1: \documentclass[aps,prl,twocolumn,showpacs,preprintnumbers,amsmath,amssymb,superscriptaddress,floatfix]{revtex4}
2: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
3:
4: % Some other (several out of many) possibilities
5:
6: %\documentclass[preprint,aps]{revtex4}
7:
8: %\documentclass[preprint,aps,draft]{revtex4}
9: %\documentclass[prb]{revtex4}% Physical Review B
10:
11: \usepackage{graphicx}% Include figure files
12: \usepackage{amsmath}
13: %\usepackage{psfrag}
14: %\usepackage{color}
15: %\usepackage{dcolumn}% Align table columns on decimal point
16: %\usepackage{bm}% bold math
17:
18: %\nofiles
19:
20: \newcommand{\un}[1]{\ensuremath{\mathrm{#1}}}
21: \newcommand{\sub}[1]{_{\mathrm{#1}}}
22: \renewcommand{\vec}[1]{\mathbf{#1}}
23: \renewcommand{\Re}{\mathrm{Re}}
24: \renewcommand{\Im}{\mathrm{Im}}
25:
26: \newcommand{\Ep}{\mathcal{E}_p}
27: \newcommand{\Ec}{\mathcal{E}_c}
28: \newcommand{\Eo}{\mathcal{E}_0}
29:
30:
31:
32: \begin{document}
33:
34: \title{Ultraslow Propagation of Matched Pulses by Four-Wave Mixing in
35: an Atomic Vapor}
36:
37: \author{V. Boyer}
38: \affiliation{Atomic Physics Division, MS 8424, National Institute of
39: Standards and Technology, Gaithersburg, Maryland 20899-8424, USA}
40: \author{C. F. McCormick}
41: \affiliation{Atomic Physics Division, MS 8424, National Institute of
42: Standards and Technology, Gaithersburg, Maryland 20899-8424, USA}
43: \author{E. Arimondo}
44: \affiliation{Atomic Physics Division, MS 8424, National Institute of
45: Standards and Technology, Gaithersburg, Maryland 20899-8424, USA}
46: \affiliation{Dipartimento di Fisica Enrico Fermi, Universit\`a di Pisa,
47: Largo Bruno Pontecorvo 3,
48: I-56127 Pisa, Italy}
49: \author{P. D. Lett}
50: \affiliation{Atomic Physics Division, MS 8424, National Institute of
51: Standards and Technology, Gaithersburg, Maryland 20899-8424, USA}
52:
53: \date{\today}
54:
55: \begin{abstract}
56:
57: We have observed the ultraslow propagation of matched pulses in
58: nondegenerate four-wave mixing in a hot atomic vapor. Probe pulses as
59: short as 70~ns can be delayed by a tunable time of up to 40~ns with
60: little broadening or distortion. During the propagation, a probe pulse
61: is amplified
62: and generates a conjugate pulse which is faster and separates from the probe
63: pulse before getting locked to it at a fixed delay. The precise
64: timing of this process allows us to determine the key coefficients of the
65: susceptibility tensor. The presence of gain in this system makes this
66: system very
67: interesting in the context of all-optical information processing.
68:
69: \end{abstract}
70:
71: \pacs{42.50.Gy, 42.65.Yj}
72:
73: \maketitle
74:
75: Slow group velocities, valuable for all-optical signal processing, are
76: obtained at a resonance peak of the transmission spectrum of a medium,
77: and a number of different implementations of this principle have been
78: demonstrated. They rely either on a reduction of the absorption, such
79: as electromagnetically induced transparency (EIT)~\cite{kasapi1995},
80: coherent population oscillations~\cite{bigelow2003}, and dual
81: absorption lines~\cite{camacho2006}, or on a gain resonance, like
82: stimulated Brillouin scattering~\cite{okawachi2005} and stimulated
83: Raman scattering~\cite{sharping2005}. To be useful in the context of
84: all-optical signal processing, an optical delay line should be able to
85: produce a fractional delay (defined as the ratio of the delay to the
86: duration of the pulse) larger than unity with only modest absorption
87: and pulse broadening. Recent
88: developments~\cite{okawachi2005,zhang2006,jiang2006} have shown the
89: benefits of using an amplifying medium to alleviate the absorption and
90: distortion issues usually associated with slow
91: light~\cite{matsko2005}.
