Single-Photon-Level Atomic Frequency Comb Storage in Room Temperature Alkali Vapour

TL;DR

We address the challenge of storing quantum states of light in a room-temperature platform by implementing the Atomic Frequency Comb protocol in a rubidium vapor. The authors use velocity-selective pumping on the D1 line to prepare an AFC in the $F=2$ ground state, achieving single-photon-level storage using a D2 readout with an echo at $\tau = 2\pi/\Delta_{AFC}$. They characterize spectral and temporal performance, observe a hyperfine-beating interference between AFCs, and demonstrate polarization and time-bin qubit compatibility with measured efficiencies up to $\sim$10% for bright pulses and $\sim$6% for single-photon inputs, along with a two-mode time-bin demonstration. The work shows that room-temperature AFC memories are feasible and potentially interoperable with quantum-dot photon sources, offering a path toward scalable, practical quantum memories and networks. The results also indicate clear routes for improvements in efficiency, bandwidth, and storage time through cavity enhancement, spin-wave storage, and line selection to mitigate interference effects.

Abstract

We have demonstrated the coherent storage and retrieval of single-photon-level light using the atomic frequency comb protocol in a room temperature rubidium vapour. Velocity-selective optical pumping is used to prepare the comb within the $F=2$ hyperfine ground state of rubidium, with the spacing between peaks coinciding with half the $F = 2 - F =3$ hyperfine splitting of the $5^2$P$_{3/2}$ excited state. Weak coherent states of average photon number $μ_\mathrm{in} = 0.083(5)$ are stored with pre-programmed recall time of $7.5\,$ns with an efficiency of $η_{\textrm{AFC}} = 6.59(5)\,\%$, while two temporally distinct modes have been stored and recalled with $η_{\textrm{AFC}} = 2.6(1)\,\%$, allowing for time-bin qubit storage. Finally, the efficiency is observed to be independent of the input pulse polarisation, paving the way for polarisation qubit storage.

Figures (9)

• Figure 1: (a) Energy level structure of $^{87}$Rb (not to scale) showing the term symbols, hyperfine $\textrm{F}$ states, and our custom state labels. Refer to text for hyperfine splitting values and definitions of state labels. The {dashed, dot-dashed, solid} arrows represent the {pump, pump-back (p-b), probe} optical modes. (b) and (c): Normalised optical density (OD) of the D$_1$ and D$_2$ lines, respectively. Zero detuning corresponds to the weighted centre of the line. The red curves $(i)$ show the OD for each of the $\textrm{F}=2 \rightarrow\textrm{F}'$ transitions while the blue curves $(ii)$ show the $\textrm{F}=1 \rightarrow\textrm{F}'$ transitions. The {dotted, dot-dashed, dashed} lines show transitions of $\Delta\textrm{F} =\{-1, 0, 1\}$. The shaded grey area shows the total OD. (d) and (e): Zoomed in spectra of the $\textrm{F}=1 \rightarrow\textrm{F}'$ transitions for the D$_1$ and D$_2$ respectively. Note the logarithmic y-axis for (d). Vertical bars indicate a frequency resonant with the $\textrm{F}=1 \rightarrow\textrm{F}' = 2$ zero-velocity class with the height corresponding to the OD of a given transition and width chosen for clarity. For (d) the heights are $\sim\{1.0,\,3.5\times10^{-4}\}$ while for (e) $\sim\{0.30,\,0.23,\,0.07\}$. This showcases the inherent advantage of using the D$_1$ line to implement velocity selective pumping where contributions from the neighbouring transition is at the $10^{-4}$ level, whereas for the D$_2$ contributions for different transitions are of a similar order. Spectra are created using the ElecSus Python Package Zentile2015Keaveney2018 with T $= 26.9~^\circ$C, cell length $= 10~\textrm{cm}$, and $^{87}$Rb fraction $= 100\%$.
• Figure 2: Simulated D$_2$$\textrm{F} = 2$ absorption spectra showing the results of velocity-selective optical pumping. (a) Velocity-selective optical pumping for the zero velocity class (blue dashed) and the unpumped thermal population (black line). The three allowed transitions, $\textrm{F}=2 \rightarrow\textrm{F}' = \{1, 2, 3\}$, appear at increasing frequencies. The detuning is defined with respect to the $\textrm{F}=2 \rightarrow\textrm{F}' = 3$ transition. (b) Three individual velocity classes are prepared with $\{-v_1, v_0, v_1\}$ with lifestyles $\{\textrm{purple dotted}, \textrm{blue dashed}, \textrm{red dot-dashed}\}$ where the velocity has been chosen such that the shift is $\Delta^{(\textrm{e}_2)}_{\textrm{F}_2'\textrm{F}_3'} / 2 \sim 133.33~\textrm{MHz}$. (c) The resulting spectrum when the three velocity classes $\{-v_1, v_0, v_1\}$ are applied concurrently showing the AFC (grey shaded). These simulations have considered an ideal pump phase that completely removes all population from the $\textrm{F} = 2$ ground state, and a pump-back phase with optical modes that have power of $100~\mu\textrm{W}$, a beam radius of $2~\textrm{mm}$, a duration of $4~\mu\textrm{s}$, a laser linewidth of $2\pi\times2~\textrm{MHz}$, T $= 26.9~^\circ$C, cell length of $10~\textrm{cm}$, and $^{87}$Rb fraction of $100\%$.
• Figure 3: (a) Experimental setup schematic, refer to text for details. RF - modulation frequency of the pump-back laser, TA - Tapered Amplifier, AOM - Acoustic-Optical Modulator, PBS - Polarising Beam Splitter, DM - Dichroic Mirror, ND - Neutral Density Filter, BF - Bandpass Filter, PPG - Programmable Pulse Generator, EOM - Electro-Optic Modulator. (b) Weak coherent state AFC storage pulse sequence. RF switches are used to turn the pump AOMs off for a duration $\tau_\textrm{off}$ in which the EOM is activated to generate an input pulse at a time of $\tau_\textrm{delay}$ after the AOMs have been turned off. The AOMs provide at least four orders of magnitude power reduction. An echo is seen at a time $\tau_\textrm{AFC}$ after this when the AFC is present. The AOMs and EOM are triggered using a Standford Research Systems Digital Delay Generator (DG645).
• Figure 4: Measured and simulated spectra for the initial thermal distribution (black-dashed and gray-shaded) and AFC (blue-dashed and blue shaded), with detuning defined with respect to the $\textrm{F}=2 \rightarrow\textrm{F}' = 3$ transition. For the optimisation of the simulated parameters, we subtract the residual population left in the ground state due to imperfect pumping (purple shaded area) from the AFC data. The resulting AFC simulation curve then has this residual population added to it for this plot. See main text for further details.
• Figure 5: AFC storage. Input (light blue solid line), transmitted and AFC echo (dark red solid line). The $3\,\mathrm{ns}$ integration windows are indicated as shaded regions. The detector ring-down results in negative signal values; the background (black dashed line) is estimated from a fit to the signal, see text for details. The inset shows a background subtracted signal with a Gaussian fit (pink dots).