Coupled Lugiato–Lefever Equations

• Coupled Lugiato–Lefever equations are a family of mean-field models that represent interacting intracavity envelopes in driven-dissipative optical cavities.
• They encompass diverse settings—including hybrid branches from avoided crossings, bidirectional propagation, linear intermode coupling, and quadratic frequency conversion—to capture essential nonlinear interactions.
• The models reveal how mechanisms such as nonlinear cross-phase modulation, backscattering, and localized forcing shape soliton formation and frequency comb generation in microresonators.

Coupled Lugiato–Lefever equations are mean-field driven-damped cavity equations for two or more interacting intracavity envelopes, introduced when a scalar Lugiato–Lefever equation is not sufficient to represent the relevant resonant degrees of freedom. In the literature covered here, the coupled formulation appears in several distinct settings: hybrid branches created by avoided crossings of guided-mode families in Kerr microresonators, counter-propagating clockwise and counterclockwise fields in bidirectionally pumped rings, linearly mixed cavity modes with one-sided localized forcing, coupled electric and magnetic cavity envelopes in nonlinear materials, and fundamental/second-harmonic envelopes in $\chi^{(2)}$ resonators. Across these settings, the common structure is the coexistence of cavity loss, detuning, dispersion or diffraction, coherent drive, and interfield coupling, but the coupling mechanism itself may be nonlinear cross-phase modulation, linear backscattering, direct linear intermode tunneling, or quadratic frequency conversion (D'Aguanno et al., 2017, Skryabin, 2020, Cardoso et al., 11 Feb 2025, Mártin et al., 2012, Skryabin, 2020).

1. Conceptual scope and defining structure

The scalar Lugiato–Lefever equation describes a single intracavity envelope subject to coherent drive, dissipation, detuning, and Kerr nonlinearity. A coupled formulation becomes necessary when the cavity dynamics cannot be reduced to one envelope without discarding essential physics. The avoided-crossing study states this explicitly: near a strong interaction of two nearly degenerate transverse mode families, a scalar single-mode LLE is no longer sufficient because the system must represent two hybrid branches, their opposite dispersions, independent pumping, and nonlinear cross-coupling (D'Aguanno et al., 2017).

In the coupled setting, each field carries its own detuning and dispersive operator, while the interaction appears through terms such as self-phase modulation and cross-phase modulation, direct linear mixing, or parametric conversion. The bidirectional Kerr model gives a particularly clear example. Its exact coupled-mode equations contain the Kerr structure $|Q_\pm|^2Q_\pm$ for self-phase modulation and $2|Q_\mp|^2Q_\pm$ for cross-phase modulation, together with a linear backscattering coefficient $R$ coupling clockwise and counterclockwise waves (Skryabin, 2020). By contrast, the localized-pump self-defocusing model uses direct linear intermode coupling $i\kappa E_{3-j}$ between two one-dimensional LLE fields, with only one component directly driven by a Gaussian pump (Cardoso et al., 11 Feb 2025). The $\chi^{(2)}$ microresonator model replaces cubic self-action by quadratic terms $Q_o^*Q_e$ and $Q_o^2$, yielding a pair of coupled Lugiato–Lefever-like equations for the fundamental and second-harmonic envelopes (Skryabin, 2020).

A useful classification is therefore by physical origin rather than by algebra alone. One class is hybrid-branch coupling, in which the fields are the avoided-crossing eigenbranches of a linearly mixed modal system. A second is propagation-direction coupling, in which the fields are counter-propagating waves. A third is component or mode coupling, in which two cavity modes are directly linearly mixed. A fourth is frequency-conversion coupling, in which the fields occupy different spectral bands linked by $\chi^{(2)}$ interactions. This suggests that “coupled Lugiato–Lefever equations” is not a single canonical model but a family of mean-field reductions adapted to different cavity symmetries and nonlinearities.

2. Core equation classes and coupling mechanisms

The papers considered here realize several non-equivalent coupled-LLE structures.

Setting	Fields	Dominant coupling
Avoided crossing in Kerr resonator	Two hybrid branches	Nonlinear cross-coupling
Bidirectional Kerr ring	CW and CCW envelopes	Backscattering and XPM
One-sided localized pump	Two linearly mixed cavity modes	Linear coupling
$\chi^{(2)}$ ring resonator	Fundamental and second harmonic	Quadratic conversion

In the avoided-crossing case, the starting point is a general two-field CLLE with envelopes $|Q_\pm|^2Q_\pm$0, normalized detunings $|Q_\pm|^2Q_\pm$1, normalized GVD coefficients $|Q_\pm|^2Q_\pm$2, pump amplitudes $|Q_\pm|^2Q_\pm$3, and nonlinear overlap integrals $|Q_\pm|^2Q_\pm$4. The fields are the two hybrid branches formed after diagonalizing the linear guided-mode coupling, so the remaining interfield interaction in the mean-field equations is nonlinear rather than an explicit residual linear term (D'Aguanno et al., 2017).

In the bidirectional Kerr case, the exact envelope equations are written for $|Q_\pm|^2Q_\pm$5 and $|Q_\pm|^2Q_\pm$6, with opposite-sign drift and odd-order dispersion because the waves propagate in opposite directions. The linear coupling is backscattering-induced and acts on reflected envelopes $|Q_\pm|^2Q_\pm$7. The nonlinear terms contain the standard Kerr SPM and XPM structure, with the expected factor of $|Q_\pm|^2Q_\pm$8 for co-polarized fields (Skryabin, 2020).

In the linearly coupled self-defocusing model, the equations are

$|Q_\pm|^2Q_\pm$9

$2|Q_\mp|^2Q_\pm$0

with a Gaussian pump applied only to the first field,

$2|Q_\mp|^2Q_\pm$1

Here the sign $2|Q_\mp|^2Q_\pm$2 is explicitly identified as self-defocusing, and the localized structures are driven-dissipative states pinned by the inhomogeneous pump rather than conservative bright solitons (Cardoso et al., 11 Feb 2025).

In the $2|Q_\mp|^2Q_\pm$3 microresonator setting, the mean-field equations are explicitly Lugiato–Lefever-like but not Kerr-type: $2|Q_\mp|^2Q_\pm$4

$2|Q_\mp|^2Q_\pm$5

These equations encode detuning, cavity loss, dispersion, quasi-phase matching, walk-off, and self-steepening-like corrections due to nonlinear-coefficient dispersion (Skryabin, 2020).

3. Avoided crossings and hybrid-branch CLLEs in Kerr microresonators

The avoided-crossing formulation is one of the clearest instances in which a genuine two-field mean-field model is required. The physical system is a whispering-gallery or integrated microresonator supporting two guided modes with different transverse profiles, exemplified by $2|Q_\mp|^2Q_\pm$6 and $2|Q_\mp|^2Q_\pm$7. When their eigenfrequencies become nearly degenerate, strong linear guided-mode interaction prevents an actual crossing and produces two hybridized resonances with eigenfrequencies

$2|Q_\mp|^2Q_\pm$8

This resonance repulsion alters the local free spectral range and therefore the effective local group-velocity dispersion (D'Aguanno et al., 2017).

A central result is that the hybrid branches acquire opposite-sign dispersions near the crossing. In the illustrated example, both uncoupled mode families are locally in the normal-dispersion regime, yet after hybridization one branch remains normal while the other becomes anomalous. This is precisely the regime in which bright solitons and broadband frequency combs can be excited when both branches are pumped for a suitable choice of pump powers and detuning parameters, and the paper reports a deterministic path for soliton generation (D'Aguanno et al., 2017).

The simplified avoided-crossing CLLE is obtained under three assumptions: negligible group-velocity mismatch $2|Q_\mp|^2Q_\pm$9, equal-magnitude opposite-sign GVD $R$0, and cross-coupling twice the self-coupling. With the rescalings

$R$1

the first branch satisfies

$R$2

The second branch has the complementary sign structure associated with the opposite GVD branch (D'Aguanno et al., 2017).

An important point of interpretation is basis choice. The paper writes the dynamical equations after hybridization, in the branch basis rather than the original guided-mode basis. For that reason, there is no explicit $R$3 term in the final reduced equations; the avoided crossing has already been diagonalized, and the remaining coupling is nonlinear cross-coupling. This distinction matters because it prevents the coupled LLE from being misread as a residual linearly mixed two-mode model. The CLLE here is a mean-field description of two hybrid branches with independent pumping, distinct detunings, and opposite effective dispersions.

4. Bidirectional CLLEs: counter-propagating combs, backscattering, and washout

In bidirectionally pumped ring microresonators, the natural coupled fields are the clockwise and counterclockwise intracavity waves. The coupled-mode hierarchy introduced for this problem distinguishes several time scales: cavity linewidth and nonlinear shifts, dispersion-induced modal mismatch, and repetition-rate oscillations $R$4. The modal frequencies are expanded as

$R$5

with $R$6, and the exact envelope equations for $R$7 read

$R$8

$R$9

The reflected fields are defined by $i\kappa E_{3-j}$0 (Skryabin, 2020).

These equations differ from the standard single-field LLE in three identified ways: there are two fields, odd-order dispersion and drift terms have opposite signs, and linear coupling involves the reflected counter-propagating envelope. The derivation proceeds from modal expansions of the intracavity field, slowly varying amplitude approximations, rotating-wave separation, and Fourier reconstruction in the azimuthal angle. Kerr nonlinear terms are retained only in near-resonant contributions consistent with modal momentum conservation (Skryabin, 2020).

The technical centerpiece of this theory is the washout of repetition-rate frequencies. In the raw nonlinear mode sums, phase-sensitive CW–CCW interaction terms carry factors proportional to $i\kappa E_{3-j}$1. Since $i\kappa E_{3-j}$2 is typically much larger than $i\kappa E_{3-j}$3, $i\kappa E_{3-j}$4, and nonlinear shifts, these terms average out unless $i\kappa E_{3-j}$5. After averaging, the phase-sensitive nonlinear mixing between counter-propagating comb teeth is removed, leaving only the phase-insensitive cross-Kerr contribution proportional to the integrated power in the opposite direction (Skryabin, 2020).

This leads to reduced coupled LLE-like equations for $i\kappa E_{3-j}$6 and $i\kappa E_{3-j}$7 with no $i\kappa E_{3-j}$8 term: $i\kappa E_{3-j}$9 with the corresponding reflected equation for $\chi^{(2)}$0 (Skryabin, 2020). The reduced model implies that without backscattering the two propagation directions behave almost like independent LLE resonators, except for a nonlinear detuning shift proportional to opposite-direction power. With nonzero $\chi^{(2)}$1, coherent CW/CCW coupling and locking become possible.

A practical caution stated explicitly is that the reduced equations rely on the scale separation $\chi^{(2)}$2 nonlinear shifts. If that separation is not strong enough, the exact $\chi^{(2)}$3 equations must be used rather than the averaged $\chi^{(2)}$4 system.

5. Linear intermode coupling, one-sided pumping, and induced localization

A distinct CLLE regime arises when two cavity modes are linearly mixed and only one is directly forced. In the self-defocusing coupled LLE model with localized Gaussian pumping, the equations are presented directly in dimensionless mean-field form and describe a passive nonlinear optical cavity supporting two interacting intracavity modes. The first field $\chi^{(2)}$5 is directly driven by the Gaussian $\chi^{(2)}$6, while the second field $\chi^{(2)}$7 is excited solely by linear coupling $\chi^{(2)}$8 (Cardoso et al., 11 Feb 2025).

The main conceptual result is induced localization in the unpumped component. Because the cubic terms are self-defocusing, the system does not support bright self-trapped states through conservative balance alone. The localized structures are therefore driven-dissipative localized states pinned by the inhomogeneous pump. Once $\chi^{(2)}$9 reaches a pinned stationary profile, the term $Q_o^*Q_e$0 acts as an effective localized source in the second equation. This is the mechanism behind the paper’s “induced localized coupled modes” (Cardoso et al., 11 Feb 2025).

The stationary localized solutions reported numerically are single-hump profiles centered at $Q_o^*Q_e$1. They are asymmetric between the two fields because only $Q_o^*Q_e$2 is pumped. For the representative parameter set

$Q_o^*Q_e$3

the powers of the two fields become equal at about $Q_o^*Q_e$4, and for larger coupling the partner field can surpass the directly pumped field in power (Cardoso et al., 11 Feb 2025).

The parameter study gives a detailed picture of how coupled-LLE localization transfer depends on detuning, losses, coupling, and pump width. Increasing the symmetric detuning from $Q_o^*Q_e$5 to $Q_o^*Q_e$6 shifts the equal-power point from $Q_o^*Q_e$7 to about $Q_o^*Q_e$8. With $Q_o^*Q_e$9, the equal-power point occurs around $Q_o^2$0. In the symmetric-loss case, increasing $Q_o^2$1 lowers the power in both fields, and by $Q_o^2$2 the induced-mode power is reported as $Q_o^2$3. If only $Q_o^2$4 is increased, $Q_o^2$5 while $Q_o^2$6 saturates near $Q_o^2$7 (Cardoso et al., 11 Feb 2025).

Detuning dependence is more structured. In the symmetric case $Q_o^2$8, the powers are larger for negative detuning. For $Q_o^2$9, maxima occur at approximately $\chi^{(2)}$0 and $\chi^{(2)}$1, while the induced field $\chi^{(2)}$2 has a single peak around $\chi^{(2)}$3. When only $\chi^{(2)}$4 varies with $\chi^{(2)}$5, the induced-field power is maximized around $\chi^{(2)}$6 (Cardoso et al., 11 Feb 2025). This establishes that the detuning of the unpumped mode materially controls the induced localized state.

The numerical method is a pseudo-spectral method in space with fourth-order Runge–Kutta in time, using $\chi^{(2)}$7 spectral points, domain size $\chi^{(2)}$8, and time step $\chi^{(2)}$9. Stability evidence is dynamical convergence from vacuum initial conditions rather than spectral linear-stability analysis (Cardoso et al., 11 Feb 2025). A plausible implication is that this model occupies a minimal corner of coupled-LLE theory: it isolates how direct linear mode mixing and spatially localized forcing can generate vector-like dissipative localization even in a self-defocusing cavity.

6. Related mean-field generalizations, reductions, and neighboring models

Several closely related works clarify what coupled Lugiato–Lefever equations are, and what they are not.

A dual-pumped scalar LLE provides a mathematically rigorous route to multi-mode interaction without introducing multiple envelopes. The equation

$\chi^{(2)}$0

describes simultaneous pumping of two different modes $\chi^{(2)}$1 and $\chi^{(2)}$2. After the traveling-wave reduction $\chi^{(2)}$3, the problem becomes a stationary periodic boundary-value problem. The paper proves existence and uniqueness theorems, establishes global continuation from one-mode states into the two-mode-pumped regime, and shows that the second pump selects specific phase shifts of preexisting patterns through explicit solvability conditions (Gasmi et al., 2022). This is not an explicit vector CLLE, but it is directly relevant to coupled-LLE thinking because the two externally driven Fourier modes are nonlinearly linked by Kerr convolution throughout the spectrum.

A second neighboring model is the single-field LLE with localized pump and loss in a ring resonator: $\chi^{(2)}$4 Although it is not a coupled-LLE system, it shows how a single cavity field with broad Fourier forcing can support multiple tilted nonlinear resonances, multistability or tristability, and coexisting solitons with different amplitudes and widths associated with spectrally distinct combs (Kartashov et al., 2017). For coupled-LLE theory, its significance is indirect: some branch-coexistence phenomena often attributed to explicit multi-envelope models can already arise in a single-field multimode mean-field description.

A more literal coupled precursor system appears in the cavity model for electric and magnetic amplitudes in a nonlinear material. The equations are

$\chi^{(2)}$5

$\chi^{(2)}$6

The key result, however, is reduction rather than persistent two-field dynamics: defining $\chi^{(2)}$7, the paper proves $\chi^{(2)}$8, so $\chi^{(2)}$9 after a transient and the system collapses to a scalar generalized LLE with effective nonlinear coefficient $|Q_\pm|^2Q_\pm$00 (Mártin et al., 2012). This is therefore a coupled-LLE precursor that reduces exactly to a one-field model under the stated assumptions.