92:
93: We have examined the group velocity reduction effects
94: due to nondegenerate four-wave mixing (4WM) in hot rubidium vapor, and
95: have obtained large fractional delays with almost no distortion. The
96: presence of gain in this system makes it in principle possible to stack such delay
97: lines and achieve fractional delays only limited by pulse broadening.
98: %
99: Another notable feature of the amplification in the 4WM process is the
100: generation of a conjugate pulse which is coupled to the probe and which
101: propagates alongside it, similar to matched pulses in EIT
102: systems~\cite{harris1997}. We have studied the interplay between the
103: 4WM coupling of the probe and conjugate, and the Raman coupling
104: of the probe and pump. This interplay leads to the ultraslow propagation of
105: matched probe and conjugate pulses and to the enhancement of the 4WM
106: gain. Such an enhancement, when applied to cross-phase modulation, is
107: the key element of recent optical quantum information processing
108: proposals~\cite{lukin2000b,lukin2001,wang2006}.
109:
110:
111: \begin{figure}[htb]
112: \begin{center}
113: \includegraphics[width=\linewidth]{exp.eps}\\
114: \includegraphics[height=.9\linewidth,angle=-90]{trans.ps}
115: \end{center}
116: \caption{(color online). (a) Experimental setup. PBS: polarizing beamsplitter.
117: (b) Energy-level diagram of the
118: D1 line of $^{85}$Rb, showing the double-lambda scheme. Note that
119: the pumps $\Omega_1$ and $\Omega_2$ are in fact the same laser beam. (c) Probe transmission
120: profile versus two-photon detuning $\delta$.}
121: \label{fig:exp}
122: \end{figure}
123:
124:
125: Our apparatus, which is essentially the same as the one described in
126: Ref.~\cite{mccormick2007}, consists of a linearly polarized,
127: continuous, strong (up to 280~mW) pump and a cross-polarized, pulsed, weak
128: (0.5 mW) probe propagating at a small angle ($0.5^\circ$) through a
129: 2.5~cm-long $^{85}$Rb cell heated to $90^\circ$C-$140^\circ$C
130: (Fig.~\ref{fig:exp}a). The pump and the probe are near resonant with
131: a Raman transition between the two hyperfine electronic ground states
132: of $^{85}$Rb, with a controllable detuning $\delta$
133: (Fig.~\ref{fig:exp}b), and have $1/e^2$ radii of 600~$\mu$m and
134: 350~$\mu$m, respectively. The residual 2-photon Doppler broadening due
135: to the small angle is a few MHz. The detuning from the 5P$_{1/2}$
136: excited state is $\Delta_1 /2\pi \approx 850$~MHz, and the peak pump
137: intensity (up to $45$~W/cm$^2$) is high enough to excite
138: off-resonant Raman transitions with a detuning $\Delta / 2\pi\approx 4$~GHz
139: from the excited state. The double-lambda~\cite{lukin1998} is closed
140: by the conjugate beam which emerges on the other side of the pump from
141: the probe, with the same polarization as the probe. The combination
142: of the beam polarizations and the Zeeman substructure makes the system
143: a four-level system (the two virtual excited states are orthogonal).
144: The probe
145: amplification is sharply resonant in $\delta$, as shown in
146: Fig.~\ref{fig:exp}c, with a gain that can reach 30. The gain feature leads to a
147: strong dispersion of the index of refraction and thus a low group
148: velocity for the probe.
149:
150:
151: We measure the group velocity delay by recording the arrival time of a
152: 70~ns-long (full width half maximum [FWHM]) gaussian probe pulse with
153: and without the atomic medium (reference pulse).
154: Figure~\ref{fig:slow}a shows an example in which the parameters are
155: set to provide a large probe gain ($G=13$). By varying the two-photon detuning
156: $\delta$ and the pump intensity, one can tune the probe delay and
157: achieve a fractional delay larger than 0.5 such that the pulse remains
158: gaussian and is broadened by less than 10\% of its original width.
