Quantum-dot single photon source performance with off-resonant pulse preparation schemes

TL;DR

This work analyzes three off-resonant driving schemes for a quantum-dot single-photon source in a waveguide QED setting, incorporating LA phonon coupling to assess key metrics $η$, $g^{(2)}(0)$, and $\mathcal{I}$. Using a TLS model for the QD, a weak phonon coupling framework, and quantum-trajectory simulations, the authors compare dichromatic, NARP, and SUPER pulses under experimentally relevant conditions. They find that NARP and SUPER deliver near-ideal photons with high coherence and indistinguishability, while the dichromatic approach is hampered by phonon-induced dephasing, particularly for longer drives; the NARP scheme also proves robust to parameter variations, whereas SUPER is highly sensitive to laser parameters. The results support using off-resonant, polarization-filter-free schemes to realize efficient, on-demand SPS implementations, with potential for optimization and inverse-design of pulses to maximize robustness and performance.

Abstract

The preparation of photonic qubits in the excited state is an integral part of the performance of an on-demand single photon source (SPS). Conventional resonant excitation, an excellent approach to maximize the coherence and indistinguishability of the SPS, often requires polarization filtering to remove the pump signal and isolate the qubit emission, but this results in an inherent 50\% hit to the efficiency. Recent excitation schemes strategically try to exploit pulses that excite the qubit while avoiding spectral overlap to bypass this required filtering. In this work, we compare three such pumping schemes to quantify the important SPS figures-of-merit for off-resonant quantum dot schemes, using: (i) a symmetrically detuned dichromatic pulse, (ii) a notch-filtered adiabatic rapid passage (NARP) pulse, and (iii) a swing up of the quantum emitter population (SUPER) pulse. Due to large instantaneous pulse strengths, the dichromatic pulse suffers from phonon-induced dephasing which can lower the SPS performance by up to 50\%. In contrast, the NARP and SUPER pulses are shielded from phonon coupling to differing degrees but both maintain excellent SPS performance. The SUPER pulse can lose significant efficiency if there is variance in its constituent pulses' amplitude, pulse width, or frequency, while the NARP pulse, though potentially more difficult to realize in experiments, is robust against variance in the pulse preparation.

Figures (7)

• Figure 1: Schematic of the single photon emission process. The qubit (with LA acoustic phonons and coupled to the waveguide with rate $\gamma$) is driven by one of the three pulse schemes and (ideally) emits a single photon in the waveguide.
• Figure 2: The laser profiles and resulting phonon rates for the four driving schemes in each column, left to right: long dichromatic pulse, short dichromatic pulse, NARP, and SUPER. By row, (a)-(d) the time domain profile of the pulse, (e)-(h) the spectral profile (in blue) with the qubit emission spectrum (in red), (i)-(l) the phonon-induced dephasing rate, and (m)-(p) the phonon-induced excitation rate.
• Figure 3: The population dynamics ($N_x(t) = \braket{\sigma^+ \sigma^- (t)}$) without dissipation in the first row, and with waveguide output and phonon coupling in the second row. Each column is a particular driving scheme: (a)-(b) long dichromatic pulse, (c)-(d) short dichromatic pulse, (e)-(f) NARP, and (g)-(h) SUPER.
• Figure 4: (a)-(c) SPS performance (through the (a) efficiency, (b) coherence, and (c) indistinguishability) as a function of the spectral separation of the two pulses for the long dichromatic pulse. The overlap function for each separation is in (d). Note that the overlap results in (d) are only dependent on the pulse form.
• Figure 5: (a)-(c) SPS performance (through the (a) efficiency, (b) coherence, and (c) indistinguishability) as a function of the spectral separation of the two pulses for the short dichromatic pulse. The overlap function for each separation is in (d). Each data point in (a)-(c) is an average of 50,000 stochastic trajectories with very low event rates which leads to the noise in the data.