Franson Interference

• Franson interference is a nonlocal two-photon phenomenon where energy–time entangled photons exhibit interference only in coincidence measurements from spatially separated unbalanced interferometers.
• It relies on a regime where single-photon coherence is much shorter than the interferometer delay, ensuring that only joint detection events (via post-selection) reveal phase-dependent interference.
• The technique underpins applications in quantum communication, high-dimensional QKD, and quantum imaging, with implementations spanning integrated photonics to long-distance fiber networks.

Franson interference is a nonlocal two-photon interference effect, and, in another standard formulation, a fourth-order interference effect, in which interference appears in joint coincidence measurements rather than in single-photon intensities. It is the canonical interferometric signature of energy–time entanglement: two photons are distributed to spatially separated unbalanced interferometers, the short–short and long–long two-photon alternatives remain indistinguishable, and the coincidence rate acquires a phase dependence of the form $P_c \propto 1 + V \cos(\phi_1+\phi_2)$ (Jogenfors et al., 2017, Gao et al., 2019, Xiang et al., 25 Feb 2026).

1. Canonical geometry and physical mechanism

The canonical Franson configuration uses a source of correlated photon pairs, typically energy–time entangled, and two spatially separated unbalanced interferometers, one per photon. Each interferometer has a short arm $S$, a long arm $L$, and a controllable phase. The defining regime is

$\tau_c \ll \Delta T \ll \tau_p,$

where $\tau_c$ is the coherence time of each individual photon, $\Delta T$ is the interferometer delay between short and long arms, and $\tau_p$ is the pump coherence time or, more generally, the coherence time of the emission process (Jogenfors et al., 2017, Xiang et al., 25 Feb 2026).

These inequalities encode the basic logic of the effect. When $\Delta T$ is much larger than the single-photon coherence time, each photon alone cannot interfere between its short and long paths, so the singles remain phase independent. When $\Delta T$ is still much smaller than the pump coherence time, the alternatives in which both photons are emitted “early” and both take long paths, and both are emitted “late” and both take short paths, remain coherent. The coincidence histogram then contains three peaks: side peaks from $(S,L)$ and $S$0, and a central peak in which $S$1 and $S$2 overlap and interfere (Jogenfors et al., 2017, Xiang et al., 25 Feb 2026).

The coincidence probability in that central peak has the standard sinusoidal form

$S$3

with $S$4 the two-photon fringe visibility. This phase dependence, together with the absence of single-photon interference, is the operational hallmark of Franson interference (Jogenfors et al., 2017).

A standard contrast is Hong–Ou–Mandel interference. In HOM interference, two photons must meet at the same beam splitter at the same time; in Franson interference, the photons never meet at a common optical element. The interference is therefore nonlocal in the sense that it resides in correlations between distant measurements, not in local intensities. This nonlocality does not imply superluminal signaling, because the phase dependence appears only in coincidence statistics, not in either party’s marginal counts (Gao et al., 2019).

2. Visibility, post-selection, and Bell-test subtleties

Franson visibility is usually defined from the maximum and minimum coincidence counts as

$S$5

In idealized treatments, a visibility above $S$6 is associated with violation of a CHSH inequality for sinusoidal correlations. However, the ordinary Franson geometry includes intrinsic post-selection: the $S$7 and $S$8 events are discarded, and only the overlapping $S$9 and $L$0 events are retained (Jogenfors et al., 2017).

That post-selection is not a minor technicality. In the standard Franson interferometer, even with perfect optics and perfect detectors, the effective detection efficiency is at best $L$1, because roughly half of the events are rejected by construction. For this reason, the familiar 70.7% threshold is not the relevant criterion for a loophole-free Bell test in the ordinary Franson geometry. The 2017 Comment on “Franson Interference Generated by a Two-Level System” states that the 70.7% bound only applies under the assumption $L$2, and that this assumption is impossible in a Franson interferometer because of the inherent 50% post-selection (Jogenfors et al., 2017).

The same Comment also rejects the use of a universal “50% classical limit” for Franson visibility. In Franson-type setups with strong post-selection, tailored classical light pulses and local-hidden-variable models can reproduce visibilities above 50%, even approaching unity, in the post-selected data. High visibility by itself is therefore not evidence against all classical descriptions in the usual Franson geometry (Jogenfors et al., 2017).

The corresponding foundational point extends into quantum cryptography. In high-dimensional energy-time-entanglement QKD, a single pair of Franson interferometers is vulnerable to attacks that localize photons to several temporally separated locations while preserving the specific coherences tested by one fixed delay. Multiple Franson interferometers with different delays improve sensitivity to such attacks, but this remains short of a full security proof (Brougham et al., 2013).