159: This low level of
160: distortion is remarkable in comparison with that seen in some EIT
161: experiments~\cite{matsko2005}. The maximum delay corresponds to a
162: group velocity of $c/500$, where $c$ is the speed of light in vacuum.
163: It is in general possible to tune the pump intensity, the pump
164: and probe detunings, and the temperature to achieve an overall gain of
165: unity. Figure~\ref{fig:slow}a also shows the record of the conjugate
166: intensity. A striking feature is the emergence of the conjugate pulse
167: \emph{before} the probe pulse. This relative delay is a fundamental property
168: of the dynamics of the system, and was predicted and observed in
169: Refs.~\cite{vanderwal2003,andre2005} in the case of resonance
170: on the ``lower'' lambda transition ($\Delta_1=0$).
171:
172:
173: \begin{figure}[htb]
174: \begin{center}
175: \includegraphics[height=.9\linewidth,angle=-90]{tune.ps}
176: \end{center}
177: \caption{(color online). (a) Slow-light effect near the peak
178: of the resonance.
179: The reference pulse is magnified 13 times.
180: Thin black solid and dashed lines:
181: probe pulses for the parameters ($\delta/2\pi$, pump power) equal to
182: (10~MHz, 280~mW) and (22~MHz, 200~mW) respectively.
183: Color lines: corresponding
184: conjugate pulses. The pulses (probe and matching conjugate) are
185: broadened by 5\% in the less
186: retarded case, and 10\% in the more retarded case. (b) For a
187: detuning $\Delta_1$ closer to the Raman absorption dip in
188: Fig.~\ref{fig:exp}c, large delays and
189: pulse breakups are observed (here for the conjugate).}
190: \label{fig:slow}
191: \end{figure}
192:
193:
194: Larger delays, shown in Fig.~\ref{fig:slow}b, can be achieved by setting $\delta$ between the gain peak
195: and the Raman absorption dip present at $\delta \lesssim 0$ (see
196: Fig.~\ref{fig:exp}c).
197: The competition between large amplification and large absorption
198: leads to some complex dynamics which can result in pulse
199: breakup, in a similar fashion to the dual-field solitons predicted to
200: exist in three-level systems~\cite{konopnicki1981}.
201: In this paper, we restrict ourselves to the regions of low absorption
202: where, in spite of complications associated with the use of hot atoms,
203: our system can be consistently described over a broad range of
204: parameters by the simple concepts developed in the theory of
205: Refs.~\cite{lukin2000,andre2005}.
206: %Although our system is complicated by factors
207: %associated with the use of hot atoms, it can be consistently
208: %described over a broad range of
209: %parameters in the low absorption regime by the simple
210: %concepts developed in the theory of Refs.~\cite{lukin2000,andre2005}.
211:
212: We neglect the hyperfine splitting of the excited state. Averaged over
213: the Zeeman substructure, the dipole moments of all four transitions
214: are equal, which gives $\Omega_1$ and $\Omega_2$, the peak resonant
215: Rabi frequencies of the pump for the ``lower'' and the ``upper''
216: lambda respectively, the same value, denoted $\Omega$. The
217: double-lambda system in its ideal incarnation operates in the limit
218: $\Delta_1 \ll \Delta$ and has the following crucial features. First,
219: a ground state coherence is established by the ``lower'', more
220: resonant lambda. The coherence has a lifetime $1/\gamma_c$, limited by
221: magnetic fields, collisions and the transit time in the laser beams,
222: and corresponds to a dark state in which the absorption of the probe
223: is reduced (EIT). Second, the ``upper'', less resonant lambda slightly
224: perturbs this coherence and creates a resonant atomic polarization at
225: the probe and conjugate frequencies via 4WM, while keeping the
226: population in the excited state near zero. The dynamics of the system
227: can thus be broken down into two intertwined processes: EIT and 4WM.
