Frictional power dissipation in a seismic ancient fault

The frictional power per unit area Q˙ (product of frictional traction τ and slip rate u˙ in MW m−2) dissipated during earthquakes triggers fault dynamic weakening mechanisms that control rupture nucleation, propagation and arrest. Although of great relevance in earthquake mechanics, Q˙ cannot, with rare exceptions, be determined by geophysical methods. Here we exploit theoretical, experimental and geological constraints to estimate Q˙ dissipated on a fault patch exhumed from 7-9 km depth. According to theoretical models, in polymineralic, silicate rocks the amplitude (< 1 mm) of the grain-scale roughness of the boundary between frictional melt (pseudotachylyte) and host rock decreases with increasing Q˙. The dependence of grain-scale roughness with Q˙ is due to differential melt front migration in the host rock minerals. This dependence is confirmed by friction experiments reproducing seismic slip where pseudotachylytes were produced by shearing tonalite at Q˙ ranging from 5 to 25 MW m−2. In natural pseudotachylytes across tonalites, the grain-scale roughness broadly decreases from extensional to compressional fault domains where lower and higher Q˙ are expected, respectively. Analysis of the natural dataset calibrated by experiments yields Q˙ values in the range of 4-60 MW m−2 (16 MW m−2 average value). These values, estimated in small fault patches, are at the lower end of broad estimates of Q˙ (3-300 MW m−2) obtained from frictional tractions (30-300 MPa) and fault slip rates (0.1-1 m/s) assumed as typical of upper crustal earthquakes.