[petsc-dev] access to finite element second derivatives

[Geoffrey Irving]

For simulations of bending shells, I'm going to need access to second
derivatives of both primal and dual basis functions.  This means
versions of f0,f1 that take second derivatives, plus f2 for inner
products with the dual second derivatives.

Have you thought about how to fit this into the current code without
interfering with speed when the second derivatives aren't required?
It also increases the relevance of Jed's note that material models and
material field evaluations might ideally be shared across evaluations
at the same quadrature point.  I.e., merging f1 and f2 into a single
call might avoid redundancy.

I haven't used actual shell elements before (all the bending
simulations I've done used ad-hoc hinge forces), so any advice about
element choice would be great.  Are the default quadratic elements a
good choice?

Thanks,
Geoffrey