228:
229: In the limit of a strong pump and low pump depletion, most of the
230: atomic population is in the ground state $F=3$, and the Fourier
231: components $\Ep(\omega)$ and $\Ec(-\omega)$ of the
232: slowly-varying
233: envelopes of the probe and conjugate fields (of
234: wavevectors $\mathbf{k}_p$ and $\mathbf{k}_c$) obey the
235: equations~\cite{lukin2000,andre2005}:
236: %
237: \begin{eqnarray}
238: \label{prop1}
239: (i\omega + c\partial_z)\Ep &=&i
240: \eta\Delta_R \Ec^* -\eta\left[i\left(\delta + \omega -
241: {\textstyle\frac{\Omega^2}{4\Delta}}\right) +
242: \gamma_c\right]\Ep\;\quad\\
243: \label{prop2}
244: (i\omega + c\partial_z)\Ec^* &=& -i \eta\Delta_R \Ep.
245: \end{eqnarray}
246: %
247: We assume perfect phase matching. In the limit of $\Omega^2/4\Delta_1
248: \gg \delta$, $\gamma_c$, and $\Delta \gg
249: \Delta_1$, $\gamma$
250: (where $\gamma/2\pi = 6$~MHz is the linewidth of
251: the atomic transition), the coefficients in Eqs.~(\ref{prop1}) and
252: (\ref{prop2}) are~\cite{lukin2000} $\eta = g^2 N/[\Omega^2/4 +
253: \Delta_1(\delta + \omega + i\gamma_c)] \approx 4g^2 N/\Omega^2 = c/v_g \gg 1$
254: and
255: %$\eta = g^2 N/\Omega^2 = c/v_g \gg 1$ and
256: $\Delta_R=\Omega^2/4\Delta$. Here, $g^2 =
257: ck\wp^2/(2\varepsilon_0\hbar)$, $k = k_p \approx k_c$, $N$ is the
258: atomic density, and $\wp$ is the average dipole moment acting on the
259: probe and the conjugate.
260:
261: The interpretation of Eqs.~(\ref{prop1}) and
262: (\ref{prop2}) is straightforward. The probe field $\Ep$ is slowed down
263: by a factor $\eta$ due to the EIT interaction with the pump [second
264: drive term on the right-hand side of Eq.~(\ref{prop1})]. The EIT
265: resonance is light-shifted by the pump on the ``upper'' lambda and
266: occurs at $\tilde\delta \equiv \delta - \frac{\Omega^2}{4\Delta} = 0$.
267: In addition, $\Ep$ and $\Ec$ are cross-coupled with a coupling
268: constant $\alpha=\eta\Delta_R$, responsible for the 4WM amplification.
269: The presence of $\eta$ highlights the role of the longer interaction
270: time due to the slow-down effect in obtaining a sizeable nonlinear
271: coupling~\cite{lukin2001}. The other factor in $\alpha$, the so-called
272: Raman bandwidth $\Delta_R$, is the Rabi frequency of a fictitious
273: resonant Raman transition driven on both legs by the pump field and
274: with an intermediate Raman detuning $\Delta$. As shown by the absence
275: of any dependence on $\Delta_1$, the 4WM dynamics is dominated by the
276: ``upper'' lambda, which acts as a bottleneck. The propagation
277: equations are asymmetrical. The imaginary part of the coefficient of
278: the direct term for the conjugate [Eq.~(\ref{prop2})],
279: corresponding to a slow-light effect, is
280: negligible compared to the same term for the probe because $\Delta \gg
281: \Delta_1$\cite{lukin2000}. The real part, corresponding to a
282: Raman amplification,
283: scales as $1/\Delta^2$ and is negligible compared to the cross-term
284: $\alpha$ which scales as
285: $1/\Delta$~\cite{lukin2000,andre2005}. As a result, in the absence
286: of the 4WM coupling (``bare''
287: fields), the probe and the conjugate propagate at velocities $v_g$ and
288: $c$, respectively. The finite decoherence $\gamma_c$ translates into a
289: small absorption of the probe.
