Physically-Based Fault Zone Constitutive Responses and Consequences for Earthquake Dynamics

基于物理的断层本构响应描述正被用来解决地震动力学中的关键问题。主要有两个问题：(1)在目前设计地壳地震序列模型的尝试中，利用实验引起的速率和状态本构参数的温度变化(从而深度变化)，已经注意到随着状态演化滑动距离L向实验室范围内的值减小，出现了重要的新特征。这些是聚集在孕震带底部的一批小事件的出现，以及这些事件产生的不均匀残余应力模式对大事件中地震辐射的早期阶段的影响。这一过程似乎很有希望解释许多大型事件中最初犹豫不决的辐射，也就是我们所知的“地震成核阶段”。为了在目前的计算机上适应这样的计算，L必须比实验室的值大得多，后者的数量级为10微米，因为所需的数值网格大小是(10^5数量级)乘以L的大因子。然而，随着L向这些值的减小，有趣的行为正在出现。本项目通过新的数值研究和简化模型的渐近分析相结合的方法解决了L的小范围问题，因为L的尺寸减小了，为解释和外推提供了基础。(2)热弱化效应被认为是在大地震的快速滑动期间发生的，导致有效摩擦系数从传播的破裂前锋首次到达断层上的一点时的实验室值减小到发生快速和大滑动时的低得多的值。这一问题的解决是在最初研究的基础上进行的，这些研究分别侧重于滑移的最早阶段、熔融发生之前和活跃的假玄武岩发育的成熟阶段。当滑移率较高(1m/S)时，粗糙接触处的闪蒸加热将是主要的热减弱过程，但总滑移量仍然很小，从某种意义上说，这是可以量化的。初步分析捕捉到了可用实验的一些特征。在发生部分熔化时，如何解决更大的滑移范围，正在形成一些想法。其中包括一种观点，即颗粒断层泥通过少量部分熔体的发展而液化，并随着这种高度加压的相渗透到相邻的断裂壁而发展出速度减弱特征的自我调节过程，这一观点得到了假硅质岩观测的支持。开发这些概念的目的是将它们与关于破裂模式如何取决于本构响应的数值模拟和理论相结合，以检验破裂动力学的后果。这将有助于理解主要断层系统如何在实际较低的总体驱动应力下运行，即使在局部引发滑动所需的应力要大得多的情况下也是如此，并有助于量化破裂的最低平均应力水平，一旦破裂在局部高剪应力或低有效正应力的位置开始，就可以传播很长距离。

Physically based descriptions of fault constitutive response are being used to address key problems in the dynamics of earthquakes. There are two main topics: (1) In current attempts to devise models of crustal earthquake sequences, using experimentally motivated temperature variation (hence depth variation) of rate and state constitutive parameters, it has been noticed that important new features emerge as the state-evolution slip distance L is decreased towards values in the laboratory range. These are the emergence of a population of small events that is clustered towards the base of the seismogenic zone, and the effect of the resulting heterogeneous residual stress patterns from those events on the earliest phases of seismic radiation in large events. This process seems promising to explain the initially hesitant radiation in many large events, known as the "seismic nucleation phase". To fit such calculations on present computers, L must be made much larger than laboratory values, which are of order of magnitude 10 microns, since the required numerical grid size scales with a large factor (of order 10^5) times L. Yet interesting behavior is emerging as L is reduced towards those values. This project addresses the small L range by a combination of new numerical studies, coordinated with asymptotic analysis of simplified models, as L is decreased in size, to provide a basis for interpretation and extrapolation.(2) Thermal weakening effects are thought to occur during rapid slip in major earthquakes, causing the effective friction coefficient to diminish from lab-like values, present when a propagating rupture front first reaches a point on a fault, to much lower values when rapid and large slip occurs. This problem is being addressed by building on initial studies that focus separately on the earliest phases of sliding, before melting occurs, and on the mature stage of active pseudotachylyte development. Flash heating at asperity contacts is expected to be the primary thermal weakening process when slip rates are high ( 1 m/s) but total slip is still small, in a sense that can be quantified. A preliminary analysis captures some features of available experiments. Some ideas are being developed on how to address the much larger slip range when partial melting occurs. These include a view, supported by pseudotachylyte observations, that a granular fault gouge becomes liquefied through development of small amounts of partial melt, and that a self-regulated process of velocity-weakening character develops as this highly pressurized phase permeates into the adjoining fault walls. These concepts are being developed for purposes of integrating them with numerical simulations and theory on how the mode of rupture depends on constitutive response, to examine consequences for rupture dynamics. That will contribute to understanding how major fault systems can operate at realistically low overall driving stresses, even when the stress needed locally to initiate slip is much larger, and to quantifying the minimum average stress level for which a rupture, once initiated at a location of locally high shear stress or low effective normal stress, can propagate over large distances.