On-Demand Millisecond Storage of Spectro-Temporal Multimode Telecom Photons

Abstract

The realization of scalable quantum networks for distribution of entanglement over long distances hinges on quantum repeaters. To outperform the exponential transmission loss in optical fibers, quantum repeaters must employ multiplexing schemes in the temporal, spectral, or spatial domain. The performance of such a multiplexed scheme is contingent on efficient quantum memories offering both extended storage times and large multimode capacities. In this work, we experimentally demonstrate such a memory operating at telecom wavelength using an Er$^{3+}$:Y$_{2}$SiO$_{5}$ crystal. Using single-photon detectors, we record on-demand storage and recall of weak coherent pulses for up to $1$ ms, exceeding that of previously reported quantum memories based on Er$^{3+}$. The memory exhibits an efficiency of 10.36\% at 300 $μ$s storage time with a signal-to-noise ratio of $10.9$. We further showcase its multimode capacity by storing 20 temporal and 3 spectral modes simultaneously with on-demand and selective recall capabilities, essential for a scalable quantum repeater architecture.

Figures (3)

• Figure 1: (A) Illustration of a spectro-temporal multiplexed quantum repeater with two elementary links. The protocol begins with the generation of $M_t$ entangled photon pairs simultaneously across $M_s = 3$ spectral modes using probabilistic non-linear sources. The signal photons (s) are stored in an on-demand QM, while the idler photons (i) travel through the channel for entanglement swapping. Idler photons from adjacent nodes interfere on a 50:50 beam splitter and are spectrally resolved for individual detection. A successful BSM triggers a feedback signal to recall the corresponding spectral mode from the QM, followed by an appropriate frequency shift for the subsequent swapping between signal photons of neighboring links. (B) Entanglement distribution success probability $P_s$ as a function of distance. The solid curves show the performance of a quantum repeater for different total multimode capacity $M = M_t \times M_s$ given memory parameters $\eta_o=65\%$ and $T_2=3$ ms. The pink dashed curve represents direct transmission through SMF-28 fiber. The success probability is optimized for the number of elementary links ($n_l$), and the squares represent an increment in $n_l$ starting from $n_l=1$. Additionally, the minimum QM storage time ($T_s$) required for each scenario is estimated. See Supplemental Material SM for details on simulation parameters.
• Figure 2: (A) Schematic of the experimental setup. The two Acousto-Optic Modulators (AOMs) before the crystal prepare input modes and the chirp pulses (CP). The AOMs after the crystal, with combined efficiency of $\eta_\textrm{AOM}=35\%$, allow only the input and desired echo to pass through for detection via Superconducting Nanowire Single Photon Detector (SNSPD), featuring $\eta_\textrm{D}=67\%$ detection efficiency. (B) Illustration of the pulse sequence for a storage cycle of the CPPE memory protocol. The storage and recall of input modes is performed using CP1 and CP2, as shown in the dashed box. (C) Timing histogram of photon counts from SNSPD for $N_c = 9000$ cycles. The leftmost panel showcases the off-resonance input, with an estimated mean photon number of 720 before the crystal, and the unabsorbed input during the memory sequence. Other panels display the recalled echo and the noise (measured without the input) for experiments with different storage times. The SNR of the recalled echo is 10.90 for $300\,\mu$s and 1.52 for $1$ ms. (D) The experimental memory retrieval efficiency is plotted against storage time and fitted with an exponential decay REI_Multimode_lago-rivera-2021. The fit's decay constant yields the coherence time, $T_2 = 858.4 \pm 80.4\,\mu$s and $\eta_o = 23.05\%$.
• Figure 3: Demonstration of spectral-temporal multimode capacity. (A) A total of 25 temporal modes, with an average input mean photon number of $2489$, are stored for 800 $\mu$s and subsequently retrieved with an average efficiency of 2.67% and SNR of 9.18. (B) Spectral-temporal multimode capacity with selective recall. The left panel shows a train of 20 temporal modes, where each mode constitute a various combination of 3 distinct spectral modes ($\omega_1, \omega_2$, and $\omega_3$) as depicted in the left panel. All prepared modes are stored simultaneously for 800 $\mu$s, while either of the spectral modes is selectively recalled as shown in the right panels. The storage efficiency of the three spectral modes is 1.20%, 1.64%, and 1.66%, respectively. (C) Sequential recall of two spectral modes. The left panel shows a 10 weak coherent pulse input sequence prepared with two spectral modes ($\omega_1$ and $\omega_2$). After simultaneous storage, they are recalled independently in the same cycle where $\omega_1$ is retrieved at 450 $\mu$s and $\omega_2$ is retrieved at 850 $\mu$s yielding efficiencies of 9.33 % and 2.08 %, respectively.