Joint Spectral Intensity (JSI)

• JSI is the squared magnitude of the joint spectral amplitude that maps frequency correlations in photon pairs produced by SPDC and SFWM.
• It guides the diagnosis of photon-pair sources by indicating factorability and heralded-state purity through its distinct spectral shape.
• Experimental techniques such as scanning monochromators, Fourier-transform spectroscopy, and stimulated-emission tomography are used to measure and engineer the JSI.

Joint spectral intensity (JSI) is the experimentally accessible, nonnegative frequency-domain map that summarizes how likely it is to detect a signal photon at one frequency together with an idler photon at another frequency. In photon-pair sources based on spontaneous parametric downconversion (SPDC) and spontaneous four-wave mixing (SFWM), it is the squared magnitude of the joint spectral amplitude (JSA), so it captures intensity correlations in the two-photon spectrum while discarding spectral phase. In its simplest and most widely used form,

$$\mathrm{JSI}(\omega_s,\omega_i)=|\mathrm{JSA}(\omega_s,\omega_i)|^2,$$

or, in the notation of one SPDC analysis,

$$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$

with $\gamma$ a normalization factor related to the squeezing strength. Because it condenses spectral correlation structure into a single two-dimensional distribution, the JSI is central to source diagnosis, factorability analysis, heralded-state purity estimation, and the engineering of integrated and bulk quantum-light sources; however, it is not a full characterization of the photon-pair state because phase information resides in the complex JSA rather than in the JSI alone (Bell et al., 2020).

1. Definition, state description, and basic structure

In low-gain SPDC and SFWM, the two-photon state is commonly written in terms of a biphoton wavefunction or JSA. For SPDC,

$$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$

with

$$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$

where $\alpha$ is the pump envelope amplitude and $\phi$ is the phase-matching amplitude. The JSI is then

$\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$

Equivalent formulations appear across SPDC, SFWM, and cavity-enhanced sources, including

$$\left| \psi \right\rangle ={\left| 0 \right\rangle }_{s}{\left| 0 \right\rangle }_{i}+\beta \iint d\omega_{s}d\omega_{i}\,F(\omega_{s},\omega_{i})\, \hat{a}_{s}^{\dagger }(\omega_{s})\hat{a}_{i}^{\dagger }(\omega_{i}) {\left| 0 \right\rangle }_{s}{\left| 0 \right\rangle }_{i},$$

followed by

$$\mathrm{JSI}(\omega_s,\omega_i)=|F(\omega_s,\omega_i)|^2.$$

In physical terms, the JSI is the two-dimensional probability distribution for detecting signal and idler photons at particular frequencies (Ye et al., 2024).

The shape of the JSI is controlled by the product of pump-envelope and phase-matching structure. A standard SPDC model writes

$$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$0

with energy conservation $$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$1 in SPDC and $$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$2 in degenerate SFWM. In periodically poled crystals, the pump-envelope intensity can form a ridge with a fixed tilt of $$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$3 in the $$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$4 plane, while the phase-matching intensity is set by

$$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$5

The orientation and widths of the resulting JSI determine whether the source appears nearly separable, strongly correlated, anti-correlated, resonance-filtered, or multi-peaked (Zielnicki et al., 2018).

Although the modulus-squared definition is standard, there are regimes in which the JSI is defined operationally rather than from a single-pair wavefunction. In a strongly driven continuous-wave microring, many photon pairs can be present at once, and a zero-bandwidth cw pump makes naive frequency-domain coincidence expressions ill-defined. In that setting the JSI is defined from a realistic coincidence-counting scheme with narrow monochromators and a short resolution time $$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$6, yielding a measured coincidence rate rather than the low-pair expression $$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$7 (Vernon et al., 2015).

2. Spectral correlations, factorability, and heralded-state purity

The principal importance of the JSI is that spectral correlations determine whether heralding one photon leaves its partner in a nearly pure state or in a mixed state. If the JSI is factorable, detecting one photon gives little information about the frequency of the other; if it is elongated or structured, the heralded photon typically occupies multiple spectral modes. This connection is commonly formalized by Schmidt decomposition. For a biphoton state with Schmidt coefficients $$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$8,

$$P(\omega_s,\omega_i)=\gamma\,|\psi(\omega_s,\omega_i)|^2,$$9

and, for idealized factorable states, $\gamma$0. The inverse Schmidt number is the heralded single-photon purity,

$\gamma$1

In a diagonal Gaussian-ellipse parameterization,

$\gamma$2

These relations make the JSI a direct diagnostic of source usability for quantum networking and multi-photon interference (Zielnicki et al., 2018).

