Using the correct part of strain rate for dislocation creep rheology and plastic weakening

[YiminJin]

In the V(E)P models, we are using Maxwell-type rheology, i.e. $\varepsilon = \varepsilon^v + \varepsilon^e + \varepsilon^p$. Thus, if we use strain-rate-dependent creeping models, e.g. the dislocation creep, then the viscosity should be calculated using $\varepsilon^v$. However, in the current implementation of model ViscoPlastic, we are using the full strain rate to calculate the pre-yielding viscosity, which will cause spurious softening in the regions where $\varepsilon^p \gg \varepsilon^v$.

The same goes for plastic weakening: strictly speaking, we should use $\varepsilon^p$ to calculate the weakening factor, but in practice we are using the full strain rate. I think its effect is small, since in most cases the plastic strain rate takes the most part of the full strain rate, but it does result in a discontinuous plastic strain field at the initial stage of yielding when the plastic strain is relatively small.

I have not figured out a solution to this problem: if we use local Newton iterations to obtain the viscous strain rate, as in model DiffusionDislocation, then the computational cost would increase significantly. And things become more complicated when considering the material averaging. I open this issue as a memo for possible future improvements.