290:
291: Our system departs from the ideal case described by the expressions of
292: $\eta$ and $\alpha$ given above in many respects. Unlike the
293: experiments described in
294: Refs.~\cite{lukin2000,vanderwal2003,andre2005}, which were performed
295: with a resonant probe and a weak pump, our probe is tuned to the side
296: of the Doppler profile, in an already almost transparent region. As a
297: result, the position of the gain peak is not tied as closely to a
298: narrow EIT window. It depends on the balance between the losses, which
299: include the Raman absorption dip and the absorption from the Doppler
300: broadened 1-photon transition, and the 4WM gain. Factors influencing
301: the peak position are the spread of values for $\Delta_1$ due to the
302: Doppler broadening, the spread of values for $\Omega_1$ and
303: $\Omega_2$ due to the Zeeman degeneracy, %
304: %
305: %\footnote{Perfect optical pumping would pump the system into a single
306: %superposition of Zeeman sublevels which is perfectly dark for the
307: %``lower'' lambda, even when taking the hyperfine structure of the
308: %excited state into account~\cite{zanon2005}.},
309: %
310: the contribution of the usual dispersion of the Doppler broadened
311: vapor to the slow-down of the
312: probe, and the only approximate phase matching. In practice, this
313: means that the position of the gain peak varies by up to 20 MHz
314: depending on parameters like the temperature, $\Delta_1$ and the probe
315: intensity.
316:
317: In spite of the added complexity and the difficulty of directly
318: calculating $\eta$ and $\Delta_R$, we assume that the propagation
319: equations (\ref{prop1}) and (\ref{prop2}) are still valid over most of
320: the resonance peak, provided that the peak is at $\tilde\delta \approx 0$,
321: and that $\gamma_c \ll 2\Delta_R$.
322: Solving them in the limit
323: $\eta \gg 1$ and starting from a probe field $\Eo(\omega)$ and no
324: conjugate field, one finds:
325: %
326: \begin{eqnarray}
327: \label{sol1}
328: \Ep(\omega, z) & = & \Eo(\omega)
329: \exp\left(i\sigma(\omega)\frac{z}{c}\right)\times\nonumber\\
330: & &\left[ \cosh \left(\xi(\omega)\frac{z}{c}\right)
331: + i\frac{\sigma(\omega)}{\xi(\omega)} \sinh
332: \left(\xi(\omega)\frac{z}{c}\right) \right]\\
333: \label{sol2}
334: \Ec^*(\omega, z) & = & \Eo(\omega)
335: \exp\left(i\sigma(\omega)\frac{z}{c}\right)
336: \frac{\alpha(\omega)}{i\xi(\omega)} \sinh
337: \left(\xi(\omega)\frac{z}{c}\right)\quad
338: \end{eqnarray}
339: %
340: where $\xi(\omega) = \sqrt{\alpha(\omega)^2 -
341: \sigma(\omega)^2}$ and $\sigma(\omega) =
342: \frac{\eta(\omega)}{2}(\tilde\delta + \omega + i
343: \gamma_c)$. Since $\gamma_c \ll 2\Delta_R$, $\xi(\omega)$
344: is real. As expected, past an initial linear
345: growth of the conjugate, both fields grow exponentially in distance
346: with $(\xi- \frac{\eta}{2}\gamma_c)/c$ as the linear gain
347: coefficient. In the limit $\tilde\delta = 0$, one has
348: $\xi \approx \alpha$.
349: %, and the Raman bandwidth $\Delta_R$
350: %can then be interpreted as Einstein's $A$ coefficient for the optical
351: %amplifier.
352:
353: The solutions also contain information about the propagation delay.
354: Equations (\ref{sol1}) and (\ref{sol2}) show that
355: the fields accumulate a phase
356: across the medium, denoted $\theta(\omega)$.