Several source-engineering programs aim explicitly at a nearly separable JSI. In group-velocity-matched PPKTP, the phase-matching ridge can be oriented so that it nearly compensates the pump-envelope correlation, making the JSI close to round or “subcircular,” i.e. nearly uncorrelated. In that system the numerical simulation predicts that the purity of joint spectral intensity ($\gamma$3) and the purity of joint spectral amplitude ($\gamma$4) can be kept higher than $\gamma$5 and $\gamma$6, respectively, when the wavelength is tuned from $\gamma$7 nm to $\gamma$8 nm, and directly measured JSIs at $\gamma$9 nm, $$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$0 nm and $$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$1 nm yielded $$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$2 of $$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$3, $$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$4 and $$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$5, respectively (Jin et al., 2013).

Microring sources provide a complementary route to factorability through resonance engineering and pump-duration control. In a silicon microring resonator, the anti-diagonal width is set by the product of the signal and idler cavity resonances, while the diagonal width is set by the convolution of the pump spectrum with the pump cavity resonance. Matching the pump pulse duration to the microring photon lifetime broadens the pump spectrally relative to cw pumping and moves the state toward a factorable form. In one implementation using an electronic step-recovery diode to carve optical pump pulses of roughly $$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$6–$$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$7 ps duration at $$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$8 GHz, the reported values were $$\left| \psi_{si} \right\rangle = \int_0^\infty \int_0^\infty d\omega_s\, d\omega_i\, f(\omega_s,\omega_i)\, \hat a_s^\dagger(\omega_s)\hat a_i^\dagger(\omega_i)\left|0\right\rangle,$$9 for the measured deconvolved JSI and $$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$0 theoretically, so the source was nearly separable though not perfectly so (Savanier et al., 2016).

The JSI is also used to distinguish separable and entangled regimes in more structured integrated resonators. In a silicon micro-ring resonator with split resonances, a separable-state regime produced an essentially single-peaked or compact measured JSI with $$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$1 and spectral purity $$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$2, whereas entangled-state regimes produced two-peaked or four-peaked JSIs in two-dimensional frequency space. The paper interprets the single-peaked JSI as weak spectral correlations and the multi-peaked JSIs as structured entanglement and richer frequency-domain wavefunction engineering (Ye et al., 2024).

3. What the JSI does not contain: spectral phase, coherence, and hidden correlations

A central limitation of the JSI is that it is only the modulus-squared of the full JSA. If

$$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$3

then the JSI discards the phase $$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$4, namely the joint spectral phase (JSP). Two sources may have similar intensity maps but different phase structure, which changes entanglement properties and the degree of spectral purity. One consequence stated explicitly is that without the JSP one can only estimate a lower bound on the Schmidt number, whereas including the phase gives a more accurate characterization of spectral correlations (Borghi, 2020).

This limitation is not merely formal. A nearly factorable JSI can coexist with non-separable phase. A simple example is

$$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$5

where the bilinear term $$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$6 represents a non-separable spectral phase correlation generated by a chirped pump laser pulse. Because $$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$7, this phase does not change the JSI directly, but it changes the multiphoton interference structure and the coherence properties of individual photons. The paper’s main message is therefore that JSI is useful but incomplete: it tells you the amplitude distribution in the joint spectrum, but not the spectral phase correlations that can reduce heralded single-photon purity, alter multi-photon interference, and limit indistinguishability even when the JSI looks factorable (Bell et al., 2020).

The phase sensitivity of higher-order observables makes this incompleteness experimentally consequential. For four-photon events, the probability

$$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$8

contains two pathways that interfere. For the chirped-pump form above,

$$f(\omega_s,\omega_i)=\phi(\omega_s,\omega_i)\,\alpha(\omega_s+\omega_i),$$9

so the four-photon coincidence pattern contains fringes whose period depends on the phase-correlation strength $\alpha$0. The same paper shows that phase correlations in the JSA induce spectral intensity correlations between two signal photons, even when the corresponding idler photons are not detected, through

$\alpha$1

with

$\alpha$2

A JSI measurement cannot access $\alpha$3, because $\alpha$4 depends on the complex amplitude overlap, not just intensities (Bell et al., 2020).

