion of a vanishing group neous switchoff would lead only to a loss of the very
velocity of light has here the same justification as the small electromagnetic component of the polariton 共Liu
notion of a group velocity in the case of ultraslow pulse et al., 2001; Matsko, Rostovtsev, Kocharovskaya, et al.,
propagation, in which likewise only a tiny fraction of the 2001; Fleischhauer and Mewes, 2002兲. Reversing the
original excitation remains in the form of photons. adiabatic rotation of ␽ by increasing the strength of the
A key conceptual advance with respect to the stop- coupling field leads to a reacceleration of the dark-state
ping of a light pulse occurred when it was realized by polariton associated with a change of character from
Fleischhauer and Lukin 共2000兲 that the bandwidth limi- spinlike to electromagnetic. In Fig. 13 a numerical simu-
tation 关Eq. 共53兲兴 of EIT can be overcome if the group lation of the stopping of a dark-state polariton and its
velocity is adiabatically reduced to zero in time. This can successive reacceleration is shown. The transfer of exci-
be achieved, for example, by reducing the Rabi fre- tation from the probe pulse to the Raman spin excita-
quency of the drive field. In this case the spectrum of the tion and back is apparent. In this way the excitation and
probe pulse narrows proportional to the group velocity, all information contained in the original pulse can be

FIG. 14. 共Color in online edition兲 Storage and

retrieval of light in ultracold Na atoms: top,
experimental setup for observing light storage
and retrieval; bottom, measurements of de-
layed 共a兲 and revived 共b兲 probe pulse after
storage. Dashed line shows intensity of cou-
pling fields. For details see Liu et al., 2001.
From Liu et al., 2001.
reversibly transferred and stored in long-lived spin co- 70– 90 ° C was used. To eliminate two-photon Doppler
herences. It should be noted that the spin excitation shifts, degenerate Zeeman sublevels were used for states
does not in general store the photon energy, as most of it 兩1典 and 兩2典. Moreover, to reduce the effects of atomic
is transferred to the coupling field by stimulated Stokes motion a He buffer gas was employed. Here storage
emission. times of up to 0.5 ms could be obtained for a pulse
The classical aspects of stopping and reaccelerating of length of several ␮s. The first successful demonstration
light pulses have been experimentally demonstrated by of light “storage” in a solid was achieved by Turukhin
Liu and co-workers in ultracold Na 共Liu et al., 2001兲, by
and co-workers in Pr-doped Y2SiO5 after proper prepa-
Phillips et al. in hot Rb vapor 共2001兲, and in solids by
ration of the inhomogeneously broadened material by a
Turukhin and collaborators 共Turukhin et al., 2001兲. The
special optical hole- and antihole-burning technique 共Tu-
experiment of Liu et al. used a similar setup to that used
for the demonstration of the group velocity reduction to rukhin et al., 2001兲. A direct proof of the coherence of
17 m / s 共Hau et al., 1999兲. Here a light pulse was slowed the storage mechanism was provided by Mair et al.
and spatially compressed in an ultracold and dense va- 共2002兲, when the phase of the stored light pulse was co-
por of Na atoms at a temperature just above the point of herently modified during the storage time by a magnetic
Bose condensation and then stopped by turning off the field. Experimental studies of light storage under condi-
coupling laser. Storage times of up to 1.5 ms for a pulse tions of large single-photon detuning were moreover
length of a few ␮s were achieved. Figure 14 shows the performed by Kozuma et al. 共2002兲. All of these experi-
experimental setup as well as the storage and retrieval of ments have to be considered as a proof of principle since
a light pulse. The use of ultracold gases has advantages the transfer efficiency or fidelity of the storage is still
such as the reduction of two-photon Doppler shifts and rather small due to a variety of limiting effects such as
high densities. Ultracold gases, however, are not neces- dark-state dephasing, atomic motion, and mode mis-
sary to achieve light stopping. In the experiment of Phil- match. An application that does not suffer from these
lips et al. a Rb vapor gas cell at temperatures of about effects is the use of the 共partially dissipative兲 stopping of

a light pulse at a spatial discontinuity of the drive field to entangled many-particle states 共Lukin, Yelin, and
create well-defined shock waves in an atomic Bose- Fleischhauer, 2000兲. An alternative and often experi-
Einstein condensate, as demonstrated by Dutton et al. mentally simpler way to achieve the same effect is to use
共2001兲. absorptive or dispersive interaction of continuous-wave
The most important potential application of light light fields with atomic ensembles and detection.
stopping is certainly in the field of quantum information. Kuzmich et al. proposed the generation of spin-squeezed
The dark-state polariton picture of light stopping also ensembles by absorption of squeezed light 共Kuzmich et
holds for quantized radiation. Here quantum states of al., 1997兲, which was experimentally demonstrated later
photons are transferred to collective excitations of the by Hald et al. 共1999兲. Employing similar ideas Polzik sug-
medium. For example, a single-photon state of a single gested the creation of two Einstein-Podolsky-Rosen en-
radiation mode 兩1典ph is mapped to an ensemble of N tangled atomic ensembles by absorption of quantum-
three-level atoms according to correlated light fields 共Polzik, 1999兲 and Kozhekin et al.
proposed a memory for quantum states of cw light fields
兩1典ph 丢兩1,1, . . . ,1典, along these lines 共Kozhekin et al., 2000兲. Recently ex-
perimental progress toward the implementation of the
cw quantum memory has been reported 共Schori et al.,
1
N 2002兲. A very intriguing technique for creating nonclas-
兩0典ph 丢
冑N 兺兩1,1, . . . ,2j, . . . ,1典. 共57兲 sical or entangled atomic ensembles is the measurement
j=1 of the collective spin using off-resonant dispersive inter-
In this way the EIT medium can act as a quantum actions. Kuzmich et al. performed in this way quantum
memory for photons. Compared to other proposals for nondemolition measurements to prepare quantum states
quantum memories, such as single-atom cavity systems with sub-shot-noise spin fluctuations 共Kuzmich et al.,
共Cirac et al., 1997兲, continuous Raman scattering 共Schori 1999, 2000兲. Duan et al. 共2000兲 suggested a detection
et al., 2002兲, or storage schemes based on echo tech- scheme in which coherent light simultaneously probes
niques, the EIT-based system is capable of storing indi- two atomic ensembles to prepare an Einstein-Podolsky-
vidual photon wave packets with high fidelity and with- Rosen correlation between them, with potential applica-
out the need for a strongly coupling resonator. It should tions in continuous teleportation of collective spin
be mentioned that a time-symmetric photon-echo tech- states. Recently, Julsgaard and co-workers have demon-
nique has been proposed by Moiseev and Kroll 共2001兲, strated this technique in a remarkable experiment 共Juls-
that is also capable of storing single photons but has not gaard et al., 2001兲.
been experimentally implemented. Several limitations of Finally, Imamoglu 共2002兲 and James and Kwiat 共2002兲
the EIT storage technique have been studied, including have suggested transferring photons to atomic en-
finite two-photon detuning 共Mewes and Fleischhauer, sembles to build single-photon detectors with high effi-
2002兲, decoherence 共Fleischhauer and Mewes, 2002; ciency, which are able to distinguish between single and
Mewes and Fleischhauer, 2005兲, and fluctuations of cou- multiple photons. Such a device has particular impor-
pling parameters 共Sun et al., 2003兲. An extension of the tance for quantum information processing with linear
model to three spatial dimensions was given by Duan et optical elements 共Knill et al., 2001兲.
al. 共2002兲.
An interesting extension of the light-stopping scheme C. Nonlinear response: Adiabatic pulse propagation and
was very recently experimentally demonstrated by adiabatons
Bajcsy et al. 共2003兲. After the coherent transfer of a light
pulse to an ensemble of Rb atoms in a vapor cell, two So far we have discussed the interaction of light pulses
counterpropagating coupling lasers were applied rather with ensembles of three-level atoms only in the linear
than one, forming a standing-wave pattern. At the nodes response limit, i.e., assuming a weak probe field. Now
of this pattern, a periodic structure of spatially narrow we consider the nonperturbative situation of probe and
absorption zones emerged. The two counterpropagating coupling pulses with comparable strength, which shows
components of the coupling laser regenerated two a number of interesting additional features. New effects
Stokes pulses with opposite wave vectors. Thus the re- arise, in particular, because in this case the medium has
generated light pulses also formed a standing-wave pat- an effect on the propagation of both pulses.
tern, which matched that of the counterpropagating In almost all theoretical treatments a quasi-one-
drive fields. In this way a stationary pulse of light was dimensional situation is considered in which all fields
created and stored for several ␮s. Although the regen- propagate parallel to the z direction, a homogeneous
erated pulse contained only a small fraction of the origi- medium is assumed, and atomic motion is disregarded.
nal photon number, the electromagnetic field was non- The properties of the atoms are described by a state
zero during the storage period in contrast to the above- vector with slowly varying probability amplitudes
mentioned experiments. This generation of stationary Cn共z , t兲, n 苸兵1,2,3其. For simplicity let us assume that the
light pulses may be an important tool for nonlinear op- carrier frequencies of coupling and probe pulses are on
tical processes with few photons. resonance with the corresponding atomic transitions,
Stopping of light can in principle also be used to which allows us to treat the slowly varying field ampli-
prepare atomic ensembles in specific nonclassical or tudes as real. The evolution of the atomic state is de-

scribed by a Schrödinger equation for Cn共z , t兲 with the

three-level Hamiltonian given in Eq. 共29兲. Because de-
cay of the polarization plays an important role for the
propagation of the fields, a decay out of the excited state
兩3典 with rate ⌫3 is included by adding an imaginary term
iប⌫3 / 2 to the energy of the excited state. The propaga-
tion of the field amplitude is most easily written in terms
of the corresponding Rabi frequencies. In a slowly vary-
ing amplitude and phase approximation one finds the
shortened wave equations for the probe 共⍀p兲 and cou-
pling pulse 共⍀c兲,
冉 ⳵
⳵t
+c
⳵
⳵z
冊 ␣p
⍀p共z,t兲 = i C2共z,t兲C*1共z,t兲,
2
共58兲
冉 ⳵
⳵t
+c
⳵
⳵z
冊 ␣c
⍀c共z,t兲 = i C2共z,t兲C*3共z,t兲,
2
共59兲
where the effect of the atoms on the fields is param-

etrized by the resonant absorption coefficients FIG. 15. Comparison of probe field and coupling field:
top, amplitude of probe field as function of time in arbitrary
␻p兩␮12兩2 ␻c兩␮32兩2 units; dotted line, at medium entrance z = 0; solid line, for
␣p = , ␣c = , 共60兲 z = 1000; bottom, the same for the coupling field. Distance is
2 ⑀ 0ប 2 ⑀ 0ប
measured in units of absorption length and peak values are
with ␮ij being the dipole moments of the corresponding ⍀c = ⍀p = 10⌫.
transitions.
If the amplitudes of the fields vary sufficiently slowly
that the adiabatic approximation holds, and if the appro- numerical calculations are used to show the propagation
priate initial conditions are fulfilled, the atoms stay at all of two matched pulses in an unprepared medium.
Apart from the preparation at the front end, matched
times in the instantaneous dark state, 兩a0共t兲典 = cos ␪共t兲兩1典
pulses are stable solutions of the propagation problem.
− sin ␪共t兲兩2典, Eq. 共17兲, with tan ␪共t兲 = ⍀p共t兲 / ⍀c共t兲. As men-
They should therefore be formed whenever pulses of
tioned before, an important property of this state is the
arbitrary shape are applied to optically thick three-level
absence of a component in state 兩3典. As a consequence of
media. In fact, it has been shown by Harris 共1993兲 that
this there is also no dipole moment on either of the two
nonequal pairs of pulses with a strong cw carrier,
transitions coupled by the probe and coupling pulses.
Atoms in the instantaneous dark state therefore have no ⍀p共z,t兲 = 关1 + f共z,t兲兴⍀pe−i␻p共t−z/c兲 , 共61兲
effect at all. Light and matter are exactly decoupled.
As first noted by Harris 共1993兲, there is another case ⍀c共z,t兲 = 关1 + g共z,t兲兴⍀ce−i␻c共t−z/c兲 , 共62兲
in which bichromatic pulses are exactly decoupled from
the atoms even if they do not vary slowly. The dark state tend to adjust their amplitude modulations f共z , t兲 and
corresponding to a pair of pulses with identical enve- g共z , t兲 in the course of propagation, i.e.,
lope, so-called matched pulses, i.e., ⍀p共z , t兲 = ⍀pf共z , t兲 and
⍀c共z , t兲 = ⍀cf共z , t兲, is time independent and thus a true
eigenstate of the system. After an appropriate prepara-
冏冏
f共z,t兲
g共z,t兲 z→⬁
= 1. 共63兲
tion of the medium, matched pulses will thus remain Fluctuations in the difference of f and g lead to non-
exactly decoupled from the interaction for all times and adiabatic transitions out of the dark state, defined by the
propagate with the vacuum speed of light. cw components, and will thus be absorbed. Eventually
As shown by Harris and Luo, a pair of matched pulses pulses of identical envelope are formed. The tendency to
applied to an atomic ensemble in the ground state, generate pulses with identical envelopes is not restricted
which in this case does not correspond to the dark state, to fields with a strong cw carrier but is also found for
will prepare the atoms by stimulated Raman adiabatic arbitrary pulses 共pulse matching兲. The phenomenon of
passage 共STIRAP; Harris and Luo, 1995兲. During the pulse matching causes a correlation of quantum fluctua-
first few single-photon absorption lengths, the front end tions in both fields, as discussed, for example, by Agar-
of the probe pulse experiences a small loss. In this way a wal 共1993兲, Fleischhauer 共1994兲, and Jain 共1994兲.
counterintuitive pulse ordering is established. This pro- As noted above, an EIT medium does not couple to
vides asymptotic connectivity of the dark state to the the fields at all in the adiabatic limit. In order to cover
initial state of the atoms. The leading end of this slightly effects like the reduction in group velocity, it is neces-
deformed pair of pulses will prepare all atoms in the sary to include first-order nonadiabatic corrections in
pathway via STIRAP and the pulses can propagate un- the dynamics of the atoms, which lead to small contribu-
affected through the rest of the medium 共Eberly et al., tions to the excited-state amplitude. Taking these into
1994; Harris, 1994a兲. This is illustrated in Fig. 15 where account, one finds for the state amplitudes

