But from the big picture standpoint, that isn’t always a bad thing.
In the early part of the 20th century psychologist Karl Lashley set out to locate and study the engram, the memory storage center for the human brain. He never found it. In fact, he ended up disproving the theory that an engram even exists, which was far more important to the understanding of the brain than if he had proven the existence of an engram.
The results of more than six decades of research in thermonuclear fusion have been equally serendipitous. While physicists can split atoms, joining them is a much tougher problem. In the 1950s, physicists predicted that cars would be powered by tiny nuclear reactors. They were wrong. But as a result of these efforts, we have learned enormous amounts about physics, nuclear acceleration, subatomic particles and how they interact, and much, much more.
The same is true in semiconductor lithography. EUV has been on the drawing board since the mid-1990s as a replacement for 193nm deep ultraviolet lithography based on ArF excimer lasers. The big issue so far has been the power source—enough to create a powerful dose of plasma.
Plasma is matter, but it’s not a solid, liquid or gas. If you had to check off a box, it would fit into the “other” category. Because plasma contains a mass of electrons, it also is highly conductive and responsive to magnetic fields.
In a nutshell, the way EUV works is that the laser beam is focused, blasted through the melted tin drop to create plasma, which in turn is reflected from a concave surface onto the photomask. This is physics/photonics at its most awe-inspiring, and the fact that it works at all is jaw-dropping. The fact that it doesn’t work consistently enough, or at least not yet, is not surprising.
But researchers are getting closer. ASML/Cymer has produced power sources that, on a good day, can achieve a base level of sufficient throughput of wafers per hour to make foundries like TSMC and Intel take it seriously. It still takes longer than commercial foundries would like, which adds to the cost of each chip, so there is work underway to make the photoresist materials more responsive to lower doses of light. And whether this approach will bear fruit in a narrow enough window of time is a matter of public debate. Moore’s Law is a formula based on time, and physics problems aren’t always solvable on a schedule.
But the massive amount of research being done in this area has opened up all sorts of new possibilities, ranging from new ways to pattern chips to new directions in self assembly, packaging and integration of wafers, dies and packages. Whether EUV works may be a short-term question. The bigger question, and one that could have much more far-reaching effects, is what else can be done with the research that has gone into EUV?