ANTECEDENTES – ESTADO DEL ARTE

The implementation of this novel memory protocol was performed in the ANU cold atomic ensemble, whose preparation is described in 2.3.4. The atoms are pre- pared in |gi =

5S1/2, F = 2, mF = +2

E

and the storage state is chosen to be |si =

5S1/2, F = 1, mF = 0 E

. The excited state is taken as |ei=

5P1/2, F

0 = 1_{, m}

F = +1

E

, on the D1-line. The configuration in frequency and beam geometry is identical to the

Raman memory described in chapter 4: each signal and control pair forms a 6 mrad angle to match the hyperfine splitting of rubidium-87, so they both address the same spin wave, as shown on figure 5.10. The forward and backward pairs are symmetrically detuned from the excited state, blue and red detuned respectively. The detuning was chosen as 160 MHz or 230 MHz in different experimental trials. The OD on the signal transition is measured on the absorption on the Raman lines formF ∈ {−1,0,+1} with

varying control power, as described in 2.3.5. OD as high as 500±100 were obtained.

Interference between counter-propagating signals

Counter-propagating probes with a profile shaped as a rising exponential are sent simultaneously in the prepared atomic ensemble. Photodiodes are placed at both ends

5.4 Time-Reversed And Coherently-Enhanced memory 133 0 5 10 15 20 0 5 10 15 20 25 30 Time (μs) Intensity (arb) Backward Detector 0 5 10 15 20 0 10 20 30 40 50 Time (μs) Intensity (arb) Forward Detector

Figure 5.11: Photodiode signals on forward and backward detectors during the realisa- tion of the TRACE protocol. A rising-exponential-shaped probe is sent from both sides of the cold atomic ensemble. Depending on the relative phase of the counter-propagating probes, the absorption process of each one interfere either constructively (in blue) or destructively (orange). Here the forward and backward control beams are sent continuously and the probes are immediately retrieved, with their time-reversed profile. The reference probes without atoms is shown for reference (green).

of the atomic ensemble, in the same configuration as for the backward-retrieval Raman memory experiment presented in chapter 4.

Initially only one probe is sent at a time, the control fields are kept on, so that each probe experiences storage in the atomic ensemble as a spin wave immediately followed by its retrieval. The forward and backward control intensity are tuned to result in the distribution of each probe pulse in four equal parts: leakage to the detector in the same direction, reflection to the detector in the opposite direction and retrieval in both directions.

Both probes are sent simultaneously and their relative phase becomes random from run to run from the interferometric instability between their different paths. Figure 5.11 shows the forward and backward photodiode signals for the cases of minimum and maximum absorption on the probe pair.

The relative phase between control-probe pairs is almost completely random after each loading phase, of the order of a second. However, it varies only by a small amount over each memory sequence, which lasts about 400 µs. Therefore a series of 17 pairs of probe pulses are sent within the same sequence, with constant control field intensity, in order to investigate the phase dependence of the observed interference. The phase of each successiveE+ pulse is incremented by 0.3π while in a second experiment, the

phase is not incremented. The spin waves produced by each counter-propagating pair interfere. Constructive interference results in an enhanced mapping to the spin wave while destructive interference corresponds to a dark state which inhibits the absorption. A complete interference fringe is obtained when a total of 2π is incremented to the phase over the length of the train of pulses. The results are shown in figure 5.12. Interference

is also visible when the frequency of the control fields are detuned from the two-photon resonance.

Visibilities were calculated by comparing the most constructive and destructive interference into the spin wave relative to the total optical power of the input pulses. Visibilities of 70% are achieved which testifies to the spatio-temporal matching of the forward and backward pairs. The production of temporal interference fringes indicates that deterministic storage could be achieved by measuring the relative phase of input probes and adjusting a control field phase or a path length accordingly.

Performance as a memory scheme

The interference fringe shows that some pulses are efficiently stored while others destructively interfere into the ensemble. By monitoring the leakage from the atomic ensemble as the phase shifts between and within experimental runs we could herald the efficient writing of constructively-interfering spin waves and recall them on-demand. We have measured the amount of probe light retrieved from the stored spin waves and characterised this heralded probabilistic optical memory and compared it to gradient echo and Raman memory schemes using the same preparation of the atomic ensemble.

At large ODs around 500, we could not conclude favourably on the relative performance of the TRACE memory scheme compared to GEM or Raman memory: the schemes performed equivalently in the 50 −65% storage-and-

retrieval efficiency range. With an OD decreased to a few tens, TRACE performed comparatively better in the only experimental attempt of performing the three memory schemes at high and low OD within a few hours. However, we cannot assert that the GEM and Raman parameters and alignment were optimised to reach the highest efficiency. In particular, the configuration with an angle between the probe and control does not minimise the spin-wave wave vector and is therefore sub-optimal for these memory schemes.

The decay of the TRACE memory with storage time was investigated by varying the duration for which both controls are kept off after absorption of the probe pulses. Figure 5.13 shows the dependence of the retrieval efficiency against storage time, which is fitted to exponential decay of characteristic time (130±10) µs, which corresponds to a

temperature (270±40) µK, compatible with an independent time-of-flight temperature

measurement.