Overcoming the Hong-Ou-Mandel interference limitation for the second photon from a two-photon cascade

AI summary

Overview

• Research area: Quantum photonics – solid-state quantum emitters (semiconductor quantum dots), entangled photon-pair sources, and two-photon (Hong-Ou-Mandel) interference for quantum networks.
• Technical level: Intermediate (accessible to graduate students with some background in quantum optics; the theoretical treatment is advanced).
• Scope (one sentence): The paper theoretically derives and experimentally confirms that the reduced Hong-Ou-Mandel visibility of the second (late, exciton) photon from a biexciton–exciton cascade is caused only by a temporal mismatch at the interference beam splitter, and that the full visibility can be recovered by post-emission synchronization.

What This Paper Is About

Quantum dots can emit polarization-entangled photon pairs through the cascaded biexciton–exciton (XX–X) decay, but the two photons are emitted at correlated, jittery times because the intermediate state has a finite lifetime. This timing jitter has been thought to fundamentally cap how well the late photons can interfere with each other, which is a problem if a quantum network wants to use entanglement and indistinguishability from the same source. The authors show that, for the late photon, there is no fundamental limit: the visibility loss is purely a matter of the two photons arriving at the beam splitter at slightly different times, and it disappears once that temporal offset is corrected.

Key Contributions

1. A theoretical model of Hong-Ou-Mandel (HOM) interference for two late photons emitted in two consecutive cascade events, expressed through a four-photon probability density function p⁽⁴⁾(τ_E, τ_L) that explicitly includes the early–late temporal correlation and an adjustable input delay δt₀.
2. A derivation showing that the standard (averaged) late-photon HOM visibility is simply the time-resolved visibility V_HOM,L(τ_E, δt₀) weighted by the exponential decay of the early photon; for δt₀ = 0 and zero spectral broadening this reduces to the known result V = 1/(1 + T_E/T_L).
3. The first measurement of time-resolved four-photon coincidences between two successively emitted photon pairs after HOM interference of the late photons, using a conditioning scheme that reduces a three-dimensional G⁽⁴⁾ measurement to an intuitive two-dimensional histogram.
4. Experimental demonstration, on two different quantum-dot sources, that the maximum time-resolved visibility exceeds the standard HOM visibility, and that the maximum is reached at τ_E = −δt₀, i.e. exactly under the condition of restored temporal overlap.

Main Findings

• The limitation is temporal overlap, not decoherence: For the late photons the reduced visibility follows V_HOM,L(τ_E, δt₀) = V_spec(σ, T_L) · exp(−|τ_E + δt₀|/T_L). The exponential factor is a pure timing mismatch; the only remaining visibility ceiling is the spectral broadening factor V_spec, which is independent of the cascade.
• Synchronization recovers the maximum: The whole visibility curve is horizontally shifted by δt₀, so unity visibility (for Fourier-limited photons) is reached at τ_E = −δt₀. A deliberate input delay of δt₀ = 300 ps moves the best-overlap point to τ_E = −300 ps, as seen both in the two-dimensional histograms and in the extracted visibility curves.
• QD1 (telecom C-band InAs QD): Time-resolved visibilities of V_HOM,L(0, 0) = 6.6 (±0.5)% and V_HOM,L(300 ps, −300 ps) = 7.6 (±0.7)% exceed the standard averaged HOM visibilities of 4.8 (±0.3)% and 4.9 (±0.3)%. The overall low values are attributed to strong spectral broadening (σ = 21.8 (±0.2) GHz for δt₀ = 0 and σ = 18.8 (±0.2) GHz for δt₀ = 300 ps, the small difference likely from a change in the charge environment during the measurement).
• Theoretical–experimental agreement: The measured p⁽⁴⁾ data match the model after convolution with a 75 ps bin width in τ_E and a 50 ps two-channel detector resolution.
• QD2 (droplet-etched GaAs QD at ~780 nm, p-i-n diode): V_HOM,L(0, 0) = 69 (±2)% clearly exceeds the measured standard value of 54.4 (±1.5)% and even surpasses the Fourier-limited threshold of 1/(1 + 152 ps/275 ps) = 64.4% set by the XX and X lifetimes.
• Remaining gap is spectral, not cascaded: A separate HOM measurement on QD2 under longitudinal-acoustic-phonon-assisted excitation of the X state gives 73.4 (±0.3)%, close to V_HOM,L(0, 0), indicating that the residual limitation is line broadening and no additional cascade-induced effect.
• Wider relevance: Without detecting the early photon, the cascade is indistinguishable from incoherent excitation of the intermediate state, so other excitation schemes limited by temporal jitter (e.g. relaxation from higher electronic or vibrational states) should show the same time-resolved visibility fingerprint.

