Abstract

Geodetic observations of megathrust events at plate boundaries (e.g. Chile, Japan) show weeks- to months-long slow-slip preceding some great earthquakes. Comparable long precursory phases are generally absent for smaller intraplate events. Does a long nucleation somehow forecast a large magnitude earthquake? Scaling laws tell us that earthquakes should be self-similar at all scales, yet, systematic differences emerge in how they nucleate, accelerate, and arrest. Motivated by contrasting observational results and a long-standing debate, I first review evidence from natural earthquakes, laboratory experiments and conflicting views on whether the initial growth of moment rate is scale-invariant across magnitudes or, instead, shows longer characteristic rise times for larger earthquakes. To interpret these observations, I revisit Griffith’s critical crack and explore a simple analytical solution for rupture acceleration based on the energy balance. This solution predicts a universal, self-similar breakout phase in which rupture velocity and acceleration scale with the initial nucleation size. I test and extend this result using 2D and 3D dynamic rupture simulations, including heterogeneous stress and complex nucleation geometries, and show that the scaling persists when expressed in terms of a characteristic length and a geometrically averaged limiting velocity. Laboratory photoelastic experiments on transparent materials further support the predicted scaling between nucleation size and acceleration. Finally, I explore how fault segmentation and self-affine stress heterogeneity may couple nucleation size to overall fault dimension, and discuss implications for probabilistic forecasting and earthquake early warning, particularly whether early moment-rate growth may encode information about eventual event size.