Two papers, separated by twenty-seven years and three fields, converge on a structural claim: the lifetime of a trapped state depends on the geometry of the exit, not the depth of the well.
Heller and Limmer study spatial patterns in reaction-diffusion systems — the stripes, spots, and spirals that form when chemical species react and diffuse on surfaces. These patterns compete. One pattern can transition to another. The classical instinct is that the more “energetically favorable” pattern wins — the deeper minimum in some free energy landscape persists longer. But these systems are far from equilibrium. There is no free energy to minimize.
Instead, Heller and Limmer develop a nonequilibrium instanton framework and find that pattern stability depends on path entropy: the number and geometry of transition paths connecting competing patterns. A pattern with many narrow exits dies faster than one with few wide exits, regardless of how “deep” the pattern's basin looks by other measures. At finite particle numbers, entropic effects can reverse the stability ordering predicted by deterministic analysis. The exit topology controls the lifetime.
Van Nimwegen and Crutchfield study how evolving populations escape metastable states — plateaus in the fitness landscape where the population sits on a neutral network of equally fit genotypes. Escape could happen by jumping directly across a fitness barrier (the depth metaphor: higher barrier, longer wait). It could also happen by diffusing along a neutral path until the population finds a narrow passage to a higher-fitness region.
They prove that neutral-path escape is exponentially faster than barrier crossing, and that the escape time depends on the width of the entropy barrier — the logarithm of the number of genotypes at the bottleneck of the neutral path — not on the height of any fitness peak. Barrier width, not barrier height. The geometry of the bottleneck, not the thermodynamics of the well.
The structural parallel: both papers replace depth with exit geometry as the variable that controls persistence. In the reaction-diffusion case, exit geometry is path entropy (many transition routes = fast escape). In the evolutionary case, it is barrier width (narrow neutral bottleneck = slow escape). Both are topological properties of the exit structure, not energetic properties of the basin interior. This matters because basin depth is the default metaphor for stability throughout science. Deep wells are stable. High barriers protect. The entire language of "energy landscape" carries this assumption: you persist because you're deep, and you escape because something gives you enough energy to climb out. Both papers show that this metaphor fails when the landscape has nontrivial topology — when exits vary in number, shape, and connectivity rather than just in height. The practical implication is that measuring depth tells you less than you think. A deep basin with many exits is less stable than a shallow basin with one narrow exit. The evolutionary population escapes not by climbing but by threading. The chemical pattern collapses not by surmounting a barrier but by finding that too many paths lead away from it. In both cases, the engineer or biologist who measures only depth will be wrong about lifetime. The question that lingers: is depth ever the right coordinate for stability, or is it always a proxy for exit geometry that happens to correlate in simple landscapes? In one dimension — a ball in a valley — depth and exit geometry are the same thing. In high dimensions, they decouple. Most real systems are high-dimensional.