[victorious_BIG]

Hi,

I have what may be a very mundane question. I'm trying to solve a system of equations to very tight tolerance. For my numerical experiments, I've been using the laplacianFoam solver and running to steady-state.

I am using a linear domain (one row of cells) with BC

x = 0 , T = 1
x = 10, T = 0

I then run the solver with fvSolution dictionaries:

solvers
{
T
{
solver PCG;
preconditioner DIC;
tolerance 1e-20;
relTol 0;
}
}

I want to see my initial residual fall below 1e-20, but this appears not to be possible. No matter how many cells I use or what tolerance I specify, the smallest absolute residual I see is ~1e-16. The final residual will go as low as I specify in tolerance, but always I see the initial residual jump back up to about 1e-16.

My question is this:

Why can't I solve to a tighter tolerance? How do I accomplish this?

Thanks!