Skip to main content
Version: Latest

Rate-and-state friction

DES3D models faults as finite-width shear zones in a continuum. A rate-and-state friction (RSF) law sets the friction angle of the Mohr–Coulomb plastic model in every element. The friction coefficient then depends on how fast the element slips and on its slip history, so a fault can creep, heal, nucleate an instability and rupture within one simulation. The approach follows Tong and Lavier (2018); Herrendörfer et al. (2018) describe a related invariant formulation.

This page gives the friction law, how DES3D measures slip rate in an element, and how the law enters the plastic update. It then describes global velocity scaling, which lets one model run through slow loading and fast rupture, and the other numerical controls. A benchmark checks the law in simple shear, and the tutorials show it at work in earthquake cycles and slip events in a shear band.

Friction law​

The friction coefficient is

where is the slip rate, the state variable (a time), a reference slip rate, the characteristic slip distance, and and the direct-effect and evolution-effect parameters (Dieterich, 1979; Ruina, 1983). The reference coefficient is the tangent of the element's Mohr–Coulomb friction angle, so when the element slides steadily at .

The state variable evolves by one of two laws. The aging law (Dieterich, 1979; Ruina, 1983) lets a fault heal while it is stationary:

The slip law (Ruina, 1983) changes the state only while the fault slips:

Under both laws, steady sliding at rate gives and the steady-state friction

A material with is velocity strengthening: it strengthens as it slides faster and creeps stably. A material with is velocity weakening and can become unstable (Marone, 1998; Scholz, 1998).

mat.state_var_model selects the law:

ValueLawFriction
0 (default)No state evolution at the current slip rate
1Aging lawfull law
2Slip lawfull law

With 0 the friction follows the current slip rate instantly. This is steady-state friction and needs no .

Slip rate in a continuum element​

The law needs a slip rate, but an element carries a strain rate. DES3D offers two measures, chosen with control.rsf_slip_rate_projection_option.

Option 1: strain-rate invariant (recommended). The slip rate is the shear across one element width,

where is the deviatoric strain rate and is the element's smallest altitude: twice the area divided by the longest edge in 2D, three times the volume divided by the largest face in 3D. For simple shear at engineering rate , and , the velocity jump across the element. A shear band wider than one element shares its slip among its elements, each with its own rate and state. The measure is independent of rigid motion and of the reference frame. The DES v2 benchmarks use it.

Option 0: velocity projection (default, historical). The element's mean velocity is projected onto the direction of maximum shear stress, which lies at 45° to the principal stress axes. In 3D the projection is onto the plane of maximum shear. This measures the velocity of the element, not a velocity difference, so it depends on the reference frame: a block that translates rigidly reports slip. The option is kept so that earlier models reproduce.

Each rate is floored at m/s before it enters the law, so the logarithms stay finite in a locked element.

Coupling with the plastic model​

RSF is available as mat.rheology_type = elasto-plastic-rate-state-friction (alias elasto-plastic-rsf) and elasto-visco-plastic-rate-state-friction (elasto-visco-plastic-rsf). They are the elasto-plastic and elasto-visco-plastic models of the constitutive update with one change: the friction angle in the Mohr–Coulomb yield function is

A higher friction coefficient raises the yield stress. A slipping velocity-weakening element lowers its own strength, which concentrates the strain further. Cohesion, dilation angle and tension cutoff are unchanged. They still weaken with plastic strain between mat.pls0 and mat.pls1. The reference friction angle also weakens: follows mat.friction_angle0 to mat.friction_angle1 as plastic strain grows.

The normal stress that resists slip is the one in the Mohr–Coulomb criterion. It is not an input. It comes from the stress field, including lithostatic load and tectonic stress.

Order of the update​

In each element and each step, DES3D:

  1. computes from the current strain rate (option 1) or velocity (option 0);
  2. advances the state explicitly with the step , ;
  3. evaluates and ;
  4. performs the elasto-plastic (or elasto-visco-plastic) stress update with that friction angle.

With several materials in an element, , , and are averaged over the element's markers, as are the friction angles.

Bounds​

The update keeps the law finite in extreme states:

QuantityBound
to s
in the lawat least m/s
at least
in the slip lawat least ; a step that would make resets it to

The state starts at steady state at the reference rate, , so at the start. The initial PT equilibrium (ic.has_body_force_adjustment with control.has_PT) holds the state and friction, since it is an instantaneous response with no time for healing. The state is saved in checkpoints as friction state variable; the friction coefficient is recomputed from it on restart.

