PHYSICA L R EVIEW LET T ERS week ending

VOLUME 90, N UMBER 11 21 MARCH 2003

Observation of Ultraslow Light Propagation in a Ruby Crystal at Room Temperature

Matthew S. Bigelow, Nick N. Lepeshkin, and Robert W. Boyd
The Institute of Optics, University of Rochester, Rochester, New York 14627
(Received 31 October 2002; published 21 March 2003)
We have observed slow light propagation with a group velocity as low as 57:5
0:5 m=s at room
temperature in a ruby crystal. A quantum coherence effect, coherent population oscillations, produces a
very narrow spectral ‘‘hole’’ in the homogeneously broadened absorption profile of ruby. The resulting
rapid spectral variation of the refractive index leads to a large value of the group index. We observe slow
light propagation both for Gaussian-shaped light pulses and for amplitude modulated optical beams in a
system that is much simpler than those previously used for generating slow light.

DOI: 10.1103/PhysRevLett.90.113903 PACS numbers: 42.65.–k, 42.50.Gy

There has recently been a flurry of interest in tech- herence technique than that of the EIT concept, namely,
niques that can allow the precise control of the velocity the creation of a spectral hole due to population oscilla-
vg of propagation of light pulses through material sys- tions. This hole is created by the periodic modulation
tems [1–3]. These techniques allow for the production of of the ground state population at the beat frequency
‘‘slow’’ light (vg  c) [4 –7], ‘‘fast’’ light (vg > c or vg between the pump and the probe fields applied to the
negative) [8,9], and ‘‘stored’’ or ‘‘stopped’’ light [10,11]. material sample. The spectral hole created by this tech-
Interest in these techniques lies at the level of both the nique can be extremely narrow, 36 Hz in our experi-
fundamental understanding [12] of the physical laws that ment, and leads to a rapid spectral variation of refractive
govern how fast or slow a light pulse can be made to index and, consequently, to a group velocity as low as
propagate and the promise of new applications such as 57:5
0:5 m=s.
controllable optical delay lines, optical data storage, op- These spectral holes due to coherent population oscil-
tical memories, and devices for quantum information. lations were first predicted in 1967 by Schwartz and Tan
It has been appreciated for quite some time that the [20] from the solution of the density matrix equations of
group velocity vg of an optical pulse propagating in a motion and have been described in greater detail by
resonant optical material can differ significantly from the subsequent authors [21–23]. In 1983, Hillman et al. [24]
velocity c of light in vacuum. In these cases, the rapid observed such a spectral hole in ruby. In their experiment,
change in the refractive index near a material resonance they used an argon-ion laser operating at 514.5 nm to
leads to a large value of the group index ng  c=vg pump population from the ground state to the broad 4 F2
[13,14]. However, in these situations strong absorption absorption band. Population decays from this level within
accompanies the low group velocity, making the experi- a few picoseconds to the metastable 2A and E levels and
mental observation of these effects difficult although not eventually returns to the ground level with a decay time
altogether impossible [15]. Most of the recent work on T10 of a few milliseconds. As a result of this long lifetime,
slow light propagation has made use of the technique of a spectral hole is created, centered at the laser frequency
electromagnetically induced transparency (EIT) to ren- with a width of approximately the inverse of the popula-
der the material medium highly transparent while still tion relaxation time. Hillman et al. [24] used modulation
retaining the strong dispersion required for the creation spectroscopy to observe this feature and measured its
of slow light [16 –18]. Using this technique, Kasapi et al. width to be 37 Hz.
[4] observed a group velocity of vg  c=165 in a 10 cm Our experimental setup is shown in Fig. 1. We use a
Pb vapor cell. More recently, Hau et al. [5] observed a single-line argon-ion laser operating at 514.5 nm as the
group velocity of 17 m=s in a Bose-Einstein condensate laser source. The beam passes first through a variable
and Kash et al. [6] used a modulation technique to mea- attenuator and then an electro-optic modulator. The
sure a group velocity of 90 m=s in rubidium vapor. Using modulator is driven by a function generator which allows
similar techniques, Budker et al. [7] inferred group ve- us to either place a 6% sinusoidal amplitude modulation
locities as low as 8 m=s. Also quite recently, Turukhin on the beam or produce long ( ms) pulses with almost
et al. [19] demonstrated the propagation of slow light no background intensity. For all pulse lengths, these
with a velocity of 45 m=s through a solid-state material, pulses had a peak power of 0.28 W with a background
Pr doped Y2 SiO5 , maintained at a cryogenic temperature that was only 4% of that of the peak. A glass slide sent 5%
of 5 K. of the beam to one detector for reference. The beam was
In this Letter, we report a new method that produces then focused with a 40-cm-focal length lens to a beam
slow propagation of light in a solid-state material at room waist of 84 m near the front surface of a 7.25-cm-long
temperature. It entails the use of a different quantum co- ruby rod. Because the center of the beam experienced less

