Conjugate–Franson Interferometry (CFI)

• Conjugate–Franson Interferometry (CFI) is a two-photon interferometric technique that measures joint temporal intensity to certify time-energy entanglement.
• CFI uses complementary time-basis measurements to probe arrival-time correlations, making it sensitive to spectral phase variations unlike standard Franson or HOM interferometry.
• Experimental demonstrations of CFI show near-unity visibility and its integration in quantum networks, highlighting its potential in high-dimensional quantum key distribution and photonic communication.

Conjugate–Franson interferometry (CFI) is a nonlocal two-photon interferometric technique that is complementary to standard Franson interferometry. In its modern formulation, it measures photon-pair arrival-time correlations in the time basis, so that the interference visibility is determined by the biphoton’s joint temporal intensity (JTI); this makes CFI an alternative method for certifying time-energy entanglement and, unlike standard Franson interferometry or Hong–Ou–Mandel interferometry, renders it sensitive to spectral phase variation (Chen et al., 2021). In dual-basis time-energy protocols, CFI supplies the arrival-time-correlation measurement that complements Franson visibility data on frequency correlations (1311.0825). The literature also contains an earlier usage in which “CFI” denotes a Franson-like configuration obtained by inserting a Mach–Zehnder interferometer in a short-coherence continuous-wave pump beam, thereby restoring only part of the lost two-photon interference (Liang et al., 2010).

1. Conceptual scope and nomenclature

Standard Franson interferometry uses two spatially separated unbalanced Mach–Zehnder interferometers to probe photon-pair frequency correlations. By contrast, modern CFI is explicitly described as the complementary technique in the time basis: it probes photon-pair arrival-time correlations and yields a visibility that is a function of the biphoton’s JTI rather than its joint spectral intensity (JSI) (Chen et al., 2021). This complementarity is operational in high-dimensional quantum key distribution, where Franson and conjugate-Franson visibilities are used together to bound frequency-difference and arrival-time-difference variances (1311.0825).

A distinct, older usage appears in work on pump coherence, where a short-coherence continuous-wave pump is passed through an unbalanced Mach–Zehnder interferometer before spontaneous parametric down-conversion. In that setting, the pump interferometer induces coherence between certain two-photon amplitudes, producing an intermediate regime between long-coherence Franson experiments and pulsed time-bin experiments; the same paper characterizes this regime as CFI, but with a fundamentally different architecture and a theoretical visibility ceiling of $50\%$ because incoherent background amplitudes cannot be gated away (Liang et al., 2010).

Usage in the literature	Core architecture	Principal observable
Standard Franson interferometry	Two unbalanced Mach–Zehnder interferometers with path delay $\Delta T$	Frequency correlations
Modern CFI	Balanced interferometers with opposite frequency shifts and opposite-sign dispersion	Arrival-time correlations / JTI
Earlier CFI usage	Short-coherence CW pump plus pump Mach–Zehnder and matched PDC interferometer	Partial restoration of two-photon interference

This terminological split is consequential. In one lineage, CFI is a conjugate-basis analyzer for time-energy entanglement; in the other, it is an induced-coherence variant of Franson-type interferometry under short-coherence pumping. The two share nonlocal two-photon interference logic, but they should not be treated as identical experimental objects.

2. Measurement principle in the time basis

The central theoretical statement of modern CFI is that its coincidence probability is governed by the biphoton’s joint temporal intensity. For spontaneous parametric down-conversion, the normalized biphoton state in the time domain is written as

$$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$

With the signal and idler sent to two spatially separated interferometers, the CFI coincidence probability is

$$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$

where $\phi_T=\phi_S+\phi_I$ is the sum of the interferometer phases and $\Delta\Omega$ is the relative frequency shift introduced in the analyzer. The corresponding visibility is

$$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$

These expressions make explicit that CFI is a time-domain measurement of arrival-time correlations, and that its visibility depends on the JTI rather than solely on spectral intensity (Chen et al., 2021).

In the dual-basis security formalism for time-energy entanglement quantum key distribution, the same idea is written as

$$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$

A small-argument expansion yields an upper bound on the arrival-time-difference variance,

$$\langle (\Delta\hat{t}_S-\Delta\hat{t}_I)^2 \rangle \leq \frac{2[1 - V_{\rm CFI}]}{\Delta\Omega^2} + \frac{\langle(\widetilde{t}_S - \widetilde{t}_I)^4\rangle}{12} \Delta\Omega^2 ,$$

which is the time-domain counterpart of the Franson bound on frequency-difference variance (1311.0825).

