Gaussian Phase Matching in Nonlinear Optics

• Gaussian phase matching is the nonlinear optics principle that ensures coherent field addition by overlapping spatially structured Gaussian beams or guided modes.
• It integrates effects like Gouy phase, spatial mode overlap, and engineered phase-matching functions to optimize processes such as second-harmonic generation and SPDC.
• Practical designs, including dual-waveguide resonators and chirped quasi-phase-matched media, demonstrate significant efficiency gains and improved control of nonlinear frequency conversion.

Gaussian phase matching is the form of phase matching that arises when the interacting waves are approximately Gaussian beams or Gaussian-like guided modes. In nonlinear optics, phase matching is the requirement that all the infinitesimal nonlinear sources that generate a new frequency radiate in phase so that their fields add constructively; in the Gaussian setting, that requirement is expressed through overlap integrals of mode profiles together with longitudinal and transverse phase factors rather than by a purely plane-wave condition alone (Liang et al., 22 Dec 2025). Across contemporary photonics and quantum optics, the term therefore denotes a family of closely related ideas: the usual Gaussian-beam generalization of the plane-wave mismatch $\Delta k$, the inclusion of Gouy phase and spatial mode overlap, the engineering of Gaussian-like or apodized phase-matching functions in spontaneous parametric down-conversion (SPDC), and, in broader Gaussian-state settings, the alignment of phases needed for coherent superpositions and interferometric optimality (Brito et al., 2020).

1. Classical formulation and Gaussian-beam generalization

For second-harmonic generation, the nonlinear polarization is written as

$$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$

and, in the simplest plane-wave model, the generated field after a length $L$ scales as

$$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$

Perfect phase matching corresponds to $\Delta k=0$, whereas Gaussian beams and Gaussian-like guided modes replace this one-dimensional picture by overlap integrals of transverse profiles and longitudinal phase factors such as $\exp(i\Delta kz)$ (Liang et al., 22 Dec 2025).

The essential physical statement is that each point of the nonlinear medium acts like a dipole emitter with a certain phase. Gaussian phase matching is therefore not limited to matching effective indices; it also requires that the spatially distributed source produced by the Gaussian pump maintain phase coherence over the region that carries appreciable amplitude. This broader interpretation appears explicitly in guided-mode resonators, dense atomic vapors, and SPDC models where the generated field is a coherent sum over spatially localized contributors rather than a single plane-wave amplitude (Liang et al., 22 Dec 2025).

A common simplification is to equate Gaussian phase matching with the vanishing of a scalar $\Delta k$. The published treatments indicate that this is incomplete. In dense rubidium vapors, for example, the emitted four-wave-mixing field acquires a Gaussian angular factor from the transverse beam envelopes and a sinc factor from longitudinal propagation through the finite cell; the observable output is governed by the overlap of those two factors, not by a bare longitudinal condition alone (Leszczyński et al., 2016).

2. Gaussian modes, Gouy phase, and spatial coherence

Gaussian phase matching is also shaped by the phase structure intrinsic to paraxial Gaussian propagation. In the Madelung–Bohm treatment of a paraxial Gaussian field, the envelope is written as $E(x,z)=A(x,z)e^{iS(x,z)}$, and the propagated Gaussian acquires a longitudinal phase term identified with the Gouy phase. For the free-propagation solution discussed in the literature, the phase contains

$\phi_G(z)=-\frac{1}{2}\arctan z,$

with the transverse-independent term interpreted as the Gouy phase required to maintain Gaussian self-similarity under propagation (Moya-Cessa et al., 2021).

That result gives a precise sense in which Gaussian phase matching must include Gouy-phase compatibility. The same analysis ties the Gaussian phase to an Ermakov–Lewis invariant and to an effective Bohm index of refraction that generates a GRIN medium producing the focusing needed for the Gouy phase. A plausible implication is that, for interacting Gaussian modes, phase matching is constrained not only by propagation constants but also by the compatibility of the width evolution and focusing dynamics that determine their Gouy phases (Moya-Cessa et al., 2021).

In SPDC, this extension becomes explicitly biphotonic. A double-Gaussian approximation represents the two-photon wavefunction as

$$\Psi(x_1,x_2)= \frac{1}{\sqrt{\pi\,\sigma\,\Omega}} \exp\!\left[-\frac{(x_1-x_2)^2}{4\sigma^2}\right] \exp\!\left[-\frac{(x_1+x_2)^2}{4\Omega^2}\right],$$

where $$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$0 governs the center coordinate and $$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$1 the correlation width. In the cited treatment, $$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$2 is related to crystal length and pump wavelength by

$$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$3

so the Gaussian approximation to the phase-matching function is built directly into the spatial correlations of the biphoton (Brito et al., 2020). Upon propagation, the two collective Gaussian modes acquire Rayleigh lengths $$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$4 and $$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$5, and the biphoton Gouy phase depends on both. This establishes a direct link between Gaussian phase matching, Rayleigh-length engineering, and entanglement structure (Brito et al., 2020).

