Poltorak's Landau-Ginzburg model treats sleep-stage switches like phase transitions

Physicist Alexander Poltorak has released a new theoretical framework on arXiv (2608.03000) that models sleep-stage transitions using Landau-Ginzburg phenomenology, a toolkit borrowed from condensed-matter physics. Rather than treating sleep stages as discrete labels, the paper proposes a continuous cortical-ordering coordinate, phi, inferred from EEG/PSG observables via a measurement model designed to avoid circularity. Each canonical boundary—wake-to-NREM, N1-to-N2, N2-to-N3, and NREM-to-REM—is represented as motion in an effective potential of a noisy, dissipative neural field. Poltorak's model yields distinct signatures: sleep onset emerges as a fold-like loss of wake stability, deeper NREM stages behave as continuous ordering crossovers, and REM appears as a candidate first-order desynchronizing switch, possibly with a tricritical mixed state within N3. The Ginzburg term adds spatial predictions absent from scalar models, such as growing correlation length and local-to-global recruitment across the cortex.

To validate these ideas, Poltorak ran time-dependent Ginzburg-Landau simulations that reproduce the proposed signature classes, then performed a synthetic classification experiment. The classifier distinguished six archetypes with a cross-validated accuracy of 0.49 ± 0.005, nearly three times the balanced baseline of 0.17, and showed little change under a noise-regime shift. The author is careful to note these results establish internal consistency and testability, not a proven taxonomy for human sleep. He specifies the exact evidence needed to distinguish bifurcations, coexistence, noise-driven escape, smooth crossovers, and scoring-induced discontinuities—questions that require transition-centered EEG validation before clinical or neuromodulation applications can be considered. The 23-page paper includes 5 figures and 7 tables, and is available as an HTML experimental preprint on arXiv.