From stable periodic orbits to many-body chaos: doubly tunable prethermalization via engineering of an emergent band structure
arXiv:2607.12355
2026
Architecture
2 ideas extracted · analyzed Aug 30, 2026
What the math gives to ML
The paper constructs stable periodic orbits whose perturbations behave as quasiparticles with a tunable dispersion near a gapless momentum point. By combining interactions at several spatial ranges, coefficients cancel successive derivatives of the dispersion, producing a high-order flat band with dispersion proportional to |k-k_0|^W. If perturbations occupy a momentum window of width R, the prethermal lifetime scales as R^{-W}, creating a sharp quantitative link between spectral concentration, dispersion flatness, and long-lived dynamics. The transferable neural-network mechanism is spectral band engineering: construct residual, convolutional, graph, or state-space updates whose linearized mode dispersion has a deliberately high-order flat point, then test whether long-horizon retention follows the predicted power law.
Ideas from this paper
△ Mechanism confirmed, baseline not beaten
2026
Replace standard nearest-neighbor residual or recurrent mixing with a learned multi-range shift operator whose coefficients cancel low-order derivatives of its Fourier symbol at a selected momentum. This creates slow modes with dispersion of order W, which should preserve low-frequency information over longer horizons while retaining an explicitly measurable spectral signature.
Useful8/10
Difficulty5/10
Novelty7/10
✓✓ Beats tuned baseline
2026
Use the paper's lifetime law as a controller for training or rollout difficulty. Estimate the active perturbation bandwidth R of hidden states or forecast errors and reduce the residual gain, increase the dispersion order W, or inject controlled bandwidth whenever the estimated prethermal lifetime becomes too short.
Useful7/10
Difficulty5/10
Novelty8/10