113903-1 0031-9007=03=90(11)=113903(4)$20.00  2003 The American Physical Society 113903-1

 PHYSICA L R EVIEW LET T ERS week ending
VOLUME 90, N UMBER 11 21 MARCH 2003

Digital ground state recovery time, T10 is the lifetime of level c,

Oscilloscope T2 is the dipole moment dephasing time, and weq is the
Reference Detector population inversion of the material in thermal equilib-
or
Function rium. The distinction between T1 and T10 has been dis-
Generator
Signal Detector
cussed by Sargent [21]. The interaction Hamiltonian in
40 cm
the rotating-wave approximation is given by Vba 
Argon Ion Laser
Ruby  ba E1 ei!1 t  E3 ei!3 t , where E1 and E3 are the
EO modulator pump and probe field amplitudes, respectively, ba is
the dipole matrix element, and !3  !1  . We seek a
FIG. 1. The experimental setup used to observe slow light solution to the density matrix equation that is correct to
in ruby.
all orders in the amplitude of the strong pump field and is
correct to lowest order in the amplitude of the probe field.
absorption than the edges due to saturation, the beam did In this order of approximation, we represent the popula-
not expand significantly in traversing the ruby. Ruby is a tion inversion as
uniaxial crystal, and we rotated the rod to maximize the
interaction. The beam exiting the ruby is incident on a wt  w0  w eit  w eit : (2)
detector, and the detected signal is stored along with that Rather complicated expressions for the quantities w0 ,
of the input beam on a digital oscilloscope. The resulting w , and w are given explicitly in Ref. [22], but the
traces were compared on a computer to calculate the important conclusion is that the amplitudes of the popu-
relative delay and amplitude of the two signals. lation oscillations are appreciable only for  & 1=T1 . The
To analyze this situation mathematically, we refer to response at the probe frequency can next be represented
the ground state as level a, the 4 F2 absorption band as as [22]
level b, and the levels 2A and E as level c, as illustrated in
inset (i) in Fig. 2. Because of the rapid decay of level b, it ba E3 w0  E1 w
ba !1    : (3)
is possible to reduce this system to a two-level system h !1  !ba  i=T2
shown in inset (ii). The density matrix equations of
motion for this system are given by [25] Note that the first term corresponds to the interaction of

 the probe wave with the static part of the population
1 i difference, whereas the second term represents the scat-
_ ba   i!ba  ba  Vba w; (1a) tering of the pump wave off the temporally modulated
T2 h
ground state population. This contribution leads to de-
w  weq 2i
w_    Vba ab  Vab ba ; (1b) creased absorption at the probe frequency, that is, to a
T1 h spectral hole in the probe absorption profile. To demon-
strate this effect, we simplify Eq. (3) by assuming that
where w is the population inversion, T1  2T10 is the
!1  !ab , that T21 is large compared to the Rabi fre-
quency  2j ab jjE1 j=h and to the beat frequency 
P = 0.1 W !3  !1 , and that weq  1, to find
0
i ba E3 T2
Relative Modulation Attenuation

ba !1   
h
-0.2

1 2 T2 1  i=
  ;
-0.4 P = 0.25 W 1  2 T1 T2 T1 2  2
1
~
T2
(4)
(i) (ii)
b Γbc b
-0.6 where the linewidth of the spectral hole (HWHM in
c
ω1 Γba = 1 angular frequency units) is given by
1 T1
Γca =
-0.8 T'1
a a 1 4T j j2 1
  2 2 ab E21  1  2 T1 T2 : (5)
T1 h T1
-1
0 200 400 600 800 1000 We can determine the linear susceptibility through use of
Modulation Frequency (Hz)
Eq. (4) by means of the relation   N ba ba !3 =E3
FIG. 2. The relative modulation attenuation A with respect and, consequently, find expressions for the refractive in-
to the pump as a function of modulation frequency for input dex and the absorption experienced by the probe field as
pump powers of 0.1 and 0.25 W. The solid lines represent the

 cT I 
theoretical model of Eq. (10). The insets show the two coupling n  1  0 1 0 ; (6)
schemes discussed in the text. 2!1 1  I0 T1  2  1  I0 2

113903-2 113903-2
 PHYSICA L R EVIEW LET T ERS week ending
VOLUME 90, N UMBER 11 21 MARCH 2003