double-lambda systems and double pairs of pulses, were

studied by Cerboneschi and Arimondo 共1995兲 as well as
by Hioe and Grobe 共1994兲. The effect of nonequal cou-
pling strength on adiabaton propagation was studied by
Grigoryan and Pashayan 共2001兲. The adiabaton solutions
are approximately stable over many single-photon ab-
sorption lengths. However, as shown by Fleischhauer
and Manka 共1996兲, they eventually decay to matched
pulses after sufficiently long propagation distances.
For time scales short enough that decays can be ig-
nored, and for equal coupling strength, exact solutions
FIG. 16. Propagation of adiabatons: amplitudes of probe and of the nonlinear propagation problem in V- and ⌳-type
coupling fields as a function of time in a comoving frame for systems exist even beyond the adiabatic approximation.
ngrz / c are 共a兲 0, 共b兲 25, 共c兲 50, and 共d兲 100 from numerical
Konopnicki and Eberly 共1981兲 and Konopnicki et al.
solution of propagation equations 共Fleischhauer and Manka,
共1981兲 found soliton solutions with identical pulse shape,
1996兲. Arbitrary space and time units with c = 1.
called simultons. An interpretation of simultons in terms
of solitons corresponding to an effective two-level prob-
2i ˙ lem was given by Fleischhauer and Manka 共1996兲.
C1 = cos ␪, C2 = ␪, C3 = − sin ␪ , 共64兲 Malmistov 共1984兲 and Bol’shov and Likhanskii 共1985兲
⍀
derived simulton solutions in three-level systems using
where ⍀ = 冑⍀p2 + ⍀2c . It was shown by Grobe et al. 共1994兲 the inverse-scattering method. Several extensions of si-
that the nonlinear equations of propagation, Eqs. 共58兲 multons have been discussed. N-soliton solutions were
and 共59兲, together with Eq. 共64兲 are adiabatically inte- given by Steudel 共1988兲. Inhomogeneously broadened
grable even for fields of comparable strength and with systems were considered by Rahman and Eberly 共1998兲
arbitrary shape. To see this, we adopt the method of and by Rahman 共1999兲 and extensions to five-level sys-
Fleischhauer and Manka 共1996兲 and transform the field tems were discussed by Hioe and Grobe 共1994兲.
equations for ⍀p and ⍀c in propagation equations for
the rms Rabi frequency ⍀ and the nonadiabatic coupling
V. ENHANCED FREQUENCY CONVERSION
␪˙ . Under nearly adiabatic conditions, 兩␪˙ 兩 Ⰶ ⍀, a weak-
coupling approximation is justified. In this limit, assum- A. Overview of atomic coherence-enhanced nonlinear
ing furthermore equal coupling strength ␣p = ␣c = ␣, one optics
finds that the total Rabi frequency fulfills the free-space
propagation equation Harris and co-workers identified the enhancement of
冉冊
nonlinear optical frequency conversion as a major ben-
⳵ ⳵ efit of electromagnetically induced transparency in their
+c ⍀共z,t兲 ⬇ 0. 共65兲
⳵t ⳵z paper of 1990 共Harris et al., 1990兲. Hemmer and co-
workers 共Hemmer et al., 1995兲 introduced the double-⌳
No photons are lost by absorption and there is only a scheme as an important tool in EIT-based resonant four-
coherent transfer from one field to the other. At the wave mixing. Coherent preparation of a maximal coher-
same time ␪˙ obeys the equation ence was likewise found by a number of researchers,
冉冊
including the groups of Harris 共Jain et al., 1996兲 in the
冉 ⳵
⳵t
+c 冊
⳵ ˙
⳵z
␪共z,t兲 = − ␣
⳵ ␪˙
⳵t ⍀2
, 共66兲
USA and Hakuta in Japan 共Hakuta et al., 1997兲, to sig-
nificantly improve the conversion efficiency in four-wave
mixing. While destructive interference reduces the linear
which is exactly integrable. The corresponding solutions, susceptibility of a laser-dressed system, this is not so for
called adiabatons 共Grobe et al., 1994兲, are particularly the nonlinear susceptibility in four-wave mixing, which
simple if ⍀ is approximately constant over the time in- in fact undergoes constructive interference. To see how
terval of interest. In that case the probe and coupling this leads to improved frequency-mixing efficiency we
pulses have complementary envelopes and ␪˙ propagates need to consider the factors that determine how a gen-
without changing form with velocity vgr = c / 共1 + ␣ / ⍀2兲. erated field can grow effectively in a four-wave mixing
The quasi-form-invariant propagation of an adiabaton is process.
shown in Fig. 16. If the Rabi frequency of the coupling Laser fields applied close to resonance in an atomic
field is much larger than that of the probe field, i.e., in medium drive various frequency components of the po-
the perturbative limit, the adiabaton shows no notice- larization in that medium, both at the frequency of the
able dip in the coupling-field strength. This case thus driving fields and at new frequencies given by combina-
resembles the propagation of a weak pulse in EIT with tions of the applied ones. Figure 17 illustrates four
reduced group velocity. schemes in which three- or four-level atoms are driven,
The first experimental evidence of adiabaton propaga- at single- or two-photon resonance, by laser fields. The
tion in Pb vapor was reported by Kasapi et al. 共1996兲. polarization component at the new frequencies will act
Adiabatons in more complicated configurations, such as as a source of new electromagnetic fields. Thus in the

FIG. 17. Different schemes for resonantly en-

hanced four-wave mixing processes based on
EIT and related interference phenomena: 共i兲
sum-frequency generation 共␻4兲 out of two
pump pulses 共␻1 and ␻2兲 and one strong co-
herent drive 共␻3兲; 共ii兲 up-conversion of pump
共␻3兲 into generated field 共␻4兲 due to maxi-
mum coherence prepared by two strong drive
fields 共␻1 and ␻2兲; 共iii兲 off-resonant variant of
共ii兲; 共iv兲 parametric generation of two fields
共␻3 and ␻4兲 out of two pump fields 共␻1 and
␻2兲 in forward- or backward-scattering con-
figuration. See text for details.
schemes illustrated in Fig. 17 the material polarization Thus we can express the linear response of the atom to
P共␻4兲 is driven at frequency ␻4 by fields applied at ␻1, the field at frequency ␻4 as
␻2, and ␻3, and will in its turn drive a new electromag-
netic field at ␻4. The growth of the new field depends P共1兲共␻4兲 = ⑀0␹共1兲E4 , 共69兲
upon the magnitude of the nonlinear source term Pnl共␻4兲
and upon the linear polarization Pl共␻4兲 of the medium at and the nonlinear response as
this frequency, which will determine the absorption and
dispersion of the generated field. The equation that de- 3
P共3兲共␻4兲 = ⑀0␹共3兲E1E2E3 , 共70兲
scribes the growth of this new field E4 is derived from 2
Maxwell’s equations within the slowly varying envelope
approximation 关see, for example, Reintjes 共1984兲兴 and where Ej is the electric-field amplitude associated with
can be written in terms of the positive frequency com- the frequency component ␻j. The forms of the suscepti-
ponent of the field as bilities that can be derived by a full quantum treatment
of the atom are extensively discussed in the literature on
⳵E4 1 ⳵E4 i ␻4 nl nonlinear optics 关see for, example, Reintjes 共1984兲兴.
+ = 共P + Pl兲. 共67兲
⳵z c ⳵t 2 c⑀0 In general, for nonresonant nonlinear mixing the suscep-
tibility will be characterized by the appearance of three
Here the vector character of the field and polarization detuning terms in the denominator while the linear
was suppressed for simplicity. The term on the right- susceptibility has only a single detuning term. As was
hand side contains the effect of the linear polarization shown in Sec. III in the case of EIT, it is of course
upon the propagation of the generated field and the necessary to use the dressed susceptibilities that already
nonlinear polarization, which is the source of the new include the effect of the strong coupling field to all
field. orders. It is the difference of these susceptibilities from
The response of the medium to an electric field is gov- the normal form that give the advantages for nonlinear
erned by its polarization. In the framework of the den- mixing. The forms of the dressed susceptibilities for
sity matrices for a three-level atom we can write the EIT are given in Sec. III. These reflect the resonant
positive frequency component of the polarization as nature of the fields involved and now contain the
important interference character due to the dressing
P = Tr关␮
ˆ ␳兴 = 共␮12␳21 + ␮13␳31 + ␮23␳32兲. 共68兲 field. Inclusion of collisional dephasing and integration
over the Doppler profile are usually required to
We can express the total polarization P in terms of sus- yield effective susceptibilities that can be used in a
ceptibilities using the familiar polarization expansion. calculation.

The form of the equation that describes the 共station- without loss or distortion in what would otherwise be an
ary兲 growth of the 共slowly varying兲 field E4 can now be extremely optically dense gas. This enables a dark state
recast in terms of these dressed and Doppler integrated with maximum amplitude and equally phased atomic co-
susceptibilities: herence to be formed along the entire propagation
length. Four-wave mixing from the maximal coherence
dE4 ␻4 ␻4
= i ␹共3兲E1E2E3ei⌬kz − Im关␹共1兲兴E4 leads to large up-conversion efficiency. This requires the
dz 4c 2c use of fields strong enough to ensure adiabatic evolution
␻4 of the dark state and so is restricted to high-power
+i Re关␹共1兲兴E4 , 共71兲 pulsed lasers or to Doppler-free cw laser experiments
2c 共Harris and Jain, 1997; Merriam et al., 1999, 2000兲.
where ⌬k = k4 − 共k1 + k2 + k3兲 is the wave-vector mismatch 共iii兲 If two very strong fields are applied close to Ra-
due to the refractive-index effect of all other resonances man resonance they will still lead to the formation of a
in the atom and can also include any other phase- dark-state maximal coherence even if they are detuned
matching effects such as beam geometry and plasma dis- far from single-photon resonance 关detuning ⌬; Fig.
persion. It is easy to see from Eq. 共67兲 that if the non- 17共iii兲兴 provided that they are strong enough so that the
linear polarization reaches a large value then the source condition ⍀1⍀2 ⬎ ⌬␥21 is satisfied 共where ␥21 is the
of the new field will be strong, and thus the production dephasing rate of the 兩1典-兩2典 coherence兲. In this case EIT
of the new field will be enhanced. The absorption and plays no role, in so far as absorption is concerned, as the
refraction of the generated field are determined by the fields are far from resonance. Nevertheless, the choice of
linear polarization of the medium at the generated fre- the correct magnitude of Raman detuning can lead to
quency. The refraction has a very important effect upon cancellation of the material refraction for the pair of
the growth of the field by determining whether or not applied fields 共Harris et al., 1997兲. The maximal coher-
the generated field remains in phase with the polariza- ence, however, greatly boosts the conversion efficiency
tion that drives it. This phase-matching condition is es- in four-wave mixing 共Hakuta et al., 1997兲 as for 共ii兲. If the
sential and dictates 共at resonance where Re关␹共1兲兴 = 0兲 that maximal coherence is excited between molecular vibra-
we also require ⌬k = 0 for efficient growth of the field. tional or rotational states the medium is very suited to
For a generated field on resonance both absorption and high-order stimulated Raman sideband generation
dispersion can be problematic. In general ⌬k will be fi- 共Sokolov et al., 2000兲. The requirement of high power to
nite, resulting in a limited length over which the field achieve adiabatic evolution restricts this to pulsed lasers.
grows before slipping out of phase with the driving po- 共iv兲 The third field may also be close to resonant with
larization lcoh = 1 / ⌬k. But in the presence of EIT, and another transition. In this case the generated field com-
with the elimination of the linear susceptibility to pletes a double-⌳ scheme, which can enhance the mixing
vacuum values, the result is no absorption and no refrac- further. An important variant of the double-⌳ scheme
tion due to the resonant level, and only nonresonant with three applied fields is that explored by Hemmer et
transitions to other levels contribute to the wave-vector al. 共1995兲 and by Zibrov and co-workers, in which only
mismatch. two fields are injected 关Zibrov et al., 1999; Fig. 17共iv兲兴.
Let us examine the four-wave mixing schemes shown These fields are applied close to single-photon reso-
in Fig. 17 in which EIT or related effects have been nance with transitions in each of the two ⌳ systems
found to be advantageous. There are several intercon- formed in alkali-metal atoms. This scheme with cw laser
nected mechanisms through which atomic coherence fields has remarkable properties including, if the applied
prepared by the applied laser fields can increase the con- fields are copropagating, the growth of both pairs of
version efficiency of these wave-mixing processes: fields and, if the applied fields are counterpropagating,
共i兲 We have here a four-wave mixing scheme in which mirrorless oscillation, as experimentally demonstrated
a two-photon and a single-photon resonance are used by Zibrov et al. 共1999兲.
for the applied fields, generating a field in resonance The advantages of all of these schemes lie in the re-
with an atomic transition 关Fig. 17共i兲兴. The generated field duction of unwanted absorption and dispersion for the
and the single-photon resonant applied field act as probe generated and the driving fields while the source term
and coupling fields, respectively, in a lambda EIT for new fields is boosted. As a consequence of the elimi-
scheme. EIT then leads to an elimination of absorption nation of drive-field dispersion and absorption, large
and refraction for the generated field. The nonlinear sus- values of coherence can be driven in the case of 共ii兲, 共iii兲,
ceptibility is enhanced by proximity to resonance, since and 共iv兲, in fact to the maximal value 共i.e., ␳12 = 0.5兲 in all
this term is subject to constructive interference between the atoms 共molecules兲 within the interaction region. The
the pair of dressed states. This mechanism is relevant to frequency-mixing process then proceeds with the atomic
both cw and pulsed lasers fields 共Harris et al., 1990; coherence acting as the local oscillator with which the
Zhang et al., 1993兲. final drive field beats to produce a generated field with a
共ii兲 Here a pair of strong fields is applied resonantly in high overall conversion efficiency.
a lambda-configured three-level system close to both In addition to the different four-wave mixing schemes
Raman and single-photon resonance 关Fig. 17共ii兲兴. The discussed in the remainder of this section, other exciting
presence of EIT eliminates absorption and dispersion in applications of EIT to nonlinear optics have been pro-
the propagation of both fields. They can then propagate posed and experimentally investigated. One recent de-

velopment is the demonstration of high-efficiency mul- 3n␻2

tiorder Raman sideband generation through the I4 = 兩␹共3兲兩2兩E1兩2兩E2兩2兩E3兩2
8Zoc2
冋冉冊册
modulation of optical fields in a molecular medium pre-
pared in a maximally coherent state of the v = 0 and v ␻R
= 1 vibrational states 共Hakuta et al., 1997; Sokolov et al., 1 + e−␻Iz/c − 2e−␻Iz/2c cos ⌬k + z
2c
冋册冋册
2000兲. The possibility of using the broad-bandwidth ⫻ .
spectrum of phase-coherent sidebands to synthesize ul- ␻I 2
␻R 2
trashort pulses was first predicted by Harris and Sokolov + ⌬k +

共1997兲 and then demonstrated by the Stanford group 2c 2c
共Sokolov et al., 2001兲. Another interesting effect hap- 共73兲
pens when in an EIT medium the light velocity matches
the speed of sound. As shown by Matsko and collabora- It is implicit that ␹共3兲, Re关␹共1兲兴, and Im关␹共1兲兴 are the laser-
tors, a new type of stimulated Brillouin scattering should dressed expressions introduced in Sec. III with any re-
occur in this case, which in contrast to the usual situation quired inclusion of collisions and Doppler integration
allows efficient forward scattering 共Matsko, Rostovtsev, having been carried out. In the limit of a large propaga-
Fleischhauer, and Scully, 2001兲. tion length z, the term in the numerator will go to unity,
and if also ⌬k = 0, the intensity is simply related to the
susceptibilities by
B. Nonlinear mixing and frequency up-conversion with

electromagnetically induced transparency
I4 ⬃ 冏冏 ␹共3兲
␹共1兲
2
, 共74兲
as was pointed out first by Harris et al. 共1990兲 and sub-

The enhancement of four-wave mixing in a two-
sequently 共in the form used here兲 by Petch et al. 共1996兲.
photon resonant scheme of the type illustrated in Fig.
It is very clear that EIT increases this ratio by both in-
17共i兲 was first treated by Harris et al. 共1990兲. The idea of
creasing the value of the numerator and reducing the
improving four-wave mixing using an additional control
denominator. It should be noted that a dramatic change
field primarily to adjust the refractive index of the me- in conversion efficiency as a function of coupling
dium was discussed by Tewari and Agarwal 共1986兲. In strength is only apparent in an inhomogeneously broad-
their scheme, however, the control field is not incorpo- ened medium. In a homogeneously broadened medium,
rated directly into the mixing fields and there is no con- although EIT increases the efficiency, the ratio of the
structive interference of the nonlinear susceptibility. susceptibilities remains independent of the coupling
Taking the linear and nonlinear susceptibilities de- strength.
rived in Sec. III, we can proceed with an analysis of the In the case of four-wave mixing with EIT we usually
case of four-wave mixing in a lambda EIT scheme. Pro- assume a weak probe field so that excited-state popula-
ceeding from Eq. 共71兲 and integrating with appropriate tions and coherences remain small 关see Fig. 17共i兲兴. The
boundary conditions 共as is usual in all four-wave mixing two-photon transition need not be strongly driven 共i.e., a
processes兲 leads to an expression for the generated in- small two-photon Rabi frequency can be used兲 but a
tensity of the form “strong” coupling laser is required. The coupling laser
amplitude must be sufficient that the Rabi frequency is
comparable to or exceeds the inhomogenous spectral
3n␻2 共3兲 2 widths in the system 共e.g., Doppler width兲. For example,
I4 = 兩␹ 兩兩E1兩2兩E2兩2兩E3兩2F共⌬k,z,R,I兲, 共72兲 a laser intensity of above 1 MW cm−2 is required for a
8Zoc2
typical Doppler-broadened visible transition. This is
trivially achieved even for unfocused pulsed lasers, but
does present a substantial barrier for the use of
where z is the propagation distance, Zo = 376.7 ⍀ is the
continuous-wave lasers unless a specific Doppler-free
impedance of free space, n is the refractive index at the
configuration is employed. The latter is not normally
generated frequency 共typically close to unity兲, R and I suitable for a frequency up-conversion scheme if a large
are shorthand for Re关␹共1兲兴 and Im关␹共1兲兴, respectively, and frequency up-conversion factor is to be achieved, e.g.,
c is the speed of light in vacuum. Here we have intro- to the vacuum ultraviolet 共VUV兲. This is in large part
duced the phase-matching factor F共⌬k , z , R , I兲, which because atomic species with high-ionization potentials
depends upon the details of the mixing scheme and the 共⬎10 eV兲 that are suitable for frequency up-conversion
focal geometry. This general result indicates the impor- to the VUV are not easily trapped and cooled at high
tance of phase matching 共and hence ⌬k兲 for the gener- density. Recent experiments, by, for instance, Hemmer,
ated field intensity. report significant progress in cw nonlinear optical pro-
To obtain clear insight into the effects of EIT we shall cesses using EIT 共Hemmer et al., 1995; Babin et al.,
follow the solution provided by Petch et al. 共1996兲. For 1996兲. A number of other possibilities, e.g., laser-cooled
the case of plane waves propagating in a homogeneous atoms and Doppler-free geometries, have also been ex-
medium this treatment yields plored.