Global velocity scaling​

Why RSF needs it​

An earthquake cycle spans loading at plate rates, about m/s for hundreds of years, and slip near 1 m/s for seconds. With physical density, an explicit step is limited by the elastic wave speed to milliseconds, and a 300-year interseismic period would need about steps. Ordinary mass scaling fixes the scaled wave speed from the boundary velocity, . That keeps loading quasi-static, but a fault slipping times faster than the boundary then moves faster than the scaled waves, and inertia dominates the event.

Global velocity scaling (GVS) ties the mass scaling to the fastest motion in the model at every step. control.use_global_velocity_scaling = yes enables it, and DES3D turns it on for any RSF rheology.

Scaled wave speed and density​

At each step DES3D finds the largest speed (nodal or boundary) and sets

where is control.inertial_scaling, the bulk modulus, the physical density and the modulus chosen by control.mass_scaling_reference_speed (shear: , the default; bulk: ). The nodal masses are rebuilt from every step.

Below the ceiling, the prescribed scaled speed is . This helps separate elastic relaxation from deformation, but does not establish mechanical equilibrium during acceleration or an instantaneous load. Check force residuals, deformation and physical-step convergence for the loading path. The ceiling is a speed used to construct fictitious mass, not a guarantee of the complete physical wave spectrum. At the bulk ceiling the dry inertial density reaches ; at the default shear ceiling it remains .

Time step​

For automatic timesteps (control.fixed_dt = 0), the elastic contribution is

where is the element's smallest altitude and . Other limits, including motion, Maxwell relaxation, heat/fluid transport and the RSF state limit, can reduce the final timestep; control.dt_fraction multiplies that minimum. Coupled fluid inertia has an additional undrained-wave limit.

Only while the first term controls does equal . Once the floor controls, that displacement estimate no longer holds. Neither the floor nor a velocity ceiling guarantees an accurate or stable coupled solution for arbitrary loading.

For a homogeneous dry example with m, , GPa and kg/m³ (shear speed 3.3 km/s), ignoring the other timestep limits and taking dt_fraction = 1:

StageCapped speed used for mass
Interseismic loading m/s m/s s (6 days)
Fast slip1 m/s3.3 km/s s

The automatic step can therefore span many orders of magnitude. GVS still uses its selected fictitious mass and damping; reaching its ceiling is not an automatic switch to undamped physical coseismic inertia.

Damping​

Force damping (control.damping_option, default 1 with control.damping_factor = 0.8) acts in every step, including during events. Ruptures in DES3D are therefore damped, as in the quasi-dynamic simulations of Tong and Lavier (2018). control.damping_option = 0 removes the damping.

Output during events​

With GVS, regular output follows sim.output_time_interval_in_yr or sim.output_step_interval as usual, and events add frames every sim.earthquake_output_step_interval steps (see earthquake output). A long cycle model therefore saves sparse interseismic frames and dense coseismic frames.

Limiting the state change per step​

The explicit state update is accurate only if the slip in one step is small compared with . While the inverse-speed term controls, the GVS displacement estimate is , set by the element size rather than . For the aging law, the update is stable only if . When is not small compared with , set control.rsf_dtheta_max . It adds two bounds over all elements:

The first limits the slip per step to . The second limits the fractional healing per step to . must lie in ; 0 (default) disables the bounds. The bounds require the aging law (mat.state_var_model = 1), slip-rate option 1, adaptive stepping (control.fixed_dt = 0), and no PT solver, body-force adjustment or isostasy adjustment.

Resolving a velocity-weakening fault​

A velocity-weakening fault is unstable only if it is softer than a critical stiffness. For a spring-slider of stiffness under effective normal stress , steady sliding is unstable when

(Ruina, 1983). A fault in an elastic medium has stiffness of order over a slipping length , so instability needs a patch larger than a nucleation length. For the aging law two length scales are useful (Rubin and Ampuero, 2005):

with for antiplane and for in-plane slip. is close to the size of the cohesive zone at a rupture tip and sets the element size. Cycle simulations need several elements across (Lapusta et al., 2000). A velocity-weakening patch shorter than creeps or slips slowly instead of rupturing. In DES3D the slip of an element is spread over its width , so element size controls both the length scales above and the width of the shear zone.

