Inventor's Day Series: Intel Fellow Kelin Kuhn -- A 45nm saga

posted by Alison Wesley on February 11, 2009

As I said in my blog yesterday, Feb. 11 is National Inventor's Day. In honor of this celebration, Fellow Kelin Kuhn is sharing her thoughts on what it takes to be an inventor.

Kelin is an Intel Fellow and director of Advanced Device Technology. In her role as the 45nm device group leader, she was deeply involved in helping to deliver one of Intel's biggest leaps in processor technology in decades. Intel reinvented the hundreds of millions of tiny transistors at the heart of the Intel® Core™ mircoarchitecture with its revolutionary Hafnium-infused Intel 45 nanometer-sized Hi-k metal gate transistor technology.

This formula improves processor performance and reduces leakage. Through this invention, Intel was able to continue to deliver on Moore's Law, enabling consumers and business users to enjoy smaller, more powerful and energy-efficient laptops and PCs, and servers whose power consumption doesn't break IT budgets.

I stopped reading science fiction when I joined Intel's 45nm development team in late 2003. I stopped not because I had no time to read--although that was somewhat true as well--but because our work on 45nm Hi-k metal gate silicon technology made science fiction seem boring in contrast. 

Process development in 45nm started--as it does in all technology generations--with the release of preliminary design rules. Design rules are the list of maximum and minimum dimensions for key process features.

One thing I've become accustomed to in technology development is that new design rules always look impossible--and those for 45nm were no exception. When the first set of design rules was released, I had my usual "HUMANS CAN'T DO THIS" reaction to them. Admittedly, I've grown used to ignoring this reaction because we DO make these rules work--generation after generation. However, I can always count on having this gut reaction when I first see the design rules.  

Why do I always have this reaction?

Think about the sizes we are talking about. Before I joined Intel, I was a university professor teaching lasers and optics. I taught that you couldn't print something smaller than the wavelength of light used to print it. Today, in 45nm, we routinely build transistors whose size is more than five times smaller than the wavelength of ultraviolet (UV) light used to print them.

A 45nm SRAM memory cell is so small that it is dwarfed by a human red blood cell.
Innovation matters

With 45nm, we faced a critical problem, one that sounds more appropriate on the bridge of the Starship Enterprise than in real life. Our task was to invent a transistor manufacturing process that would significantly reduce the quantum mechanical tunneling of electrons through a barrier. While you may think, "who cares?", the fact is that without taming this undesirable behavior, our chips would get too hot and be too power-hungry to be of any practical use.

In the late 1990s, a team led by Robert Chau in Intel's Components Research group began investigating solutions to this problem.  They identified that replacing the conventional gate dielectric material with a hafnium-based Hi-k dielectric material would significantly reduce this quantum mechanical tunneling.

They also discovered that effective manufacture of hafnium-based Hi-K materials required replacing the poly-silicon gate electrode with a metal gate electrode. Let me step back for a moment and place in context just how scary this was.

At this time (late in 2003 and early in 2004), the only working Hi-k metal gate transistors were being built in research programs, not being driven by technology development. Our seminal research publication [Datta et al., IEDM 2003,  pp. 653-655] is an example of this phase of Hi-k metal gate development, where key features (integration of strained Si and three orders of magnitude reduction in gate leakage) were being demonstrated; but the basic scientific facts about these materials were still in contention. In this time frame, public-domain discussions at conferences and panels were vigorous debates between scientific and academic experts on very fundamental aspects of the physics of these materials.

The idea that Intel could transform research experiments into revenue-producing material in less than four years seemed like science fiction.

Today I look at the 45nm yield roadmap and marvel that we were able to deliver this innovative and complex transistor architecture (with its challenging manufacturing requirements) while still maintaining a Moore's law two-year off-set from 65nm.

Success is lots of little steps

One of the challenges of technology development is that innovation is typically received with a flood of "This won't work" criticisms. Hi-k metal gate was no exception. 

My father used to say, "An amateur expects things to work and is surprised when they don't; a professional expects things to not work and is surprised when they do."

While humorous, there is some real truth to this. In solving difficult problems, such an enormous amount of energy goes into working through all the things that are wrong that it is sometimes difficult to recognize success when it finally happens. 

In science fiction (even the best science fiction, I'm afraid), the protagonist has a brilliant idea and--"Wow, presto!"--a few days later, it happens. It would be great if real innovations worked like that, but unfortunately they don't. If I am to leave you with one key idea, it is that 45nm was an excellent example of Thomas Edison's quote, "Genius is 90 percent perspiration and 10 percent inspiration."  

While 45nm was built on the 10 percent "inspiration" of one key innovation (the integration of hafnium-based gate dielectrics with a metal gate), it was the 90 percent "perspiration" by a large number of talented people that drove continued incremental improvements in a number of areas--including yield improvement, reliability enhancement, and transistor performance improvement --that really delivered the 45nm Hi-k metal gate technology. 

While Kelin jokes that she still has no time to read science fiction, her top three favorite books are: "The Moon is a Harsh Mistress," by Robert A. Heinlein, "The Genesis Machine," by James P. Hogan, and Janet Kagan's "Mirabile."