Research

Rate-and-state friction theory describes how friction evolves with slip. It has now been widely applied to explain frictional behaviors in both the lab and the tectonic scale. State variable is a key parameter in rate-and-state friction that describes how friction depends on the slip history. The underlying physics of the state variable, however, remains an active area of debate, because its current mathematical form was obtained by empirically fitting laboratory data.

In this project, we revisit the theory that explains the state variable as the real area of contact of the rough interfaces. The real area of contact Ar is usually way smaller than the apparent total contact area. The theory states that state dependence is due to creep or rejuvenation at the micro-contacts. We developed the existing theory, brought up a quantitative relation, and tested the relation with two published laboratory friction experiments: Dieterich and Kilgore (1994) and Svetlizky, Bayard, and Fineberg (2019). Our models explain the experiment data reasonably well, suggesting a valid constitutive relationship between the real area of contact and the size and age of micro-asperities at contact junctions. Our findings demonstrate a direct link between the state-variable and an observable quantity in the laboratory, providing new insights into the physical mechanisms underlying rate- and state-dependent friction laws.

Learn more in my May-2023 presentation slides @UCLA: 

Last updated Aug-28-2023

Slow earthquakes manifest as unstable fault slips, but their slip rates, rupture velocities, and radiation efficiencies are significantly lower than fast (regular) earthquakes. The classic model derived from fast earthquake observations cannot explain the first-order characteristics of slow earthquakes. 

Many recent studies suggest that these slower characteristics may be associated with structural heterogeneity within the fault zone, rather than specific friction or rheological properties. In this study, I theoretically investigate why a heterogeneous setup may lead to slow rather than fast earthquakes. I identify three features of a general frictional-viscous heterogeneous model that help generate slow earthquake characteristics.

First, the frictional components in the fault zone, even if only being a small portion, could pin the fault zone due to their spatially heterogeneous distribution. Consequently, the fault can still exhibit stick-slip behaviors as a whole. 

Second, in-between slip events, the frictional components that pin the fault bear most of the shear loading. The local shear stresses on the frictional components would be greater than the average shear stress due to stress concentration. As a result, the fault might fail more easily as a whole, leading to a shorter inter-event interval and a smaller average stress drop during events. 

Third, the seismic radiation from the frictional components may be too small and incoherent to lead the rupture propagation, even though seismometers may detect them. As a result, the large-scale rupture propagation is primarily facilitated by slip cascading of the bulk fault zone, as opposed to dynamic stress perturbation, leading to slow rupture speed and occasional "stress diffusion" behaviors. 

With analytical and numerical models, I provide some quantitative relations between the heterogeneous setup and its resulting slip behaviors. The results imply that a frictional-viscous heterogeneous model can potentially reconcile multiple geophysical observations to first order.

Learn more in my PhD dissertation and my JPGU 2022 presentation. 

Last updated Aug-28-2023

Earthquake scaling relations are commonly used in seismic and tsunami hazard analysis. These scaling relations implicitly assume the self-similarity of earthquakes regardless of their magnitude. However, for earthquakes that rupture close to the Earth's surface (a.k.a., the free surface), the self-similar assumption doesn't hold. Due to the Earth's surface-fault stress interaction, shallow earthquakes tend to slip more.

Shallow earthquakes pose great threats to humans because of their proximity to our society. It is, therefore, important to characterize the Earth's surface effect on earthquake scaling relations. Such effects have rarely been systematically quantified for megathrust earthquakes, probably due to a lack of computation power and a high-quality finite fault model database. We bridge this gap by conducting a large number of numerical models and analyzing the recently compiled finite-fault-model database SRCMOD (Mai & Thingbaijam, 2014). 

Our results may have implications for explaining the apparently depth-dependent source parameters, and may also help reduce the uncertainties in the seismic and tsunami hazard assessment for large megathrust earthquakes.

Learn more in our publication Wu et al., 2023, GRL and my Oct-2022 presentation slides @USC. 

Last updated Aug-28-2023

With the deployment of continental scale seismic arrays, seismologists can quickly locate the high-frequency seismic radiation sources and track the earthquake rupture propagation using a technique called back-projection. It is a signal beamforming technique application in seismology, and similar applications can be found in fields such as radar, wireless communication, and radio astronomy. Recent studies have proposed multiple advancements in improving the back-projection location. However, the physical interpretation of the amplitude of stacked high-frequency source radiations, which is commonly referred to as beam power, is still challenging since the analysis is not based on a forward model. 

