Our Results

1. Experiment Setup

Experiment setup. TL1-2: tunable lasers; I1-3: optical isolators; FPC1-4: fiber polarization controllers; MZM1-2: Mach-Zehnder modulators; PPG: pulse pattern generator; EDFA1-4: Erbium-doped fiber amplifiers; 50/50: 3-dB coupler; BPF1-2: bandpass filters (bandwidth 0.2 nm and 0.06 nm, respectively); C1-2: optical circulators; ATT1-2: tunable attenuators; HNLF: highly nonlinear fiber; D1-2: 13-GHz-bandwidth detectors; OSC: digital sampling oscilloscope.

The figure above shows the schematic of our experiment setup for oberserving stored light. We use a short length (5-m) of commercially available highly nonlinear optical fiber optical fiber (HNLF, from OFS) as the stored-light medium. We use two tunable cw diode lasers operating at wavelengths near 1.55 μm, with output frequencies offset by the Brillouin frequency shift of the HNLF (ΩB/2π ~9.6 GHz). The outputs of the lasers are amplitude-modulated to create the write, read and data pulses. We cascade three Erbium-doped fiber amplifiers to increase the peak power of the write and read pulses. The fiber loop formed by two 3-dB couplers, a 1-km-long single-mode fiber (SMF28e) and a fiber polarization controller is used to remove the weak coherent background and amplified-spontaneous-emission noise. By adjusting the data pattern in the pulse pattern generator, we position the write, read, and pulses so that they interact within the HNLF. The data and write/read pulses exiting the HNLF are respectively attenuated, detected and displayed in a digital sampling oscilloscope. The strength of the SBS process is maximized by adjusting the polarization state of the data pulses via a polarization controller and the frequency detuning between the two lasers.


2. Results

Observation of stored light. A shows experimental results for a 2-ns-long rectangular-shaped data pulse and C shows the corresponding theoretical simulations. B shows the case of a 2-ns-long smooth data pulse with the corresponding simulations in D.  The retrieved pulses are shown with a multiplication factor of 2 to the right of the dashed vertical line.
Typical write and read pulses used in the experiment.

The figure above shows the experimental results for storing and retrieving a single rectangular-shaped 2-ns-long data pulse using identical, 1.5-ns-long write and read pulses with peak powers of ~100 W (see figure on the right). The incident data pulse in the absence (presence) of the write and read pulses is shown to the left of the dashed vertical line by the blue (green) line. The energy storage efficiency of the storage process is very high — equal to ~66% — indicating that we have faithfully encoded the optical pulse information onto the acoustic excitation. The curves to the right of the vertical dashed line are the observed retrieved pulses for various storage time Ts, where we have scaled them by a factor of 2 for clarity. For Ts = 4 ns, we obtain a readout energy efficiency of 25%, defined as the energy of the released pulse divided by the energy of the incident data pulse. For Ts = 12 ns, the storage time is equal to 6 pulse widths with an efficiency of 1.8%. As expected, we see that the readout efficiency drops with increasing storage time due to the decay of the acoustic excitation.

The fast rising and falling edges of the data pulse in the time domain contribute to its high-frequency content in the frequency domain. The edges are not fully stored because the write-pulse spectrum is not broad enough to fully encompass the data-pulse spectrum. To see this, we also stored and retrieved a smoothed data pulse (shown in B). In this case, the energy storage efficiency increases to 86% and the read out efficiencies are improved slightly (~29% for Ts = 4 ns).

To verify the interpretation of our results, we have solved numerically the equations governing the interaction between the optical and acoustic waves in an optical fiber. The simulations (C and D) are in good agreement with observations.

The figure below shows our observations for the storage and retrieval of sequences of 2 and 3 data pulses, respectively. For clarity, we show only the input data pulses to the left of the dashed vertical line, and to the right we show the retrieved pulses scaled by a factor of 5. The light released from the fiber clearly replicates the input data stream with reasonable fidelity. The numerical simulations (C and D) are again in good agreement with the observations (A and B).

Storage of pulse sequences. A shows experimental results for two 2-ns-long data pulses separated by 1 ns and C shows the corresponding theoretical simulations. B shows the case of three 2-ns-long data pulses separated by 1 ns with the corresponding simulations D.  The retrieved pulses are shown with a multiplication factor of 5 to the right of the dashed vertical line. For clarity, the depleted data pulses are not shown.