Return to Basics


Hand in hand with the earlier discussion of precision and accuracy goes the need for completeness. It’s always amazed me when an engineer is surprised when simplifying a simulation results in predictions that don’t match reality, even when they know that the same change in the actual process would have changed the results. If it didn’t, the simulator would actually be deficient. But the typical response is often to blame it on shortcomings of the simulator. Simulators can have enough shortcomings on their own, they don’t need any additional.

An example of this can be cleans. Very often these are completely ignored in a simulation. Depending on the type of clean and the materials present, this may be cause for errors. A situation that may occur is a surface oxide is cleaned following an implant. It’s a mild clean that the etch setup test says only removes a small amount of the oxide, so it is ignored. If it was a heavy implant though, the set up test numbers may be off by more than an order of magnitude. This could drastically change the prediction of a follow up implant or diffusion.

In setting up a simulation to be as predictive as possible, it is important for the simulation to be complete, including as many steps as possible. If a step is left out, the ramifications better be understood. Usually the steps left out are things like cleans, which are cheap from a computational point of view, so why leave them out? When a simulation is simplified by leaving out a step, best practice is to mark it with a comment saying it was skipped, and why. If possible a note on the limitations caused would be really good.

The other motivation for this is to avoid issues in future use. Typically a process simulation is set up and calibrated for a few core devices. The predictability for these devices may be really good. But what happens when a simulation is done for a new device, one that is sensitive to the skipped operations? At best tcad takes a hit on it’s reputation, at worst it delays a production line coming back up and costs huge amounts of money.