Quantifying the Role of 3D Fault Geometry Complexities on Slow and Fast Earthquakes

Abstract

Traditional models of slow slip events (SSEs) often oversimplify fault geometry, yet imaging studies show that real subduction faults are segmented and complex. We investigate how fault interactions influence slip behavior using 3D quasi-dynamic earthquake sequence simulations of two parallel faults with uniform rate-weakening friction, accelerated with hierarchical matrices. Our results identify four slip regimes-periodic earthquakes, coexisting SSEs and earthquakes, only SSEs, and complex sequences-while a single planar fault under the same conditions produces only earthquakes. We quantify fault interaction using the maximum Coulomb stress induced on a target fault by unit, spatially uniform stress drop on a neighboring fault. Because the source stress drop is normalized, the metric depends only on geometry and is independent of friction and nucleation length, and it can be extended to arbitrary fault configurations. The occurrence of SSEs is confined to an intermediate range of interaction strength. We also reproduce the observed moment-duration scaling and show that it depends on event detection thresholds. These results demonstrate that complex fault geometry can naturally generate both slow and fast earthquakes through evolving traction heterogeneities.

Figures (10)

• Figure 1: (a) Step-over fault configuration. The mesh is shown exaggerated for visualization purposes; all simulations use a grid spacing of $L_b/3$ (b-c) Example of SSE-dominant regime and earthquake-dominant regime. Time evolution of the maximum slip rate. The bottom x-axis shows real time, while the top x-axis shows time normalized by the recurrence interval of a single fault with the same frictional properties. Green curve is on fault 1 and yellow curve is on fault 2. Grey dash lines are slip rate threshold for earthquakes and slow slip events (d) Spatial distribution of nomalized slip for slip sequences with selected events A, B, C, D in simulation showing in (b) and E, F, G, H in simulation showing in (c). Blue represents slow slip events and red represents earthquakes. Event A is slow slip event with full rupture. Event D, F and G are slow slip events with partial rupture. Event B and E are earthquakes with partial rupture. Event C and H are earthquakes with full rupture. Red and blue color shows normalized slip for earthquakes and slow slip events.
• Figure 2: Moment-Duration scaling across all simulations. Red denotes earthquakes exhibiting cubic scaling. Blue and black represent slow slip events identified with slip rate thresholds of $10^{-6} m/s$ and $10^{-8} m/s$, respectively, showing varying but predominantly linear scaling from $M \sim T^{1.5}$ to $M \sim T^{1.2}$
• Figure 3: (a) Definition of $\Lambda$ in relation to the geometric parameters L, D, and $L_f$ (b) The relationship between the fault interaction metric $\Lambda$ and the SSEs ratio $\chi_{sse}$. Different colored dots represent various friction parameters $a/b$. The dashed lines indicate that SSEs emerge within a specific range of $\Lambda$. Grey area shows the results with SSE-dominant slip catalog. (c) Fault interaction metric $\Lambda$ as a function of overlap and the distance between two faults, expressed as a fraction of the fault length. The positions of the symbols represent the geometrical configurations in our simulations. Red triangles indicate simulations with only earthquakes, while green circles indicate simulations with coexisting SSEs and EQs, with the size of each green circle corresponding to the SSEs ratio.
• Figure 4: Traction field evolution during four events (E, F, G, and H) on fault 2 in Model 2. The colormap indicates the ratio of shear to normal traction. The first column displays the traction field before the nucleation of each event, with red stars marking earthquake hypocenters and blue stars indicating SSE hypocenters. The second column shows the traction field after each event, with black contour lines representing the slip distribution. The duration of each event and the inter-event times are noted.
• Figure S1: A single-fault model exhibits periodic earthquakes ($a/b = 0.8$, $W/L_\mathrm{nuc} = 2$) with a recurrence interval of 3.73 years.