0 I0 1  I0 In Fig. 2, we show the measured relative modulation
  1 2 2
; (7) attenuation and compare it with the numerical solution of
1  I0 T1   1  I0
Eq. (10). In the limit in which the pump field becomes
where I0  I1 =Isat 2 T1 T2 is the normalized pump very weak, the spectral hole has a width of 1=T1 2! or
intensity and 0 is the unsaturated absorption coefficient. about 35.8 Hz (HWHM). As the input power is increased,
Note that the second term in Eq. (7) is the spectral hole the hole experiences power broadening. This result is in
due to coherent population oscillations. We can calculate good agreement with the characteristics of the spectral
the group index as hole that Hillman et al. [24] found in a 1 cm ruby.
dn This spectral dip causes an amplitude modulated beam
ng  n  !1 : (8)
d to experience a large group index. We show the delay
experienced by the modulation in Fig. 3 for input pump
This equation describes the propagation of spectrally powers of 0.1 and 0.25 W. We observed the largest delay,
narrow band pulses centered at the pump frequency !1 . 1:26
0:01 ms, with an input pump power of 0.25 W,
For broad band pulses, higher-order dispersion effects which corresponds in Fig. 2 to the power where the
need to be taken into account [26]. Some of our experi- spectral hole is deepest but still very narrow. The inferred
mental results were obtained through use of modulation group velocity at this power is 57:5
0:5 m=s. Note that
techniques such that the optical field contained only a the group velocity can be controlled by changing the
carrier wave (to act as the pump that creates the hole) and modulation frequency or the input intensity. We found
two sidebands (to act as probes). Since this field contains the nature of the effect to be strongly intensity dependent
discrete frequencies rather than a continuum of frequen- in that by moving the ruby a small distance from the focus
cies, a generalized form of Eq. (8) describing the propa- we could greatly decrease the measured delay. As a check
gation of the modulation pattern through the material is of our results, we found that the time delay would com-
given by nmod  n1  !1 n  n =2, where n1 is pletely disappear if we moved the ruby far from the focus.
the refractive index experienced by the pump. Combining Moreover, we found that it is not necessary to apply
this with Eq. (6), we obtain separate pump and probe waves to the ruby crystal in
 
0 cT1 I0 1 order to observe slow light effects. A single intense pulse
nmod  n1  : (9)
2 1  I0 1  I0 2  T1  2 of light is able to provide the saturation required to
modify the group index to provide slow light propagation.
For off-resonance modulations (  1=T1 ), the modula- These relatively intense pulses can be thought of as pro-
tion index reduces to nmod  n1 . Similarly, the relative ducing their own pump field and are thus self-delayed. We
modulation attenuation [A ], defined as the difference know of no other examples where a separate pump beam
between the attenuation of the modulation intensity and is not required for generating ultraslow light. For this
the attenuation of the pump intensity, is given by [21,24] experiment, we used the programmable pulse generator
 
1 1  I0 L 2  T1  2
A  ln ; (10) 1.4
2 1  I0 0 2  T1  2
1.2 P = 0.25 W
where L is the length of the crystal. Output
To model the total group delay and modulation attenu- 1.0 Input
ation observed in our experiments, we first numerically
calculated the value of I0 throughout the length of the
Delay (ms)

0.8
crystal. The pump beam intensity depends on the propa- 612 µs
0.6
gation distance through the ruby as
-8 0 8
dI0 z I 0.4 Time (ms)
 0 0 : (11) P = 0.1 W
dz 1  I0 0.2
Using the accepted value of 1:5 kW=cm2 [27] for the
0
saturation intensity of ruby at 514.5 nm, we integrate
Eq. (11) numerically to find I0 z . Combining this func- 0 200 400 600 800 1000
Modulation Frequency (Hz)
tion with our theoretical model for the dispersion, we
could fit the total delay and the relative modulation at- FIG. 3. Observed time delay as a function of the modulation
tenuation measured in our experiment. We assumed 0 frequency for input pump powers of 0.1 and 0.25 W. The solid
and T1 to be free parameters and found the values of 0  lines represent the theoretical model of Eq. (9). The inset shows
1:17 cm1 and T1  4:45 ms. These values are in the the normalized 60 Hz input (solid line) and output (dashed
range found by Cronemeyer [28] and others. The total line) signal at 0.25 W. The signal was delayed 612 s corre-
transmission in our experiments is on the order of 0.1%. sponding to an average group velocity of 118 m=s.

113903-3 113903-3
 PHYSICA L R EVIEW LET T ERS week ending
VOLUME 90, N UMBER 11 21 MARCH 2003

ity 5 ms identical delay. Also, since the slow light can be produced
e ns in a solid and at room temperature, this technique offers
nt I
d
l i ze 10 ms the possibility of applications in photonics such as fully
a
rm integrated, controllable optical delay lines.
Input Pulse No
20 ms This work was supported by ONR Grant No. N00014-
Output Pulse 99-1-0539, ARO Grant No. DAAD19-01-1-0623, DOE
30 ms Grant No. DE-FG02-01ER15156, AFOSR Grant
No. F49620-00-1-0061, and the U.S. Department of En-
ergy Office of Inertial Confinement Fusion under Co-
operative Agreement No. DE-FC03-92SF19460 and the
300 m/s University of Rochester. The support of DOE does not
constitute an endorsement by DOE of the views expressed
159 m/s in this article.
119 m/s