4. Measurement strategies and reconstruction methods

A substantial experimental literature treats the JSI as a characterization target in its own right. One comparison of photon-pair source diagnostics distinguishes six techniques: scanning monochromator measurements, a variant of Fourier transform spectroscopy designed to extract the desired information exploiting a resource-optimized technique, dispersive fibre spectroscopy, stimulated-emission-based measurement, measurement of the second-order correlation function $\alpha$5 for one of the two photons, and two-source Hong-Ou-Mandel interferometry. The first four reconstruct the JSI directly; the latter two do not reconstruct the JSI directly, but infer heralded-single-photon purity. The same comparison states that scanning monochromator, diagonal Fourier transform spectroscopy, and dispersive fibre spectroscopy are not phase sensitive, stimulated-emission tomography reconstructs the JSI and can be phase sensitive in principle, while $\alpha$6 and two-source HOM are phase sensitive and yield purity (Zielnicki et al., 2018).

Scanning monochromator measurements directly build the JSI from coincidence counts as a function of two monochromator settings, but they are very slow and of low pair collection efficiency because only photons passing both narrow slits are counted. Dispersive fibre spectroscopy maps wavelength to arrival time using chromatic dispersion in a long optical fibre; it is relatively fast and direct, but its resolution is strongly limited by detector and electronics timing jitter. Fourier-transform spectroscopy reconstructs the JSI from a joint temporal interferogram obtained with two interferometers and a two-dimensional Fourier transform; a simplified diagonal version estimates correlation using only one-dimensional scans when the JSI is approximately Gaussian. Stimulated-emission-based measurement exploits the spontaneous–stimulated correspondence and reconstructs the JSI by scanning a classical seed over frequency and measuring the stimulated output spectrum; it offers very high count rates, excellent signal-to-noise, and does not require single-photon detectors (Zielnicki et al., 2018).

The extension from intensity-only to full complex-state reconstruction proceeds by making phase-sensitive interferometric measurements. In a silicon photonic chip containing a double-bus ring resonator source and a broadband reference spiral waveguide, stimulated emission tomography was used not only to reconstruct the JSI but also to reconstruct the complex JSA. The interferogram is fitted to

$\alpha$7

from which a relative phase is extracted and combined with the phase of the resonator’s field enhancement obtained from the complex Through transfer function $\alpha$8. The measured JSI of the ring source was concentrated near the cavity resonance pair and appeared broader than the ideal simulation, likely because of additional stimulated radiation generated in the nearby spiral waveguides. Crucially, swapping the seeded resonance order changed the phase map but not the JSI, demonstrating that intensity alone does not reveal the full structure of the state (Borghi, 2020).

An alternative phase-sensitive route uses intensity interferometry. In that method, a heralded single photon or stimulated field is mixed with a weak, characterized reference pulse on a beam splitter and frequency-resolved intensity correlations at the outputs are measured. The phase-sensitive term

$\alpha$9

is isolated by a two-dimensional Fourier transform, and the spectral density matrix $\phi$0 is reconstructed. For SPDC, a herald frequency selects a cross-section of the JSA, and repeating this for all herald bins reconstructs the full two-photon spectral mode. In the unchirped case the measured Schmidt number was $\phi$1, while in the chirped-pump case the complex JSA had Schmidt number $\phi$2 and ignoring phase gave only $\phi$3, showing that the dominant correlations were phase correlations rather than amplitude correlations (Thekkadath et al., 2021).

5. Resonators, strong driving, loss, and engineered JSI structure

In cavity-enhanced sources, the JSI inherits the spectral filtering and enhancement of the resonator rather than appearing as a smooth broadband blob like in a straight waveguide. In a double-bus silicon microring resonator, the resonator JSI is concentrated near the cavity resonance pair, reflecting the frequency-selective enhancement of SFWM by the ring, and the profile is influenced by the ring’s field enhancement, linewidth, and finite coupling to the buses. In a split-resonance silicon micro-ring resonator, the cavity response distinguishes forward and backward circulating modes, and the output JSA is treated as a coherent weighted sum over four possible pump/signal/idler propagation combinations, allowing the spectral response to become multi-lobed rather than a single cavity peak (Borghi, 2020).

