Non-Normal Phase Transitions: A New Universality in Complex Systems

Abstract

We identify a new universality class of phase transitions that emerges in non-normal systems, extending the classical framework beyond eigenvalue instabilities. Unlike traditional critical phenomena, where transitions occur when eigenvalues cross zero, we show that the geometry of eigenvectors alone can trigger qualitative changes in dynamics. Within a large-deviation framework, transient amplification intrinsic to non-normal operators renormalizes the effective noise amplitude, acting as an emergent temperature. Once the non-normality index $κ$ exceeds a critical threshold $κ_c$--the balance between restoring curvature and non-normal shear--stable equilibria lose practical relevance: fluctuations are amplified enough to induce escapes even though spectral stability is preserved. This mechanism defines a fundamentally new route to criticality (pseudo-criticality) that generalizes Kramers' escape beyond potential barriers and can dominate noise-driven transitions in natural and engineered systems. In biology, we demonstrate that DNA methylation, a cornerstone of epigenetic regulation, operates in this regime: by extending a bistable CpG dyad model to include non-normality, we reconcile long-term epigenetic memory with rapid stochastic switching observed on minute timescales. More broadly, the same mechanism underlies abrupt tipping in climate, ecological collapse, financial crises, and engineered network failures. By showing that phase transitions can arise from non-normal amplification rather than spectral instabilities, our work provides a predictive framework for sudden transitions across disciplines.