Suppression of the four-wave-mixing background noise in a quantum memory retrieval process by channel blocking

In a quantum memory scheme with the Raman process, the read process encounters noise from four-wave mixing (FWM), which can destroy the nonclassical properties of the generated quantum fields. Here we demonstrate experimentally that the noise from FWM can be greatly suppressed by simply reducing the FWM transition channels with a circularly polarized read beam while at the same time retaining relatively high retrieval efficiency.

Figure 1

(a) Write process. (b) Read process. (c) Normal optical pumping. (d) Polarized optical pumping. The circularly polarized pump field clears the population in all magnetic sublevels but the last.

Figure 2

Schematic experimental setup. W: write; S: Stokes; R: read; AS: anti-Stokes; ${\mathrm{P}}_1$: linearly polarized pump; ${\mathrm{P}}_2$: circularly polarized pump; M: mirror; QWP: zero-order quarter-wave plate; PBS: polarizing beam splitter; FP1 and FP2: Fabry–Perot cavities; SPD1 and SPD2: single-photon detectors; TAC: time-amplitude converter; SCA: single channel analyzer. The read laser counterpropagates and overlaps with the write laser.

Figure 3

The spectrum of the anti-Stokes signal from the scan of the FP cavity without write process. The FWM noise with linearly polarized read beam is shown as black squares and the circularly polarized read beam is shown as red dots. The solid lines connect the data points in a smooth fashion for visual guidance. The $x$ axis is the channel of SCA and stands for the frequency, with 18 MHz per channel. The Fabry–Perot cavity is scanned by a triangular voltage with a frequency $f$ of 5 Hz that is synchronized with the pulse sequence. The FSR is about 1.1 GHz. The time window per channel, $\Delta T_c$, is set to 0.5 ms. The pulse number can be obtained by $N_{\text{pulse}}=\Delta T_{c}ft/T$. $T$ is the period of the pulse sequence and $t$ is the integral time. The $y$ axis stands for the intensity of signal expressed in photons per pulse: $N_{\text{photon}}/N_{\text{pulse}}$. $N_{\text{photon}}$ is the total number of photons detected during the time window.

Figure 4

(a) The generation of FWM noise during the read process based on Raman scattering with normal optical pumping and using a linearly polarized read beam. (b) The suppression of FWM noise based on the scheme of Walther et al. in which atoms are prepumped into the highest magnetic sublevel and the read beam is circularly polarized. The circularly polarized read beam cannot couple $|5{{}^{2}S}_{1/2},F=2,m_F=2\rangle$ to the excited state $|5{{}^{2}P}_{1/2},F^{\prime }=1,2\rangle$.

Figure 5

The spectrum from the scan of the FP cavity for (a) the Stokes field in the write process and (b) the anti-Stokes field in the read process. The cases for linearly and circularly polarized read beams are marked as black squares and red dots, respectively. The solid lines connect the data points in a smooth fashion for visual guidance. The labels for the axes are the same as for Fig. 3.

Figure 6

(a) Stokes (squares) and anti-Stokes (solid circles) photons number per pulse generated in the write and read processes with circularly polarized write and read beams and atomic polarized population ratio of the magnetic sublevels (triangles) as a function of the power of laser ${\text{P}}_2$. (b) The corresponding retrieve rate for the case of panel (a).

Figure 7

The FWM transition channels starting from the state $5{{}^{2}S}_{1/2}(F=1)$ with (a)–(c) linearly polarized and (d), (e) circularly polarized read laser. Blue is ${\sigma }^{+}$ polarization, red is ${\sigma }^{-}$ polarization, and yellow is $\pi$ polarization.