The strongly driven cw regime produces a qualitatively different operational picture. There the JSI is defined from coincidence measurements and splits into correlated and uncorrelated parts,

$\phi$4

The correlated contribution corresponds to both photons in a detected coincidence originating from the same SFWM pair, while the uncorrelated contribution arises when the detected signal and idler photons come from different pairs. At low power, the JSI is concentrated near

$\phi$5

forming a narrow antidiagonal ridge. For nonzero pump detuning $\phi$6, the low-power JSI splits into two peaks separated by approximately $\phi$7. At high pump power, even with $\phi$8, the JSI can split because cross-phase modulation shifts the signal and idler resonances. Near optical parametric oscillation threshold the JSI narrows dramatically, and at sufficiently high power the uncorrelated JSI develops four peaks, including diagonal peaks interpreted as a direct signature of multi-pair generation (Vernon et al., 2015).

Linear scattering loss modifies the JSI during generation rather than merely attenuating the output after the fact. In lossy SPDC waveguides the reduced generated-photon density operator decomposes naturally into two-photon, one-photon, and zero-photon contributions,

$\phi$9

and the JSI is associated with the two-photon sector $\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$0. The central result is that the usual pump-envelope $\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$1 phase-matching structure acquires an additional loss-matching structure: $\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$2 If either the pump or generated photon loss is much higher than the other, the side lobes of the $\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$3 phase-matching function are washed out; if pump and generated photon loss are appropriately balanced, the lossy JSI is identical to the lossless JSI in shape; and if the generated photon loss is frequency dependent, the shape of the JSI can be altered more severely, potentially leading to generated photons that are less frequency correlated though also produced less efficiently (Helt et al., 2014).

Recent integrated engineering work has treated the JSI itself as a programmable object. In a silicon micro-ring resonator with split resonances, two nearly independent functions shape the JSI: $\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$4 which governs two-dimensional signal-idler filtering, and

$\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$5

which controls the distribution along the energy-conservation antidiagonal direction. By choosing different resonance combinations and employing on-chip optical field differentiation, the experiments moved between a nearly separable single-peaked JSI, a two-peaked JSI along the main diagonal, and a four-peaked JSI in frequency space (Ye et al., 2024).

6. Applications, diagnostics, and recurring misconceptions

The most persistent misconception is that a favorable JSI is equivalent to a complete source characterization. The literature repeatedly rejects this equivalence. A source with an excellent JSI can still perform poorly in interference-based quantum applications if its JSA has nontrivial phase structure, because many important quantum-optical effects depend on spectral phase. This is why phase-sensitive tomography, intensity interferometry, multi-photon coincidence analysis, $\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$6, and HOM-based benchmarks remain necessary complements to JSI measurements (Bell et al., 2020).

At the same time, the JSI remains a primary practical witness for source design and certification. In the context of heralded single photons and multi-photon protocols, it indicates whether the source is factorable or frequency-correlated in intensity, reveals widths, orientations, side-lobes, and resonance-centered features, and supports purity estimation under flat-phase assumptions. In integrated photonics it is used to diagnose cavity-induced correlations and to verify whether pump shaping and resonance engineering have produced a nearly separable or deliberately entangled state. In micro-ring and waveguide platforms, multi-peaked, broadened, shifted, or side-lobe-suppressed JSIs are interpreted as signatures of split resonances, finite coupling, pump chirp, cross-phase modulation, multi-pair generation, dispersion, or loss (Ye et al., 2024).

The JSI has also been used as a spectral witness outside source characterization narrowly construed. In work on entangled two-photon absorption (ETPA), the molecule is modeled as a two-photon notch filter acting on the biphoton spectrum,

$\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$7

so that the transmitted JSI becomes

$\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$8

The central claim is that a measurable ETPA signature arises only if the molecular notch overlaps the JSI asymmetrically in frequency space; that asymmetry appears as unequal marginals, distortion of the transmitted JSI, and reduction in HOM visibility. The same work emphasizes that the phenomenon of ETPA presents a persistent controversy in the literature and concludes, for RhB in methanol at $\mathrm{JSI}=|f(\omega_s,\omega_i)|^2.$9 mM, that ETPA was not observed because the measured JSI and marginals showed no clear asymmetry beyond the input spectrum (Yepiz-Graciano et al., 30 Nov 2025).

Taken together, these developments establish the JSI as the standard intensity-domain descriptor of biphoton spectral structure and as a key observable for experimental quantum optics. It is the natural starting point for evaluating source factorability, resonance filtering, loss reshaping, and frequency-bin structure, but it is not identical to the full spectral state. A plausible implication is that the modern role of the JSI is dual: it remains the most accessible map of frequency-frequency correlations, while increasingly serving as one layer in a hierarchy of spectral diagnostics whose upper layers reconstruct phase, coherence, and multimode structure when the application demands more than intensity alone.