102 m/s
[1] R.W. Boyd and D. J. Gauthier, in Slow and Fast Light,
-20 -10 0 10 20 30
Progress in Optics Vol. 43, edited by E. Wolf (Elsevier,
Time (ms) Amsterdam, 2002).
[2] P.W. Milonni, J. Phys. B 35, R31–R56 (2002).
FIG. 4. The normalized input and output intensities of a 5, 10, [3] A. B. Matsko et al., Adv. At. Mol. Opt. Phys. 46, 191
20, and 30 ms pulse. The corresponding average group veloc- (2001).
ities of the pulses are 300, 159, 119, and 102 m=s. The inset [4] A. Kasapi et al., Phys. Rev. Lett. 74, 2447 (1995).
shows a close-up of the 20 ms pulse. [5] L.V. Hau et al., Nature (London) 397, 594 (1999).
[6] M. M. Kash et al., Phys. Rev. Lett. 82, 5229 (1999).
[7] D. Budker et al., Phys. Rev. Lett. 83, 1767 (1999).
to produce Gaussian pulses with a 1=e intensity full width
[8] B. Segard and B. Macke, Phys. Lett. 109A, 213 (1985).
of 1 to 30 ms with almost no background, and we ob- [9] L. J. Wang, A. Kuzmich, and A. Dogariu, Nature
served how they were delayed in propagating through the (London) 406, 277 (2000).
ruby. We found that the longer pulses also had the longer [10] C. Liu, Z. Dutton, C. H. Behroozi, and L.V. Hau, Nature
delays with the center of mass of a 30 ms pulse delayed by (London) 409, 490 (2001).
0.71 ms with little pulse distortion. We show this result [11] D. F. Phillips et al., Phys. Rev. Lett. 86, 783 (2001).
and those for other pulse lengths in Fig. 4. While the [12] A. Kuzmich et al., Phys. Rev. Lett. 86, 3925 (2001).
theory developed above, which assumed the presence of [13] L. Brillouin, Wave Propagation and Group Velocity
distinct cw pump and probe fields, does not model the (Academic, New York, 1960).
pulsed experiment directly, that theory can be used to [14] C. G. B. Garrett and D. E. McCumber, Phys. Rev. A 1,
gain some intuition regarding the experiment. For in- 305 (1970).
[15] S. Chu and S. Wong, Phys. Rev. Lett. 48, 738 (1982).
stance, we would expect longer pulses, which contain
[16] S. E. Harris, J. E. Field, and A. Imamoglu, Phys. Rev.
lower frequency components, to experience longer delays. Lett. 64, 1107 (1990); S. E. Harris, J. E. Field, and
In conclusion, we observed slow light propagation with A. Kasapi, Phys. Rev. A 46, R29 (1992).
a group velocity as low as 57:5
0:5 m=s in a ruby [17] S. P. Tewari and G. S. Agarwal, Phys. Rev. Lett. 56, 1811
crystal. In addition to seeing the amplitude modulation (1986).
on a beam significantly delayed, we observed that single [18] R. S. Bennink et al., Phys. Rev. A 63, 033804 (2001).
pulses without a pump background can also experience [19] A.V. Turukhin et al., Phys. Rev. Lett. 88, 023602 (2002).
long delays. We have shown theoretically that these large [20] S. E. Schwartz and T. Y. Tan, Appl. Phys. Lett. 10, 4
group indices originate from a very narrow spectral hole (1967).
in the homogeneously broadened absorption spectrum [21] M. Sargent III, Phys. Rep. 43, 223 (1978).
caused by coherent population oscillations. This tech- [22] R.W. Boyd et al., Phys. Rev. A 24, 411 (1981).
[23] A. D. Wilson-Gordon, Phys. Rev. A 48, 4639 (1993).
nique for producing slow light is very easy to implement.
[24] L.W. Hillman et al., Opt. Commun. 45, 416 (1983).
It requires the use of only a single laser. In addition, since [25] See, for instance, R.W. Boyd, Nonlinear Optics
the location of the spectral hole follows exactly any drift (Academic, San Diego, 1992).
in the laser frequency, the laser does not need to be [26] J. Peatross, S. A. Glasgow, and M. Ware, Phys. Rev. Lett.
frequency locked to any particular transition frequency. 84, 2370 (2000).
In fact, the laser does not even have to operate in a single [27] P. F. Liao and D. M. Bloom, Opt. Lett. 3, 4 (1978).
longitudinal mode since all the modes will experience an [28] D. C. Cronemeyer, J. Opt. Soc. Am. 56, 1703 (1966).

113903-4 113903-4