Phase transitions are classified by order. A first-order transition has a discontinuity in the first derivative of the free energy — latent heat, a density jump. A second-order transition has a discontinuity or divergence in the second derivative — a diverging heat capacity, a diverging susceptibility. Third-order transitions, if they exist, should involve the third derivative. But third-order transitions have been theoretically contentious and empirically elusive, partly because no canonical criterion existed to identify them. The microcanonical inflection-point analysis (MIPA) detects them in the entropy surface, but whether these detections correspond to anything measurable in the canonical ensemble — the ensemble that experiments actually sample — has remained unclear.
Wang, Liu, Qi, Cui, Tang, and Di (arXiv:2603.09124, March 2026) close the gap. They construct a cumulant-ratio criterion — a combination of energy fluctuation cumulants — whose signed extrema identify third-order transitions directly from canonical measurements. The criterion requires no density-of-states reconstruction, no explicit entropy derivatives. It operates on the fluctuation statistics that any thermal measurement provides.
The central finding is what the criterion reveals about the nature of third-order transitions themselves. They come in two kinds: dependent and independent. Dependent third-order transitions are not separate phenomena. They are the fluctuation reorganization that accompanies a nearby first- or second-order transition. On the disordered side, they appear as precursors — the fluctuations begin restructuring before the main transition. On the ordered side, they appear as aftershocks — the fluctuations complete their rearrangement after the main transition has occurred. The third-order transition is the echo of the lower-order one, registered in the next derivative.
The authors validate this across three systems: the exact Onsager solution of the 2D Ising model, finite-size Potts models, and a driven nonreciprocal Ising model in a nonequilibrium steady state. In each case, the dependent third-order signatures bracket the main transition — one on each side, precursor and restructuring. The independent third-order transitions, when they occur, stand alone, unattached to any lower-order event.
The implication is structural: the hierarchy of phase transitions is not a stack of independent phenomena at successive derivatives. It is nested. A first-order transition generates second-order fluctuation signatures at its boundary (critical endpoints, spinodals), and third-order fluctuation signatures further out. Each order is a shadow cast by the orders below it — not additional physics, but the same physics viewed through a higher-order lens. The signal is nested because the transition is.
Wang, Liu, Qi, Cui, Tang, and Di, "Canonical Criterion for Third-Order Transitions," arXiv:2603.09124 (March 2026).