Scalable Optical Quantum State Synthesizer with Dual-Mode Resonator Memory

TL;DR

This work tackles the key challenge of generating highly non-Gaussian optical states for scalable, fault-tolerant quantum computing by introducing a dual-mode optical resonator memory that supports continuous-time storage and time-domain-multiplexed breeding. The authors demonstrate a complete, write–store–read memory with 93% overall efficiency and lifetimes $T_1=2.3\,\mu$s and $T_\phi=0.96\,\mu$s, and realize cat- and GKP-breeding protocols in a single-device, time-domain fashion using a single non-Gaussian source. The results show nonclassical state generation via Wigner-function negativities, a significant interference-rate improvement ($k=3.8$) that scales with memory, and a clear pathway toward multi-step breeding and fault-tolerant state preparation. Collectively, this work provides a scalable platform for generating complex non-Gaussian states with broad implications for optical quantum computing, communication, sensing, and metrology, by bridging continuous-time quantum memory with scalable entangling operations. Future work emphasizes faster switching and multi-memory architectures to push toward fault-tolerant, large-scale quantum information processing.

Abstract

Optical quantum computing is a promising approach for achieving large-scale quantum computation. While Gaussian operations have been successfully scaled, the inherently weak nonlinearity in optics makes generating highly non-Gaussian states a critical challenge for universality and fault tolerance. Here, we propose and experimentally demonstrate a scalable method to generate optical non-Gaussian states with a resonator-based quantum memory that supports continuous-time storage and retrieval, in contrast to conventional loop-based memories. We introduce a dual-mode operation of the memory, enabling both storage and entangling functionalities within a single device. By employing a time-domain-multiplexed approach, we successfully demonstrate both cat and Gottesman-Kitaev-Preskill (GKP) breeding protocols in a scalable fashion, marking a key step toward quantum error correction. Our experiment also marks the first full demonstration of an optical resonator memory performing writing, storage, and readout operations. We validate the memory by storing squeezed single-photon states with up to 93% total efficiency, and measure an energy relaxation time $T_1 =$2.3$μ$s and dephasing time $T_φ =$0.96$μ$s. These results establish a scalable pathway to generating complex non-Gaussian states required for fault-tolerant optical quantum computing. Beyond computation, our techniques provide new tools for enhancing quantum communication, sensing, and metrology.

Figures (8)

• Figure 1: Dual-mode resonator quantum memory and TDM breeding protocol. a Dual-mode operation of the resonator quantum memory, modeled as an optical resonator with a variable beamsplitter (VBS) of tunable coupling $\gamma(t)$. In memory mode, an incoming wavepacket (label "1") is absorbed and later released; in beamsplitter mode, the input (1) interferes with the intra-resonator mode (2), producing outputs (1',2'). b Conventional breeding protocol requiring many parallel non-Gaussian sources and iterative beamsplitters. c Time-domain-multiplexed (TDM) breeding protocol using a single source and the dual-mode memory, sequentially implementing the equivalent circuit of b.
• Figure 2: Detailed experimental setup. OPA: optical parametric amplifier; VBG: volume Bragg grating; PBS: polarization beam splitter; QAWG: quantum arbitrary waveform generator QAWG; SNSPD: superconducting nanowire single-photon detector.
• Figure 3: Performance of the resonator memory. a Schematic of the experimental setup. b Pulse sequence applied during the experiment. c Temporal-mode functions of the input and output wavepackets for different storage times $t_0$. d Wigner functions of the input and output states for varying squeezing levels and storage times (axis range $x,p \in [-5,5]$, with $[\hat{x},\hat{p}] = i$). e Memory efficiency and coherence, obtained from storing a single-photon state (first row of d) and a squeezed single-photon state (last row of d), respectively; the solid lines are exponential fits, yielding a total efficiency of 93%, an energy relaxation time $T_1 = 2.3µ s$, and a pure dephasing time $T_\phi = 0.96µ s$.
• Figure 4: Implementation of the TDM breeding protocol using the resonator memory. a Schematic of the experimental setup. b Pulse sequence applied during the experiment. c Experimentally obtained temporal-mode functions: (first and second rows) input wavepackets, and (third row) output wavepackets. The second output wavepackets are shown for two different storage times, $t_2 = 0$ and $t_2 = 40ns$. d Wigner functions of (left) the input squeezed cat state, (middle) output state of the breeding protocol with $t_2 = 0$, and (right) output state with $t_2 = 40ns$. Results for both cat and GKP breeding are presented.
• Figure 5: Implementation of the TDM GKP breeding protocol using resonator quantum memories. a Configuration with a single quantum memory, as implemented in the main text. b Configuration with two quantum memories. c Schematic representation of a protocol equivalent to that implemented by the system in a. d Evolution of the Wigner function during the breeding protocols.