Polarization-Controlled Single Photons

Vacuum-stimulated Raman transitions are driven between two magnetic substates of a ${}^{87}\mathrm{Rb}$ atom strongly coupled to an optical cavity. A magnetic field lifts the degeneracy of these states, and the atom is alternately exposed to laser pulses of two different frequencies. This produces a stream of single photons with alternating circular polarization in a predetermined spatiotemporal mode. MHz repetition rates are possible as no recycling of the atom between photon generations is required. Photon indistinguishability is tested by time-resolved two-photon interference.

Figure 1

Scheme and setup: (a) and (b) Relevant energy levels in ${}^{87}\mathrm{Rb}$ for photon production. (a) The ${\sigma }^{-}$ polarized component of the pump laser combines with the cavity to drive a resonant Raman transition which takes the atom from $|+1⟩$ to $|-1⟩$, producing a ${\sigma }^{+}$ photon. Although the ${\sigma }^{+}$ component of the pump laser (shown in gray) is present as well, this transition (which would produce an additional ${\sigma }^{-}$ photon) is far off-resonant and thus unlikely. (b) A second pump pulse of different frequency returns the atom to $|+1⟩$, while producing a ${\sigma }^{-}$ photon. Again, the off-resonant return transition is unlikely. (c) As atoms fall from a MOT through the cavity we shine in the pump laser from the side to generate photons. The photons are directed through one of two fibers by a PBS, the long 270 m fiber acting as a delay line. Photons emerging from the fibers can interfere at a BS, and are detected by a pair of avalanche photodiodes (APDs).

Figure 2

Photon characteristics observed with only one path to the beam splitter open. (a) Pump pulse sequence. The first pulse, labeled ${\omega }_{+}$, has a detuning of ${\Delta }_{pc}/2\pi =28\mathrm{MHz}$, the second (${\omega }_{-}$) ${\Delta }_{pc}/2\pi =-28\mathrm{MHz}$. (b),(c) Photon detection-time distribution for ${\sigma }^{+}$/${\sigma }^{-}$ photons, showing they are overwhelmingly detected during the ${\omega }_{+}$/${\omega }_{-}$ pulses. (d) Detection-time distributions for photons conditioned on a detection during the previous pulse. The probability of detecting a ${\sigma }^{+}$ photon after a ${\sigma }^{-}$ photon, is much larger than the probability of the opposite case. (e) Measurement of the intensity correlation between the two detectors (both polarizations are detected). The missing peak at $\tau =0$ results from the single-photon nature of the source. Data are binned in 150 ns intervals.

Figure 3

Two-photon interference. Inset: coincidences as a function of the time difference between detections for photons with parallel and perpendicular polarizations. For perfect spatial mode matching, the dip around $\tau =0$ for parallel polarizations would reach zero coincidences. The observed dip is consistent with an interferometer visibility of 98%. $({\Omega }_0/2\pi ,t_p)=(12\mathrm{MHz},0.72\mu \mathrm{s})$. An integrated visibility $V=77\%$ is observed. Main figure: visibility and photon-generation efficiency as a function of pump laser peak Rabi frequency ${\Omega }_0$ and various pulse durations $t_p$. The open symbols refer to the integrated visibility $V$ of the two-photon interference, with different shapes for different pulse durations. $V$ increases with lower ${\Omega }_0$ and shorter $t_p$. The filled symbols show the conditional probabilities $p({\sigma }^{+}|{\sigma }^{-})$ and $p({\sigma }^{-}|{\sigma }^{+})$ for generating photons with a pulse duration $t_p=1.42\mu \mathrm{s}$.