Methodology in Plain English

The team used one decay channel of a quantum dot's XX–X cascade (selected by polarization filtering) as an effective three-level ladder system, excited resonantly by a pulsed laser through two-photon excitation. The early and late photons were separated with volume Bragg grating filters and projected onto H-polarizers. The early photons went to a Hanbury Brown and Twiss interferometer, while late photons from consecutive laser pulses met at an unbalanced Mach-Zehnder interferometer for HOM interference, with wave plates in one arm to switch between parallel and orthogonal polarization. Superconducting nanowire single-photon detectors and the laser reference clock fed a TimeTagger.

The key trick is conditioning: coincidence events were only counted when one early–late pair was detected within the same repetition period. That turns the usual three-dimensional four-photon correlation into a two-dimensional histogram whose (0,0) point means all four photons arrived simultaneously. Cuts through this histogram at fixed early-photon delay τ_E give the interference dip for a given amount of temporal mismatch, and the visibility is obtained from the counts in each cut integrated over a ±3292 ps window, with the orthogonal-polarization model as the distinguishable-photon reference. The experiments were repeated with a second, different quantum dot and with an added 300 ps input delay to test whether shifting the arrival times restores overlap.

Why This Matters

The result reframes a widely cited limitation: the late photon of a cascade is not intrinsically less indistinguishable than an isolated photon, which means quantum-dot sources can in principle deliver entanglement and indistinguishability simultaneously rather than trading one for the other.

• Entanglement swapping between remote nodes – efficient Bell-state measurement between two entangled photon pairs is a core requirement of quantum repeater architectures.
• Quantum repeaters and large-scale networks – sources whose second photon interferes well remove a bottleneck in connecting distant nodes.
• Quantum memories as synchronizers – memories already needed in a network can absorb the cascade timing jitter at essentially no extra cost, avoiding complex imbalanced cavity structures.
• Broader emitter platforms – the same time-resolved visibility fingerprint applies to other excitation schemes that introduce timing jitter, giving a diagnostic tool for source engineering.

Industry relevance: telecom C-band emission (QD1) and GaAs-based devices (QD2) are both relevant to fiber-link and integrated-photonics deployment, and the result lowers the performance bar that entangled-light sources must meet for network hardware, including deterministic-memory-based synchronization schemes.

Future Directions

• Implement the post-emission synchronization experimentally using a real quantum memory rather than a fixed optical delay, and quantify the added loss and fidelity.
• Apply the time-resolved visibility measurement to emitters excited via relaxation from higher electronic or vibrational states, where no heralding early photon exists and the measurement is expected to be harder.
• Raise the spectral-overlap factor V_spec, since it is the only remaining ceiling on the late-photon visibility in both dots (σ ≈ 19–22 GHz for QD1, residual broadening for QD2) – for instance through improved charge-environment stability and narrower emission lines.
• Demonstrate full entanglement swapping using late photons from two cascaded sources under synchronized arrival, the application the paper identifies as the key next milestone.

Target Audience

Graduate students and researchers in quantum photonics and quantum information, particularly those working on semiconductor quantum-dot sources, photon indistinguishability, and Hong-Ou-Mandel interference. It is also valuable to quantum-network architects interested in how source-level timing jitter can be absorbed by synchronization infrastructure, and to source engineers who need a practical diagnostic for distinguishing spectral from temporal limitations. Beginners will follow the qualitative argument but will need background in cascade emission and HOM interference for the theoretical sections.

Authors’ abstract

Quantum emitters such as semiconductor quantum dots can provide entangled photon pairs emitting from a system of cascaded states. However, the resulting temporal correlation between the two photons from the three-level ladder system has been shown to set a limit on the achievable two-photon interference. In this work, we show that such limitation, for photons emitted by the second transition, is only due to reduced temporal overlap at the interference beam splitter. We investigate this theoretically and experimentally by determining time-resolved four-photon coincidences between two successively emitted photon pairs. Post-emission synchronization shows the recovery of maximum interference visibility; in prospective quantum networks this could be achieved via deterministic quantum memories.