Whiteboard Session: computational lithography
Here at ASML, the whiteboard is our favorite forum. We sketch our ideas on them, and only the best ones stay put. We are eager to share our knowledge and want to learn new things, and that’s exactly how The Whiteboard Sessions was born. In every episode, one of our experts takes a complex technical topic and aims to explain it in simpler terms.
In this episode: “What is computational lithography?”
When you look at a chip under a microscope, you will see an interconnected web of geometric patterns. Chipmakers use our machines to print these patterns. In our machines, we project light through a mask, which contains one such pattern, shrink that pattern down with optics and then print it on photosensitive silicon wafer. But we’ll let you in on a little secret: the machine alone actually never prints exactly what we want it to print.
This has to do with chemical effects on the photosensitive layer and diffusion of the light. A decade ago, when chip patterns were about four times bigger than they are now, this deformation didn’t necessarily lead to a faulty chip. But as chipmakers continue to shrink the patterns down to make smaller and more powerful chips possible, the problem becomes bigger. The intended patterns may convolute and deform to the point that we ask: will it print at all?
So this is where computational lithography comes in. We gather all lithography data in computational models that learn from the printing effects. In fact, we use this data to flip the script and design an optimized mask that does print what we want. Only a couple of years ago, we designed these optimized patterns by hand, but today we’re using fully automated free-form modeling software. In the end, our machines are only as good as our computational models.