The standard remedy for genuine Bell violation in a post-selected Franson or time-bin setting is not ordinary CHSH but a modified test. The Comment points to a three-setting chained Bell inequality, introduced by Pearle, as a viable route, with a required visibility of at least $L$3 for a genuine violation of local realism in that scenario (Jogenfors et al., 2017).

3. Variants and generalizations

Franson interference has generated a substantial family of generalizations. A polarization-based Franson-type interferometer can reproduce HOM-type two-photon peak and dip fringes even though the photons do not meet at a common beam splitter; the relevant interference is again between indistinguishable two-photon amplitudes leading to a coincidence event (Kim et al., 2017). Closely related polarization-Sagnac implementations have been used to demonstrate nonlocal quantum erasure and correction of phase objects in coincidence imaging, where inserting matching phase objects into distant interferometers restores interference that had been destroyed in one arm alone (Gao et al., 2019).

A complementary construction is conjugate-Franson interferometry, which probes arrival-time correlations rather than the frequency-domain correlations emphasized by standard Franson interferometry. In that scheme, balanced Mach–Zehnder interferometers, frequency shifts, and opposite dispersions produce a visibility

$L$4

so the interference depends on the joint temporal intensity and becomes sensitive to spectral phase, unlike standard Franson or Hong–Ou–Mandel interferometry. Experimentally, a conjugate-Franson visibility of $L$5 without background subtraction was reported, surpassing the quantum-classical threshold by 25 standard deviations (Chen et al., 2021).

Franson interference has also been resolved directly in the spectral domain. In spectrally resolved Franson interference, the joint spectral intensity after the interferometer takes the form

$L$6

which modulates the biphoton spectrum along both the signal and idler frequency axes. This spectral viewpoint connects Franson interferometry to high-dimensional frequency entanglement and time-frequency grid states (Jin et al., 2023).

Multipartite extensions are possible. A generalized three-photon Franson interferometer with energy-time-entangled photon triplets yields a third-order coincidence probability

$L$7

The reported three-photon interference visibility was $L$8 in the abstract, and the experiment was specifically designed so that no single-photon or two-photon interference accounted for the effect (Agne et al., 2016).

A useful boundary case is thermal light. Two classically correlated beams of thermal light sent through two unbalanced Mach–Zehnder interferometers can exhibit genuine second-order interference with visibility $L$9, and that interference depends on the difference of the UMZI phases rather than their sum. The paper presenting this result emphasizes that it differs substantially from entangled-photon Franson interferometry, whose interference depends on the sum of the phases and vanishes as the path imbalance exceeds the pump coherence length (Ihn et al., 2017).

4. Sources, platforms, and implementation regimes

The traditional Franson source is cw-pumped SPDC, though SFWM is also standard in modern fiber and integrated platforms (Xiang et al., 25 Feb 2026). The basic requirement is a long-coherence-time emission process combined with individual photons whose own coherence time is much shorter than the interferometer imbalance (Jogenfors et al., 2017).

Semiconductor quantum emitters constitute an important alternative. A two-level semiconductor quantum dot source used in a Franson-type interferometer produced a visibility of 66%, which motivated the 2017 Comment about the interpretation of “classical limits” and Bell thresholds (Jogenfors et al., 2017). A resonantly driven biexciton cascade in a quantum-dot three-level system later achieved a maximum visibility of $\tau_c \ll \Delta T \ll \tau_p,$0, surpassing the nominal $\tau_c \ll \Delta T \ll \tau_p,$1 Bell-violation threshold by more than one standard deviation, while explicitly noting that the setup could not satisfy a loophole-free violation because of post-selection (Hohn et al., 2023).

Telecom and fiber-network implementations have progressed toward deployment-oriented architectures. Over 50 km of single-mode fiber, nonlocal Franson interferometry with a co-propagating Radio-over-Fiber clock signal achieved a raw visibility of $\tau_c \ll \Delta T \ll \tau_p,$2 without dedicated timing infrastructure, and the same platform yielded $\tau_c \ll \Delta T \ll \tau_p,$3 under a common clock reference. The work emphasizes passive synchronization with picosecond precision, cross-band O-band/L-band allocation for Raman-noise suppression, and metropolitan-scale quantum-network relevance (Xiang et al., 25 Feb 2026).