In the case of frequency up-conversion to short wave-

lengths, Doppler shifts arising from the Maxwellian ve-
locity distribution of the atoms or molecules in a gas at
finite temperature lead to a corresponding distribution
in the detunings for the atomic ensemble. The calcula-
tion of the response of the medium, as characterized by
the susceptibilities, must therefore include the Doppler
effect by performing a weighted sum over possible de-
tunings. From this operation the effective values of the
susceptibilities at a given frequency are obtained, and
these quantities can be used to calculate the generated
field. Interference effects persist in the dressed profiles
provided the coupling laser Rabi frequency is compa-
rable to or larger than the inhomogeneous width. This is
because the Doppler profile follows a Gaussian distribu-
tion, which falls off much faster in the wings of the pro-
file than the Lorentzian profile arising from lifetime
broadening. Indeed, in the case of Doppler broadening, FIG. 18. Experimental demonstration of sum-frequency gen-
the effects of increasing ⍀c manifest themselves most eration in H: left, level and coupling scheme; right, calculated
clearly. As discussed by Harris et al. 共1990兲 there is ex- 共dashed兲 and measured 共solid line兲 nonlinear susceptibility as
pected to be a steplike increase in the conversion effi- function of detuning for the 3p-1s transition for different
ciency when the magnitude of ⍀c approaches the Dop- density-length products of the medium 共nL兲. Only under the
pler width of the transition. For coupling strengths least dense condition is the Autler-Townes split structure
below this value the atomic sample remains opaque, as it apparent in the generated signal. At higher density strong
would under ordinary resonant conditions; in contrast, resonant enhancement is seen. From Zhang et al., 1993.
for a coupling strength exceeding the Doppler width the
full benefit of the quantum interference is obtained.
In the Doppler-broadened case, therefore, the wave mixing process that was enhanced by this EIT. In
coupling-field Rabi frequency must remain greater than these experiments the opening of photoionization chan-
the Doppler width for a significant fraction of the inter- nels from both of the involved excited atomic states
action time. Pulsed lasers, because of the high peak pow- leads to loss of the conditions for perfect transparency.
ers that arise, are required. Moreover, a transform- Nevertheless, because the photionization cross section
limited single-mode laser pulse is essential for the for the 2s metastable level is about one order of magni-
coupling laser fields that drive EIT, since a multimode tude less than that from the 3p state, the effect of EIT is
field will cause an additional dephasing effect on the only partially reduced. It is in general important to keep
coherence, resulting in a deterioration of the quality of the photoionization rate sufficiently small so as not to
the interference, as discussed in Sec. III. quench the coherence through electron impact broaden-
The work of the Stanford group 共Sokolov et al., 2001兲 ing 关see Buffa et al. 共2003兲兴. In this experiment and sub-
has highlighted that when pulsed laser fields are used sequent work by the same authors, conversion efficien-
additional considerations must be made. Kasapi and co- cies up to 2 ⫻ 10−4 were reported 共Zhang et al., 1995兲.
workers showed 共Kasapi et al., 1995兲 that the group ve- A limit to the conversion in an EIT-enhanced four-
locity of a 20-ns pulse can be modified for pulses propa- wave mixing scheme is set by the Doppler width. A large
gating in the EIT; large reductions, e.g., by factors down Doppler width dictates that a comparable Rabi fre-
to ⬍c / 100, in the group velocity have been observed. quency is required to create good transparency. This in
Another consideration beyond that found in the simple turn leads to a reduction in the nonlinear susceptibility.
steady-state case is that the medium can only become This is one of the most important factors limiting the
transparent if the pulse contains enough energy to dress conversion efficiency in the hydrogen scheme 共Hakuta,
all of the atoms in the interaction volume. This prepara- 2004兲. The requirements on a minimum value of ⍀c
tion energy condition was discussed in Sec. IV. This puts ⬎ ⌬Doppler constrain the conversion efficiency that can be
additional constraints on the laser pulse parameters. achieved because of a scaling factor by 1 / ⍀2c , which ul-
Up-conversion to the UV and vacuum UV has been timately leads to diminished values of ␹共3兲. The use of
enhanced by EIT in a number of experiments. The first gases of higher atomic weight at low temperatures is
experiment to show EIT in a resonant scheme, where therefore highly desirable in any experiment utilizing
both the EIT effect on opacity and phase matching were EIT for enhancement of four-wave mixing to the VUV.
important, was reported by Zhang et al. 共1993兲. They A combination of the low mass of the H atom and
employed a four-wave mixing scheme in hydrogen the elevated temperatures required in the dissociation
关equivalent to Fig. 17共i兲兴 in which the EIT was created process leads to especially large Doppler widths
on the 3p-1s transition at 103 nm in a lambda scheme by 共⬎10 GHz兲. To overcome such a severe limit, investiga-
the application of a field at 656 nm on the 2s-3p transitions were undertaken by one of the authors in krypton
tion 共Fig. 18兲. A field at 103 nm is generated by the four- gas, in which the relevant Doppler width is only 1 GHz.

Conversion efficiencies approaching 10−2 were obtained Under the conditions of STIRAP, or for matched
共Dorman et al., 2000b兲. In this case, the conversion effi- pulse propagation, large populations of coherent popu-
ciency appears to be limited by the finite value of ⌬k lation trapped states are created. To achieve this situa-
due to the other atomic levels rather than absorption in tion the laser electric-field strength must be large
the medium 共Dorman et al., 2000a兲. enough to create couplings that will ensure adiabatic
In a Doppler-free medium the effect of constructive atomic evolution and pulses that are sufficiently ener-
interference for the nonlinear term is most apparent. In getic to prepare all of the atoms in the beam path. The
a system without inhomogeneous broadening perfect ␳12 coherence thus created will have a magnitude 兩␳12兩
transparency can be induced with ⍀c Ⰶ ␥31. As ⍀c is = 0.5 共i.e., all the atoms are in a coherent state兲 and nega-
small relative to ␥31, the nonlinear susceptibility will, be- tive sign 共i.e., all the atoms are in the population trapped
cause of the constructive interference, have a value es- dark state兲. The complex coherence varies in space and
sentially identical to that of the unmodified resonant time and can be written
atomic system. The transparency dip will be very narrow
共see Fig. 7兲. The spectral widths of these features are ⍀1⍀*2
␳12 = −
typically subnatural and are therefore accompanied by 冑⍀21 + ⍀22 exp兵i共␻1 − ␻2兲t − 共k1 − k2兲z其
very steep normal dispersion, which corresponds to a
= 兩␳12兩e−i␦kzei共␻1−␻2兲t . 共75兲
much reduced group velocity. As was pointed out by
several authors, including Schmidt and Imamoglu 共1996兲 Under these circumstances mixing of additional fields
and Harris and Hau 共1999兲, the nonlinear susceptibilities with the atom can become extremely efficient 共Harris et
in this case can be extremely large, as there is a con- al., 1997兲.
structive interference. Nonlinear optics at very low light The preparation energy condition for the creation of
levels, i.e., at the few-photon limit, are possible in this maximal coherence requires that the pulse energy ex-
regime. Associated with their first dramatic measure- ceed the following value:
ments of ultraslow light, Hau et al. 共1999兲 reported a
nonlinear refractive index in an ultracold Na vapor that f13
Eprep = ALប␻ , 共76兲
was 0.18 cm2 W−1. This is already 106 times larger than f23
that measured for cold Cs atoms in a non-EIT system,
which itself was much greater than the nonlinearity in a where fij are the oscillator strengths of the transitions
typical solid-state system. A discussion of this process in and L is the product of the density and the length
the few-photon limit is given in Sec. VI. 共Harris and Luo, 1995兲. Essentially the number of pho-
tons in the pulse must exceed the number of atoms in
the interaction volume to ensure that all atoms are in
the appropriate dressed state. It should be noted that
C. Nonlinear optics with maximal coherence
the preparation energy is much smaller for nonlinear
conversion processes not using maximal coherence. Fur-
thermore, this preparation energy is not lost by dissipa-
Harris demonstrated an important extension of the
tion if the pulses evolve adiabatically and can be recov-
EIT concept that occurs if the atomic medium is strongly
ered.
driven by a pair of fields in Raman resonance in a three-
An additional field applied to the medium can partici-
level system. Due to the elimination of absorption and
pate in sum- or difference-frequency mixing with the
refraction, both fields can propagate into the medium
two Raman resonant fields. The importance of the large
and large values of the material coherence are created
value of coherence is that it is the source polarization
on the nondriven transition. Considering the system il-
that drives the new fields generated in the frequency-
lustrated in Fig. 17共ii兲, we can imagine that both applied
mixing process. This is described by the following equa-
fields are now strong. Under appropriate adiabatic con-
tion:
ditions 共see Sec. III兲 the system evolves in the dark state
to produce the maximum possible value for the coher- dE4
ence ␳12 = 0.5. Adiabatic evolution into the maximally = − i␩4ប␻4n关共a4␳11 + d4␳22兲E4 + b4␳12E3兴, 共77兲
dz
coherent state is achieved by adjusting either the Raman
detuning or the pulse sequence 共i.e., to “counterintui- with coefficients a4, b4, and d4 that depend upon atomic
tive” order兲. The pair of fields may also be in single- factors. The second term on the right-hand side
photon resonance with a third level, in which case the 共b4␳12E3兲, the source of the new field, is comparable in
EIT-like elimination of absorption will be important. magnitude to the first term that describes the dispersion
This situation is equivalent to the formation of a dark and loss. Complete conversion can occur over a short
state, since neither of the two strong fields is absorbed distance if the factor ␳12 is large. This significantly re-
by the medium. For sufficiently strong fields the laxes the constraints usually set by phase matching in
single-photon resonance condition need not be satisfied nonlinear optics. Recently near-unity conversion effi-
and a maximum coherence can still be achieved 关see ciencies to the far UV were reported in an atomic lead
Fig. 17共iii兲兴 provided that the condition ⍀1⍀2 ⬎ 兩⌬兩␥21 is system where maximum coherence had been created
met, where ⌬ is the detuning of the strong fields from 共Jain et al., 1996; Harris et al., 1997; Merriam et al., 1999,
state 兩3典. 2000兲. An example of high-conversion-efficiency four-

FIG. 19. Parametric up-conver-

sion with maximum coherence:
共a兲 a pair of 共near兲 resonant
coupling 共⍀c兲 and probe pulses
共⍀p兲 generate maximum coher-
ence between states 兩1典 and 兩2典.
From Harris et al., 1997. 共b兲 The
coherently prepared medium
then leads to efficient up-
conversion of second coupling
pulse ⍀c into the ⍀h. Also
shown is the conversion effi-
ciency from 425 to 293 nm as a
function of coupling laser inten-
sity. From Jain et al., 1996.
wave mixing is shown in Fig. 19 共adapted from Jain et al., Since the 兩1典-兩2典 dephasing rate is very small in appropri-
1996兲. ately prepared samples of solid hydrogen, interference
An alternative way to look at the origin of the large that causes the dispersion to become negligible can oc-
four-wave mixing efficiency is to consider the suscepti- cur. Because of the removal of the usual phase mis-
bilities. Recalling the density-matrix results of a ⌳ sys- match, efficient operation of these frequency-conversion
tem, Eq. 共12兲, one recognizes that the polarization at ␻4 schemes over a broad range of frequencies 共infrared to
proportinal to ␳31 is a sum of two terms, the first being vacuum UV兲 has been shown.
proportional to the linear susceptibility and the second A recent prediction is of broadband spectral genera-
to the nonlinear susceptibility. The drive fields E1 and E2 tion associated with strong-field refractive index control
are now implicitly included in ␳12, which has the maxi- 共Harris and Sokolov, 1997; Kien et al., 1999兲. The obser-
mal value of 0.5. These drive fields are resonant so the vations of very efficient high-order Raman sideband
structure of ␹共3兲 simplifies to a single nonresonant term generation point the way to synthesizing very-short-
in the denominator. The nonlinear susceptibility there- duration light pulses, since the broadband Raman side-
fore shares this single nonresonant term in the denomi- band spectrum has been proved to be phase locked
nator with the linear susceptibility and so these two 共Sokolov et al., 2001兲. It is anticipated that high-power
quantities are of comparable size. This is a very unusual sources of sub-femtosecond-duration pulses might be
situation in nonlinear optics, and the length over which achieved through this technique.
efficient frequency conversion can occur is now reduced
to values comparable to the coherence length. A conse- D. Four-wave mixing in double-⌳ systems
quence of this is that complete nonlinear frequency con-
version of the fields can occur in a distance of the order Four-wave mixing with all the optical fields close to
of the coherence length 共determined by the real part of resonance is achieved in atoms with four levels in the
the susceptibility兲. This is equivalent to having near so-called double-lambda configuration 共Hemmer et al.,
vacuum conditions for the refraction 共and absorption兲 of 1995兲, which is detailed in Fig. 20. The fully resonant
the medium while the nonlinearity remains large. character of the light-matter interaction for all of the
In a molecular medium large coherence between vi- fields 共both those applied and those generated兲 involved
brational or rotational levels has also been achieved us- in four-wave mixing leads potentially to very high con-
ing adiabatic pulse pairs in gas-phase hydrogen and deu- version efficiencies and to a number of important char-
terium 共Sokolov et al., 2000兲. Efficient multiorder acteristics. Earlier work on double-lambda schemes in
Raman sideband generation has been observed to occur. which all four fields are applied has shown, for instance,
In this case a pair of lasers that are slightly detuned from that the formation of dark states depends upon the rela-
Raman resonance are used to adiabatically establish a tive phase of the fields 共Buckle et al., 1986; Arimondo,
superposition of two molecular states. The large energy 1996兲. Experiments in the cw limit have been carried out
separation to the next excited electronic state in these in sodium vapor that show the phase sensitivity of EIT
molecules 共10 eV兲 means that the situation is similar to in a double-lambda scheme 共Korsunsky et al., 1999兲.
that illustrated in Fig. 17共iii兲. This superposition then First we shall consider the case in which three reso-
mixes with the applied fields to form a broad spectrum nant fields are applied such that the fourth field is gen-
of sidebands through multiorder mixing. Molecular co- erated in resonance between the highest of the excited
herence in solid hydrogen has recently also been used to states and the ground state. Under appropriate condi-
eliminate phase mismatch in a Stokes or anti-Stokes tions the presence of the three applied fields eliminates
stimulated Raman frequency-conversion scheme not only the resonant absorption or refraction of the
共Hakuta et al., 1997兲. In this scheme, v = 0 and v = 1 vi- generated field but also that in the applied fields them-
brational states of the hydrogen molecule electronic selves. The situation of four-wave mixing in a double-
ground state form the lower states of a Raman scheme. lambda system has been treated theoretically in detail