3. Green’s-function phase matching in guided-mode resonators

A rigorous generalization of Gaussian phase matching to nanophotonic guided resonators is provided by the two-dimensional Green’s function integral method (GFIM), which expresses the generated harmonic field as

$$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$6

In the lithium-niobate structure analyzed in that work, the relevant scalar form is effectively an integral of $$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$7 over the nonlinear region, making the local phase $$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$8 the decisive quantity rather than a single propagation mismatch (Liang et al., 22 Dec 2025).

The method introduces a pointwise phase-matching factor and an overall phase-matching factor (PMF). The PMF is the magnitude of the coherent sum of all constituent second-harmonic contributions divided by the incoherent sum of their magnitudes. It is bounded between 0 and 1, equals 1 when all microscopic sources add with identical phase, and reduces to the familiar $$\mathbf{P}^{(2)}(2\omega,\mathbf{r})=\varepsilon_0\chi^{(2)}:\mathbf{E}_\omega(\mathbf{r})\,\mathbf{E}_\omega(\mathbf{r}),$$9 in a one-dimensional plane-wave limit. In this sense, the PMF is a rigorous generalization of the Gaussian and plane-wave phase-matching factor to realistic guided-mode resonators (Liang et al., 22 Dec 2025).

GFIM visualization shows that conventional single-layer guided-mode resonators can suffer severe phase mismatch because the harmonic Green’s function oscillates spatially while the fundamental standing-wave mode fixes the nonlinear source phase. The product $L$0 then rotates in the complex plane, and constructive and destructive regions alternate across the nonlinear layer. The cited work reports that unoptimized guided-mode resonators can have PMF values as low as 0.008, whereas a dual-waveguide design with a high-index silicon layer beneath lithium niobate increases the PMF from 0.008 to 0.63 and raises the second-harmonic-generation efficiency to $L$1, a $L$2-fold improvement over the conventional single-layer guided-mode resonator. With additional duty-cycle tuning, the PMF exceeds 0.91 and the device achieves a record SHG efficiency of 26.7% at a low pump intensity of $L$3 (Liang et al., 22 Dec 2025).

The central design rule is to confine the fundamental field to regions where the harmonic Green’s-function phase varies slowly, or equivalently to suppress nonlinear material in out-of-phase regions. In Gaussian language, this is the statement that Gaussian-like confined fields should overlap only those parts of the structure where the harmonic response preserves spatial phase coherence.

4. Engineered Gaussian-like phase-matching functions in SPDC and dense media

In chirped quasi-phase-matched media for SPDC, the phase-matching function becomes an explicitly engineered object. For multilayer chirped photonic-like crystals and aperiodically poled crystals, the biphoton spectral amplitude factorizes into a pump envelope and a longitudinal phase-matching function $L$4, and the latter is expressed as a discrete sum over layer contributions with quadratic phase in the layer index. The resulting amplitudes are discrete Gauss sums, and the cited analysis shows that small numbers of layers produce well-resolved spectral lines, whereas larger numbers of layers yield broadband smooth spectra (Antonosyan et al., 2012).

That framework does not explicitly derive a Gaussian phase-matching function, but it does identify the route toward one. The discussion states that, in the large-$L$5 limit with slowly varying layer parameters, the discrete sum approaches a continuous chirped integral. A plausible implication is that suitable chirp and apodization allow the effective phase-matching envelope to approach a Gaussian or chirped-Gaussian profile, since the same formalism is described as a natural starting point for synthesizing a desired target envelope such as a Gaussian (Antonosyan et al., 2012).

Dense atomic vapors provide a complementary real-space example. In the rubidium double-ladder four-wave-mixing geometry, each input beam is treated as Gaussian, and the generated far-field amplitude takes the form

$L$6

The first factor is the angular Fourier transform of the Gaussian nonlinear polarization; the second is the longitudinal phase-matching factor of the finite medium. The cited work shows that refractive-index changes on the order of $L$7 can shift the phase-matching cone enough to alter both the emission direction and the detuning at which the four-wave-mixing maximum occurs, so the observed optimum need not coincide with the maximum of $L$8 alone (Leszczyński et al., 2016). Gaussian phase matching in this setting is therefore the overlap between a broader Gaussian angular distribution and a much narrower phase-matching ring.

