Projects |

Earthquake complexity and scaling laws

Mon, 03 Aug 2026 00:00:00 +0000

Universal scaling laws lie at the core of earthquake science, and they provide an essential bridge across subdisciplines. Are large earthquakes simply “scaled up” versions of smaller earthquakes - as implied by self-similar models - or are there processes or rupture patterns that only emerge at larger scales?

We’ve been addressing this question from the perspective of fracture mechanics, by developing theoretical models that describe how earthquake initiation, propagation and arrest arise from the interplay of friction and elasticity. We demonstrated that energy balance criteria predict that the timing and size of seismic events is controlled by the dimension of a fault relative to a characteristic length arising from frictional and elastic properties; for sufficiently large faults, this leads to power-law distributions commonly observed in earthquake catalogs ( ). On the other hand, small faults can rupture in simple, quasi-periodic sequences of identical events. We have developed theoretical arguments predicting the recurrence intervals and its scaling with magnitude, in agreement with observations of small repeating earthquakes worldwide ( ).

More recently, we demonstrated that fault dimension also has a profound effect on the rupture process of individual earthquakes, with an increase in rupture complexity and high-frequency radiation for larger earthquakes ( ). A remarkable finding of this study is that rupture complexity is an intrinsic property of laterally propagating earthquakes: the widely used earthquake crack model, which assumes uniform residual stress and predicts a simple rupture evolution, is fundamentally incompatible with lateral propagation on velocity-weakening faults. Instead, ruptures that exceed ~100 nucleation lengths experience back-propagating fronts, similar to the “boomerang earthquakes” increasingly reported in observational studies.

These efforts complement our work on fault geometry and heterogeneity, which provide a distinct mechanism for slip complexity - see .

Media coverage: on Eos; on MIT News; on Eos ·

Related publications: · · ·

Seismic cycles on geometrically complex faults

Mon, 03 Aug 2026 00:00:00 +0000

Faults are not planar features, but instead exhibit geometrical roughness at all scales, and are surrounded by highly fractured rock, which hosts small earthquakes and modifies the stress state on the main fault. In my group, we have worked both on the physical origin of off-fault damage ( ), and on the effect of fault roughness and geometrical complexity on fault slip and seismicity patterns. We showed that roughness primarily controls fault dynamics by introducing spatial variability in normal stress, which modulates both the characteristic timescale for frictional failure and slip stability ( ). We have also explored the role of pore-fluid pressure perturbations in this setting ( ), and the interplay between roughness and surface wear ( ). The effect of fault roughness is further amplified by the interplay between slip and friction, resulting in complex slip patterns during slow slip events ( ).

Accounting for this structural complexity significantly improves physics-based earthquake forecasts ( ; ; see also ), and recent advances in earthquake detection techniques create opportunities to probe the physical processes driving seismic swarms, foreshocks and aftershock sequences. With support from an NSF CAREER award, we have implemented new numerical techniques to efficiently and accurately model earthquake cycles of a major fault embedded in a damage zone by employing a 2.5D representation of damage faults, that captures stress interactions in a 3D volume at a reduced computational cost; as well as implementing local proxies for inelastic stress relaxation due to plastic yielding and viscoelasticity, under which the system evolves toward a well-defined limit cycle without unphysical stress concentrations.

Media coverage: on phys.org · on EAPS News

Related publications: · · · · · ·

How do earthquakes start?

Mon, 03 Aug 2026 00:00:00 +0000

The seismological signature of earthquake nucleation. Elasto-frictional theories, confirmed by laboratory experiments, predict the existence of a finite nucleation dimension over which slow slip transitions to dynamic rupture. However, seismological observations remain rare and controversial, largely because small earthquakes are typically observed in the far field. In order to better constrain the nucleation process from far-field observations, I have developed a source model for earthquakes near the nucleation dimension ( ), which predicts systematic deviations from earthquake scaling laws typically assumed in seismology. This model offers the opportunity to interpret deviations from self-similar scalings in terms of the nucleation process, and potentially produce in-situ estimates of the nucleation dimension.

Linking interseismic friction evolution to earthquake nucleation. Classical nucleation models fail to account for the evolution of friction and sliding arising from heterogeneity in fault properties and loading conditions. To this end, I am developing a more comprehensive model of the seismic cycle that explicitly accounts for heterogeneity and time-dependent loading, and treats earthquake nucleation as the final stage of a gradual, interseismic slip evolution (currently under review). I find that the energy budget that controls earthquake initiation, propagation and arrest is profoundly affected by the gradual evolution of friction during the interseismic period, and a friction evolution law supported by recent laboratory experiments (the “slip law”; Bhattacharya et al., 2022) produces a feedback whereby friction increases rapidly ahead of a propagating rupture. This stabilizes slow slip and delays nucleation; on the other hand, once an earthquake begins, this feedback is suppressed, producing large earthquakes.

Finally, fault geometry provides another mechanism for precursory slip: on rough faults, normal stress heterogeneity can produce accelerating aseismic creep and foreshocks ahead of large ruptures ( ). See for more on this mechanism.

Related publications: ·

Slow and fast slip at subduction plate boundaries

Mon, 03 Aug 2026 00:00:00 +0000

The world’s largest earthquakes occur in subduction zones, and their size is affected by depth-dependent rheological transitions as well as structural heterogeneity. The factors controlling their initiation and final magnitude remain largely unknown. In my group, we have explored the complementary role of depth-dependent rheology ( ) and stress heterogeneity ( ) in determining the interplay of slow deformation and earthquakes in subduction zones.

Earthquakes are mostly confined to the uppermost few percent of the Earth, due to a transition from brittle to viscous deformation with increasing pressure and temperature at depth. The extent of the “seismogenic zone” is key to seismic hazard, as it determines the maximum area and landward reach of earthquakes in subduction zones. Taking a global perspective on shallow (0-70 km) seismicity, we show that the seismogenic zone consistently extends to approximately 50 km, independent of temperature ( ). This unexpected finding challenges prevailing models of a temperature-controlled brittle-viscous transition.

Within the seismogenic zone, slip along subduction plate boundaries spans a continuum from fast earthquakes to steady creep and slow transients. We show that geometrical heterogeneity along the plate interface, such as seamounts on the downgoing plate, can generate a broad range of slip behaviors ( ).

Related publications: · · .

Earthquake triggering and operational earthquake forecasting

Mon, 03 Aug 2026 00:00:00 +0000

Aftershock sequences take place after all moderate and large earthquakes, and are a significant source of hazard. Physics-based models for Operational Earthquake Forecasting bring our physical knowledge of elasticity and friction into time-dependent, probabilistic earthquake forecasts, but face significant challenges ( ).

Our approach to developing physics-based earthquake forecasts has been guided by an attempt to construct physically consistent and realistic models of the processes involved, by including time-dependent (aseismic) fault slip and a realistic fault geometry. We found that stress heterogeneity due to the geometrical complexity of a fault system has a first-order impact in model behavior ( ), and it significantly improves performance ( , ). Moreover, deep afterslip following large subduction earthquakes significantly contributes to triggering seismicity on shallow crustal faults ( ). We have also explored earthquake triggering due to other transient stress sources, from viscoelastic postseismic relaxation in Southern California ( ) to rainfall-driven, tidally modulated triggering at Mt. Hochstaufen, Germany ( ).

Media coverage: on phys.org

Related publications: · · · · · · · · ·