[Nek5000-users] Problem using gradm1

[nek5000-users at lists.mcs.anl.gov]

Dear nek developers,

I am running a very simple Conjugate Heat Transfer case. It is a 2D channel with a solid wall. The inlet temperature of the fluid is 1 and I initialize the temperature in the entire domain to 1. I set a boundary condition 't' on the outer surface of the wall and I impose TEMP = 1.1.
I run nek in order to solve the transient prolem (using cvode) and at the interface between solid and fluid I check the energy balance:

lambda*gradT]solid = lambda*gradT]fluid

I have already checked many cases and, after the changes implemented by Stefan (thanks again Stefan!) it seems that the energy balance is fine.

This time (and also other times) the error in the energy balance is quite big compared to the fluxes but, what is strange, is that the fluxes have different sign.
In order to compute the fluxes I put in userchk the following lines of code:

call gradm1(g1,g2,g3,t)
call col2(g2,vdiff(1,1,1,1,2),ntot)

I don't use opgrad because it works only on the fluid mesh, isn't it?
I checked the temperature and, as expected, it grows from the fluid to the solid, going towards the outer surface of the wall (I checked the t array at every grid point in debug mode). In order to be sure that I was taking the right points, I checked also ym1 of the same points.
I took two contiguous elements in y-direction, one on the fluid side and one on the solid side. I varied iy from 1 to ny1. In the fluid, the gradient is always positive, as expected. In the solid, the gradient is negative for iy = 1 (interface with the fluid), but then it becomes positive for iy = 2 to ny1.

I also tried to compute a rough estimation of the gradient between two grid points as dt/dym1 and the values are not very different from those computed by gradm1 but for the first point of the solid gradm1 is completely wrong.

Is there any other subroutine to compute the gradient in the solid?

Thanks a lot in advance.

Regards,
Andrea.