The project BREAK started with the development of novel experimental techniques that couples rock deformation and the use of X-rays to image deformations or measure directly stress and deformation inside samples. The main goal is to reproduce processes that occur at depth in the Earth’s crust where earthquake start, propagate and then stop. The work has been carried out through a collaboration between the University of Oslo and three beamlines at the European Synchrotron Radiation Facility (ID11 for X-ray diffraction, ID19 for ultra-fast imaging of shock wave in rocks, and BM18 for 4D in-situ imaging of rock deformation processes leading to system-size failure). We spent the first two years to develop the experimental techniques and publish the first results in a dozen of articles released in the best journals in geophysics. An important achievement has been to use machine learning techniques to predict the proximity to failure of rock cores while they are deformed until they break. Deformation occurs by the nucleation of microcracks in the samples and, as loading is increased, these cracks grow and coalesce until a system-size spanning fracture develop, leading to a laboratory earthquake. Using images of these fracture networks and a deep-learning algorithm we developed, we could show that not only the number of fractures controls the proximity to failure, but also how they are organized in space. Another important result is the development of a theoretical physical model of earthquakes with a minimal number of parameters. Such an approach is useful to unravel the basic mechanisms of earthquake propagation and arrest, and therefore identify which physical process provide a primary control on these processes.