This time-basis description also explains why CFI is sensitive to spectral phase variation. The JTI encodes information that is inaccessible to measurements whose visibilities depend only on the JSI. Consequently, two biphoton states with the same JSI but different spectral phases can be indistinguishable to standard Franson or Hong–Ou–Mandel interferometry and yet distinguishable to CFI (Chen et al., 2021).

3. Experimental demonstrations and performance regimes

The experimental demonstration identified explicitly as CFI used time-energy entangled photon pairs generated by spontaneous parametric down-conversion in a type-II phase-matched periodically-poled potassium titanyl phosphate waveguide pumped by a $780$ nm continuous-wave laser. The photon pairs were nondegenerate, with approximately $\Delta T$0 GHz center-frequency offset and $\Delta T$1 GHz FWHM bandwidth. Each analyzer arm included a frequency shifter realized by quadrature phase-shift keying modulators, with signal and idler shifts of $\Delta T$2 and $\Delta T$3, respectively, and $\Delta T$4 GHz. Fiber Bragg-grating dispersive elements with equal magnitude but opposite sign dispersions of $\Delta T$5 mapped frequency information onto arrival time, and superconducting nanowire single-photon detectors provided high-efficiency, low-jitter detection (Chen et al., 2021).

That experiment obtained a conjugate-Franson interference visibility of $\Delta T$6 without background subtraction. The measured result surpassed the quantum-classical threshold of $\Delta T$7 by $\Delta T$8 standard deviations and validated the conjugate-Franson interferometer as an alternative method for certifying time-energy entanglement. The same study then prepared two biphoton states with the same $\Delta T$9 GHz flat-top intensity spectrum but different spectral phases, applying a phase step $$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$0 to frequencies $$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$1 between $$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$2 and $$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$3 GHz. For $$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$4 and $$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$5, the CFI visibility was about $$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$6; for $$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$7, it dropped to $$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$8, with theoretical predictions changing from $$|\psi\rangle_{SI} \propto \int dt_-\,\psi_{SI}(t_-) |t_+ + t_-/2\rangle_S |t_+ - t_-/2\rangle_I .$$9 to $$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$0. The observed $$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$1 reduction directly exhibited spectral-phase sensitivity (Chen et al., 2021).

The earlier short-coherence continuous-wave implementation occupies a different performance regime. There, a type-I BBO crystal was pumped by a $$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$2 nm diode laser with short coherence length, and a pump Mach–Zehnder interferometer was used to induce coherence between otherwise incoherent two-photon amplitudes. When the pump-interferometer delay $$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$3 matched the PDC interferometer delay $$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$4, only two of four amplitudes were mutually coherent, while the remaining two formed an incoherent background. The coincidence rate took the form

$$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$5

so that the theoretical maximum visibility was $$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$6. Experimentally, the restored two-photon interference fringe had visibility of approximately $$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$7, close to that limit (Liang et al., 2010).

Taken together, these demonstrations show that “CFI” in the literature can refer either to a genuinely complementary time-basis analyzer with near-unity visibility or to an induced-coherence intermediate regime with an intrinsic $$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$8 ceiling. The difference is architectural and not merely terminological.

4. Relation to Franson interferometry, HOM interferometry, and time-bin methods

The most direct comparison is between standard Franson interferometry, Hong–Ou–Mandel interferometry, and modern CFI. Standard Franson interferometry probes frequency correlations, and its visibility is determined by the JSI. Hong–Ou–Mandel interferometry is likewise sensitive only to the JSI. CFI, by contrast, probes arrival-time correlations, and its visibility depends on the JTI; it is therefore sensitive to spectral phase variation in a way that Franson and HOM interferometry are not (Chen et al., 2021).

Technique	Visibility governed by	Sensitive to spectral phase?
Franson interferometry	JSI / frequency correlations	No
Hong–Ou–Mandel interferometry	JSI	No
CFI	JTI / arrival-time correlations	Yes

The older pump-interferometer usage of CFI is best understood through comparison with Franson and time-bin experiments. In canonical Franson interferometry, a long-coherence continuous-wave pump allows all relevant multi-path amplitudes to remain phase coherent, and visibility can reach $$P_{\rm CFI}(\phi_T) = \frac{\eta^2}{8}\left(1 + \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_- + \phi_T)\right),$$9. In pulsed time-bin experiments, the pump Mach–Zehnder interferometer creates two well-defined emission times, and the incoherent amplitudes can be discarded by timing gates, again allowing $\phi_T=\phi_S+\phi_I$0 visibility. In the short-coherence continuous-wave CFI regime, coherence is induced by the pump Mach–Zehnder interferometer, but the continuous-wave source provides no absolute clock, so the incoherent background cannot be removed; the maximum visibility is therefore capped at $\phi_T=\phi_S+\phi_I$1 (Liang et al., 2010).

