When a brittle material is loaded to the limit of its strength, it fails by nucleation and propagation of a crack. The conditions for crack propagation are created by the concentration of a long-range stress field at an atomically sharp crack tip, creating a complex and strongly coupled multiscale system. This thesis reports the results of multiscale simulations of the brittle fracture of silicon on the (111) cleavage plane. The simulations are made possible by combining a quantum mechanical description of the processes taking place near the crack tip with a classical atomistic model that captures the long-range elastic relaxation. The ‘Learn on The Fly’ technique is used to couple the quantum and classical models, allowing accurate quantum forces to be combined with classical forces using a simple adjustable potential to give stable dynamics. The simulations predict that fracture is unstable on the (111) plane at low speeds; conventionally this has been thought of as the most stable crack plane. The instability is caused by a crack tip reconstruction which triggers a positive feedback ‘sinking’ mechanism leading to macroscopic, experimentally observable corrugations. Recent experiments have observed crack surface features consistent with these predictions. The instability is the first example in a crystalline material of a fracture instability which onsets below a critical velocity, and shows how subtle atomistic details at the crack tip can control the qualitative macroscopic fracture behaviour.