FIG. 20. Generation of fields E1 and E2 by

resonant four-wave mixing of two counter-
propagating pump fields Ef and Eb: left, level
and coupling scheme; right, calculated thresh-
old behavior of output intensity of field E1 as
a function of medium length L. The inset
shows the spatial distributions of the fields
above threshold of parametric oscillation at
the indicated point. Adapted from
Fleischhauer et al., 2000.
by Korsunsky and Kosachiov 共1999兲, who found the rela- VI. EIT WITH FEW PHOTONS
tive phases of the four fields to evolve with propagation
so as to support lossless propagation in a fashion similar In the absence of a strongly polarizable medium, pho-
to that predicted by Harris for a pair of fields propagat- tons are essentially noninteracting particles. In commu-
ing in resonance in a lambda medium 共Harris, 1993兲. The nications technology this is a desired property, and op-
phase and the amplitudes of all four fields, the generated tics has emerged over the last two decades as the
preferred method for communicating information. For
field and the applied fields, are predicted to adjust as
the same reason, attempts to use light in information
they propagate so as to support efficient transfer of en-
processing tasks such as computation have failed. For
ergy between the fields, under lossless conditions.
the latter, one needs strong, dissipation-free light-light
An experimental realization of up-conversion in a interactions. In the present section we summarize and
double-lambda scheme driven by three resonantly tuned discuss some of the potentials and limitations of electro-
pulsed fields, with the new field generated in the far UV magnetically induced transparency in this respect.
at 186 nm, was achieved by Merriam et al. 共2000兲 in lead
vapor. High conversion efficiencies, more than 0.3 from
the shortest-wavelength applied field at 233 nm to the A. Giant Kerr effect
186 nm field, were indeed observed. The experiments
confirm that the double-lambda scheme supports loss- Information processing based on light requires nonlin-
less propagation provided that the pairs of lambda reso- ear interactions that would lead, for example, to a Kerr
nant fields have matched ratios of Rabi frequency. For effect or, equivalently, a cross-phase modulation in
pulses not initially satisfying the condition of matched which the phase of a light field is modified by an amount
Rabi frequencies, absorptive loss and nonlinear energy determined by the intensity of another optical field. If
transfer occur until the condition is reached, after which the nonlinear phase shifts arising from such a Kerr effect
all fields propagate without further loss and refraction. were on the order of ␲ rad, it would be possible to
Limits to the up-converted power density were found, implement all-optical switching. Such large Kerr nonlin-
however, to arise from power broadening. earities without appreciable absorption can, however,
An interesting alternative to the schemes just de- only be obtained for intense laser pulses containing
scribed for efficient nonlinear optical conversion in a roughly 1010 photons.
double-lambda scheme was identified by Zibrov et al. One of the principal challenges of nonlinear optics is
共1999兲. In essence the scheme is very simple. Two fields the observation of a mutual phase shift exceeding ␲ rad
using two light fields, each containing a single photon. In
are applied resonantly, one in each of the two lambda
addition to new possibilities for addressing fundamental
systems. These then efficiently couple to the pair of
issues in quantum theory, such strong nondissipative in-
fields that complete the double-lambda configuration.
teractions could be used to realize a controlled-NOT gate
The generated fields correspond to Stokes and anti- between two quantum bits 共qubits兲, defined, for ex-
Stokes components of the two applied fields. In the ample, by the polarization state of a single-photon pulse.
original experiment in Rb vapor the applied fields coun- However, given the weakness of nonresonant optical
terpropagate, and the resulting nonlinear gain and effi- nonlinearities and the dominant role of absorption in
cient intrinsic feedback in this configuration lead to mir- resonant processes, the combination of a large third-
rorless self-oscillation at extremely low applied field order Kerr susceptibility and small linear susceptibility
strengths 共at the ␮W level of cw power兲. Further analysis appears to be incompatible.
of the quantum dynamics in this situation has high- In this section, we show how EIT can overcome these
lighted the possibility of using this technique to generate seemingly insurmountable obstacles. We have already
narrow-band sources of nonclassical radiation 共Lukin et seen how the nonreciprocity between linear and nonlin-
al., 1999; Fleischhauer, Lukin, et al., 2000兲. ear susceptibilities in EIT media can be used to enhance

FIG. 21. Level scheme for giant Kerr effect.

FIG. 22. Cross-phase modulation with and without giant Kerr
effect: left, level configuration for cross-phase modulation in
generation of radiation at new frequencies. We shall see
an ordinary three-level scheme; right, dressed-state represen-
in this subsection how dissipation-free strong photon- tation of giant Kerr scheme.
photon interactions can be induced using EIT, due to a
constructive interference in Kerr nonlinearity at two-
photon resonance 共␦ = 0兲, where linear susceptibility ex- Similarly, we can obtain the linear and nonlinear suscep-
periences a complete cancellation. This so-called giant tibilities at ␻s. For simplicity, we consider only the reso-
Kerr effect, proposed by Schmidt and Imamoglu 共1996兲, nant case in which ⌬ = ␦ = 0.
can be understood using the four-level system depicted First, we emphasize that the linear susceptibility for
in Fig. 21. The principal element here is once again the both probe and signal fields vanishes in the limit ⌫2 = 0.
lambda subsystem consisting of states 兩1典, 兩2典, and 兩3典. This is trivially true for ␻s, since it is assumed not to
The state 兩4典 has the same parity as 兩3典 and has an elec- couple to state 兩1典 atoms, and is the case for ␻p to the
tric dipole coupling to 兩2典; the signal field at frequency ␻s extent that ␻s remains perturbative. We also note that
is applied on this transition, with a Rabi frequency ⍀s. self-phase-modulation-type Kerr nonlinearities 关i.e.,
The cross-phase modulation nonlinearity that we are in- ␹共3兲共−␻p , ␻p , −␻p , ␻p兲 and ␹共3兲共−␻s , ␻s , −␻s , ␻s兲兴
terested in is between this signal field and the probe field for both fields are identically zero. The cross-phase
applied on the 兩1典-兩3典 transition. We assume that the ap- modulation nonlinearity ␹xpm 共3兲
= ␹共3兲共−␻p , ␻s , −␻s , ␻p兲
plied fields couple only to the respective transitions in- 共3兲
= ␹ 共−␻s , ␻p , −␻p , ␻s兲, on the other hand, is given by
dicated in Fig. 21. Throughout this section, we assume 共Schmidt and Imamoglu, 1996兲
that the coupling field with Rabi frequency ⍀c is nonper-
turbative, remains unchanged, and can be treated classi-
cally.
To calculate the third-order susceptibility correspond-
共3兲
␹xpm =
兩␮13兩2兩␮24兩2 1
2 ⑀ 0ប 3 冋
1
+i
⌫4
⍀c ⌬␻42 2⌬␻42
2 2 . 册共80兲
ing to the cross-phase modulation Kerr nonlinearity, we To evaluate the significance of this result, it is perhaps
assume both ⍀p and ⍀s to be perturbative. As we have illustrative to compare it to the Kerr nonlinear suscepti-
共3兲
discussed in Sec. III, we can use the effective non- bility of a standard three-level atomic system 共␹3−level 兲
Hermitian Hamiltonian depicted in Fig. 22共a兲. For this system consisting of the
ground state 兩a典, an intermediate state of opposite parity
ប 兩b典, and a final state 兩c典, we find
Heff = − 关⍀p␴31 + ⍀c␴32 + ⍀s␴42 + H.c.兴
2
冉冊冉冊共3兲兩␮ab兩2兩␮bc兩2 1
i i Re关␹3−level 兴= , 共81兲
+ ប ⌬ − ⌫3 ␴33 + ប ␦ − ⌫2 ␴22 2 ⑀ 0ប 3
⌬␻ab⌬␻bc
2
2 2
冉冊
where ␮ij and ⌬␻ij denote the dipole matrix element of
i
+ ប ⌬␻42 − ⌫4 ␴44 共78兲 and the detuning from the transition 兩i典-兩j典, respectively
2 共i , j = a,b,c兲. When we compare the two expressions, we
to describe the atomic dynamics in this limit. Here we observe that they have practically identical forms: in the
have introduced the detuning ⌬␻42 = ␻42 − ␻s and the EIT cross-phase modulation scheme, 1 / 4⌬␻ab 2
is re-
spontaneous emission rate ⌫4 out of state 兩4典. Once placed by 1 / ⍀c . Even though one needs to have ⌬␻ab
2 2
again, we can solve the Schrödinger equation with the 艌 ⌫b in a three-level scheme to avoid absorption, ⍀2c
2
Hamiltonian of Eq. 共78兲 to obtain the probability ampli- Ⰶ ⌫23 is possible in the EIT cross-phase modulation sys-
tude for finding the atom in state 兩3典关a3共t兲兴 and use it to tem. This in turn implies that the magnitude of Kerr
evaluate the linear and nonlinear components of the po- nonlinearity can be orders of magnitude larger in the
larization at ␻p: latter case. This is the essence of the giant Kerr effect
共Schmidt and Imamoglu, 1996兲.
P共t兲 = N␮13a3共t兲e−i␻pt + H.c. We can understand the origin of this enhancement us-
= ⑀0␹共1兲共− ␻p, ␻p兲Epe−i␻pt ing the dressed-state picture depicted in Fig. 22共b兲.
When we apply the dressed-state transformation intro-
+ ⑀0␹共3兲共− ␻p, ␻s,− ␻s, ␻p兲兩Es兩2Epe−i␻pt + H.c. duced in Sec. III, we find that the initial and final states
共79兲共which remain unchanged兲 are coupled via two interme-

diate states 兩2d典 and 兩3d典. The two paths contributing to ⍀2c 1
virtual excitation of state 兩4典 experience a constructive vp = , 共82兲
⌫3 ␴13
interference for ␻p tuned in between the two dressed
states, partially leading to the above-mentioned en- where ␴13 = 3␭p2 / 2␲ is the peak absorption cross section
hancement. Perhaps more importantly, however, the “ef- of the 兩1典-兩3典 transition 共␭p = 2␲c / ␻p兲. Given that vp Ⰶ c,
fective detuning” between the intermediate dressed the time it takes for the probe pulse to traverse the me-
dium,
states is given by ⍀c and can be much smaller than their
width. ⌫3
Alternatively, we can understand the predicted en- ␶d = ␴13L, 共83兲
⍀2c
hancement of the Kerr nonlinearity by recalling the
referred to as the group delay time 共Harris and Yama-
steep dispersion obtained at transparency for ⍀2c Ⰶ ⌫23. In
moto, 1998; Harris and Hau, 1999兲, can be much longer
this limit, small changes in two-photon detuning ␦ that than the traversal time of the signal field Ts = L / c and
are caused, for example, by a shift in the energy of state the pulse widths of the probe 共␶p兲 and signal 共␶s = ␶p兲
兩2典 can give rise to a drastic increase in Re关␹共1兲兴 experi- fields. As discussed earlier, the pioneering experiment of
enced by the probe field. We can understand the role of Hau et al. 共1999兲 already demonstrated ␶D ⬎ 1 msecⰇ ␶p.
the signal field as creating an ac-Stark shift of state 兩2典, The generated nonlinear phase shift is proportional to
thereby leading to a large change in the index of refrac- the interaction length of the probe and signal fields. In
tion at ␻p that is proportional to ⍀s2. the EIT giant Kerr scheme, this is determined by
Finally, we note that by optical pumping and by min关L,Lh兴, where Lh is the probe propagation length
choosing the polarization of the lasers appropriately, one that would give rise to a time delay between the two
can realize the energy-level diagram depicted in Fig. 21 pulses that is equal to their pulse width ␶p, i.e.,
practically for all alkali-metal atoms. It has been esti- ⍀2c 1
mated that the enhancement of ␹xpm 共3兲
over a standard L h = v p␶ p = ␶p , 共84兲
⌫3 ␴13
three-level atomic medium with identical density and di-
pole matrix elements could be as much as six orders of which was originally defined as the Hau length 共Harris
and Hau, 1999兲. The nonlinear phase shift can be deter-
magnitude 共Schmidt and Imamoglu, 1996兲. Naturally, the
mined from the slowly varying envelope equation 共Har-
interest in this system arises from the possibility of doing
ris and Hau, 1999兲.
冏冉冊冏
nonlinear optics on the single-photon level, without re-
quiring a high-finesse cavity; this possibility was dis-
cussed in the original proposal 共Schmidt and Imamoglu,
⳵Ep
⳵z
+ 冉
1 1 ⳵Ep
+
vp c ⳵t ␭p
冊
␲ 共3兲
= i ␹xpm Es t −
z
c
2
Ep
1996兲, but in the limit of continuous-wave fields. Recent
experiments have already demonstrated the enhance-
ment of Kerr nonlinearities in the limit of ⌬␻42 = 0 共Braje
= 冉冊 i
4共⌬␻42 + i⌫4/2兲vp
and was discussed by Harris and Yamamoto in the con-

