Mackwheaton: That is, if performance is measured at x at 100nm, can it be assumed that there is a mathematical certainty that POET will perform at some function or ratio of x at 40nm, of does a 40nm ring oscillator need to be developed to prove its performance definitively?

I wouldn't speak of mathematical certainty, but rather of a high degree of confidence.

Dan DeSimone explained to me how they are working with the TCAD software. First they are modeling a device or a circuitry on the computer, then run a simulation. The simulation is also purely done in software in the computer. The software knows a great deal about the physical properties of various materials, it knows about the physical properties of the POET devices and how they behave – this in particular is what POET provides to and develops with Synopsys. Based on that knowledge the software can "calculate" how the circuitry will behave.

So they are running the simulator with the software-modeled ring oscillator (or whatever it is) and get some results. If these results do not meet their expectation, they'll change the model or certain parameters, then run the simulator again and so on, until the results are fine.

Only then the circuitry will be manufactured as a real chip, and the performance of that chip will be measured. However, since a model is always an approximation of the real world only, actual results may deviate from calculated results, e.g. the frequency the ring oscillator is oscillating at. Such deviations are measured and (should) get back into the device models. So the modeling foundation improves over time.

Regarding the ring oscillator, I don't know how exactly the simulation results will match the real device, but the deviations should be small enough to leave no doubt that the POET device is several times faster than the equivalent silicon device.

(Another issue is repeatability, which is addressed and improved by the lab2fab transition. In this context that would mean that if you produce 1,000 ring oscillators, you would expect that they all oscillate at the same frequency – or at least that the deviations between them are be small enough to meet some quality standard. A fab can deliver better repeatability than a lab.)

All this hasn't answered your question regarding 100 nm and 40 nm, but I think you got the idea.