A common misconception is that CFI is simply “Franson in another basis” with no added measurement content. The data do not support that simplification. Modern CFI is complementary to Franson interferometry, but its dependence on the JTI means that it responds to phase structure that Franson and HOM measurements can miss (Chen et al., 2021). Another misconception is that any Franson-type correlation fringe automatically certifies entanglement; the broader Franson literature shows that the relation between fringe observation and entanglement certification is subtler, particularly when coincidence postselection and engineered input states are considered.

5. Entanglement certification, dual-basis security, and quantum communication

CFI’s most explicit protocol role appears in high-dimensional quantum key distribution based on time-energy entanglement. In that setting, security against collective attacks rests on visibility data obtained from Franson and conjugate-Franson interferometers, which probe photon-pair frequency correlations and arrival-time correlations, respectively. These measurements are translated into bounds on the time-frequency covariance matrix, and then into an upper bound on the eavesdropper’s Holevo information by adapting the Gaussian-state security analysis of continuous-variable quantum key distribution (1311.0825).

The secure-key-rate lower bound was written as

$\phi_T=\phi_S+\phi_I$2

where $\phi_T=\phi_S+\phi_I$3 is the reconciliation efficiency, $\phi_T=\phi_S+\phi_I$4 is the Alice–Bob mutual information, $\phi_T=\phi_S+\phi_I$5 is the number of raw bits per postselected frame, $\phi_T=\phi_S+\phi_I$6 is the single-pair emission fraction from decoy analysis, and $\phi_T=\phi_S+\phi_I$7 is the upper bound on Eve’s Holevo information for the allowed time-frequency covariance matrices (1311.0825).

This dual-basis use of CFI materially tightens security certification. The same work states that visibility data from just the Franson interferometer provides a weaker, but nonetheless useful, secure-key-rate lower bound; adding CFI constrains arrival-time correlations as well, thereby closing attacks that would not be visible in frequency-only monitoring. With decoy states incorporated to handle multiple-pair emissions, the analysis showed that over $\phi_T=\phi_S+\phi_I$8 transmission distance in optical fiber, time-energy entanglement HDQKD could permit a $\phi_T=\phi_S+\phi_I$9 bit/sec secure-key rate and a photon information efficiency of $\Delta\Omega$0 secure-key bits per photon coincidence in the key-generation phase using receivers with $\Delta\Omega$1 system efficiency (1311.0825).

Outside QKD, CFI has been presented as useful for photonic entanglement, quantum communications, and quantum networking, precisely because it is phase-sensitive in a way complementary to standard JSI-based analyzers. The reported implementation was also described as nonlocal and reference-free, in that it does not require a classical reference or pulsed operation (Chen et al., 2021).

6. Network implementation, synchronization, and hardware constraints

Distributed CFI inherits the same practical constraints that govern high-visibility nonlocal Franson interference: matched delays, low accidental counts, robust phase stability, suppression of classical-to-quantum crosstalk, and control of differential dispersion. A fiber-network experiment on nonlocal Franson interferometry addressed the synchronization problem directly by co-propagating a classical Radio-over-Fiber clock signal with energy-time entangled photon pairs in the same fiber, using cross-band allocation with the clock in the O-band at $\Delta\Omega$2 nm and the quantum signal in the L-band at $\Delta\Omega$3 nm. Over $\Delta\Omega$4 km of single-mode fiber, the differential delay between the clock and the quantum signal had a standard deviation of $\Delta\Omega$5 ps over $\Delta\Omega$6 hours, and nonlocal quantum interference was observed with visibility $\Delta\Omega$7 without dedicated timing infrastructure; the same work explicitly states that its technical solutions pave the way for scalable, operational CFI in practical networks (Xiang et al., 25 Feb 2026).