共3兲
et al., 2003兲; this is the purely dissipative limit of ␹xpm
冏冉冊冏
⫻ ⍀s t −
z
c
2
Ep , 共85兲
text of a two-photon absorber 共Harris and Yamamoto,
where we have used Eq. 共80兲.
1998兲.
When the group delay is negligible compared to ␶p
As we have already discussed above, the steep disper- 共i.e., L Ⰶ Lh兲, and we assume Gaussian input pulses, we
sion curve of the probe field is responsible for the en- find for the peak 共complex兲 phase shift for the maximum
hancement in the Kerr effect. It is interesting that it is of the pulse
冉冊冑
this steep dispersion, or the slow group velocity that
arises from it, that at the same time gives rise to a strin- ⌫4 ns ␴24 ln 2 ␶d
␾xpm → , 共86兲
gent limit on the available single-photon nonlinear ⌬␻42 + i⌫4/2 A 2 ␲ ␶p
phase shifts for pulsed fields. This limitation arises from
where ns is the number of photons in the pulse and A its
the fact that the probe and signal pulses travel at vastly
cross-sectional area. Equation 共86兲 is, as expected, the
different group velocities in the limit ⍀2c Ⰶ ⌫23, and their result one would obtain using Eq. 共80兲 without taking
interaction time will be limited by the fact that the slow into account the group-velocity mismatch between the
probe pulse will separate spatially from the fast signal two pulses. In the opposite limit of L Ⰷ Lh, we have
冉冊
pulse 共Harris and Hau, 1999兲.
To make quantitative predictions on the limitation on ⌫4 ns ␴24
␾xpm → . 共87兲
nonlinear phase shifts arising from disparate group ve- ⌬␻42 + i⌫4/2 A 8
locities, we assume a nearly ideal EIT scheme where This result shows that the maximum nonlinear phase
⌫3⌫2 Ⰶ ⍀2c Ⰶ ⌫23, ⌬ = 0 and concentrate on the limit of shift that can be obtained in the EIT system is indepen-
small 兩␦兩. For this system, we have a probe-field group dent of the system parameters, such as ⍀c, , and L
velocity 共Harris and Hau, 1999兲, and is on the order of 0.1 rad for

single-photon pulses focused into an area A ⬃ ␴24. This ␶d 冑

remarkably simple result assumes that ⌬␻42 ⬃ ⌫4 / 2, in 艋 ␴13L. 共89兲
␶p
which case we get a two-photon absorption coefficient
that is comparable to the nonlinear phase shift. Since the other factors on the right-hand side of Eq. 共88兲
can be of order unity, we can obtain ␾xpm 艌 ␲, provided
that we have ␴13L 艌 10. The presence of an optically
B. Cross-phase modulation using single-photon pulses thick medium is essential for large photon-photon inter-
with matched group velocities action. At the same time, we note that this interaction is
necessarily slow and has a ␾xpm-bandwidth product that
From the perspective of applications in quantum in- is independent of the optical depth 共Lukin and Imamo-
formation processing, the result obtained in the previous glu, 2000兲.
subsection has a somewhat negative message: While EIT Next, we examine the validity of this conclusion using
allows for weak optical pulses containing n ⬃ 10 photons a full quantum field-theoretic approach 共Lukin and Ima-
to induce large nonlinear phase shifts, it cannot provide moglu, 2000兲. To this end, we replace the classical elec-
strong dissipation-free interactions between two single- tric fields describing the signal and probe fields with op-
photon pulses. The limitation, as the presented analysis erators Êi共z , t兲 = Ê共+兲共−兲
i 共z , t兲 + Êi 共z , t兲, where
has revealed, is due to the disparate group velocities of
the two interacting pulses. The natural question to ask,
then, is whether one can obtain larger nonlinear phase i 共z,t兲 = 兺
Ê共+兲
k
冑 បkc
2 ⑀ 0V
âki共z,t兲eik共z−ct兲 . 共90兲
shifts by assuring that both probe and signal pulses
travel at the same ultraslow group velocity. Here i = p,s and V denotes the quantization volume. We
Before proceeding with the analysis, we emphasize suppress the polarization index for convenience. In or-
that it is possible to consider an atomic medium with der to eliminate dissipation and simplify the Heisenberg
two coherently driven atomic species, each establishing equations for the quantized fields, we assume ⌫2 = 0 and
EIT at different frequencies. If the two atomic species 兩⌬␻42兩 Ⰷ ⌫4. To ensure, however, that we can adiabati-
are different isotopes of the same element, then the fre- cally eliminate atomic degrees of freedom, we also need
quencies at which the medium becomes transparent can to impose a finite bandwidth ⌬q on the quantized field.
differ by an amount that is on the order of the hyperfine With this assumption, the equal space-time commutation
splitting. We can then envision a scenario in which two relations explicitly depend on this bandwidth 共Lukin and
different light pulses, centered at the two transparency Imamoglu, 2000兲. Finally, to guarantee negligible pulse
frequencies, travel with the identical ultraslow group ve- spreading, we take ␶p−1 ⬍ ⌬q Ⰶ ⌬␻trans and arrive at a pair
locity through the medium containing an optically dense
mixture of these two isotopes 共A and B兲 of the same of operator equations for Êp共z , t兲 and Ês共z , t兲 that have
element. By adjusting the external magnetic field, it is the identical form to that given in Eq. 共85兲, provided we
also possible to ensure that one of these light pulses that replace c by vs on the right-hand side. These equations
sees EIT induced by species B will also be 共nearly兲 reso- can then be solved to yield
nant with the 兩2典-兩4典 transition of species A. By choosing
Êp共+兲共z,t兲 = Êp共+兲共0,t⬘兲
冋册
this pulse to be the signal pulse, one can realize a giant
Kerr interaction between two ultraslow light pulses ␮24
2
␶d
共Lukin and Imamoglu, 2000兲. An alternative scheme ⫻ exp i Ê共−兲共0,t⬘兲Ês共+兲共0,t⬘兲 ,
ប 4␲⌬␻42 s
2
based on only a single atomic species was recently pro-
posed by Petrosyan and Kuritzki 共2002兲.
Even though the analysis of nonlinear optical pro- Ês共+兲共z,t兲 = Ês共+兲共0,t⬘兲
cesses using single-photon pulses requires a full quan-
tum field-theoretical approach, we first consider the clas-
sical limit, following up on the analysis previously
⫻ exp i 冋 ␮13
2
␶d
册
Ê共−兲共0,t⬘兲Êp共+兲共0,t⬘兲 ,
ប 4␲⌬␻42 p
2
presented for a fast signal pulse. In the limit of a slow 共91兲

signal pulse with group velocity vs we obtain
where t⬘ = t − z / vp.
␾xpm =
⌫4 ns ␴24
⌬␻42 + i⌫4/2 A 2
冑 ln 2 ␶d
␲ ␶p
, 共88兲
Using the expressions in Eq. 共91兲, we can address the
question of nonlinear phase shift obtained during the
interaction of two single-photon pulses. To this end, we
which is identical in form to that given in Eq. 共86兲. In assume an input state of the form
兩1i典 = 兺 ␰kâki
contrast to that earlier result, however, the expression in †
兩0典共92兲
Eq. 共88兲 is valid even when ␶d Ⰷ ␶p 共Lukin and Imamo- k
glu, 2000兲.
To determine the maximum possible phase shift, we for both fields, where we require 兺k兩␰k兩2 = 1 to ensure
recall that for an optically dense medium the EIT trans- proper normalization. The spatiotemporal wave func-
mission peak has a width given by ⌬␻trans tion of each pulse is given by the single-photon ampli-
= ⍀2c / ⌫3冑␴13L. If we set ␶p = ⌬␻trans
−1
, we find tude, ⌿i共z , t兲 = 具0兩Ê共+兲
i 共z , t兲兩1i典. To evaluate the correla-

FIG. 23. Four-wave mixing schemes for the

generation of 共a兲 one field of frequency ␻4,
and 共b兲 two fields ␻2 and ␻4.
tions induced by the interaction, we concentrate on the vacuum speed of light. Harris and Hau showed that
two-photon amplitude within an undepleted pump approximation the conver-
sion efficiency, i.e., the ratio of the number of generated
⌿sp共z,t;z⬘,t⬘兲 = 具0兩Ês共+兲共z,t兲Êp共+兲共z⬘,t⬘兲兩1s,1p典. 共93兲 photons at ␻4 to the number of input photons at ␻1, is
One finds that the equal-time 共t = t⬘兲 local 共z = z⬘兲 corre- given by a dimensionless function ⌽共␩ , ␬L兲 multiplied
lations can be written as by the number of input photons per atomic cross sec-
tion:
⌿sp共z,t;z,t兲 = ⌿s共0,t兲⌿p共0,t兲ei␾ , 共94兲
nout nin
where ␾ = ␶d⌬q␴24⌫4 / 4␲A⌬␻42. Since the equal space 4
= ⌽共␩, ␬L兲 3
␴24 . 共95兲
and time commutation relations are proportional to ⌬q, nin
1 A3
there is explicit dependence on this quantity. The non- Here ␬ is the 共E-field兲 absorption coefficient at the 兩1典-兩4典
linear phase shift ␾ is the counterpart of the classical transition and L is the medium length. ␩ = L / Lh is the
phase ␾xpm obtained earlier in Eq. 共88兲; the equivalence ratio of the length Lh introduced in Eq. 共84兲, i.e., the
is obtained by replacing ␶p−1 with the quantization band- probe propagation length corresponding to a time delay
width ⌬q 共Lukin and Imamoglu, 2000兲. equal to the pulse width, to the medium length L. In the
A large single-photon nonlinear phase shift ␾ that ex- limit of an optically thin medium ␬L → 0 and of a small
ceeds ␲ rad can be used to implement two-qubit quan- group delay within the medium ␩ → 0, the usual result of
tum logic gates 共Nielsen and Chuang, 2000兲. However, perturbative long-pulse nonlinear optics is obtained,
we have seen that creating large phase shifts between which for Gaussian pulses reads
冋册
single-photon pulses also changes the mode profile of
the signal 共probe兲 pulse, conditioned on the presence of ln共2兲 1/2
T1
⌽共␩, ␬L兲 = ␩␬L. 共96兲
the probe 共signal兲 pulse. Therefore we expect entangle- ␲ 冑 T21 + T23
ment between the internal degrees of freedom of the
pulses, such as photon number or polarization, and the Here the T␮’s denote the temporal length of the pulses.
external degrees of freedom, such as mode profile deter- In the limit of ␩ → ⬁, i.e., if the group velocity reduction
mined by 兵␰k其, invalidating the simplistic assumption of of ␻1 is large, so that the ␻3 pulse sweeps over it very
ideal single-photon switching using large Kerr nonlinear- rapidly, the function is given by
冋册
ity.
ln共2兲 1/2
␬L
⌽共␩, ␬L兲 = exp共− 2␬L兲, 共97兲
␲ ␩
C. Few-photon four-wave mixing
i.e., falls off with ␩−1. A maximum of ⌽ of about 0.074 is
Besides the resonantly enhanced Kerr effect discussed attained when ␩ = ␬L = 1, i.e., when the pulse delay is just
in the previous subsections, nonlinear interactions on one pulse length and if the medium has an optical thick-
the few-photon level have been predicted and analyzed ness of one. In this case about ten photons per atomic
in greater detail in the four-wave mixing schemes shown cross section are sufficient to obtain perfect conversion.
in Fig. 23. In the first scheme discussed by Harris and In the second scheme, where the field ␻2 is generated
Hau 共1999兲, photons from two pulses 共␻1兲 and 共␻3兲 are 关see Fig. 23共b兲兴, a finite detuning from level 兩3典 is needed
converted into an electromagnetic-field pulse at ␻4 in to suppress absorption. Furthermore, in order to cancel
the presence of a strong monochromatic drive field at ␻2 ac-Stark shifts the frequencies ␻1 and ␻2 can be tuned
关Fig. 23共a兲兴. In the second scheme 关Fig. 23共b兲兴, discussed midway between two states with appropriate transition
by Johnsson et al. 共2002兲 for copropagating pump fields matrix elements 共Johnsson et al., 2002兲. Efficient fre-
and by Fleischhauer and Lukin 共2000兲 and by Lukin et quency conversion 共Hemmer et al., 1994; Ham, Shariar,
al. 共1999兲 for counterpropagating pump fields, the field et al., 1997; Babin et al., 1999; Hinze et al., 1999兲, genera-
␻2 is initially absent and spontaneously generated along tion of squeezing 共Lukin et al., 1999兲, as well as the pos-
with ␻4. sibility of mirrorless oscillations 共in counterpropagating
In the first scheme 关see Fig. 23共a兲兴, the input pulse ␻1 geometry兲 have been predicted 共Lukin, Hemmer, et al.,
experiences a group delay due to EIT on the 兩1典-兩3典 tran- 1998兲 and in part experimentally observed 共Zibrov et al.,
sition, while the second one, ␻3, moves almost with the 1999兲. It has been shown by Fleischhauer and Lukin

共2000兲 that an extremely low input photon flux ␾thr is Let us first discuss linear phenomena associated with
sufficient to reach the threshold of mirrorless oscilla- intracavity EIT. Consider a single three-level atom
tions: placed inside a resonator where the quantized resonator
mode acts as the coupling field between the internal
␾thr = fN␥12 , 共98兲 states 兩2典 and 兩3典. As shown by Field 共1993兲, electromag-
netically induced transparency on the probe transition is
where N is the number of atoms in the beam path and
already possible with few resonator photons or even
␥12 is the coherence decay rate of the 兩1典-兩2典 transition. f vacuum, provided that the atom and the resonator are
is a numerical prefactor of order unity. Johnsson et al. strongly coupled. To see this we note that the fully quan-
have shown that in the adiabatic limit and for ␥12 = 0, an tum interaction Hamiltonian
effective interaction Hamiltonian of the four-wave mix-
ing process can be derived, which reads for a one- ⍀p g
dimensional model 共Johnsson and Fleischhauer, 2002兲: Hint = − ប 兩3典具1兩 − ប â兩3典具2兩 + H.c. 共100兲
2 2
Hint =
បgc
2⌬
冕 dz
⍀̂†1⍀̂†3⍀̂2⍀̂4 + H.c.
⍀̂†3⍀̂3 + ⍀̂†4⍀̂4
, 共99兲
separates into effective three-level systems,
⍀p g冑n+1
兩1,n典↔ 兩3,n典 ↔ 兩2,n + 1典, 共101兲
where ⍀̂␮ are the operators corresponding to the 共com- where 1,2,3 denote the internal state of the atom and n
plex兲 Rabi frequencies of the fields and where a com- and n + 1 the number of photons in the resonator mode.
mon coupling constant g = ␮2␻ / 2ប␧0c of all transitions If g, which characterizes the atom-cavity coupling, is suf-
was assumed. One recognizes that, in contrast to the ficiently large and the atom is initially in state 兩1典, EIT
case of ordinary nonresonant nonlinear optics, the sum can be achieved for the probe field even for n = 0. Equa-
of the intensities of the resonant fields appears in the tion 共101兲 suggests another interesting application. The
denominator. As a consequence, in a copropagating ge- effective three-level systems have dark states. For ex-
ometry, the conversion length, i.e., the length after which ample,
a pair of photons in the input fields ␻1 and ␻3 is com-
pletely converted into a pair of photons in the fields, ␻2 兩D典 = cos ␪共t兲兩1,0典 − sin ␪共t兲兩2,1典, 共102兲
and ␻4, decreases with decreasing input intensity. Thus
with tan ␪共t兲 = ⍀p共t兲 / g. Thus stimulated Raman adiabatic
weak input fields lead to a faster conversion than stron-
passage from state 兩1,0典 to state 兩2,1典 induced by the
ger ones.
probe field offers the possibility of a controlled genera-
Finally, we note that although there had been a num-
tion of a single cavity photon. The potential usefulness
ber of theoretical proposals on EIT-based quantum non-
of this effect in cavity QED was first pointed out by
linear optics, it is only very recently that the first experi-
Parkins et al. 共1993兲. Law and Eberly proposed an appli-
mental demonstrations of the predicted effects have
cation for the generation of a single photon in a well-
been seen. Kuzmich and co-workers 共Kuzmich et al.,
defined wave packet 共Law and Eberly, 1996兲. Recently
2003兲 and independently van der Wal and collaborators
such a “photon pistol” was experimentally realized
共van der Wal et al., 2003兲 have demonstrated quantum
共Hennrich et al., 2000; Kuhn et al., 2002兲.
correlations between Stokes and anti-Stokes photons
The process of transferring excitation from atoms to a
generated with a controllable time delay by resonant
field mode is reversible and allows the opposite process
Raman scattering on atomic ensembles. The intermedi-
of mapping cavity-mode fields onto atomic ground-state
ate storage and subsequent retrieval of correlated pho-
coherences. In this case the resonator mode takes on the
tons in atomic ensembles are essential ingredients of the
role of the probe field, i.e.,
proposal of Duan and co-workers for a quantum re-
peater 共Duan et al., 2001兲, which is an important tool for g ⍀c
long-distance quantum communication. Hint = − ប â兩3典具1兩 − ប 兩3典具2兩 + H.c. 共103兲
2 2
and we have the coupling
D. Few-photon cavity EIT g冑n ⍀c共t兲
兩1,n典↔ 兩3,n − 1典 ↔ 兩2,n − 1典. 共104兲
It is well known in quantum optics that the presence
of high-finesse cavities can be used to enhance photon- The dark state for n = 0 is identical to that given in Eq.
photon interactions, for example, for the purpose of 共102兲, except tan ␪共t兲 = g / ⍀c共t兲 in the present case. The
quantum computation 共Nielsen and Chuang, 2000兲. It is mapping provides a possibility for measuring cavity
therefore natural to consider few-photon EIT inside a fields 共Parkins et al., 1995兲. It was shown, furthermore,
cavity, i.e., when the probe or the coupling field is re- by Pellizzari et al. 共1995兲 that a quantum-state transfer
placed by a single quantized radiation mode. We shall from one atom to a second atom via a shared cavity
show in this subsection that electromagnetically induced mode can be used to implement a quantum gate 共Pelliz-
transparency combined with cavity quantum electrody- zari et al., 1995兲. Finally, Cirac et al. 共1997兲 proposed an
namics can lead to a number of interesting linear and application to transfer quantum information between at-
nonlinear effects. oms at spatially separated nodes of a network.