Integrated photonics has produced very high raw performance. A cascaded PPLN waveguide source combined with fully passive, path-imbalanced Mach–Zehnder interferometers on photonic integrated circuits achieved a two-photon interference visibility of 97.1% from sinusoidal fringe fitting, with raw visibility 95.2%, background-corrected visibility 95.6%, heralding efficiency 4.8%, and a coincidence-to-accidental ratio exceeding 1000 at only 1.7 mW of pump power (Emadi et al., 27 Mar 2026). In that implementation, the analyzers required neither on-chip phase shifters nor active stabilization; phase was scanned by thermal tuning of the chip (Emadi et al., 27 Mar 2026).

Franson interferometry is also used as a diagnostic for hybrid frequency-entangled qudits. In that context, it confirms global frequency correlations with visibility exceeding 98% and verifies continuous-variable entanglement within individual frequency modes with visibility greater than 95% (Wang et al., 1 Mar 2025). At the source-engineering level, coherent time-delayed feedback has been proposed as a way to boost Franson visibility in photon pairs emitted by a three-level ladder system by slowing the decay of the upper level (Barkemeyer et al., 2021).

5. Applications and cross-disciplinary roles

In quantum communication, Franson interferometry is a standard method for diagnosing energy-time entanglement quality and for operating time-bin and energy-time protocols over fiber. Its robustness against polarization disturbances is one reason it is central to fiber-based quantum networks (Xiang et al., 25 Feb 2026). In high-dimensional time-frequency settings, spectrally resolved Franson interference and hybrid frequency-entangled qudits connect the technique to high-dimensional frequency entanglement, time-frequency grid states, and related continuous-variable encodings (Jin et al., 2023, Wang et al., 1 Mar 2025).

In QKD, Franson interferometers have been proposed for security monitoring in high-dimensional energy-time protocols, but a single fixed-delay pair is not secure against multi-peaked time-localizing attacks. Multiple Franson interferometers with different delays can improve sensitivity to those attacks, although the cited work explicitly states that it does not constitute a full security proof (Brougham et al., 2013).

Franson-based nonlocal interference has also been used in quantum imaging. A modified Franson interferometer enabled nonlocal quantum erasure and correction of phase objects, with proposed applications in phase corrections in quantum imaging and microscopy and in user authentication of two foreign distant parties (Gao et al., 2019).

Beyond photonic communication, the Franson concept has an electronic analogue. A Franson-type setup for Cooper pairs emitted from a superconductor into two one-dimensional channels was analyzed as a way to probe both the internal coherence of emitted Cooper pairs, proportional to Pippard’s length, and the de Broglie wavelength of their center-of-mass motion via current-current correlation measurements (Giovannetti et al., 2012).

6. Foundational debates and current outlook

Two issues dominate the contemporary interpretation of Franson interference. The first is the post-selection loophole: observed Franson fringes, even at high visibility, do not by themselves certify nonlocality in the ordinary geometry. The standard CHSH threshold of $\tau_c \ll \Delta T \ll \tau_p,$4 is not the right benchmark under intrinsic 50% post-selection, and tailored classical pulses or explicit local-hidden-variable models can reproduce the post-selected statistics of ordinary Franson experiments (Jogenfors et al., 2017).

The second is broader interpretive scope. One line of analysis argues that, under coincidence-provided superposition between independent Mach–Zehnder interferometers, Franson-type correlation fringes can be created from non-entangled photons through the interferometers themselves, emphasizing the role of wave coherence and indistinguishability rather than entanglement alone (Ham, 2020). A plausible implication is that experimental claims based solely on fringe observation require unusually careful scrutiny of source assumptions, post-selection rules, and loopholes.

The technical outlook is correspondingly twofold. On the foundational side, genuine Bell tests in energy-time variables require modified inequalities or geometries, together with very high visibility, careful treatment of efficiency, and control of post-selection (Jogenfors et al., 2017). On the applied side, passive synchronization over 50 km fiber, cross-band coexistence of classical timing and quantum channels, and compact passive PIC analyzers with raw visibilities above 95% indicate that Franson interferometry is becoming a practical network component rather than only a laboratory testbed (Xiang et al., 25 Feb 2026, Emadi et al., 27 Mar 2026).

Franson interference therefore occupies a dual position. It remains a paradigmatic probe of energy–time entanglement and nonlocal correlation, while also serving as a technically mature framework for fiber distribution, integrated photonics, spectral–temporal state analysis, and high-dimensional quantum communication. Its interpretation, however, depends critically on which question is being asked: diagnosis of joint coherence, certification of entanglement, or loophole-resistant nonlocality.