In this project, we conduct synthetic experiments to investigate the physical significance of back-projection beam power. We find that beam power is mainly controlled by the spatial heterogeneity wavelength near the rupture front, rupture directivity, and the seismogram frequency. In addition, back-projection alone may be unable to distinguish which type or types of source heterogeneity are responsible for the signal. It is in contrast with some previous studies that link the beam power to the maximum slip rate (acceleration) amplitude near the rupture front. Based on the results, we develop a novel theoretical framework that can quantitatively interpret the frequency- and array-dependent back-projection results not only in our synthetic experiments, but also the 2019 bilateral rupture M7.6 New Ireland earthquake.

Learn more in our publication Li et al., 2022, and my Feb-2023 presentation slides @USC. 

Last updated Aug-28-2023

This project started in 2017. There is a debate over whether very low frequency earthquakes (VLFEs), which are considered a type of slow earthquake signal in the 0.02 - 0.05 Hz range, always coincide with and are colocated with tectonic tremors, another type of slow earthquake signal in the 2-8 Hz range. While many literature suggests this is the case, several reports have shown that VLFEs can occur independently of tremors. At that time, there were only a few models for VLFEs, and all predicted VLFEs with tremor signals. 

We were curious whether a model could generate VLFEs without tremors. We delved into a frictional-viscous model from Nakata et al. (2011), which was the only mechanical VLFE model available at the time. Using our own analytical and numerical tools, we analyzed their model and found that it is the frictional-viscous rheology in the model that causes slip to be slow, rather than the spatial mixtures of weakening and strengthening patches, as argued in the original paper. In particular, we discovered that a spatially homogeneous fault with frictional-viscous rheology could generate VLFEs without tremor signals.

Learn more in our publication Wu et al., 2019

Last updated Sept-30-2023

Seismic noises were found to drop significantly in many stations around the globe during the early stage of the COVID-19 lockdown. The UC Riverside Seismology Journal Club discussed this topic at that time and decided to have a look at the noise level of our nearby stations in southern California. 

To our surprise, many stations did not record a drop during the early stage of the COVID-19 pandemic. Of the 19 stations of the Southern California Seismic Network surveyed, we found that only five show a similar extent of drop in anthropogenic seismic noise comparable to the Christmas holiday break in 2019. This suggests that the human activity that caused seismic noise did not significantly reduce during the COVID-19 pandemic near most surveyed stations in southern California. 

We go on to explore why this is the case. A further analysis implies that the primary seismic noise source in southern California might be traffic, and the continuation of industrial traffic, such as cargo transportation, during the COVID-19 pandemic may be the reason why many stations did not record a noise drop.

Learn more in our publication Wu et al., 2021, and my Jan-2021 presentation sides @UCR. 

Last updated Aug-28-2023

The 2016 M7.8 Kaikoura, New Zealand earthquake is a puzzle in many ways. One of the outstanding questions is how the rupture propagates from the south to the north, as the initial field mapping of surface ruptures finds a large apparent gap between the sudden fault and the northern fault. Even though later studies seem to bring the gap closer, it is still large for rupture propagation. 

Several hypotheses are brought up for this question. Some workers suggested the southern faults first triggered the subduction interface beneath, and the megathrust slip triggered the northern faults (e.g., Furlong & Herman, 2017; Mouslopoulou et al., 2019). Some other workers suggested the wide jump occurred over crustal faults. One model belonging to this group suggested that the southern faults are actually connected with the northern faults through the off-shore Point Kean fault, via which the south rupture propagates to the north (Ulrich et al., 2019). Another model in this group suggested that the southern rupture jumps through the White Fault to the Upper Kowhai Fault and Jordan Fault (Ando & Kaneko, 2017).

We proposed a possibility that the southern rupture propagates to the north through crustal faults. It takes a zig-zag jump from the Hundalee-White thrust fault group to the Papatea-Point Kean fault group, and these two fault groups don't need to be connected, different from what is shown in Ulrich et al. (2019). Through dynamic rupture models, we show that such a zig-zag jump is a mechanically favored outcome when two thrust faults are parallel to each other.

Learn more in my Aug-2018 presentation slides @Brown. 

Last updated Aug-28-2023