Abstract
This paper establishes a rigorous mathematical framework for simplifying seismic fault models. By applying Gamma-convergence, the authors demonstrate that thin, high-strain fault zones can be accurately represented as 2D sharp interfaces. This reduction from volumetric deformation to a displacement jump significantly lowers the computational complexity of earthquake simulations while maintaining physical accuracy.
Introduction
Simulating ground motion during an earthquake is a classic multiscale challenge. Geologists model the Earth as an elastic solid, but the critical action occurs within "fault zones"—thin regions where rock ruptures under intense strain. To achieve accurate simulations, researchers traditionally employ incredibly fine meshes to resolve the massive deformations inside these narrow faults. When a fault is only meters wide within a simulation covering kilometers, the number of elements in a numerical model explodes, making computation prohibitively slow.
This work seeks to mathematically justify a "limit model" where the fault is treated not as a volume, but as a mathematical surface (an interface) where the displacement simply "jumps" from one side to the other. This transition from a 3D volumetric problem to a 2D interface problem offers a path toward drastically more efficient seismic modeling.
Methods
The authors employ a combination of variational analysis and numerical simulation to bridge the gap between volumetric and interface models:
- Gamma-Convergence: This core mathematical tool is used to define an energy functional for a model with a fault of width $\epsilon$. The authors prove that as $\epsilon \to 0$, this functional converges to a limit functional representing the "sharp interface" model.
- Scaling Laws: A key contribution is identifying the correct scaling for the shear modulus. They demonstrate that for a non-trivial limit to exist (one where the fault actually slides), the shear modulus within the fault must scale proportionally with the fault's width.
- Evolutionary Equations: The proof is extended from static energy to time-dependent momentum equations, which describe the actual wave propagation of an earthquake.
- Numerical Validation: The theory is verified using the Finite Element Method (FEM) with piecewise quadratic elements, comparing the "resolved" thin-fault model against the simplified "limit" interface model.
Key Insights and Results
The authors identified a "mathematically interesting" case: if the fault material is too stiff or too soft relative to its width, the model either "locks" (preventing sliding) or "decouples" (the sides lose interaction). The identified scaling allows for a displacement jump that perfectly captures the physics of a rupture. In this limit model, permanent deformation is elegantly handled as a jump condition in horizontal displacement.
Computational Efficiency
The numerical examples highlight the efficiency of the limit model. To resolve a thin fault with a standard mesh, a model might require over 1.5 million unknowns. In contrast, the limit model provides a highly accurate approximation on modest meshes with only a fraction of that computational load. The FEM implementation demonstrated optimal third-order convergence rates in $L^2$ norms, proving the robustness of the numerical scheme.
Dislocation Robustness
The model successfully captured the behavior of a singular "edge dislocation," a classic benchmark describing stress fields around crystal defects. This confirms the model's ability to handle the extreme stress concentrations typically found at the tip of a rupture.
Limitations and Future Work
While providing a rigorous foundation, the current analysis assumes the location of the fault is known a-priori and primarily treats the coefficients as linearized. The model does not yet address crack nucleation (where a rupture starts) or propagation (how it grows). Furthermore, future iterations of this variational technique will need to explicitly account for complex friction laws and the physical constraint of non-interpenetration, ensuring the two sides of a fault cannot mathematically push through each other.
Summary updated