Earthquake output​

DES3D detects events in RSF models from the largest speed and plastic activity:

  • an event starts when exceeds sim.earthquake_start_factor (default 5) times the largest boundary speed, but at least m/s, while plastic strain is growing;
  • it ends when falls below sim.earthquake_end_factor (default 2) times the boundary speed, but at least m/s, or when plastic strain stops growing.

During an event, fields are written every sim.earthquake_output_step_interval steps (default 50) in addition to the regular output. With sim.seismic_moment_calculate_output = yes DES3D writes each event's start and end times to seismic_moment_magnitude.txt, together with the seismic moment accumulated over the event,

where is the shear modulus, the element volume and the element's plastic-strain increment of the step (the second invariant, in simple shear, so each term is ). It needs no fault geometry: only elements that yield contribute, and rigid motion adds nothing. In 3D is in N·m and the file also gives (Hanks and Kanamori, 1979). In 2D is the element area, so is per unit length out of the plane (N·m/m) and no is written. In the RSF earthquake example the main events give about N·m/m, an average slip of 5 m on the 100 km layer, which matches the loading of one cycle.

The point monitor can record the friction coefficient and state at chosen points with monitor.output_dynamic_friction and monitor.output_state_variable; see the point monitor tutorial.

Parameters​

ParameterDefaultMeaning
mat.rheology_type–elasto-plastic-rsf or elasto-visco-plastic-rsf
mat.direct_a[0.020] per material
mat.evolution_b[0.025] per material
mat.characteristic_velocity[1e-6] per material (m/s), must be positive
mat.characteristic_distance[4e-3] per material (m)
mat.state_var_model00 steady state, 1 aging law, 2 slip law
mat.friction_angle0, mat.friction_angle1[30], [5]reference friction angle (degrees) before and after plastic weakening
control.rsf_slip_rate_projection_option00 velocity projection, 1 strain-rate invariant
control.rsf_dtheta_max0state-change limit on the time step, in
control.use_global_velocity_scalingnoGlobal velocity scaling; forced on for RSF
control.inertial_scaling1e5: scaled wave speed over the fastest speed
control.mass_scaling_reference_speedshearWave-speed cap: shear () or bulk ()
control.damping_option, control.damping_factor1, 0.8Force damping
sim.earthquake_start_factor, sim.earthquake_end_factor5, 2event thresholds relative to the boundary speed
sim.earthquake_output_step_interval50output interval during an event (steps)
sim.seismic_moment_calculate_outputnowrite event times and seismic moment to seismic_moment_magnitude.txt

References​

  • Dieterich, J. H. (1979). Modeling of rock friction: 1. Experimental results and constitutive equations. Journal of Geophysical Research, 84(B5), 2161–2168. https://doi.org/10.1029/JB084iB05p02161
  • Hanks, T. C., and Kanamori, H. (1979). A moment magnitude scale. Journal of Geophysical Research, 84(B5), 2348–2350. https://doi.org/10.1029/JB084iB05p02348
  • Herrendörfer, R., Gerya, T., and van Dinther, Y. (2018). An invariant rate- and state-dependent friction formulation for viscoeastoplastic earthquake cycle simulations. Journal of Geophysical Research: Solid Earth, 123, 5018–5051. https://doi.org/10.1029/2017JB015225
  • Lapusta, N., Rice, J. R., Ben-Zion, Y., and Zheng, G. (2000). Elastodynamic analysis for slow tectonic loading with spontaneous rupture episodes on faults with rate- and state-dependent friction. Journal of Geophysical Research, 105(B10), 23765–23789. https://doi.org/10.1029/2000JB900250
  • Marone, C. (1998). Laboratory-derived friction laws and their application to seismic faulting. Annual Review of Earth and Planetary Sciences, 26, 643–696. https://doi.org/10.1146/annurev.earth.26.1.643
  • Rubin, A. M., and Ampuero, J.-P. (2005). Earthquake nucleation on (aging) rate and state faults. Journal of Geophysical Research, 110, B11312. https://doi.org/10.1029/2005JB003686
  • Ruina, A. (1983). Slip instability and state variable friction laws. Journal of Geophysical Research, 88(B12), 10359–10370. https://doi.org/10.1029/JB088iB12p10359
  • Scholz, C. H. (1998). Earthquakes and friction laws. Nature, 391, 37–42. https://doi.org/10.1038/34097
  • Tong, X., and Lavier, L. L. (2018). Simulation of slip transients and earthquakes in finite thickness shear zones with a plastic formulation. Nature Communications, 9, 3893. https://doi.org/10.1038/s41467-018-06390-z