357: The main contribution to the group
358: delay $\frac{\mathrm{d}}{\mathrm{d}\omega}\theta(\omega)$ for both
359: fields comes from the first exponential and gives a common delay $\tau
360: = \left.\frac{\mathrm{d}}{\mathrm{d}\omega}\Re[\sigma(\omega)]
361: z/c\right|_{\tilde\delta = 0,\omega = 0} = \eta z/2c$. In other words,
362: the probe and the conjugate are slowed down by half the ``bare''
363: slow-down factor $\eta$. The probe experiences an extra delay
364: due to the second term in Eq.~(\ref{sol1}). At large gain, the cosh
365: and sinh functions are equal and the additional delay is $\Delta\tau =
366: \left. \frac{\mathrm{d}}{\mathrm{d}\omega}\Re[\sigma(\omega) /
367: \xi(\omega)] / (1 - \Im[\sigma(\omega) / \xi(\omega)])
368: \right|_{\tilde\delta = 0,\omega=0} = \eta/[2\xi -
369: \eta\gamma_c] \approx \eta/2\xi$. At low
370: gain, a first order expansion in $z$ gives $\Delta \tau = \eta
371: z/ 2c$. The picture emerging from this analysis is the following: the
372: conjugate pulse is created without delay by the probe and travels
373: at a velocity $2v_g$ ($\ll c$). The probe pulse travels initially at a
374: velocity $v_g$ and then locks onto the conjugate by accelerating to
375: $2v_g$ when the delay reaches $\eta/2\xi$ (at a gain
376: close to 2).
377:
378:
379: \begin{figure}[htb]
380: \begin{center}
381: \includegraphics[height=.9\linewidth,angle=-90]{delays2.ps}
382: \end{center}
383: \caption{(color online). Two-photon detuning scan at a temperature
384: of $140^\circ$C, using 120~ns-long pulses. The bare state
385: 2-photon resonance corresponds to $\delta = 0$. The calculated
386: gain, inferred
387: from the delays, assumes linear losses equal to 14\% of the peak
388: linear gain.}
389: \label{fig:delays}
390: \end{figure}
391:
392:
393: We test this interpretation by first scanning the two-photon detuning
394: $\delta$, using 120~ns-long (FWHM) probe pulses and a pump power
395: of 280~mW, which corresponds to a spread of $\omega$ of 10~MHz
396: around zero and to $\Omega/2\pi = 420$~MHz.
397: Figure~\ref{fig:delays} shows the measured gain, the probe delay $\tau
398: + \Delta\tau$, which includes the contribution of all the retardation
399: effects, and the differential delay $\Delta\tau$. The contribution to
400: the probe delay of the usual dispersion effect, which is measured
401: with the pump intensity strongly reduced, is found to be $8\pm2$~ns.
402: From the experimental data and the theoretical expressions of $\tau$
403: and $\Delta\tau$, one can deduce a value of $\eta$ and $\xi$ for each
404: $\delta$, in the limit of large gain and small $\gamma_c$. Inserting
405: these values into Eq.~(\ref{sol1}) (with $\tilde\delta \approx 0$) and
406: adjusting $\gamma_c$ to $0.5\gamma$ (making the linear losses equal
407: to $14\%$ of the
408: peak linear gain)~%
409: %
410: \footnote{This value of $\gamma_c$ is sensitive to the timing errors
411: ($\pm1$~ns) and is only an indication of the order of magnitude.
412: For $G=3$, it corresponds to 15\% of absorption,
413: which is compatible with the level of squeezing
414: measured in Ref.~\cite{mccormick2007}.}
415: %
416: , one can reproduce the gain curve with reasonable accuracy. For our
417: parameters, the EIT resonance is light-shifted to $\delta =
418: 11$~MHz$\times 2\pi \approx 2\gamma$, close to the observed gain
419: maximum ($\tilde\delta = 6$~MHz$\times 2\pi \approx \gamma$). It can be
420: shown that in the above calculations, the approximation $\tilde\delta
421: \approx 0$ is valid as long as $\tilde\delta \ll 2\Delta_R$
422: ($4\gamma$ for our beam parameters).
423: For $|\tilde\delta|$ larger than a few $\gamma$, that is to say in the
424: wings of the gain peak, the approximation is
425: expected to break down.
426: %In practice, the largest discrepancy arises
427: %near $\delta = 0$ (see Fig.~\ref{fig:delays}), where the 2-photon
428: %absorption distorts and broadens the pulses.
429:
430:
431: Next, we directly observe the locking between the probe
432: and the conjugate during the propagation. It is impractical to
433: continuously vary the distance of propagation and we instead vary the
434: atomic density $N$ via the temperature, which is equivalent.
