Jacobian question on heat_conduction module

[jsevic]

Hello,

I need a Joule heating kernel that allows coupling of instantaneous electrical conductivity so I made a new kernel that allows
sigma to be a coupled variable.

My kernel computation is identical the JouleHeatingSource kernel .C file however my Jacobian term is missing this term _delec_cond_dT[_qp] and as best as I can tell it looks like this is being computed based on constants from material properties.

Thank you!

John Sevic

[GiudGiud]

Hello

Are you saying because the material properties are constants you expect this term to be 0?
Does this term come into the definition of the jacobian?

Guillaume

[jsevic]

Hello,

Yes the material properties in this case would be constant, but the electrical conductivity is not constant it's a coupled variable.

Thank you.

For my kernel the Jacobian is

return -1*_D[_qp]*_grad_V[_qp]*_grad_V[_qp]*_phi[_j][_qp]*_test[_i][_qp];

and for the built-in Joule heating kernel, JouleHeatingSource, that does not allow for a coupled electrical conductivity, but does depend on material properties specified by mateirlaProperties, the Jacobian is

return -_delec_cond_dT[_qp] * _grad_elec[_qp] * _grad_elec[_qp] * _phi[_j][_qp] * _test[_i][_qp];

[GiudGiud]

are you using AD?
The JouleHeatingSource also has off-diagonal terms

[jsevic]

We have not tried AD, would you suggest that for a problem like this?

Thx!

[GiudGiud]

how many variables do you have?

you made this kernel

a new kernel that allows sigma to be a coupled variable.

did you add the off-diagonal term for this?
AD can let you do this easily, but it's a little slower and it requires transitioning all the material properties to AD type

[jsevic]

OK I have found out the issue I had thank you for your help!
John Sevic