Researchers have traced a long-standing interference loss in quantum-dot photon pairs to a correctable timing mismatch. In their cleaner of two emitters, selecting late photons with matched arrival times raised Hong-Ou-Mandel interference visibility to 69% from a standard time-averaged value of 54.4%.
Hong-Ou-Mandel interference is a test of whether two photons are indistinguishable. When matching photons enter opposite sides of a beam splitter at the same time, they leave together instead of taking separate outputs. That behavior supports operations such as entanglement swapping, in which photons from different pairs connect distant quantum-network links.
Semiconductor quantum dots can emit an entangled pair through a three-level cascade. An early biexciton photon is followed by a late exciton photon, but the intermediate state has a finite lifetime. The emission time of the early photon therefore predicts when the late photon will arrive. Averaging over that correlation makes late photons from consecutive excitations appear less indistinguishable, even when their underlying wave packets can overlap well.
Four detections expose the hidden timing variable
The team measured four-photon coincidences from two consecutive emission cycles: both early photons served as timing markers while the two late photons met at an interference beam splitter. Comparing parallel polarization, where interference is possible, with orthogonal polarization provided the visibility reference. This resolved visibility as a function of the early-photon time difference instead of collapsing every event into one average.
With no added offset, maximum overlap occurred when the early-photon detection times matched. The researchers then inserted a 300-picosecond delay in one late-photon path. The best overlap shifted to events whose early-photon times differed by minus 300 picoseconds, as their model predicted. The fixed delay and timestamp-conditioned analysis thus synchronized late photons that had left their emitters with an initial temporal mismatch.
Two quantum dots tested the mechanism under different optical conditions. A telecom C-band InAs dot showed maximum time-resolved visibilities of 6.6% without the added delay and 7.6% with it, above respective time-averaged values of 4.8% and 4.9%. Spectral broadening strongly limited that device, so the timing recovery did not produce high absolute visibility.
A droplet-etched GaAs dot emitting near 780 nanometers provided the clearer test. Its visibility reached 69%, with a reported uncertainty of 2 percentage points, when the early-photon times matched. The standard averaged measurement gave 54.4% with a 1.5-point uncertainty. The synchronized value also exceeded the 64.4% lifetime-based limit calculated for the unconditioned cascade and approached a separate 73.4% measurement made under phonon-assisted excitation.
From selected photon pairs to on-demand timing
The result does not mean the experiment delivered every photon on demand with corrected timing. It used an optical delay, time-resolved four-photon coincidences and conditioning on detection records to reveal which late photons overlapped. The lower-signal GaAs data also used an orthogonal-polarization reference for distinguishable photons because a reliable fit could not be obtained.
The practical proposal is to place deterministic quantum memories after emission. A memory that already synchronizes remote network nodes could hold each late photon for a duration inferred from its partner, aligning their arrivals before interference. The paper demonstrates the timing relation and its recoverability, while leaving memory efficiency, storage fidelity, feed-forward latency and usable event rate for a combined source-and-memory experiment.
Sources
- Paper record and abstract, submitted September 21, 2026
- Full paper with theory, apparatus, measurements and limitations
- Lead image: Figure 1 from the paper, showing the cascade, excitation timing and predicted visibility, published under the paper's CC BY 4.0 license.