Long-lived and Efficient Optomechanical Memory for Light

We demonstrate a memory for light based on optomechanically induced transparency. We achieve a long storage time by leveraging the ultralow dissipation of a soft-clamped mechanical membrane resonator, which oscillates at MHz frequencies. At room temperature, we demonstrate a lifetime $T_1\approx 23\mathrm{ms}$ and a retrieval efficiency $\eta \approx 40\%$ for classical coherent pulses. We anticipate the storage of quantum light to be possible at moderate cryogenic conditions ($T\approx 10\mathrm{K}$). Such systems could find applications in emerging quantum networks, where they can serve as long-lived optical quantum memories by storing optical information in a phononic mode.

Figure 1

Steady-state OMIT. (a) An optomechanical cavity mode with frequency ${\mathrm{\Omega }}_{\mathrm{cav}}$ and linewidth $\kappa$ is pumped near the red sideband $\mathrm{\Delta }={\mathrm{\Omega }}_{\mathrm{L}}-{\mathrm{\Omega }}_{\mathrm{cav}}\approx -{\mathrm{\Omega }}_{\mathrm{m}}$. An externally applied modulation at ${\mathrm{\Omega }}_{\mathrm{sig}}$ provides a signal to store. (b) The optical ($n$) and mechanical ($m$) excitations form a $\mathrm{\Lambda }$ system that can facilitate storage when the two-photon resonance condition $\delta ={\mathrm{\Omega }}_{\mathrm{sig}}-{\mathrm{\Omega }}_{\mathrm{m}}\approx 0$ is met. (c) Experimental data (circles) for coarse broadband (top) and fine narrowband (bottom) sweeps of ${\mathrm{\Omega }}_{\mathrm{sig}}$, with fits (solid blue) to OMIT theory. The broad sweep outlines the cavity response, while the narrow sweep shows interference between the mechanical sideband and applied signal around $\delta =0$.

Figure 2

Soft-clamped membrane resonator. (a) The mechanical resonator. From left to right, a micrograph of the device used, a simulated displacement field of the mode we work with, and a photograph of an exemplary mechanical resonator. (b) In black, simulated band diagram, showing a fundamental (red) and higher-order (green) band gap. Inset: first Brillouin zone. (c) Mechanical displacement spectrum around the second band gap, showing distinct localized modes, annotated with displacement profiles (the black peak at 2.36 MHz is an external phase modulation). (d) Amplitude decay of the mode of interest, exhibiting $Q=100(3)\times {10}^6$ at ${\mathrm{\Omega }}_{\mathrm{m}}/2\pi =2.4\mathrm{MHz}$.

Figure 3

Membrane-in-the-middle cavity optomechanics. (a) Simplified experimental schematic. All beam splitters are polarizing. (b) Mechanical memory bandwidth ${\mathrm{\Gamma }}_{\mathrm{eff}}$ as a function of science laser detuning $\mathrm{\Delta }$ at fixed input power. The black line is a single free parameter ($g_0$) fit to a model of dynamical backaction.

Figure 4

Optomechanical light storage. (a) Pulse sequence diagram. We drive the EOM with exponentially rising sinusoids, while the science laser is on. After waiting a time $T_{\text{delay}}$, we switch the AOM on again and look at the retrieved mechanical signal. (b) By lock-in detection around the mechanical sideband, we record the magnitude of the reflected science laser. Here, we show data for fixed ${\mathrm{\Gamma }}_{\mathrm{eff}}$ and increasing ${\mathrm{\Gamma }}_{\mathrm{sig}}$ from light to dark blue, the legend showing ${\mathrm{\Gamma }}_{\mathrm{eff}}/{\mathrm{\Gamma }}_{\mathrm{sig}}$.

Figure 5

Storage efficiency contributions. (a) Normalized retrieved amplitude subject to increasing readout delay. The gray hatched area indicates the thermomechanical noise floor. (b) Blue points: retrieved energy $E_{\mathrm{out}}$ normalized to signal energy $E_{\mathrm{in}}$ for varying ${\mathrm{\Gamma }}_{\mathrm{sig}}$ and fixed ${\mathrm{\Gamma }}_{\mathrm{eff}}$. In dashed gray, the theoretical prediction Eq. (6) corresponding to unity cavity coupling efficiency. The black curve is a fit to the same model with one free multiplicative constant, with the shaded region indicating $\pm$ twice the error on the fit. The red hatched area indicates the limit to efficiency predicted by the cavity overcoupling, in fair agreement with the fit. (c) Two-photon detuning. For fixed ${\mathrm{\Gamma }}_{\mathrm{sig}}\approx {\mathrm{\Gamma }}_{\mathrm{eff}}$ the signal frequency ${\mathrm{\Omega }}_{\mathrm{sig}}$ is tuned across the mechanical resonance. The normalized, retrieved energy (blue points) is well-modeled by a Lorentzian, as predicted by Eq. (7). The gray hatched area indicates the mechanical decay during the storage time. Accounting for the limitations of overcoupling, and the mechanical decay during the storage, we find good quantitative agreement between measurements and the expected bound. All error bars indicate twice the standard error on the mean.