High-quality single photons from cavity-enhanced biexciton-to-exciton transition

Abstract: Resonant laser excitation of a two-level system with subsequent single-photon emission can be used to generate single photons with high indistinguishability or Hong-Ou-Mandel (HOM) visibility. However, spectral overlap between excitation laser and emitted photons generally poses significant challenges. Furthermore, emitter re-excitation intrinsically limits achievable single-photon purity. Established solutions mitigate these issues at significant cost to source efficiency and with increased source complexity. This motivates the use of few-level systems with spectral separation of excitation and emission pathways. One option is a three-level cascade. However, without targeted lifetime engineering of emitting states, the cascade naturally limits achievable photon indistinguishability. Here we study a semiconductor quantum dot with resonant and selective cavity-enhancement of biexciton-to-exciton transition. Following resonant two-photon excitation of the biexciton state, we collect the emitted single photon with the cavity. This approach circumvents emitter re-excitation and naturally introduces spectral separation of excitation laser and emitted single photon. Supported by first experimental results, we demonstrate theoretically that with selective Purcell enhancement, the observed quality quantifiers of single-photon emission (purity, equivalently $g<sup>{(2)}(0)$, and HOM visibility $\mathcal{V}$, equivalently indistinguishability) are competitive with respect to high-quality deterministic quantum-dot single-photon sources. This is already achieved without systematic optimization or targeted system engineering, which firmly places the reported approach as a viable route to the next generation of highest-quality quantum-dot based deterministic single-photon sources.

• The paper demonstrates a method of generating high-quality single photons by cavity-enhancing the biexciton-to-exciton transition in semiconductor quantum dots (QD), achieving a g^(2)(0) of 0.0041 and a Hong-Ou-Mandel (HOM) visibility of 96.07%.
• Utilizing orthogonal polarized cavity modes, the method addresses issues of impurity re-excitation and laser-photon spectral overlap, leading to better excitation separation with effectively eliminating the intrinsic limitations on indistinguishability.
• Crucially, introducing engineering the lifetime ratio, the paper provides an alternative architecture for high-quality single-photon sources by selectively enhancing the XX-X transition, thus enabling indistinguishability values in a practical regime.

Motivation and scheme

Deterministic single-photon sources based on semiconductor quantum dots (QDs) face two persistent obstacles: emitter re-excitation under resonant driving, which degrades single-photon purity $g^{(2)}(0)$, and the spectral overlap of excitation laser and emitted photon, which complicates extraction. The standard workaround — initiating emission from a biexciton (XX) cascade via resonant two-photon excitation (TPE) — solves both problems but introduces a new one: the intrinsic XX-to-X lifetime ratio $\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$ imposes a fundamental ceiling on the achievable Hong-Ou-Mandel (HOM) visibility, and hence on photon indistinguishability (2602.18153).

The paper proposes to invert the usual strategy: rather than collecting the exciton-ground (X-G) photon while Purcell-enhancing the XX-X transition with an external stimulus, it collects directly the photon emitted from the selectively cavity-enhanced XX-X transition itself. A spectrally narrow H-polarized cavity mode is tuned to the XX-X transition, while the V-polarized mode is detuned by $\hbar\delta = 206.8~\mu$eV to suppress competing emission; TPE initializes the biexciton through the orthogonal V channel. This configuration simultaneously eliminates re-excitation, provides natural laser-photon spectral separation, and — crucially — replaces the natural XX lifetime with the inverse Purcell rate $4\tilde{g}^2/\kappa$, thereby engineering the lifetime ratio into the regime where high indistinguishability is possible.

Theoretical framework

The system is modeled as a four-level QD (ground state G, excitons X$_{\mathrm{H,V}}$, biexciton XX) coupled to two orthogonally polarized cavity modes, treated within the polaron master equation formalism including phonon-mediated QD-cavity interactions, phonon-induced pure dephasing, and radiative decay into non-cavity modes ($$\hbar\gamma_{\mathrm{rad}}^{\mathrm{XX}} = 1.38~\mu$$eV, $\hbar\gamma_{\mathrm{rad}}^{\mathrm{X}} = 1.5~\mu$eV). Parameters are chosen to match an experimental InGaAs QD in an open tunable microcavity: fine structure splitting 10.8 $\mu$eV, binding energy 2.9 meV, $\hbar g = 20.8~\mu$eV, $\hbar\kappa \approx 103.4~\mu$eV. Correlation functions are evaluated via the quantum regression theorem.

A central analytical result extends the Wigner-Weisskopf treatment of cascaded emission [Huang and Eberly] to Purcell-enhanced transitions: in the Purcell regime ($\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$0), the lifetime-ratio limit on HOM visibility,

$\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$1

remains valid provided each transition's decay rate is taken as the sum of its Lorentzian-masked Purcell rate and radiative rate. This justifies why cavity-enhancing the XX-X transition raises the attainable visibility, and generalizes to the experimental situation of a single tunable cavity coupled to both transitions, since spectrally separated transitions sharing one Lorentzian bath can still be treated independently. Numerical simulations agree with this analytic approximation to about 1% within its weak-coupling validity range ($\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$2).

