[pbohorquez]

Do you know if is it possible to solve implicitly the term "fvm::grad(U) & fvc::grad(nu)" in foam-extend-3.1 using the block matrix?

[schröder]

yes, of course

[pbohorquez]

Originally Posted by schröder

yes, of course

Is it straightforward? I tried to simply add this term in pUCoupledFoam but lots of errors arose during the compilation.

[schröder]

It is not straightforward. It is impossible.

[pbohorquez]

Originally Posted by schröder

It is not straightforward. It is impossible.

Is it really impossible even implementing the fvmLaplacianTranspose operator from the scratch?
Best wishes,
Patricio

[schröder]

Assuming that you mean "solve term==0" by "solving the term", your problem is ill-defined.
Describe, WHAT you want to solve. Not HOW you want to solve it. Then it may be possible to help.

Martin

[pbohorquez]

Originally Posted by schröder

Assuming that you mean "solve term==0" by "solving the term", your problem is ill-defined.
Describe, WHAT you want to solve. Not HOW you want to solve it. Then it may be possible to help.

Martin

I simply want to include into the momentum balance equation of a steady incompressible fluid of variable kinematic viscosity nu the term fvm::grad(U) & fvc::grad(nu) where U denotes the velocity field. In previous versions of openfoam we treated grad(U) explicitly by means of fvc::grad(U) & fvc::grad(nu). I want to use an implicit solver, i.e. fvm::grad(U), to improve the convergence. Is it clear enough to you now?

[blais.bruno]

I guess (? I am not sure) that using fvm::grad(U) will lead to an implicit scheme being used. However, the issue you will have is that fvc::grad(nu) will be calculated with an explicit scheme applied to (nu). If you have a non-Newtonian fluid (where nu is a function of the shear rate, etc.) that part will remain fully explicit, so you will only get a "partially implicit" scheme at best... It might be preferable to keep everything explicit, but to go with a smaller time step instead.