435: Indeed, $\sigma$ and
436: $\xi$ are proportional to $N$ through their dependence on
437: $\eta$, and changing $N$ is like renormalizing $z$ in the solutions
438: (\ref{sol1}) and (\ref{sol2}). According to (\ref{sol1}), the
439: renormalized propagation length $L$ is related to the probe intensity
440: gain $G$ by $L= \cosh^{-1}(\sqrt{G})$.
441: The detuning $\delta$ is set to 15 MHz$\times 2\pi$, near the gain
442: maximum, the pump power
443: is still set to 280~mW, and the measured
444: delays as a function of the renormalized distance for a temperature
445: scan of $50^\circ$C around $120^\circ$C are shown in
446: Fig.~\ref{fig:locking}. The two regimes of
447: propagation are very clear. First, the pulses separate in time, and second they lock
448: to each other at a fixed delay. The time offset at the origin is not
449: well understood.
450: A direct evaluation of
451: $2\Delta/\Omega^2 \approx \Delta\tau$ in the ideal case using
452: our beam parameters gives a value of 7~ns, comparable to the one measured
453: near the gain maximum (see Fig.~\ref{fig:delays}).
454: %
455: We checked qualitatively that $\Delta\tau$ increases when the pump
456: intensity decreases. It is worth noting that the detail of the
457: low-gain transient regime depends on the initial conditions. For
458: instance, swaping the frequencies of the probe and the conjugate would lead
459: to an initial propagation in which the probe travels at a velocity $c$
460: while the conjugate travels at a velocity $2v_g$.
461:
462:
463: \begin{figure}[htb]
464: \begin{center}
465: \includegraphics[height=.9\linewidth,angle=-90]{locking2.ps}
466: \end{center}
467: \caption{(color online). Probe delay $\tau + \Delta\tau$, conjugate
468: delay $\tau$, and differential delay $\Delta\tau$ as a function of
469: a pseudo propagation distance.}
470: \label{fig:locking}
471: \end{figure}
472:
473: Finally, an important feature of the model is that the gain saturates
474: with the pump intensity. For a gain peak location $\delta$ and a
475: decoherence $\gamma_c$ of the order of $\gamma$, $\xi$ saturates when
476: $\Omega \gg 2\sqrt{\Delta\gamma} = 300$~MHz $\times 2\pi$. In agreement
477: with this prediction, we observe that $G$ starts saturating at our
478: operating intensity.
479:
480: To conclude, we have observed the ultraslow propagation of probe and
481: conjugate pulses with matched shapes and group velocities through a rubidium
482: vapor. The study of the coupled propagation gives access to the atomic
483: dynamics through a simple model that reflects a few key concepts.
484: Although the hypothesis of the model does not match precisely the
485: conditions of the experiment, our findings on slow propagation and
486: delay locking of the probe and conjugate pulses are generic to the
487: double-lambda system. The quality of the retardation effect in terms
488: of fractional delay and absence of loss and distortion suggests the
489: possible existence of a dual-field soliton, which would be the result
490: of higher order terms in the propagation equations. Such a soliton has
491: been predicted in related 4WM schemes~\cite{deng2005}.
492:
493: It should also be noted that this double-lambda scheme is known
494: to generate relative-number
495: squeezed twin beams~\cite{mccormick2007} (as well as correlated
496: photons~\cite{kolchin2006}). Extending this
497: semi-classical pulse
498: theory to the quantum correlated beams raises two comments. First, as
499: pointed out in Ref.~\cite{andre2005}, the time lag between the probe
500: and the conjugate should be the limiting factor to the squeezing
501: bandwidth observed in Ref.~\cite{mccormick2007}. Second, the system
502: could be used with gain close to unity to slow light in the quantum
503: regime, while preserving nonclassical correlations,
504: possibly more efficiently than with EIT alone~\cite{akamatsu2006}.
505:
506: CFM was supported by an IC Postdoctoral Fellowship.
507: We acknowledge very helpful discussions with A. M. Marino.
508:
509: \bibliography{slow}
510:
511: \end{document}
512: