Can pure high-even-order dispersion solitons form in multimode fiber lasers?

PHEOD solitons balance nonlinearity against one high even-order term

A conventional soliton in a fiber laser balances Kerr self-phase modulation against anomalous second-order dispersion, β2. A PHEOD soliton replaces that second-order term with a negative even-order term — β4, β6, or β8 — while β2 is held small and positive and β3 is suppressed. The paper's stated motivation is scaling: PHEOD solitons follow an energy–duration law E ∝ τ^-(k-1), steeper than the ordinary soliton relation, which the authors link to the soliton area theorem that limits pulse energy in standard fiber lasers.

The obstacle in multimode fiber is control. Different transverse modes see different dispersion, and because their spatial fields overlap, a spectral pulse shaper cannot address them independently. The paper's answer is a transverse mode division control (TMDC) module: a mode multiplexer/demultiplexer physically separates the modes into distinct single-mode channels, each of which then passes through a programmable spectral pulse shaper. That separation is what makes mode-by-mode dispersion imprinting possible.

A unified phase mask pushes fiber dispersion aside

The dispersion engineering uses a two-step process. First, the shaper corrects the dispersion of the fibers and intracavity components for each spatially resolved mode, setting net β2 to a small positive value and suppressing β3 to negligible levels. Second, it superimposes strong negative high-even-order dispersion. The applied higher-order dispersion exceeds the fiber's own higher-order dispersion by several orders of magnitude, which the authors use to argue that soliton formation is governed by the intended term interacting with Kerr nonlinearity rather than by the fiber.

The hardware imposed one compromise: a single uniform phase profile was applied to all mode channels, so the intrinsic dispersion differences between modes left small residual β2 and β3. The authors report that sensitivity analysis and comparisons of spectral-phase and group-delay contributions show these residuals do not compromise PHEOD soliton formation. This is the first place where the reader should note that the mode channels were not perfectly equalized.

Sidebands and energy scaling confirm each mode separately

The authors first ran the laser with only one mode channel connected at a time. Mode-locking thresholds rose with mode order: 1.1 W for LP01, 1.19 W for LP11, 1.35 W for LP02, and 1.56 W for LP21. The clearest spectral signature was strong resonant sidebands, analogous to Kelly sidebands in ordinary soliton lasers. When the k-th power of the sideband frequency shift was plotted against sideband index m, the relation was linear across all four modes under β4, β6, and β8 dominance — the test the paper uses to argue that a single k-th order term governs the pulse.

The exception matters. Under β8, higher-order modes, particularly LP21, deviated from the linear trend at higher sideband orders. The authors attribute this to the residual β2 and β3 left by the uniform-mask strategy, citing prior work on sideband position discrepancies. Pulse energies in single-mode operation ranged from about 50 to about 400 pJ, and the measured energy–duration relation agreed with theoretical scaling; the comparison baseline was a conventional β2 soliton measured in the same cavity, with sideband energy subtracted by spectral integration. Data were kept below the pulse shaper's 3 nJ damage threshold.

Four modes lock together at a shared repetition rate

With four modes running together, the laser mode-locked at a lower pump threshold of 870 mW and was characterized at 1.6 W. The demultiplexed outputs showed stable pulse trains in every mode, and the radio-frequency spectra of the individual modes peaked within 1 kHz of one another at a repetition rate near 25 MHz. The paper describes this as synchronization to the 1-kHz level, comparable to reported spatiotemporal mode-locking results.

The multimode state is not simply the four single-mode states side by side. The authors report an abrupt change in spatial, temporal, and spectral properties on entering the pulsed state, including a change in beam profile between continuous-wave and pulsed operation, and they argue this can only be explained by nonlinear interactions between transverse modes. Mode-resolved spectra also shifted systematically: low-order modes moved to lower frequencies and higher-order modes to higher frequencies, a pattern the authors say helps equalize modal velocities.

Modal dispersion alone cannot explain the locked state

The TMDC compensates modal dispersion at the end of each round trip, but the few-mode fiber separates pulses in different modes by 5–10 ps as they traverse the cavity — more than the roughly 1 ps pulse duration. Without intermodal coupling, the modes would behave as separate cavities with distinct repetition rates and no fixed temporal relationship. The paper therefore treats nonlinear coupling as necessary, not incidental.

The proposed mechanism has two parts. Cross-phase modulation pulls the modes together in time over the limited overlap regions, and the saturable absorber, where synchronized pulses produce deeper saturation and less loss, reinforces the alignment. Numerical simulations of linear propagation and steady-state operation reproduced the measured spectra, which the authors take as support for the simulated pulse evolution. The simulations also reproduced the suppression of resonant sidebands seen in multimode operation, though less dramatically than in experiment.

The synchronized state falls short of full mode-locking

The authors are explicit that they do not claim spatiotemporal mode-locking (STML), which requires phase-locking of transverse and longitudinal modes into a coherent wave packet traversing the cavity. Because pulses in different modes separate by 5–10 ps each round trip before being recombined, temporal overlap is partial, which the paper says hinders strong phase-sensitive interactions. They also do not claim true phase-locking of the distinct modes.

What the laser does show are some STML-like signatures: single spatiotemporal pulses emerging spontaneously, a reorganization of the multimode field at the onset of synchronization, and nearly transform-limited temporal profiles. The authors position the work as a step toward STML in PHEOD systems and suggest that larger chromatic dispersion, better modal dispersion control, or normal dispersion supporting longer chirped pulses could extend it.

Multimode operation raises energy but not at higher dispersion orders

The energy comparison is the paper's most decision-relevant number set, and it cuts both ways. Four-mode pulses carried more energy than their single-mode counterparts despite a lower pump level: at β4 = −1.44 ps4, the four-mode state reached 185.2 pJ versus 101.0 pJ for LP01 alone. But within each configuration, pulse energy fell as dispersion order and magnitude increased. At a fixed pump, four-mode energy dropped from 206.8 pJ at k = 4 and −0.96 ps4 to 66.1 pJ at k = 8 and −1.96 ps8.

So the multimode advantage is real but conditional. It appears at a given dispersion setting, not as a blanket gain across the β4, β6, β8 ladder, and the highest-order, largest-magnitude regime produced the lowest four-mode energy in the table. Readers comparing PHEOD scaling claims should keep these two axes — mode count and dispersion order — separate.

A platform result, bounded by its hardware

The paper's contribution is a working platform: physically decoupling transverse modes lets a spectral pulse shaper impose PHEOD conditions that were previously confined to single-mode systems, and four modes then self-organize into a synchronized pulsed state through nonlinear coupling. The evidence for that claim rests on sideband linearity, energy–duration scaling, matched simulations, and mode-resolved synchronization measurements, with the β8 sideband deviations as the visible imperfection.

The boundaries are stated by the authors. The uniform phase mask prevents precise compensation of residual intermodal dispersion differences, which is the most likely source of the β8 deviations. Partial temporal overlap means the state is not STML and not phase-locked. And the demonstration is at 1550 nm in a four-mode erbium-doped cavity with a 24.95 MHz fundamental repetition rate, with pulse energies held below the shaper's 3 nJ limit. The priority the paper sets is therefore not maximum energy but establishing that PHEOD soliton physics survives in a multimode cavity — and the next test is whether better modal control can turn partial synchronization into genuine STML.

Sources used in this answer