The $|Q_\pm|^2Q_\pm$01 ring-resonator work, by contrast, remains genuinely two-field at the mean-field level. Starting from Maxwell’s equations and multimode coupled-mode sums, it derives envelope equations for the ordinary fundamental field $|Q_\pm|^2Q_\pm$02 and extraordinary second harmonic $|Q_\pm|^2Q_\pm$03, with arbitrary quasi-phase-matching profile $|Q_\pm|^2Q_\pm$04, walk-off through $|Q_\pm|^2Q_\pm$05, and self-steepening-like corrections. In the regime $|Q_\pm|^2Q_\pm$06, the second-harmonic field becomes slaved to the pump and induces an effective cubic self-action; for the lithium-niobate example with $|Q_\pm|^2Q_\pm$07 GHz, the estimated cascaded Kerr coefficient is $|Q_\pm|^2Q_\pm$08, about three orders of magnitude larger than the intrinsic $|Q_\pm|^2Q_\pm$09 (Skryabin, 2020). This provides a concrete bridge between explicit coupled LLE-like equations and effective scalar Kerr descriptions.

Taken together, these neighboring models delimit the notion of a coupled Lugiato–Lefever equation. Some systems require a persistent multi-envelope description because the interacting degrees of freedom remain dynamically independent at mean-field level, as in avoided crossings, bidirectional propagation, and doubly resonant $|Q_\pm|^2Q_\pm$10 cavities. Others only mimic coupled-LLE phenomena through multimode forcing or collapse onto a scalar equation after a transient. This suggests that the central criterion is not simply the presence of more than one physically distinguishable wave, but whether the mean-field reduction retains more than one indispensable dynamical envelope.

7. Physical significance, common misconceptions, and research directions

The coupled-LLE framework is often associated narrowly with Kerr microcombs, but the literature shows a broader range. It encompasses Kerr hybrid-mode problems near avoided crossings, bidirectional rings with backscattering, directly linearly mixed cavity modes, and quadratic cavities with resonant harmonic generation (D'Aguanno et al., 2017, Skryabin, 2020, Skryabin, 2020). The unifying theme is a driven-dissipative cavity with periodic geometry and nontrivial modal interaction, not a single choice of nonlinearity or coupling.

A common misconception is that every two-mode cavity requires an explicit linear coupling term between the two LLE fields. The avoided-crossing paper shows why this is not generally true: when the equations are written in the hybrid-branch basis after diagonalizing the linear guided-mode interaction, the coupled mean-field dynamics contains nonlinear cross-coupling but no explicit residual $|Q_\pm|^2Q_\pm$11 term (D'Aguanno et al., 2017). Conversely, the one-sided localized-pump model demonstrates a regime in which direct linear coupling is the defining mechanism and is precisely what creates the unpumped localized partner mode (Cardoso et al., 11 Feb 2025).

A second misconception is that two counter-propagating combs necessarily engage in strong mode-by-mode nonlinear mixing. The bidirectional hierarchy shows that phase-sensitive nonlinear mixing between clockwise and counterclockwise comb teeth is largely suppressed by fast repetition-rate dephasing, so that after averaging the dominant surviving nonlinear interaction is an average-power-induced detuning shift rather than full coherent four-wave mixing between all opposite-direction teeth (Skryabin, 2020).

A third misconception is that branch coexistence or multi-comb behavior always proves the existence of a vector mean-field model. The localized-pump single-field LLE and the dual-pumped scalar LLE both show that multiple resonances, coexistence of distinct soliton families, and rich continuation structure can arise in a scalar equation when forcing projects onto several Fourier modes or resonances (Kartashov et al., 2017, Gasmi et al., 2022). This does not diminish the role of explicit coupled LLEs, but it changes the evidentiary status of some observed phenomena.

The current literature therefore presents coupled Lugiato–Lefever equations as a methodological family with several open interfaces. One interface is between exact modal or Maxwell descriptions and reduced mean-field PDEs, as emphasized by both the bidirectional Kerr hierarchy and the $|Q_\pm|^2Q_\pm$12 derivation (Skryabin, 2020, Skryabin, 2020). Another is between genuinely vector mean-field dynamics and scalar reductions, either asymptotic or exact, as in the electric–magnetic cavity model (Mártin et al., 2012). A further interface lies between externally imposed coupling, such as linear backscattering or one-sided pumping, and coupling that is generated by basis choice and modal hybridization.

For arXiv-level research practice, the main value of the coupled-LLE framework is therefore not a universal normal form, but a disciplined way of preserving the minimum number of cavity envelopes needed to represent the experimentally or computationally relevant interaction channel. In some problems that minimum is two hybrid branches; in others, two propagation directions; in others, a pump and its second harmonic; and in still others, the correct conclusion is that no persistent coupled-LLE description is needed because the dynamics is effectively scalar.