That synchronization approach relies on a large wavelength separation of $\Delta\Omega$8 nm to reduce spontaneous Raman scattering. The reported experiment and simulations showed that SpRS noise at $\Delta\Omega$9 nm due to a $$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$0 nm O-band classical signal is about $$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$1 times lower than at $$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$2 nm in the C-band. A $$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$3 MHz clock was transmitted through the same $$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$4 km fiber link using a $$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$5 nm SFP module with $$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$6 dBm output power, and the recovered clock was used directly for timing synchronization of the time-tagging electronics. The Bell parameter was $$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$7, violating the CHSH inequality by $$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$8 standard deviations (Xiang et al., 25 Feb 2026).

Passive integrated interferometer technology addresses a different but equally important implementation bottleneck. A telecom-band experiment using cascaded PPLN waveguides and fully passive path-imbalanced Mach–Zehnder interferometers on photonic integrated circuits achieved a two-photon interference visibility of $$V_{\rm CFI} = \int dt_-\,{\rm JTI}(t_-) \cos(\Delta\Omega t_-) .$$9 from sinusoidal fringe fitting, with raw visibility $$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$0, background-corrected visibility $$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$1, heralding efficiency $$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$2, and coincidence-to-accidental ratio exceeding $$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$3 at only $$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$4 mW of pump power (Emadi et al., 27 Mar 2026). Although that work is framed as Franson interference, this suggests a hardware path for compact, fiber-integrated analyzers suitable for distributed CFI as well.

Dispersion control remains a separate issue. In fiber-based Franson interferometry, the visibility-limiting quantity is the differential dispersion between the long and short paths within each interferometer, and the condition for restoration of visibility is

$$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$5

Using opposite group-velocity dispersion in distant interferometers, nonlocal dispersion cancellation restored the visibility to $$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$6, the same value obtained with local cancellation (Zhong et al., 2013). The relevance to CFI is explicit in the summarized discussion of that work: nonlocal compensation is presented as a route to overcoming dispersion sensitivity in Franson/CFI schemes, including heterogeneous networks and photonic-chip settings (Zhong et al., 2013).

7. Interpretive issues and broader generalizations

Several papers on Franson-type interference address the physical origin of the nonlocal fringe. One analysis argues that the correlation fringe arises from coincidence-provided quantum superposition between independent Mach–Zehnder interferometers and states that nonlocal correlation can be created from non-entangled photons through the interferometers under suitable indistinguishability and spectral conditions (Ham, 2020). A related coherence-based treatment describes the Franson fringe as second-order amplitude superposition between the $$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$7 and $$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$8 nonlocal basis products, with the joint-phase relation protected against random spectral detuning by the fixed sum-phase relation of entangled photon pairs (Ham, 2021). These accounts do not eliminate the operational role of high visibility in entanglement certification; rather, they sharpen the distinction between the existence of a Franson-type fringe and the specific evidentiary standard required for entanglement claims.

The same interpretive broadening appears in work on macroscopic Franson-type nonlocal correlation, where randomness-based polarization-basis coherent superposition of laser light is used to reproduce phase-dependent correlations. That scheme replaces path randomness by polarization-basis randomness and attributes the effect to $$V_{\rm CFI}(\Delta\Omega) = \left\langle \cos\left[(\hat{t}_S-\hat{t}_I)\Delta\Omega \right] \right\rangle .$$9-type coherence, with the claim that Franson-type nonlocal correlation can be extended to macroscopic regimes via coherent superposition (Ham, 2021). A plausible implication is that CFI and related conjugate-basis interferometries sit at an interface between strict entanglement certification and a wider theory of coincidence-conditioned nonlocal interference.

Franson-type ideas have also been extended well beyond the standard energy-time setting. A polarization-based Franson-type interferometer exhibited Hong–Ou–Mandel-type peak and dip fringes without a common beam splitter, spatial quantum beating for nondegenerate photon pairs without frequency entanglement, and recovery of polarization entanglement through delayed compensation (Kim et al., 2017). Spectrally resolved Franson interference modulated the JSI along both signal and idler directions for positively correlated, negatively correlated, and non-correlated biphotons, with proposed applications to high-dimensional frequency entanglement and time-frequency grid states (Jin et al., 2023). In mesoscopic electron physics, a Franson-interferometer-based setup was analyzed for probing Cooper pairs, with current-current correlations giving access to internal coherence proportional to Pippard’s length and to the de Broglie wavelength of center-of-mass motion (Giovannetti et al., 2012).

Within this broader landscape, CFI occupies a specific and technically valuable niche: it is the conjugate time-basis interferometric probe of time-energy entanglement, distinguished by JTI dependence and spectral-phase sensitivity, yet embedded in a larger family of Franson-type measurements whose physical interpretations, security uses, and hardware realizations continue to diversify.