5. Relaxed and nonstandard geometries

Standard Gaussian phase matching in bulk crystals and conventional waveguides is normally tied to a specific arrangement of beams, usually co-propagating, because the relevant longitudinal wavevector components change strongly with angle. In zero-index waveguides, that directional dependence is dramatically relaxed. The cited experiments use silicon Dirac-cone zero-index metamaterial waveguides with effective refractive index crossing zero near $L$9 or $$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$0 nm, so that $$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$1 becomes very small and the phase mismatch becomes weakly dependent on propagation direction (Gagnon et al., 2021).

The reported measurements demonstrate four-wave mixing with spectrally distinct co- and counter-propagating pump and probe beams, backward generation of a nonlinear signal, and excitation by an out-of-plane probe beam. The work describes this as the first experimental observation of direction-independent phase matching for a medium sufficiently long for phase matching concerns to be relevant, and it estimates a coherence length of at least $$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$2 in a $$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$3 device from the observed efficiencies (Gagnon et al., 2021). The standard Gaussian-beam logic based on overlap integrals still applies, but the k-space geometry is altered because one or more participating modes have near-zero effective index.

This does not abolish Gaussian phase matching; it changes which constraints dominate. The published account emphasizes that overlap physics remains analogous to that of Gaussian guided modes, while the usual angular restrictions on $$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$4 are relaxed by the low-index dispersion. A plausible implication is that zero-index platforms move Gaussian phase matching away from directional fine-tuning and toward spectral placement near the zero-index point.

6. Broader Gaussian-state and interferometric usages

Outside nonlinear frequency conversion, the phrase “Gaussian phase matching” acquires related but distinct meanings. In Mach–Zehnder interferometry with Gaussian input states, phase matching refers to the alignment of coherent phases, squeezing angles, beam-splitter conventions, and local-oscillator phase so as to maximize the quantum Fisher information (QFI). For a squeezed-coherent plus squeezed-vacuum input, the QFI-optimal conditions are

$$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$5

leading to

$$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$6

For the most general squeezed-coherent plus squeezed-coherent input, the literature identifies three optimal phase-matching patterns, denoted PMC1, PMC2, and PMC3, each dominant in a different resource regime; with PMC2 and equal squeezing, homodyne detection saturates the QCRB (Ataman, 2019). In this setting, Gaussian phase matching is an interferometric optimization problem rather than a nonlinear propagation condition.

In Gaussian quantum field theory and hybrid continuous-variable systems, the relevant question is the matching of phase across different representations of a Gaussian state. The cited work derives the phase needed to equate the unitary representation of a zero-mean Gaussian state, its Bogoliubov or nullifier description, and its Fock-basis representation. The central result is an exact expression for the complex scalar $$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$7 that makes

$$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$8

so that superpositions of Gaussian states can be manipulated coherently rather than suffering an erroneous relative phase (Funai, 16 Apr 2025). Here phase matching means representation-theoretic consistency.

A further distinct usage appears in computer-generated holography based on Gaussian Wave Splatting. In that context, “Gaussian phase matching” concerns how the phase of each Gaussian wave primitive is defined, constrained, or deliberately decorrelated so that many Gaussians compose into a wavefield with the desired spatial and angular behavior. Smooth-phase Gaussian Wave Splatting implicitly aligns phases, whereas random-phase GWS deliberately breaks that alignment to improve bandwidth utilization, increase eyebox size, reconstruct accurate defocus blur and parallax, and support time-multiplexed rendering to suppress speckle artifacts (Chao et al., 24 Aug 2025). This is not nonlinear-optical phase matching in the usual sense, but it preserves the central theme: the performance of a Gaussian-field system is determined by how phase is distributed across Gaussian constituents.

Taken together, these formulations show that Gaussian phase matching is not a single formula but a structured principle. In its narrowest and most traditional sense, it is the Gaussian-beam extension of the plane-wave $$E_{2\omega}(L)\propto \int_0^L e^{i\Delta kz'}\,dz' = L\,\mathrm{sinc}\!\left(\frac{\Delta kL}{2}\right)e^{i\Delta kL/2}, \quad \Delta k=2k_\omega-k_{2\omega}.$$9 condition. In its more general contemporary form, it is the requirement that Gaussian or Gaussian-like contributions retain the correct relative phase under propagation, confinement, interference, or representation change, so that the intended coherent process is realized rather than degraded by hidden spatial or modal mismatch.