Parameter dependence and performance

The quality quantifiers — emission probability, purity, and indistinguishability — were mapped as functions of XX binding energy, QD-cavity coupling, and cavity loss. For sufficiently large absolute binding energy (exceeding the cavity linewidth), the cavity emits essentially only XX photons: purity approaches unity and indistinguishability approaches the lifetime-ratio limit. Near zero binding energy, the X-G transition becomes resonant with the cavity, producing a purity dip from contaminating X photons and a corresponding emission maximum. Phonon-mediated cavity feeding introduces an asymmetry favoring positive binding energies (phonon-emission processes), while pure dephasing at 4.2 K uniformly lowers indistinguishability. Notably, the finite cavity mode splitting reduces brightness but not the single-photon character of the detected H-mode output.

Regarding coupling and loss: increasing $\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$3 or decreasing $\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$4 boosts emission via the Purcell rate, but large $\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$5 also enhances phonon-assisted feeding (scaling as $\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$6), degrading purity; indistinguishability exhibits an optimum in both parameters. Cavity loss has comparatively minor influence on purity (~1% variation over the studied range).

Benchmarking against state-of-the-art sources, for a binding energy of 5 meV at $\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$7 K with moderate spectral filtering (bandwidth $\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$8), the scheme yields $\tau_{\mathrm{XX}}/\tau_{\mathrm{X}}$9 and $\hbar\delta = 206.8~\mu$0 without any optimization of $\hbar\delta = 206.8~\mu$1 or $\hbar\delta = 206.8~\mu$2. The accompanying experiment reports, on an unoptimized device and with no post-processing, $\hbar\delta = 206.8~\mu$3 and $\hbar\delta = 206.8~\mu$4, in excellent agreement with theory at $\hbar\delta = 206.8~\mu$5 K — consistent with the measured near-transform-limited linewidths of that source. These figures are competitive with deterministic resonantly excited QD sources while avoiding their re-excitation and filtering penalties.

Robustness under two-photon excitation

Simulations with realistic TPE pulses (Gaussian, durations 3–15 ps, pulse areas up to $\hbar\delta = 206.8~\mu$6) show that single-photon purity and indistinguishability are essentially unaffected by the excitation process when performed in the perpendicular polarization channel; emission drops only 5–10% relative to ideal initialization due to imperfect XX preparation in the presence of losses. The authors note this deficit is not fundamental and could be removed using schemes such as SUPER or rapid adiabatic passage. The absence of photon-number coherence in the H channel was verified numerically.

Entangled photon pairs

Extending the scheme to entangled pair generation requires high indistinguishability for both cascade photons — which the lifetime-ratio analysis supports (phonon-free indistinguishabilities near 98%) — plus high concurrence. Here the current geometry fails: the deliberate frequency detuning of the V-cavity mode imprints which-path information, yielding $\hbar\delta = 206.8~\mu$7. With zero cavity mode splitting and zero fine structure splitting, perfect entanglement ($\hbar\delta = 206.8~\mu$8) is recovered, indicating feasibility with modest design changes, though TPE itself is known to undermine entanglement and may require alternative excitation protocols.

Limitations and open questions

Several constraints qualify the results. First, the lifetime-ratio limit on indistinguishability is fundamental: neither spectral nor temporal filtering can circumvent it, nor can they remove pure dephasing; further gains require reducing $\hbar\delta = 206.8~\mu$9 and, more critically, suppressing residual X decay. Second, the reported best theoretical figures assume $4\tilde{g}^2/\kappa$0 K behavior justified by the exceptionally low noise of the specific experimental device; other realizations with charge noise would fare closer to the 4.2 K curves including dephasing. Third, phonon parameters derive from a spherical isotropic approximation of a pancake-shaped dot, and the included phenomenological dephasing deliberately overestimates effects for the present experiment. Fourth, the finite cavity mode splitting limits brightness, and larger splittings await suitably designed cavities. Finally, the entangled-pair variant remains untested experimentally, and the interplay of TPE with entanglement in the modified (zero-splitting) geometry is unresolved.

Conclusion

This work establishes selective Purcell enhancement of the biexciton-to-exciton transition, combined with orthogonal-channel TPE, as a viable architecture for deterministic single-photon generation that inherently suppresses re-excitation and laser contamination. Theory and first experimental measurements agree closely, with demonstrated $4\tilde{g}^2/\kappa$1 and HOM visibility competitive with leading resonant-excitation sources despite no device optimization. Remaining headroom lies in lower cavity loss, suppressed exciton decay, larger biexciton binding energies (potentially via GaAs dots), temporal filtering, and extension to entangled-pair operation.