initially diverging temp in one cell using convMeltFoam
 

Hello all,

I am using the convMeltFoam solver posted by Fabian Rösler with OF 2.4.0. After changing BC for p_rgh to zeroGradient and replacing in pEqn.H

/*phi = (fvc::interpolate(U) & mesh.Sf())
+ fvc::ddtPhiCorr(rAU, U, phi);
*/
with
phi = (fvc::interpolate(U) & mesh.Sf())
+ fvc::interpolate(rAU)*fvc::ddtCorr(U, phi);

the code compiled and ran but has a hard time converging in the very beginning second of the computation independent on whether fixed or variable time stepping is applied.
The problem appears at the left bottom corner cell of the domain, where temp BC are zeroGradiant from below and imposed T > T_melt from left for t>0 with initial T < T_melt everywhere. Also with variable time stepping, the time steps become very small to keep Courant # in bounds, but the velocities are nowhere really large to warrant those small dt. :confused:
I am fairly new to OF so I would appreciate good advice.
Attached is a screen shot after 0.4 s when the cell is way out of bounds. This error decreases later in the computation to non-existent.