Humans tend to believe that the same trends will continue in the same way regardless of circumstances. A trend will get everlastingly worse or better with nothing changing. The fact is that every trend is susceptible to outside factors. Here’s a case in point.
In the late 19th Century the trend was that New York City would be buried in horse manure. Of course changes in technology and the use of rapid transit curbed the usage of horses long before that happened. The track record for dire predictions is pretty dismal. Which is a good thing.
Yet people still insist on making the assumption that the way things are is the way they will always be. Seriously, literally just a couple of years after the big conference manure the problem was nonexistent because of the new el, subways, electric trolley and the onset of the automobile.
Likewise with technologies. No trend is going to go on forever. There are limits and before they are encountered they may not be viable. That certainly goes for planar lithography technology and Moore’s Law. I think even Gordon Moore will admit that all good things come to an end. That end seems to be coming for lithography.
Think of these scanners as the front line of Moore’s Law, the repeated doubling of the density of integrated circuit components that has defined more than 50 years of astounding technological progress. For decades, a steady series of remarkable breakthroughs, many of them in photolithography, have enabled chipmakers to repeatedly shrink chip features, keep the length of R&D cycles under control, and economically pack more and more transistors on a chip. Those advances have taken us from chips with thousands of transistors in the early 1970s to billions today.
But to keep the good times rolling, GlobalFoundries and other leading-edge chipmakers won’t be able to rely on the brilliant lithographic advances of the past. And so they’re contemplating another radical shift, one that could prove to be the most challenging yet.
For the entirety of its existence, semiconductor lithography has been done with electromagnetic radiation that was more or less recognizable as light. But for the change chipmakers are now weighing, the radiation is something else altogether. It’s called extreme ultraviolet (EUV) radiation, but don’t let that name fool you. Unlike the ultraviolet light used in today’s scanners, EUV can’t travel in air, and it can’t be focused by lenses or conventional mirrors.
And it’s also difficult to produce; the process begins by firing laser light at a rapid-fire stream of tiny molten tin droplets. The hope is that scanners built to use the resulting 13.5-nanometer light—a wavelength that is less than a tenth of what is used in today’s most state-of-the-art machines—will save chipmakers money by allowing them to print in a single step layers that would otherwise require multiple exposures.
But creating EUV systems that are bright and reliable enough to operate in the fab—nearly 24 hours a day, 365 days a year—has proved to be a monumental engineering challenge. For many years, EUV faced significant skepticism and repeatedly failed to live up to predictions that it was almost ready for prime time.
Now, though, the technology really does seem to be turning a corner. The brightness of the EUV light source made by Dutch lithography-tool manufacturer ASML Holding seems to be closing in on a figure long targeted for commercial production. ASML, which has emerged as the technology’s standard-bearer, is now shipping EUV scanners that it says should be ready to mass manufacture leading-edge microprocessors and memory starting in 2018. The world’s most advanced chipmakers are working hard to determine when and how these machines will be incorporated into their production lines.
The stakes are high. Moore’s Law is facing significant challenges, and no one is sure how the semiconductor industry—which grossed more than $330 billion last year—will navigate the next five or 10 years or what a post-Moore’s-Law semiconductor industry will look like. A decline in revenues might be inevitable. But if keeping the “law” in effect avoids, say, a 15 percent drop in the industry’s income, that would keep an amount of money flowing that is twice as great as the total revenues of the U.S. video game industry.
The fineness of the details that can be made with a photolithographic system depends on several factors. But a powerful way to make dramatic improvements is to shorten the wavelength of the light it uses. For decades, lithographers have done just that, shifting their wafer-exposing tools from operation at the blue edge of what’s visible to the human eye down to successively shorter wavelengths in the ultraviolet part of the spectrum.
In the late 1980s, the semiconductor industry was beginning the process [PDF] of shifting from mercury lamps to lasers as the light source of choice, reducing the wavelength from 365 nm to 248 nm in the process. But some researchers were already contemplating a far bigger jump, into the X-ray range. Hiroo Kinoshita, then at the Japanese telecommunications firm NTT, reported the results of early work on this idea way back in 1986, using 11-nm radiation. Others, at AT&T Bell Laboratories and at Lawrence Livermore National Laboratory, also explored the technology independently. In 1989, some of these researchers met and traded notes at a lithography conference. In ensuing years, research into the notion got infusions of investment from government and industry.
The optics division of ASML is essentially right across street from me. It’s the direct descendent of the old Perkin Elmer(no the current PKI) optical company here in CT. Perkin Elmer pioneered the optical stepper process back in the 1970’s and the after a couple of mergers and being acquired over the years it’s now ASML that has the technology. I’ve explored a good bit of the history here.
I also linked to an IEEE Spectrum post here.
The fact is that Moore’s Law isn’t a physical law. It’s statement of how the planer lithographic transistor was evolving at the time that Gordon Moore made his statement back in the early 1960’s when the technology was in it’s infancy. Planar lithographic technology has followed the path of every other evolving technology. first you have the clunky experiments and effort to make anything work at all, then a rapid advance of evolving and expanding products as the tools to make the products get better and better and finally a leveling off as further improvements get more costly because the tools get bigger, more complicated, and much more expensive. Planar lithography has had good run, but the technology is reaching the bigger and more complicated phase as the requirements for accuracy and precision at higher levels are running out of the easy solutions. I’ve seen a stepper first hand and they are amazing machines but there’s so much complicated engineering inside that it beggars belief. All of which has to work multiple times in nanometers and be repeatable across different machines as the chips have their resists exposed over and over. The fact that the more expensive the machine the harder you have to run it to make money. That’s where the IC industry is right now.
Richard Feynman once gave a talk about the realm of the very small and the new technologies that might come out of that.
This has led to a great deal of talk about nanotechnology and the supposedly coming technological singularity, with super intelligent machines and augmented humans essentially playing with magic to create things out of thin air. The fact is that artificial intelligence and nanotech haven’t been the things that we imagined that they would be. In the end the singularity is a product of tools and right now the tools do not exist. So no singularity any time soon.
What is happen is additive and numerically controlled machines. They’ve been around for a long time now, but in the last ten years or so the technologies have dropped dramatically in price while capabilities have skyrocketed. It’s now possible to have machines make parts while being unattended and the same machines make different parts in the same work cell.
We are just at the beginning of the understanding of what can be done with these new tools.
Where things are going to end up will be an interesting and fun ride. new tools drive new technologies and innovations. Moore’s Law is coming to the end of it’s curve. that doesn’t mean though that innovation is stopping. Technology builds on what has gone before and just as the stepper machines build on the machine tools and optics of the 1960’s so too will the new CNC additive and subtractive tools build on the computing power and precision that the IC chip has provided. If you are looking for innovation, look for the tools because innovation is using the tools you have to create what you imagine.