Summer 1999 Newsletter

Alumni Spotlight - Jeff Welser

A CIS Newsletter interview with Dr. Jeff Welser ('94), Research Staff Member Manager at IBM.

When I was in 7th grade, I was given the choice of getting either a radio-control airplane or a TRS-80 computer (with a whopping 4K of memory and cassette tape recorder for storage). I chose the computer, thus sealing my fate, I guess. As an undergraduate at Stanford, I went into EE because I wasn't sure if I wanted to be strictly computer science focused, and EE seemed broader. So starting at software, I slowly moved down the food chain. I realized that the programs I wrote would always be dependent on the computer architecture they were running on, so I looked briefly at computer systems. I rapidly decided that they were dependent on the circuits they used, and so started studying those.

But circuits were dependent on the devices that comprised them, so on to device research. These shifts took place throughout my undergrad and graduate degrees, so that my final PhD research was on strained-Si transistors, which was a combination of device physics and material science. I don't know why I was so bent on moving down toward more fundamental work. Given the way the computer world has developed, I might have been better off sticking with software in the first place (if only I had gotten that B.S. in computer science in 1988, and started working for Microsoft, hmmm...) But for some reason, I always disliked the idea of my work being dependent on something else below it that I couldn't control.

As I was finishing up my MSEE at Stanford, I decided I wanted to take a year off before continuing with my PhD. Professor Jim Harris had a colleague at the IBM T.J. Watson Research Center, Moty Heiblum, who liked to have grad students come work with him during their PhD's, so I joined his group for a year. I found the lab to be a great place where one could still do basic research, like at a university, but without the funding and tenure track issues that would come with a junior faculty position.

When I finished my PhD, IBM offered me a job, and it was clearly the best opportunity, so I returned. My department, "Exploratory Memory and Device Modeling," works on a variety of device topics for integrated circuits, from basic research on exploratory device structures, to advanced development of more near-term devices. When I first started in 1995, I was working on "quantum-dot" memory as well as a FLASH memory transistor (the kind of memory used in cell phones) where the storage gate was replaced with Silicon nano-crystals. Last year, I was working on changing the transistor in a standard DRAM cell from the traditional planar device, where current flows parallel to the surface of the chip, to a vertical device, where the current flows perpendicular to that surface. So the former projects are things that wouldn't appear in a product for at least five to ten years (if ever!), while the latter is something that could be considered for even the next generation of DRAM chips in a couple years. I've also been looking at other aspects of transistors and scaling, such as how to design devices with gate lengths less than 25nm that still function well (i.e. turn on and off when you want them to).

All the work I'm doing, whether exploratory or more applied, is focused on how to get increased performance from Si ICs. For the past 20 years, we've been able to do that by simply shrinking the devices each year, following the "Moore's Law" curve. But further shrinking is becoming increasingly infeasible, so we need to look for alternatives. While most people think about Moore's Law pertaining to an ever-increasing number of devices per chip, what it really means (at least to the marketplace) is ever-increasing function per dollar. In some ways, this makes for very interesting work because it opens up the field to all sorts of innovations that would have been dismissed as too difficult a few years ago. On the other hand, there is a lot of concern about which direction to go, or whether the field of devices is going to reach a plateau, in which case it's time to look more closely at some of the higher levels in the food chain.

My research has definitely shifted from purely exploratory to more applied over the years. Up until the late 80s, the labs were very isolated, allowing all sorts of pure, blue-sky research. One of the reasons I chose IBM was that it was one of few labs still doing really exploratory work. But I have come to enjoy the more applied science quite a bit. I'm pretty lucky in my job, because I have very few deadlines other than conference papers. Since we are a research lab, a great emphasis is placed on continued learning. We have seminars several times a week on current research topics, both with internal and external speakers, and people are encouraged to try to attend these as often as they can, even if the seminars aren't exactly in their field.

One thing I'm really glad I did during my undergraduate and graduate studies was take some time off and try working. The year I spent at IBM before starting my thesis work really taught me a lot about how to do research, and gave me insight into what research outside of a grad student environment is like.

My advice for current students is not to be limited by your thesis topic, or even your thesis area, when looking for your first job. I fell into that trap initially. I now think this is completely the wrong attitude. The time you spend in grad school learning about engineering and research in general, not to mention really honing your analytical skills, can all be applied to great advantage in any number of fields, engineering or otherwise. I'm very happy with the job I chose, but in retrospect, I limited myself a lot during the job hunt, and probably should have been more open-minded at the time.