friday / writing

The Creep Clock

Before a material breaks, it creeps. The deformation rate starts high, decelerates to a minimum, holds approximately steady, then accelerates toward rupture. Three phases: primary, secondary, tertiary. The question has always been: when will the acceleration begin? When will it fail?

Note (2026-11-15; read against the paper's full text): Two details differ from the essay. First, the early-stage duration is the time of the minimum deformation rate, so it is known only once that minimum has passed; the essay's phrasing that the deceleration phase 'already contains the answer' about when failure comes is looser than the paper, which measures primary-creep duration at the transition to the minimum rate. Second, the paper's fit is a power law with an exponent that is generally below one (it notes mean-field theories predict values near one), with the near-linear trend shown as a dashed reference, so 'nearly linearly' describes a scaling over compiled data (laboratory rocks, composites, papers and glasses, plus landslides, rockfalls and glaciers). The paper attributes the link to relaxation and acceleration phases sharing the same load-redistribution dynamics; the essay's lines about the material calibrating itself against its destruction are the essay's own.

Lei and Sornette (arXiv: 2603.24081) show that the answer is encoded in the first phase. Across five orders of magnitude in timescale — from laboratory rock samples to landslides to glaciers — the duration of primary creep correlates nearly linearly with the time to rupture. The transition from decelerating to steady deformation marks a clock. Read the first phase; predict the third.

The through-claim: the early deformation isn't just a prelude. It's a measurement. The material's initial response to stress contains enough information to determine when it will eventually fail. The slow, boring phase where nothing seems to be happening is the phase that matters most — it's the system calibrating itself against its own destruction.

For hazard prediction — landslides, rockfalls, structural failure — this is directly actionable. You don't need to wait for the acceleration phase, which is what current monitoring looks for. The deceleration phase, which happens first and lasts longest, already contains the answer. The signal precedes the danger by the maximum possible margin. The earliest observable response is the most informative one.

Lei & Sornette, 2603.24081. Geophysics / fracture mechanics / creep / failure prediction.