Nonlinear solver behavior when fault effective stress is zero (tensile stresses)

PyLith
1

Hi Brad,

I am performing 2D quasi-static simulations using the rate and state friction model. In these simulations I am applying normal tensile traction at the fault to model pore pressure effect on the fault strength.

My problem is the following:

  1. I would like to model a case with very low effective stress. In this situation, I noticed that, due to fault slip and domain deformation, normal stress at the fault decreases causing the fault to open (effective stress >= 0) at some locations.

  2. The nonlinear solver then does not converge anymore, which I suspect is related to the frictionless interface.

I have verified that the nonlinear and linear solvers converges in few iterations whenever 1) normal tractions are not applied at the fault 2) or when the effective stress is large enough to prevent it to be larger than zero.

I know you may need a working version of the problem to provide detailed feedback, which I can prepare if needed. In the mean time, would you be able to comment a little on what is the expect behavior of the nonlinear solver when the fault is in tension (effective stress >= 0) ?

Thank you,
Josimar

2

A localized region of very low effective normal stress or even tension should not cause much change in the number of nonlinear iterations and rate of convergence of the nonlinear solve. The nonlinear solve is most likely to have difficulties and a slow rate of convergence in cases with large simultaneous slip over large fractions of the domain. There is not that much that can be done to mitigate a slow rate of convergence with the current formulation when a large fraction of a through-going fault is slipping simultaneously.