Stiffness/Residual Vector Error - NaN/Inf findings

Hi,

I am especially writing this, because there is no specific explanation I found online. I checked the BCs and other variables, but could not find any mistakes.

I am working on a semiconductor device that is basically n-doped homogeneously, has sizes 50um-15nm, and with a drain voltage 1V on the left boundary, source voltage 0V on the right, and a varying gate voltage at its top with a thin insulator gate of. Although I work on a supercomputer with 36gb ram, and making a mesh not much detailed, I gather the error attached as image.

The error does not happen when I sweep the gate voltage from around -13V to 20V, but whenever I start it from around -15V it just happens. The error contains NaN/Inf findings in the solution, not a convergence error or a iteration one I guess, but somehow a calculation error. The simulation works well for -13V to 20V of gate voltage, giving a nice plot, but not for below -13V.

Sometimes it is an undefined value in stiffness vector, sometimes a residual vector error.

I want to plot the ID-VG graph accordingly, but cannot see below -13V. Can't I simulate after some voltage level?

Any comments for a better simulation?

Thanks,

~bera