Dear Friends,
I would like to bring to your attention a new CMP Process Simulation Tool "CMP Sim". Use of CMP Sim is FREE as long as all pertinent information remains in the public domain, and appropriate reference and acknowledgement is given.
CMP Sim includes Multi-scale and Multi-Physics analysis of the CMP process. You can get information at the Wafer Scale, Die Scale, Feature Scale and Particle Scale. Defectivity and Scratch Propensity prediction is one of the capabilities of CMP Sim. It was developed at Iowa State University.
For more information or just to use CMP Sim, please either visit www.cmpsim.com or just click on the banner display in the side bar in the planarization lounge.
If you will like more information that is not available at the website, please email me at [email protected]
I and the CMP Sim Team sincerely hope that you will find this simulation package useful, and enjoy using it. We look forward to receiving your feedback, and will continually strive to impove it based on the feedback we receive.
Many Happy Simulations,
Abhijit Chandra
Let me explain my position on simulations in a couple of more words. Theoretical, or based on first principles, modeling (I believe this is what you mean saying "scientific") is more than valuable in our attempts to understand the modeled process. Unfortunately, it requires perfect and detailed understanding and correct theoretical description of the particular phenomena involved, as well as their interactions, which makes modeling in case of CMP extremely complicated, if not to say impossible (which does not mean it should not be attempted). Simplifications are always associated with introduction of empirical constants, requiring experimental calibrations, which in its turn normally depends on many experimental details, and are conditions sensitive. I am sure you know this matter very well.
As one of my good friends (a heavy theoretician in fluid dynamics) usually says: " Give me a couple of free variables and I will fit an elephant". Well, to make the long story short, I am not against the simulations, however, blind use of models by an end-user who does not have deep understanding of the theoretical approaches for the particular phenomena as well as the assumptions and simplifications involved, may, and in many cases probably will, yield false conclusions.Best,
Yehiel
Dear Yehiel,
I think you and I are in agreement.
First principle based simulations are certainly useful, while curve-fits that employ zillions of fitting constants are not. The hallmark of a sound simulation is not more than a single calibration constant. I know it is difficult, for all the reasons you mentioned, but I think it is certainly worth doing, at least attempting to do. Because if successful, it gives you a wonderful ability - to extrapolate or "forecast".
As you mention, good experimental data is invaluable, and I concur. However, experimental data is like the image on your rear view mirror - that of past, in space or time - it has no ability to interpolate (for which you need a model - easier to do, modeling requirements are less rigorous), or extrapolate (impossible based on experimental data alone. Can only be done right using a self-consistent model.). In fact, my opinion is extrapolation (by using mere curve-fits) of experimental data is the root cause of many of our societal problems.
The vision of a model, however, is foggy. Because it must involve simpliying assumptions in terms of satisfying a conservation principle.
So, if experimental data is a 20/20 vision through the rear view mirror, simulation result is like that foggy outlook through the windshield. Each has its strengths and weaknesses, but by choosing the right combination of the two, we can "drive safely" !!
I will be coming to MRS in San Francisco in April. If you are there, let's get together and have al ive chat on this.
Best,
Abhijit