Running a Simulation

Running

One button, and then two or three windows. What matters afterwards is knowing where the solver tells you whether to believe the result.

01Before it solves

Run Simulation first reads the table and the settings panels. Anything it cannot make sense of stops the run with a message naming the row and the field — a thickness that is not a number, a substrate row carrying both dopings, a rear contact voltage given in two places at once. Nothing is solved until the structure is complete, so an error here costs you no time.

02What opens

A normal solve opens two windows:

• Poisson Solver Plot — the band diagram and carrier profiles, with the individual traces switchable. Carrier densities start on a linear scale; Log scale: carriers / G switches them. See The Plot Window.
• Poisson Solver Results — the numbers, in four tabs: Electrostatics, Boundary Conditions, Layer Transport and Derived. See The Results Window.

If Find Quantised States is ticked, the bound-state results follow in a window of their own. Both windows are independent of the main one, so you can leave a result open, change the structure and run again to compare.

03Did it converge?

This is the one thing to check every time, and it is the third row of the Electrostatics tab:

Solution converged           Yes
Iterations (used / limit)   10 / 1000

A converged solve is one where the electrostatics settled. Most do so in ten or twenty iterations out of the thousand allowed, so a healthy run leaves the count nowhere near the limit — which makes a failure unmistakable. If it did not settle, the row reads No — reached iteration limit with the count sitting at 1000.

Treat everything else in both windows as unreliable at that point. The plot is still drawn, and it still looks like a band diagram.

04If it says no

Reach for the structure before the settings. The solver already retries on its own — when a direct attempt fails it starts from flat bands and walks to the target in steps, without telling you — so a failure has usually survived more than one strategy.

By far the most common cause is asking for something outside a zero-current model rather than anything numerical: a junction biased into forward conduction, or two contacts of the same type with a voltage between them. Both are real currents, and this solver holds current at zero by construction. If a warning appeared when you pressed the button, read it — it is usually describing exactly this.

05C–V takes a different route

With CV calculation ticked, the button does something else. You are asked for the sweep range first, then a progress bar works through the bias points — it has a Stop button, and a long sweep is the one place in Poisson where you may want it.

What follows is the capacitance plot and its own results window. The ordinary plot, results and bound-state windows do not appear — a C–V run gives you C–V outputs and nothing else. See Capacitance–Voltage.

Because of that, it is worth running the structure once with CV calculation unticked before you sweep it. That gives you the equilibrium band diagram and the convergence check described above, on the same structure, in a couple of seconds — and a structure that will not solve at equilibrium is not going to give you a meaningful capacitance curve either.