Broadband and long-duration optical memory in Yb:YSO

Optical quantum memories are essential components for realizing the full potential of quantum networks. Among these, rare-earth-doped crystal memories stand out due to their large multimode storage capabilities. To maximize the multimode capacity in the time domain, it is key to simultaneously achieve large memory bandwidth and long optical storage time. Here, we demonstrate an atomic frequency comb optical memory in Yb:YSO, with a memory bandwidth of 250 MHz and a storage time of up to 125 $μ$s. The efficiency reaches 20 % at short storage times, and 5 % at 125 $μ$s. These results were enabled by an optimized optical pumping scheme, guided by numerical modelling. Our approach is specifically designed for future spin-wave storage experiments, with the theoretical bandwidth limit set at 288 MHz by the hyperfine structure of Yb:YSO. Additionally, we introduce an efficient method for synthesizing the optical pumping waveforms required for generating combs with tens of thousands of teeth, as well as a simple yet frequency-agile laser setup for optical pumping across a 10 GHz bandwidth.

💡 Research Summary

This paper presents a breakthrough in rare‑earth‑doped crystal quantum memories by demonstrating an atomic frequency comb (AFC) memory in ^171Yb³⁺:Y₂SiO₅ (Yb:YSO) that simultaneously offers a large bandwidth of 250 MHz and a long optical storage time of up to 125 µs. The authors achieve a short‑delay efficiency of about 20 % and a 5 % efficiency at the longest delay, representing the highest AFC coherence time (≈350 µs) reported for any rare‑earth crystal.

Key to these results is an optimized optical pumping scheme based on a newly designed class‑cleaning (CC) protocol. By selecting four specific hyperfine transitions (ν₄g‑₁e, ν₁g‑₁e, ν₃g‑₄e, ν₂g‑₃e) and scanning them over a 250 MHz window, the authors isolate a single frequency class that couples only to the desired Λ‑system (|4g⟩–|1e⟩ and |1g⟩–|1e⟩). Numerical simulations confirm that this CC scheme creates a flat, highly absorbing anti‑hole of 250 MHz width while pumping ≈80 % of the total population into the |4g⟩ ground state, thereby increasing the effective optical depth by a factor of 3.3.

The AFC preparation itself is refined through a fast waveform‑synthesis algorithm that generates a square‑shaped power spectral density (PSD) matching the theoretical optimum finesse (F_opt = π/ arctan(2π/d)). The authors implement this with a single frequency‑locked laser and a single electro‑optic modulator (EOM), achieving frequency agility over a 10 GHz range. This minimalist setup replaces more complex multi‑laser schemes and enables the creation of combs with tens of thousands of teeth.

Material improvements also contribute: the Yb³⁺ doping concentration is reduced from 5 ppm to 2 ppm, extending the homogeneous optical coherence time to 1.05 ms at ~3 K and reducing spin‑spin flip‑flop relaxation, which enhances optical pumping efficiency. The hyperfine structure of ^171Yb³⁺ provides a ground‑state splitting of 3.025 GHz, setting a theoretical bandwidth ceiling of 288 MHz for a spin‑wave AFC memory. The demonstrated 250 MHz bandwidth approaches this limit.

The work outlines a clear path toward a spin‑wave AFC memory: the 3025.5 MHz spin transition can be efficiently driven with a lumped‑element microwave resonator, and the optical control pulse (Rabi frequency ≈2 MHz) is sufficient for adiabatic population inversion across the full bandwidth. Future experiments will need to integrate microwave control and optimize the optical Rabi frequency to fully exploit the 100 MHz+ bandwidth predicted for a spin‑wave implementation.

Overall, this study delivers a practical solution to the longstanding trade‑off between bandwidth and storage time in rare‑earth quantum memories. By combining a sophisticated class‑cleaning protocol, efficient waveform synthesis, and material engineering, the authors achieve a memory that is both broadband and long‑lived, paving the way for multimode quantum repeaters that can store many temporal and spectral modes simultaneously. The techniques introduced are readily adaptable to other Kramers ions (e.g., Er³⁺, Nd³⁺), suggesting broad impact on the development of scalable quantum networks.