While using PC1D for simulation of silicon surface recombination, i very often encounter the problem of steady state or transient convergence failure. What could be the possible reason.
There is no easy answer for this one as several factors can come into play. The first is the size of your device simulation relative to how small the regions are where there are high fields. The element size factor allows the nodes to grow in size by the size factor and 1/size factor on either side. So if you have a big device with lots of small field regions the algorithm can't mesh fine enough to give a good solution. Second, is the normalized error limit being to low. In some instances the solution of the problem is such a small minimum that it is easily over stepped so the solution steps keep moving back and forth over the minimum. Alternatively the answer could be so far away from where the algorithm started it can't get there in the allotted time. This is especially true for large bias changes per transient steps. Finally, the normalized clamp applied to either the electrostatic potential or the carrier quasi Fermi levels could be limiting the solution. If the bias change between transient steps is large but the clamp is to small then it will take a long time for the boundary conditions to percolate through the nodes. Alternatively, if the clamp is to large the algorithm could overshoot the solution over the correct value then on the next iteration undershoot and get into a loop that the error does not decrease. Play with the numbers and the size of your device to find what works best. I am not a total expert on this as some times it will not converge for me on some really simple structures. For instance on a fairly thick Silicon cell with fixed bias and a spectral response curve from 400 to 1100 nm it will run successfully up to 800 nm then not converge. If I run from 1100 to 400 nm it will not even start. Yet if I decrease the cell thickness by 10 microns it always works.