All of the above discussed effects require, however, example, to build a controllable single-photon source or
the strong coupling of the single atom to the resonator a controlled-phase gate 共Duan and Kimble, 2004兲. Ima-
mode. This experimentally challenging requirement can moglu et al. suggested the use of a large ensemble of
be partly met if an optically thick ensemble of three- atoms in the resonator to enhance the photon blockade
level atoms is used. When an ensemble of N three-level effect. It was later shown 共Gheri et al., l999兲, however,
atoms interacts with a single resonator 共probe兲 mode that the large dispersion of the EIT medium does not
and a classical 共coupling兲 field according to allow for a collective enhancement of the nonlinear
N atom-cavity coupling. While photon blockade can be
ប
Hint = − 兺关gâ兩3典jj具1兩 + ប⍀c兩3典jj具2兩 + H.c.兴
achieved using either single-atom or multiatom EIT sys-
共105兲
2 j=1 tems 共Werner and Imamoglu, 1999兲, the strong-coupling
limit of cavity QED appears to be required in both
only symmetric collective states are coupled to the initial cases. It must be emphasized, however, that in contrast
state 兩1,1, . . . , 1典, where all N atoms are in the ground to a many-atom cavity-EIT medium, a collection of N
state 兩1典: Ⰷ 1 two-level atoms in a cavity is a nearly ideal linear
兩1N典 ⬅ 兩1, . . . ,1典, system, exhibiting a vanishingly small photon blockade
effect. The survival of nonlinearity in N Ⰷ 1 atom cavity
N EIT is related to the reduction of the cavity linewidth, or
1
兩1N−12典 ⬅
冑N 兺兩1, . . . ,2j, . . . ,1典, equivalently the width of the atom cavity dark state
j=1 共Lukin, Fleischhauer, et al., 1998; Werner and Imamoglu,
1999兲.
N
1
兩1 2 典⬅
N−2 2
兺兩1, . . . ,2i, . . . ,2j, . . . ,1典,
冑2N共N − 1兲 i⫽j=1 VII. SUMMARY AND PERSPECTIVES
etc. 共106兲 EIT has had a profound impact upon optical science.
Hopefully we have succeeded in explaining the relation-
If the initial photon number is n = 1, again a simple ship between EIT and the earlier related ideas of coher-
three-level coupling scheme emerges: ent preparation of atoms by fields and especially linked
g冑N ⍀c共t兲 it with the concept of the dark state. We stress again that
兩1N,1典 ↔ 兩1N−13,0典 ↔ 兩1N−12,0典. 共107兲 the distinct feature of EIT and related phenomena, in
contrast to the earlier spectroscopic tools such as coher-
Due to the symmetric interaction of all N atomic dipoles ent population trapping, is that they occur in media that
with the resonator mode, the coupling constant is, how- are optically dense and cause not only a modification of
ever, collectively enhanced: the excitation state of the matter but also significant
g → g冑N.
changes to the optical fields themselves. It is this latter
共108兲
point that is crucial to understanding the importance of
Thus controlled dark-state Raman adiabatic passage EIT in optics. EIT provides a new means to change the
from a state with a single photon in the resonator to a optical characteristics of matter, for instance, the degree
single collective excitation and vice versa is possible of absorption or refraction or the group velocity, and so
without the requirement of a strong-coupling limit of provides a way to alter the propagation of optical fields
cavity QED, defined here as g2 Ⰷ ␬⌫3, where ␬ is the and to enhance the generation of new fields.
cavity decay rate. For higher photon numbers, more At the time of writing 共2004兲 we are some 14 years on
complicated coupling schemes emerge. All of them do from the first experimental demonstration of the phe-
possess, however, dark states which allow a parallel nomenon of EIT. Much research activity has been un-
transfer of arbitrary photon states with n Ⰶ N to collec- dertaken in the period since this discovery, in an effort
tive excitations, which has important applications for to understand the phenomenon and its various implica-
quantum memories for photons and quantum network- tions. More importantly there has been a concerted ef-
ing 共Lukin, Yelin, and Fleischhauer, 2000兲. fort in many laboratories to harness the effect for appli-
As discussed in previous subsections, EIT can be used cations. It is informative to review the extent to which
to achieve a resonant enhancement of nonlinear interac- this field has grown, and has remained a healthy and
tions leading, for example, to a cross-phase modulation. active subject with no sign of declining activity. There
Large phase shifts produced by the interaction of indi- continue to be new and interesting proposals and dem-
vidual photons are very appealing for their potential use onstrations that extend EIT to fresh applications. It has
in quantum information processing. Imamoglu et al. become a standard tool for the study of optical proper-
共1997兲 showed that if a medium with sufficiently large ties of atomic and molecular gases. For these reasons of
Kerr nonlinearity is put into an optical cavity, it can lead durability and utility we believe that EIT has earned a
to a photon blockade effect. When a photon enters a significant place in optical physics that make it a valu-
previously empty cavity, the induced refractive index de- able topic of study for new research students and expe-
tunes the cavity resonance. If this frequency shift is rienced researchers alike.
larger than the cavity linewidth, a second photon cannot We have tried in this review to cover the main ideas
enter and will be reflected. This could be employed, for behind the application of EIT and related dark-state

phenomena to optics. Some omissions were inevitable. combined with a second longer-wavelength tunable laser
Much work has been published in this field, especially in to excite two-photon transitions to highly excited states,
the theoretical area, and it was simply not possible to or perhaps in Raman spectroscopy. The fixed frequency
cover or cite every single contribution. We humbly nature of the output would appear to preclude a wider
apologize to anyone we have failed to cite. Our inten- range of applications.
tion was primarily to provide the reader new to the field While EIT-type four-wave mixing enhancement may
with an accessible and comprehensive framework that be of limited utility, since it results in a fixed frequency
would support further, more detailed, explorations of output, this is not, in principle, a limit for off-resonant
this topic. schemes with maximal coherence. The high conversion
Where possible we have tried to connect the theory efficiencies already demonstrated in the far UV may be
with the results of real experiments. Again we have been extended into the VUV. There is a problem with
selective, choosing illustrative examples rather than at- Raman-like excitations: a limit to the degree of up-
tempting a comprehensive review of all work. Because conversion as the initial excitation is via difference-
of space constraints we have not explored in any struc- frequency mixing of the applied fields. The inherent
tured way the experimental techniques used in these ex- limit is set at around 170 nm for the shortest wavelength
periments, either to achieve robust EIT or to investigate that can be generated due to the energy-level structure
EIT phenomena. This omission is regrettable, but the of the atoms and the shortest available wavelengths
situations in which EIT can be observed are too diverse from lasers. For shorter-wavelength extensions an im-
to be covered in depth here. The reader is urged to con- portant problem that must be solved is how to excite
sult the cited literature and where necessary other re- maximal coherence between the ground state and a very
views and textbooks to obtain a fuller understanding of highly excited upper state. One approach is to use sum-
the experimental methods. frequency excitation by the applied lasers. The creation
One feature of the work in this field is that the vast of maximal coherence in this case cannot use the con-
majority of investigations have been carried out in the ventional STIRAP-type adiabatic excitation schemes.
gas phase. This trend runs somewhat contrary to a gen- Recently a promising alternative has been demon-
eral tendency for nonlinear optics to move towards em- strated, Stark chirped rapid adiabatic passage 共SCRAP兲,
ploying solid-state media. The reason for this is that gas- with the potential for efficient excitation of maximal co-
phase media still offer the physicist unique features; herence between the ground state and a very highly ex-
small homogeneous and inhomogeneous linewidths cited state of an atom 共Rickes et al., 2000兲. A recent first
compared to most solid-state media, transparency over demonstration of the use of this technique showed a de-
much of the IR-VUV range, the possibility of using laser gree of enhancement to extreme-UV generation via
cooling, and high thresholds against optical damage. third-harmonic generation 共Rickes et al., 2003兲, but fur-
Moreover, a problem normally identified with gas-phase ther work is needed in this area.
nonlinear optical media, the low value of the nonlinear Besides the potential for efficient generation of coher-
susceptibility compared to many solid-density optical ent radiation in new frequency domains, resonant non-
crystals, is circumvented by EIT since it permits the use linear optics based on EIT will be of growing impor-
of resonance to enhance the susceptibility. tance for controlled nonlinear interactions at the few-
Having reviewed the subject and examined a number photon level. The above-mentioned limitations
of applications that have already been thoroughly inves- concerning tunability and accessible frequencies are of
tigated, we now turn to the future. While predictions are no relevance here. Photons are ideal carriers of quantum
always risky, we offer some thoughts on what new appli- information, yet processing of this information, for ex-
cations are likely to emerge in this field in the hope of ample, in a quantum computer, is quite difficult. The
stimulating further ideas. We shall confine ourselves to reason for this is the weakness of photon-photon inter-
areas where there is at least some current activity and so actions in the usual nonlinear media. Here EIT may of-
some evidence upon which to base our extrapolations. fer new directions. Although some theoretical proposals
The potential of EIT and maximal coherence to gen- exist, the full potential of EIT-based nonlinear optics for
erate high-brightness coherent fields in the short- these applications has still not been explored, especially
wavelength range remains only partially explored. on the experimental side.
EIT has been shown to enhance frequency-conversion Another feature of EIT, which makes it very attractive
efficiencies to the 0.01 level in up-conversion to the for future applications in quantum information, is the
100-nm frequency range. By careful choice of the atomic possibility of transferring coherence and quantum states
system and through judicious optimization of the density from light to collective atomic spin excitations. Quan-
and length of the medium and the use of transform- tum memories for photons 共Lukin, Yelin, and Fleisch-
limited lasers to drive all the fields, it may be possible to hauer, 2000; Fleischhauer and Lukin, 2001兲, as discussed
improve the conversion efficiency by a substantial fur- in Sec. III, are just one potential avenue; quantum net-
ther factor. This could result in the generation of works based on them, including quantum repeaters, are
transform-limited nanosecond pulses 共spectral line- another 共Kuzmich et al., 2003; van der Wal et al., 2003兲.
widths of ⬃100 MHz兲 of VUV radiation with energies of Combining the state mapping techniques with atom-
tens of microjoules. These could be of use in a number atom interactions in mesoscopic samples may provide
of applications in nonlinear spectroscopy, for instance, if new tools for generating photon wave packets with tai-

lored quantum states or for manipulating these states. A ion-doped 共insulating兲 crystals 共Kuznetsova et al., 2002兲.
first proposal in this direction using a dipole blockade of In both cases, the ratio of the coherence relaxation rates
Rydberg excitations was given by Lukin et al. 共2001兲. of the dipole-allowed and spin-flip transitions can be as
The controlled coupling of a photon to many atomic large as 104, indicating that EIT could be implemented
spins in EIT could be of interest for the preparation or efficiently.
the probing of entangled many-particle states. We have Most of the experimental efforts aimed at demonstrat-
just begun to explore the role of entanglement in many- ing EIT in solid-state spins have so far focused on rare-
body physics and quantum information. Laser manipu- earth ion-doped crystals 共Ham, Hemmer, and Shahiar,
lation of spin ensembles via EIT could provide a very 1997; Turukhin et al., 2001兲 and nitrogen vacancy centers
useful tool in this quest. in diamond 共Wei and Manson, 1999兲. In experiments car-
EIT also offers new possibilities in linear and nonlin- ried out using Pr-doped Y2SiO5 at cryogenic tempera-
ear matter-wave optics. As pointed out in Sec. IV, slow tures 共T = 5 K兲, Turukhin and co-workers have observed
light in atomic gases corresponds to a quasiparticle, group velocities as slow as 45 m / s and a group delay of
which is a mixed electromagnetic-matter excitation. For 66 ␮s for light pulses with a pulse width of 50 ␮s 共Tu-
very low group velocities almost all excitation is concen- rukhin et al., 2001兲. Storage and retrieval of these pulses
trated in atomic spins while the propagation properties have also been demonstrated 共Turukhin et al., 2001兲.
are still mostly determined by the electromagnetic part. While the ratio of the group delay to the pulse width in
Thus slow light in ultracold atomic gases provides a new these experiments is on the order of unity, this is an
way to control the propagation of matter waves with impressive and promising development for solid-state
potential applications in matter-wave interferometry EIT.
共Zimmer and Fleischhauer, 2004兲 or if atom-atom inter- Optoelectronics and photonics technology are largely
actions are included, in nonlinear matter-wave optics based on semiconductor heterostructures formed out of
共Masalas and Fleischhauer, 2004兲. column-III–V semiconductors. The conduction-band
Ultrafast measurements have recently entered the at- electrons of these elements have predominantly s-type
tosecond domain 共Drescher et al., 2002兲. The possibility wave functions with small spin-orbit coupling. As a re-
of employing the highly efficient multiorder Raman gen- sult, the spin coherence times are four orders of magni-
eration that results from an adiabatically prepared maxi- tude longer than the radiative recombination rate of ex-
mal vibrational coherence has already been investigated. citons. The possibility of observing efficient EIT and
A key problem to solve before this approach can be slow light propagation in a doped GaAs quantum well in
widely employed for ultrafast measurements will be the the quantum Hall regime has been discussed 共Imamoglu,
synchronization of the train of subfemtosecond pulses 2000兲. While this particular realization requires low tem-
that are generated with external events. One possibility peratures and high magnetic fields 共B ⬃ 10 T兲, long spin
共suggested by Sokolov, 2002兲 is to use the pulse train to coherence times in GaAs have been observed even at
“measure” processes in the modulated molecules them- room temperature and with vanishing magnetic fields
selves so that the synchronization is automatically satis- 共Wolf et al., 2001兲. In fact, there is substantial activity in
fied. the emerging field of spintronics 共Wolf et al., 2001; Zutic
As we have emphasized throughout this review, coher- et al., 2004兲, and it is plausible to argue that EIT could
ent preparation techniques that are at the heart of EIT play a role by providing long optical delays and/or en-
can be easily implemented in optically dense atomic abling “spin-photon” information transfer.
samples in the gas phase. It is generally argued, however, Yet another interesting extension of gas-phase EIT to
that the range of applications can be significantly en- plasmas was suggested by Harris 共1996兲. Here classical
larged if EIT is implemented in solid-state media. The interference effects in parametric wave interactions
obvious roadblock in this quest are the ultrashort coher- which are analogous to interference effects in three-level
ence times of optical transitions in solids. Even for tran- atoms lead to an induced transparency window below
sitions that are not dipole allowed, the coherence times the plasma frequency. This could have interesting appli-
are typically on subnanosecond time scales. Among the cations for magnetic fusion, plasma heating, and plasma
many different physical mechanisms contributing to this diagnostics. For some recent work on this see, Litvak
fast dephasing, perhaps the most fundamental one is the and Tokman 共2002兲 and Shvets and Wurtele 共2002兲.
interaction of electrons with lattice vibrations 共phonons兲. Even though EIT is by now a well-established tech-
While cooling the samples to liquid-helium tempera- nique, new applications continue to appear regularly. It
tures reduces the average number of phonons and was recently proposed, for example, that a “dynamical
thereby phonon-induced dephasing, the coherence times photonic band-gap” structure could be realized in an
still remain significantly shorter than their atomic coun- EIT medium by spatially modifying the refractive index
terparts. with a standing-wave optical field 共Andre and Lukin,
Fortunately, there is an exception to the general rule 2002兲. This could be achieved by making use of the large
of ultrashort dephasing times in solids: electron spin de- absorption-free dispersion around the transparency fre-
grees of freedom in a large class of solid-state materials quency. EIT may also allow for the realization of a high-
have relatively long coherence times, ranging from 1 ␮s efficiency photon counter. After mapping the quantum
in optically active direct-band-gap bulk semiconductors state of a propagating light pulse onto collective hyper-
such as GaAs 共Wolf et al., 2001兲 to 1 ms in rare-earth fine excitations of a trapped cold atomic gas, it is pos-

