Spatial clustering of microscopic dynamics governs the slip avalanche of sheared granular materials

Abstract

Establishing quantifiable links between individual-particle dynamics and macroscopic response of granular materials has been a longstanding challenge, with implications in material science, geology and industry. Despite sustained efforts in uncovering generic features in both macroscopic flow and microscopic dynamics, further advance on the subject matter demands quantitative correlations to be established. We propose a 3D convolution neural network (CNN) to quantify the link between microscopic dynamics and macroscopic stress fluctuations, including both stress recharge (stick regime) and stress drop (slip regime). Through the model interpretation, microscopic dynamics is found to demonstrate distinctive spatial patterns in the stick and slip regimes, which root in the result of free volume-induced structural rearrangements and contact network dynamics, respectively. We conclude that the spatial clustering of microscopic dynamics governs the occurrence of slip avalanches and acts as the “fingerprint” of macroscopic stress fluctuation. The data-driven framework developed in this paper can be readily extended to other amorphous solids for building cross-scale relations, paving a new way to understand the complex behavior of amorphous solids.