The Ecology of Physical Laws: A Map of Coexisting Mechanisms ✛ ✎

Original: "Physical Law Ecology: mapping multi-mechanism ecologies as the zeroth step of data-driven scientific discovery"
arXiv:2609.08536v1 · 2026-09-08 · arXiv non-exclusive · independent analysis · 2 min
Before searching for the equation, find out how many there really are.
In a rubber band stretched in the lab, three different elasticity mechanisms operate simultaneously. At small strains, Hooke's law prevails; at intermediate strains, entropic chain elasticity; at large strains, strain-induced crystallization. Each mechanism is described by its own equation, and they hand over influence smoothly rather than abruptly.

Since Newton's time, the scientific method has sought a single equation for a single phenomenon. Even modern symbolic regression, which automatically derives formulas from data, silently inherits this assumption: it looks for one optimal curve. But real systems switch between mechanisms as conditions change, much like an ecosystem shifts its dominant species with climate. The authors of the new work abandoned the quest for the "best" equation and instead asked: how many laws coexist at once?

The answer turned out to be remarkably consistent. Across four very different systems—elastomers, boiling, galaxy rotation, and droplet evaporation—the algorithm identified exactly three independent mechanisms. For rubber, these are the power law of crystallization, the root law of entropic elasticity, and Hooke's linear law; prediction improved by 72% over gradient boosting. For boiling, accounting for monotonic regime changes added 20% accuracy.

The most striking result came from 3269 measurements of 163 galaxies. The system independently recovered three gravitational laws: Newtonian, transitional, and a "constant acceleration floor." The characteristic acceleration a0, at which the Newtonian regime gives way to the flat one, differs by a factor of 2.2 between gas-poor and gas-rich galaxies. This rules out a single modified Newtonian dynamics curve with a significance of p < $10^{-34}$, turning the debate over MOND's universality from a philosophical one into a computational test.

In an evaporating droplet, the algorithm automatically detected the moment of contact line detachment, when the liquid stops wetting the surface. This regime transition was not specified in advance; the system inferred it from the data as a change in the dominant equation.

The method opens a new direction: not improving individual equations, but describing their coexistence. Instead of hard boundaries between regimes, a continuous field of weights is constructed, where each law has its own "territory" of influence. This approach applies to turbulence, material fatigue, and stellar evolution, as well as to debates about one versus several laws in biology and climatology. The next step is to test whether the number of mechanisms changes over time, like species succession in an ecosystem.

A candle, which we see as a single flame, is actually governed by four different equations: heat conduction in the wax, diffusion combustion in the blue zone, radiation in the bright part, and convection in the tongue above the flame.

w_i(s,c) = \frac{\exp(-n \cdot MSE_i / 2\sigma^2)}{\sum_j \exp(-n \cdot MSE_j / 2\sigma^2)}
The smaller the equation's error on the data, the higher its weight; normalization ensures the weights sum to one.
K^* = \arg\min_K \left[ \sum_{s,c} n_{s,c} \ln(MSE^*_{s,c}) + K \cdot p_{max} \cdot \ln(N) \right]
Balances descriptive accuracy against model complexity; the minimum determines the number of equations.
Concepts
principle of least action Kepler's first law quantum law ecological pyramid rule Cope's rule bayesian evidence neural network scientific discovery physics physics informed neural network control theory biology Gause's competitive exclusion principle geodesic equation statistical mechanics evolution quantum mechanics initial conditions Bayesian inference computational fluid dynamics time evolution