sible to monitor the resonance fluorescence induced by tum Optics: Lectures Presented at the Université Libre de
an additional laser field that only couples to the hyper- Bruxelles, October 28 to November 4, 1991 共Springer, Berlin兲.
fine excited state 共Imamoglu, 2002; James and Kwiat, Cerboneschi, E., and E. Arimondo, 1995, Phys. Rev. A 52,
2002兲. Even with a fluorescence collection/detection ef- R1823.
ficiency as low as 10%, it should be possible to achieve Cirac, J. I., P. Zoller, H. J. Kimble, and H. Mabuchi, 1997,
photon counting with an efficiency exceeding 99%. Such Phys. Rev. Lett. 78, 3221.
a device could be of great use in quantum optical infor- Cohen-Tannoudji, C., J. Dupont-Roc, and G. Grynberg, 1992,
mation processing 共Knill et al., 2001兲. Atom-Photon Interactions 共Wiley, New York兲.
Csesznegi, J. R., and R. Grobe, 1997, Phys. Rev. Lett. 79, 3162.
Dalibard, J., Y. Castin, and K. Molmer, 1992, Phys. Rev. Lett.
ACKNOWLEDGMENTS 68, 580.
Dalton, B. J., and P. L. Knight, 1982, Opt. Commun. 41, 411.
We would like to express our gratitude to Steve Har- Dorman, C., I. Kucukkara, and J. P. Marangos, 2000a, Opt.
ris, who has been a source of inspiration through his Commun. 180, 263.
innovative ideas, insights, and encouragement. Our nu- Dorman, C., I. Kucukkara, and J. P. Marangos, 2000b, Phys.
merous discussions and collaborations with Klaas Berg- Rev. A 61, 013802.
mann, Peter Knight, Misha Lukin, Marlan Scully, Moshe Drescher, M., M. Hentschel, R. Kienberger, M. Uiberacker, V.
Shapiro, and Bruce Shore have played a key role in Yakovlev, A. Scrinizi, T. Westerwalbesloh, U. Kleineberg, U.
shaping our understanding of various aspects of EIT. Heinzmann, and F. Krausz, 2002, Nature 共London兲 419, 803.
A.I. would also like to thank John Field for truly en- Duan, L. M., J. I. Cirac, and P. Zoller, 2002, Phys. Rev. A 66,
lightening discussions dating from the very early days of 023818.
EIT. Finally, we acknowledge K. Boller, K. Hakuta, L. Duan, L. M., J. I. Cirac, P. Zoller, and E. S. Polzik, 2000, Phys.
Hau, O. Kocharovskaya, and H. Schmidt. Rev. Lett. 85, 5643.
Duan, L. M., and H. J. Kimble, 2004, Phys. Rev. Lett. 92,
127902.
REFERENCES
Duan, L. M., M. D. Lukin, J. I. Cirac, and P. Zoller, 2001,
Affolderbach, C., M. Stahler, S. Knappe, and R. Wynands, Nature 共London兲 414, 413.
2002, Appl. Phys. B: Lasers Opt. 75, 605. Dutton, Z., M. Budde, C. Slowe, and L. V. Hau, 2001, Science
Agarwal, G. S., 1993, Phys. Rev. Lett. 71, 1351. 293, 5530.
Alzetta, G., A. Gozzini, L. Moi, and G. Orriols, 1976, Nuovo Eberly, J. H., M. L. Pons, and H. R. Haq, 1994, Phys. Rev.
Cimento Soc. Ital. Fis., B 36, 5. Lett. 72, 56.
Andre, A., and M. Lukin, 2002, Phys. Rev. Lett. 89, 143602. Faist, J., F. Capasso, C. Sirtori, K. W. West, and L. Pfeiffer,
Arimondo, E., 1996, Prog. Opt. 35, 259. 1997, Nature 共London兲 390, 589.
Arimondo, E., and G. Orriols, 1976, Lett. Nuovo Cimento Soc. Fano, U., 1961, Phys. Rev. 124, 1866.
Ital. Fis. 17, 333. Fano, U., and J. Cooper, 1965, Phys. Rev. 137, 1364.
Armstrong, J. A., and J. J. Wynne, 1974, Phys. Rev. Lett. 33, Field, J. E., 1993, Phys. Rev. A 47, 5064.
1183. Field, J. E., K. Hahn, and S. E. Harris, 1991, Phys. Rev. Lett.
Armstrong, Lloyd, Jr., and Brian Lee Beers, 1975, Phys. Rev. 67, 3733.
Lett. 34, 1290. Fleischhauer, M., 1994, Phys. Rev. Lett. 72, 989.
Autler, S. H., and C. H. Townes, 1955, Phys. Rev. 100, 703. Fleischhauer, M., 1999, Opt. Express 4, 107.
Babin, S., U. Hinze, E. Tiemann, and B. Wellegehausen, 1996, Fleischhauer, M., C. H. Keitel, M. O. Scully, and C. Su, 1992a,
Opt. Lett. 21, 1186. Opt. Commun. 87, 238.
Babin, S. A., E. V. Podivilov, D. A. Shapiro, U. Hinze, E. Fleischhauer, M., C. H. Keitel, M. O. Scully, C. Su, B. T. Ul-
Tiemann, and B. Wellegehausen, 1999, Phys. Rev. A 59, 1355. rich, and S. Y. Zhu, 1992b, Phys. Rev. A 46, 1468.
Bajcsy, M., A. S. Zibrov, and M. D. Lukin, 2003, Nature Fleischhauer, M., and M. D. Lukin, 2000, Phys. Rev. Lett. 84,
共London兲 426, 638. 5094.
Bergmann, K., H. Theuer, and B. W. Shore, 1998, Rev. Mod. Fleischhauer, M., and M. D. Lukin, 2001, Phys. Rev. A 64,
Phys. 70, 1003. 022314.
Boller, K. J., A. Imamoglu, and S. E. Harris, 1991, Phys. Rev. Fleischhauer, M., M. D. Lukin, A. B. Matsko, and M. O.
Lett. 66, 2593. Scully, 2000, Phys. Rev. Lett. 84, 3558.
Bol’shov, L. A., and V. V. Likhanskii, 1985, Sov. J. Quantum Fleischhauer, M., and A. S. Manka, 1996, Phys. Rev. A 54, 794.
Electron. 15, 889. Fleischhauer, M., and C. Mewes, 2002, in Proceedings of the
Braje, D. A., V. Balic, G. Y. Yin, and S. E. Harris, 2003, Phys. International School of Physics “Enrico Fermi”: Experimental
Rev. A 68, 041801共R兲. Quantum Computation and Information, edited by F. De
Buckle, S. J., S. M. Barnett, P. L. Knight, M. A. Lauder, and D. Martini and C. Monroe 共IOS, Amsterdam/Washington, D.C.兲,
T. Pegg, 1986, Opt. Acta 33, 2473. pp. 511–529.
Budker, D., D. F. Kimball, S. M. Rochester, and V. V. Yash- Fleischhauer, M., and M. O. Scully, 1994, Phys. Rev. A 49,
chuk, 1999, Phys. Rev. Lett. 83, 1767. 1973.
Budker, D., V. Yashchuk, and M. Zolotorev, 1998, Phys. Rev. Fry, E. S., X. Li, D. Nikonov, G. G. Padmabandu, M. O.
Lett. 81, 5788. Scully, A. V. Smith, F. K. Tittel, C. Wang, S. R. Wilkinson,
Buffa, R., M. P. Anscombe, and J. P. Marangos, 2003, Phys. and S.-Y. Zhu, 1993, Phys. Rev. Lett. 70, 3235.
Rev. A 67, 033801. Gao, J. C., C. Guo, G. Jin, P. Wang, J. Zhao, H. Zhang, Y.
Carmichael, H. J., 1993, An Open Systems Approach to Quan- Jiang, D. Wang, and D. Jiang, 1992, Opt. Commun. 93, 323.

Gardiner, C. W., A. S. Parkins, and P. Zoller, 1992, Phys. Rev. Jain, M., 1994, Phys. Rev. A 50, 1899.
A 46, 4363. Jain, M., A. J. Merriam, A. Kasapi, G. Y. Yin, and S. E. Harris,
Garton, W. R. S., 1966, Adv. At. Mol. Phys. 2, 93. 1995, Phys. Rev. Lett. 75, 4385.
Gaubatz, U., P. Rudecki, S. Schiemman, and K. Bergmann, Jain, M., H. Xia, G. Y. Yin, A. J. Merriam, and S. E. Harris,
1990, J. Chem. Phys. 92, 5363. 1996, Phys. Rev. Lett. 77, 4326.
Gheri, K. M., W. Alge, and P. Grangier, 1999, Phys. Rev. A 60, James, D. F. V., and P. Kwiat, 2002, Phys. Rev. Lett. 89, 183601.
R2673. Johnsson, M., E. Korsunsky, and M. Fleischhauer, 2002, Opt.
Gornyi, M. B., B. G. Matisov, and Y. V. Rozhdestvenskii, 1989, Commun. 212, 335.
Sov. Phys. JETP 68, 728. Johnsson, M. T., and M. Fleischhauer, 2002, Phys. Rev. A 66,
Gray, H. R., R. M. Whitley, and C. R. Stroud, 1978, Opt. Lett. 043808.
3, 218. Julsgaard, B., A. Kozhekin, and E. S. Polzik, 2001, Nature
Grigoryan, G. G., and Y. T. Pashayan, 2001, Phys. Rev. A 64, 共London兲 413, 400.
013816. Juzeliunas, G., and H. J. Carmichael, 2002, Phys. Rev. A 65,
Grobe, R., F. T. Hioe, and J. H. Eberly, 1994, Phys. Rev. Lett. 021601共R兲.
73, 3183. Juzeliunas, G., M. Masalas, and M. Fleischhauer, 2003, Phys.
Hahn, K. H., D. A. King, and S. E. Harris, 1990, Phys. Rev. Rev. A 67, 023809.
Lett. 65, 2777. Kasapi, A., M. Jain, G. Y. Yin, and S. E. Harris, 1995, Phys.
Hakuta, K., 2004, private communication. Rev. Lett. 74, 2447.
Hakuta, K., M. Suzuki, M. Katsuragawa, and J. Z. Li, 1997, Kasapi, A., G. Y. Yin, M. Jain, and S. E. Harris, 1996, Phys.
Phys. Rev. Lett. 79, 209. Rev. A 53, 4547.
Hald, J., J. L. Sorensen, C. Schori, and E. S. Polzik, 1999, Phys. Kash, M. M., V. A. Sautenkov, A. S. Zibrov, L. Hollberg, G. R.
Rev. Lett. 83, 1319. Welch, M. D. Lukin, Y. Rostovtsev, E. S. Fry, and M. O.
Ham, B. S., P. R. Hemmer, and M. S. Shahiar, 1997, Opt. Scully, 1999, Phys. Rev. Lett. 82, 5229.
Commun. 144, 227. Kien, F. E., J. Q. Liang, M. Katsuragawa, K. Ohtsuki, K.
Ham, B. S., M. S. Shariar, M. K. Kim, and P. R. Hemmer, 1997, Hakuta, and A. V. Sokolov, 1999, Phys. Rev. A 60, 1562.
Opt. Lett. 22, 1849. Knight, P. L., 1990, Phys. Rep. 190, 1.
Hänsch, T. W., and P. Toschek, 1970, Z. Phys. 236, 213. Knill, E., R. Laflamme, and G. J. Milburn, 2001, Nature
Harris, S. E., 1989, Phys. Rev. Lett. 62, 1033. 共London兲 409, 46.
Harris, S. E., 1993, Phys. Rev. Lett. 70, 552. Kocharovskaya, O., 1992, Phys. Rep. 219, 175.
Harris, S. E., 1994a, Phys. Rev. Lett. 72, 52. Kocharovskaya, O., Y. Rostovtsev, and M. O. Scully, 2001,
Harris, S. E., 1994b, Opt. Lett. 19, 2018. Phys. Rev. Lett. 86, 628.
Harris, S. E., 1996, Phys. Rev. Lett. 77, 5357. Kocharovskaya, O. A., and Y. I. Khanin, 1986, Sov. Phys. JETP
Harris, S. E., 1997, Phys. Today 50 共7兲, 36. 63, 945.
Harris, S. E., 2000, Phys. Rev. Lett. 85, 4032. Kocharovskaya, O. A., and Y. I. Khanin, 1988, JETP Lett. 48,
Harris, S. E., J. E. Field, and A. Imamoglu, 1990, Phys. Rev. 630.
Lett. 64, 1107. Konopnicki, M. J., P. D. Drummond, and J. H. Eberly, 1981,
Harris, S. E., J. E. Field, and A. Kasapi, 1992, Phys. Rev. A 46, Opt. Commun. 36, 313.
R29. Konopnicki, M. J., and J. H. Eberly, 1981, Phys. Rev. A 24,
Harris, S. E., and L. V. Hau, 1999, Phys. Rev. Lett. 82, 4611. 2567.
Harris, S. E., and M. Jain, 1997, Opt. Lett. 22, 636. Korsunsky, E. A., and D. V. Kosachiov, 1999, Phys. Rev. A 60,
Harris, S. E., and Z.-F. Luo, 1995, Phys. Rev. A 52, R928. 4996.
Harris, S. E., and A. V. Sokolov, 1997, Phys. Rev. A 55, R4019. Korsunsky, E. A., N. Leinfellner, A. Huss, S. Baluschev, and
Harris, S. E., and Y. Yamamoto, 1998, Phys. Rev. Lett. 81, L. Windholz, 1999, Phys. Rev. A 59, 2302.
3611. Kozhekin, A. E., K. Molmer, and E. Polzik, 2000, Phys. Rev. A
Harris, S. E., G. Y. Yin, M. Jain, H. Xia, and A. J. B. Merriam, 62, 033809.
1997, Philos. Trans. R. Soc. London, Ser. A 355, 2291. Kozlov, V. V., S. Wallentowitz, and S. Raghavan, 2002, Phys.
Hau, L. V., S. E. Harris, Z. Dutton, and C. H. Behroozi, 1999, Lett. A 296, 210.
Nature 共London兲 397, 594. Kozuma, M., D. Akasmatsu, L. Denf, E. W. Hagley, and M. G.
Hemmer, P. R., K. Z. Cheng, J. Kierstead, M. S. Shariar, and Payne, 2002, Phys. Rev. A 66, 031801共R兲.
M. K. Kim, 1994, Opt. Lett. 19, 296. Kuhn, A., M. Hennrich, and G. Rempe, 2002, Phys. Rev. Lett.
Hemmer, P. R., D. P. Katz, J. Donoghue, M. Cronin-Golomb, 89, 067901.
M. S. Shariar, and P. Kumar, 1995, Opt. Lett. 20, 769. Kuklinski, J. R., U. Gaubatz, T. F. Hioe, and K. Bergmann,
Hennrich, M., T. Legero, A. Kuhn, and G. Rempe, 2000, Phys. 1989, Phys. Rev. A 40, 6741.
Rev. Lett. 85, 4872. Kuzmich, A., W. P. Bowen, A. D. Boozer, A. Bocha, C. W.
Hinze, U., L. Meyer, B. N. Chichkov, E. Tiemann, and B. Chou, L. M. Duan, and H. J. Kimble, 2003, Nature 共London兲
Wellegehausen, 1999, Opt. Commun. 166, 127. 423, 731.
Hioe, F. T., and R. Grobe, 1994, Phys. Rev. Lett. 73, 2559. Kuzmich, A., L. Mandel, and N. P. Bigelow, 2000, Phys. Rev.
Imamoglu, A., 1991, Ph.D. thesis 共Stanford University兲. Lett. 85, 1594.
Imamoglu, A., 2000, Opt. Commun. 179, 179. Kuzmich, A., L. Mandel, J. Janis, Y. E. Young, R. Ejnisman,
Imamoglu, A., 2002, Phys. Rev. Lett. 89, 163602. and N. P. Bigelow, 1999, Phys. Rev. A 60, 2346.
Imamoglu, A., and S. E. Harris, 1989, Opt. Lett. 14, 1344. Kuzmich, A., K. Molmer, and E. S. Polzik, 1997, Phys. Rev.
Imamoglu, A., H. Schmidt, G. Woods, and M. Deutsch, 1997, Lett. 79, 4782.
Phys. Rev. Lett. 79, 1467. Kuznetsova, E., O. Kocharovskaya, P. Hemmer, and M. O.

