Cortical bursts of gamma activity that lock to the phase of slower theta oscillations — a pattern called PAC — are a frequent signature in memory tasks and are altered in epilepsy. The authors set out to explain how PAC can arise from generic, collective dynamics rather than from specialized cells or dedicated microcircuits. They began from a generalized Langevin equation describing a few collective variables; in this picture friction and random forces stand in for unobserved, microscopic neural activity. The investigators then allowed the model’s stiffness and friction to depend on the current theta and gamma state, and they expanded those dependencies to first order to enumerate possible interactions.
Expanding stiffness and friction to first order produced ten possible interaction terms. The study team reported that only the two friction-type terms change the energy of the gamma oscillator and therefore can produce PAC. The sign and form of those friction-type couplings determine whether gamma bursts align with theta peaks, troughs, or the rising or falling phases. In other words, a simple change in a friction coefficient can reposition the preferred phase of gamma bursts relative to theta.
The authors reduced the model analytically so that the PAC strength becomes a closed-form function of a single dimensionless number they call kappa. Noise and damping transform, via the fluctuation dissipation theorem, into an effective theta temperature that enters this expression. Because every coefficient in the reduced equations enters linearly, the authors noted these coefficients could, in principle, be estimated from electrophysiological recordings by linear regression — a pragmatic point for experimentalists seeking data-driven parameter fits.
As a worked example, the team studied a friction term in which gamma friction follows theta velocity. Since the time integral of theta velocity is the theta coordinate itself, gamma amplitude acquires a multiplicative factor that scales as exp(gvvxθ/2) in the simplified derivation reported. That analytic dependence ties gamma amplitude directly to theta coordinate, and it provides an explicit mechanism for bursts to concentrate at particular theta phases without invoking specific neural subpopulations.
The preprint documents extensive numerical validation. Without fitting free parameters, the derived closed-form prediction accounted for 51,095 simulated records with a reported R2 = 0.993. The same formula also explained the median PAC measured across 5,160 combinations of coupling and damping, with a reported R2 = 0.989. Investigators presented these fits as evidence that the single-dimensionless-number description captures the core relationship between velocity-dependent friction and PAC strength across a broad parameter range.
Using the fluctuation dissipation theorem, the authors turned noise and damping into an effective theta-temperature that governs gamma amplitude fluctuations. They also reported that the coupling drives the gamma oscillator out of equilibrium: its kinetic temperature, as defined by the model, grows much higher than the temperature inferred from linear response, indicating excess kinetic energy injected by the coupling. That nonequilibrium signature parallels observations in other driven oscillator systems and offers an interpretable metric for departures from thermal-like behavior.
Because the reduced equations contain coefficients that enter linearly, the study team argued these parameters can be estimated directly from data using linear regression on appropriately chosen collective variables. The authors suggested this route would let experimentalists test whether recorded PAC reflects the friction-type interactions isolated in the model, although the preprint does not present a direct application to empirical recordings.
The article is a theoretical preprint and has not undergone peer review; the authors declared no competing interests. The source did not report empirical fits to biological recordings or detailed procedures for extracting the collective variables from experimental data. The companion paper is referenced as linking the same dimensionless number to a stability boundary, but methodology and results from that follow-up are not reported here. Remaining open questions in the preprint include how to map microscopic circuit elements to the effective friction and how robust the linear-regression parameter estimation will be when applied to noisy, real-world recordings.