Scully, 2002, Phys. Rev. A 66, 063802. Nielsen, M., and I. Chuang, 2000, Quantum Computation and
Law, C. K., and J. H. Eberly, 1996, Phys. Rev. Lett. 76, 1055. Quantum Information 共Cambridge University Press, Cam-
Leonhardt, U., and P. Piwnicki, 1999, Phys. Rev. A 60, 4301. bridge, England兲.
Leonhardt, U., and P. Piwnicki, 2000a, Phys. Rev. Lett. 84, 822. Nottlemann, A., C. Peters, and W. Lange, 1993, Phys. Rev.
Leonhardt, U., and P. Piwnicki, 2000b, Phys. Rev. Lett. 85, Lett. 70, 1783.
5253. Oreg, J., F. T. Hioe, and J. H. Eberly, 1984, Phys. Rev. A 29,
Litvak, A., and M. D. Tokman, 2002, Phys. Rev. Lett. 88, 690.
095003. Padmabandu, G. G., G. R. Welch, I. N. Shubin, E. S. Fry, D. E.
Liu, C., Z. Dutton, C. H. Behroozi, and L. V. Hau, 2001, Na- Nikonov, M. D. Lukin, and M. O. Scully, 1996, Phys. Rev.
ture 共London兲 409, 490. Lett. 76, 2053.
Lounis, B., and C. Cohen-Tannoudji, 1992, J. Phys. II 2, 579. Parkins, A. S., P. Marte, P. Zoller, O. Carnal, and H. J. Kimble,
Lukin, M. D., 2003, Rev. Mod. Phys. 75, 457. 1995, Phys. Rev. A 51, 1578.
Lukin, M. D., M. Fleischhauer, R. Cote, L. M. Duan, D. Parkins, A. S., P. Marte, P. Zoller, and H. J. Kimble, 1993,
Jaksch, J. I. Cirac, and P. Zoller, 2001, Phys. Rev. Lett. 87, Phys. Rev. Lett. 71, 3095.
037901. Pellizzari, T., S. A. Gardiner, J. I. Cirac, and P. Zoller, 1995,
Lukin, M. D., M. Fleischhauer, M. O. Scully, and V. L. Veli- Phys. Rev. Lett. 75, 3788.
chansky, 1998, Opt. Lett. 23, 295. Petch, J. C., C. H. Keitel, P. L. Knight, and J. P. Marangos,
Lukin, M. D., M. Fleischhauer, A. S. Zibrov, H. G. Robinson, 1996, Phys. Rev. A 53, 543.
V. L. Velichansky, L. Hollberg, and M. O. Scully, 1997, Phys. Petrosyan, D., and G. Kurizki, 2002, Phys. Rev. A 65, 033833.
Rev. Lett. 79, 2959. Phillips, D., A. Fleischhauer, A. Mair, R. Walsworth, and M.
Lukin, M. D., P. Hemmer, and M. O. Scully, 2000, Adv. At., D. Lukin, 2001, Phys. Rev. Lett. 86, 783.
Mol., Opt. Phys. 42, 347. Polzik, E. S., 1999, Phys. Rev. A 59, 4202.
Lukin, M. D., P. R. Hemmer, M. Löffler, and M. O. Scully, Rahman, A., 1999, Phys. Rev. A 60, 4187.
1998, Phys. Rev. Lett. 81, 2675. Rahman, A., and J. H. Eberly, 1998, Phys. Rev. A 58, R805.
Lukin, M. D., and A. Imamoglu, 2000, Phys. Rev. Lett. 84, Rathe, U., M. Fleischhauer, S. Y. Zhu, T. W. Hänsch, and M.
1419. O. Scully, 1993, Phys. Rev. A 47, 4994.
Lukin, M. D., and A. Imamoglu, 2001, Nature 共London兲 273, Reintjes, J. F., 1984, Nonlinear Optical Parametric Processes in
273. Liquids and Gases 共Elsevier Science/Academic, Boston兲.
Lukin, M. D., A. B. Matsko, M. Fleischhauer, and M. O. Rickes, T., J. P. Marangos, and T. Halfmann, 2003, Opt. Com-
Scully, 1999, Phys. Rev. Lett. 82, 1847. mun. 277, 133.
Lukin, M. D., S. F. Yelin, and M. Fleischhauer, 2000, Phys. Rev. Rickes, T., L. P. Yatsenko, S. Steuerwald, T. Halfmann, B. W.
Lett. 84, 4232. Shore, N. V. Vitanov, and K. Bergmann, 2000, J. Chem. Phys.
Madden, R. B., and K. Codling, 1965, Astrophys. J. 141, 364. 113, 534.
Mair, A., J. Hager, D. F. Phillips, R. L. Walsworth, and M. D. Ruostekoski, J., and D. F. Walls, 1999a, Eur. Phys. J. D 5, 335.
Lukin, 2002, Phys. Rev. A 65, 031802共R兲. Ruostekoski, J., and D. F. Walls, 1999b, Phys. Rev. A 59,
Malmistov, A. I., 1984, Sov. J. Quantum Electron. 14, 385. R2571.
Mandel, P., 1993, Contemp. Phys. 34, 235. Sautenkov, V. A., M. D. Lukin, C. J. Bednar, I. Novikova, E.
Masalas, M., and M. Fleischhauer, 2004, Phys. Rev. A 69, Mikhailov, M. Fleischhauer, V. L. Velichansky, G. R. Welch,
061801共R兲. and M. O. Scully, 2000, Phys. Rev. A 62, 023810.
Matsko, A. B., O. Kocharovskaya, Y. Rostovtsev, A. S. Z. G. Schmidt, H., K. L. Chapman, A. C. Gossard, and A. Imamo-
R. Welch, and M. O. Scully, 2001, Adv. At., Mol., Opt. Phys. glu, 1997, Appl. Phys. Lett. 70, 3455.
46, 191. Schmidt, H., and A. Imamoglu, 1996, Opt. Lett. 21, 1936.
Matsko, A. B., Y. V. Rostovtsev, M. Fleischhauer, and M. O. Schmidt, O., R. Wynands, Z. Hussein, and D. Meschede, 1996,
Scully, 2001, Phys. Rev. Lett. 86, 2006. Phys. Rev. A 53, R27.
Matsko, A. B., Y. V. Rostovtsev, O. Kocharovskaya, A. S. Zi- Schori, C., B. Julsgaard, J. L. Sorensen, and E. S. Polzik, 2002,
brov, and M. O. Scully, 2001, Phys. Rev. A 64, 043809. Phys. Rev. Lett. 89, 057903.
Mazets, I. E., and B. G. Matisov, 1996, JETP Lett. 64, 515. Scully, M. O., 1991, Phys. Rev. Lett. 67, 1855.
Merriam, A. J., S. J. Sharpe, M. Shverdin, D. Manuszak, G. Y. Scully, M. O., 1992, Phys. Rep. 219, 191.
Yin, and S. E. Harris, 2000, Phys. Rev. Lett. 84, 5308. Scully, M. O., and M. Fleischhauer, 1992, Phys. Rev. Lett. 69,
Merriam, A. J., S. J. Sharpe, H. Xia, D. Manuszak, G. Y. Yin, 1360.
and S. E. Harris, 1999, Opt. Lett. 24, 625. Scully, M. O., S. Zhu, and A. Gavrielides, 1989, Phys. Rev.
Mewes, C., and M. Fleischhauer, 2002, Phys. Rev. A 66, Lett. 62, 2813.
033820. Scully, M. O., and S. Y. Zhu, 1992, Opt. Commun. 87, 134.
Mewes, C., and M. Fleischhauer, 2005, Phys. Rev. A 共to be Scully, M. O., and M. S. Zubairy, 1997, Quantum Optics 共Cam-
published兲. bridge University Press, Cambridge, England兲.
Moiseev, S. A., and N. S. Kroll, 2001, Phys. Rev. Lett. 87, Serapiglia, G. B., E. Paspalakis, C. Sirtori, K. L. Vodopyanov,
173601. and C. C. Phillips, 2000, Phys. Rev. Lett. 84, 1019.
Moseley, R. R., S. Shepherd, D. J. Fulton, B. D. Sinclair, and Shapiro, M., 1970, J. Chem. Phys. 56, 2582.
M. H. Dunn, 1995, Phys. Rev. Lett. 74, 670. Shore, B. W., 1967, Rev. Mod. Phys. 39, 439.
Müller, G., M. Müller, A. Wicht, R.-H. Rinkleff, and K. Shore, B. W., 1990, Theory of Coherent Atomic Excitation
Danzmann, 1997, Phys. Rev. A 56, 2385. 共Wiley, New York兲.
Nagel, A., L. Graf, A. Naumov, E. Mariotti, V. Biancalana, D. Shvets, G., and J. S. Wurtele, 2002, Phys. Rev. Lett. 89, 115003.
Meschede, and R. Wynands, 1998, Europhys. Lett. 44, 31. Sokolov, A. V., 2002, KITP Workshop on Quantum Optics

共University of California Santa Barbara, Santa Barbara, CA兲. Whitley, R. M., and C. R. Stroud, 1976, Phys. Rev. A 14, 1498.
Sokolov, A. V., D. R. Walker, D. D. Yavuz, G. Y. Yin, and S. E. Wilson-Gordon, A. D., and H. Friedman, 1992, Opt. Commun.
Harris, 2000, Phys. Rev. Lett. 85, 562. 94, 238.
Sokolov, A. V., D. R. Walker, D. D. Yavuz, G. Y. Yin, and S. E. Wolf, S. A., D. D. Awschalom, R. A. Buhrman, J. M. Daugh-
Harris, 2001, Phys. Rev. Lett. 87, 033402. ton, S. von Molnar, M. L. Roukes, A. Y. Chtchelkanova, and
Stahler, M., S. Knappe, C. Affolderbach, W. Kemp, and R. D. M. Treger, 2001, Science 294, 1488.
Wynands, 2001, Europhys. Lett. 54, 323. Xiao, M., Y. Li, S. Jin, and J. Gea-Banacloche, 1995, Phys. Rev.
Steudel, H., 1988, J. Mod. Opt. 35, 193. Lett. 74, 666.
Strekalov, D., A. B. Matsko, Nan Yu, and L. Maleki, 2004, Zhang, G. Z., K. Hakuta, and B. P. Stoicheff, 1993, Phys. Rev.
Phys. Rev. Lett. 93, 023601. Lett. 71, 3099.
Sun, C. P., S. Yi, and L. You, 2003, Phys. Rev. A 67, 063815. Zhang, G. Z., M. Katsuragawa, K. Hakuta, R. I. Thompson,
Tewari, S. P., and G. S. Agarwal, 1986, Phys. Rev. Lett. 56, and B. P. Stoicheff, 1995, Phys. Rev. A 52, 1584.
1811. Zhao, Y., C. Wu, B. S. Ham, M. K. Kim, and E. Awad, 1997,
Turukhin, A. V., V. S. Sudarhanam, M. S. Shahriar, J. A. Phys. Rev. Lett. 79, 641.
Musser, B. S. Ham, and P. R. Hemmer, 2001, Phys. Rev. Lett. Zibrov, A. S., M. D. Lukin, D. E. Nikonov, L. Hollberg, M. O.
88, 023602. Scully, V. L. Velichansky, and H. G. Robinson, 1995, Phys.
van der Wal, C., M. D. Eisaman, A. Andre, R. L. Walsworth, Rev. Lett. 75, 1499.
D. F. Phillips, A. S. Zibrov, and M. D. Lukin, 2003, Science Zibrov, A. S., M. D. Lukin, and M. O. Scully, 1999, Phys. Rev.
301, 196. Lett. 83, 4049.
Visser, M., 2000, Phys. Rev. Lett. 85, 5252. Zibrov, A. S., A. B. Matsko, O. Kocharovskaya, Y. V. Rostovt-
Vitanov, N. V., M. Fleischhauer, B. W. Shore, and K. Berg- sev, G. R. Welch, and M. O. Scully, 2002, Phys. Rev. Lett. 88,
mann, 2001, Adv. At., Mol., Opt. Phys. 46, 55. 103601.
Vitanov, N. V., and S. Stenholm, 1997, Phys. Rev. A 56, 1463. Zimmer, F., and M. Fleischhauer, 2004, Phys. Rev. Lett. 92,
Wei, C., and N. B. Manson, 1999, Phys. Rev. A 60, 2540. 253201.
Werner, M., and A. Imamoglu, 1999, Phys. Rev. A 61, Zutic, Igor, J. Fabian, and S. Das Sarma, 2004, Rev. Mod. Phys.
011801共R兲. 76, 323.

Ref 5

Uploaded by

Copyright:

Available Formats

Ref 5

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ref 5

Uploaded by

Copyright:

Available Formats

REVIEWS OF MODERN PHYSICS, VOLUME 77, APRIL 2005

Electromagnetically induced transparency: Optics in coherent media

CONTENTS B. Cross-phase modulation using single-photon pulses

0034-6861/2005/77共2兲/633共41兲/$50.00 633 ©2005 The American Physical Society

co-workers 共Harris et al., 1990兲. The importance of EIT

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

FIG. 3. Fano interferences of excitation channels into a con-

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

The physics underlying the cancellation of absorption

Here ⌬1 = ␻31 − ␻p and ⌬2 = ␻32 − ␻c are the detunings of

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

FIG. 7. EIT absorption spectrum for different values of cou-

rather than a simple line splitting. Before proceeding

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

FIG. 9. Contour plot of imagi-

This strong dependence on reciprocal ⍀c is a direct con-

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

two dressed-state resonances, with a perfect cancellation sfg

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

acteristic of EIT, which distinguishes it from coherent

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

form limited. The essence of STIRAP is that a proper

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

If an empty-cavity resonance ␻c0 is sufficiently close

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

As mentioned earlier, lasing without inversion is

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

FIG. 13. Numerical simula-

time delay of a pulse by EIT requires an optically thick vgr共t兲

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

FIG. 14. 共Color in online edition兲 Storage and

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

scribed by a Schrödinger equation for Cn共z , t兲 with the

where the effect of the atoms on the fields is param-

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

double-lambda systems and double pairs of pulses, were

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

FIG. 17. Different schemes for resonantly en-

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

velopment is the demonstration of high-efficiency mul- 3n␻2

trashort pulses was first predicted by Harris and Sokolov + ⌬k +

B. Nonlinear mixing and frequency up-conversion with

as was pointed out first by Harris et al. 共1990兲 and sub-

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

In the case of frequency up-conversion to short wave-

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

FIG. 19. Parametric up-conver-

Rev. Mod. Phys., Vol. 77, No. 2, April 2005

FIG. 20. Generation of fields E1 and E2 by

Rev. Mod. Phys., Vol. 77, No. 2, April 2005