Monday, March 19, 2007

Long-term research, companies, and universities

I've posted about this topic before, but Gordon Watts' recent post on the subject of long-term research makes me want to throw this out there again. That, and the disturbing news I heard at the APS March Meeting about a round of layoffs of some of the few remaining physical sciences researchers at Bell Labs. It's terribly depressing: since my time in high school, long-term industrial R&D has been gutted in this country (and in most of the world). "Long-term" now means two years. Companies are under so much pressure to have year-over-year quarterly revenue increases that they blanch at the idea of spending money on something risky that may not lead to a big revenue stream quickly. Maybe that's always been true to some extent, and places like Bell Labs and IBM Research (and RCA and GE Research and GM and Ford Scientific and Westinghouse Research) were all effectively accidental monopolies or near-monopolies when they had major research labs. It's demonstrably much worse now.

More distressing to me is the tacit assumption, mentioned by Gordon, that university research will somehow pick up the slack. That is, federal dollars are more appropriate for this kind of basic work, and companies can always fund university labs to do work for them, too. Anyone who knows how university research actually works can tell you many reasons why this is a bad idea. Apart from low-level practical considerations (publish vs patent? foreign vs. domestic students? export controls?), the big killer here is just one of resources. Back when I was at Bell, if they wanted to they could have put a dozen condensed matter PhDs to work on a problem, along with technical support staff. Given how universities work, with teaching commitments, administrative tasks, student timescales, etc., no university achieve that kind of critical mass.

10 comments:

borat said...

Lucent technologies (research arm of Bell labs) has been sold to french firm Alcatel recently. Can it be the reason of the layoffs you are mentioning?

Incoherent Ponderer said...

Layoffs of fundamental (long-term payoff) scientists have been happening at Lucent and other places long before Alcatel deal. IBM, Hitachi, HP, Exxon have been gutting their long-term R&D groups.

Universities are not likely to fill the vacancy, for reason Doug mentions - university research happens in small groups, collaborations are often discouraged - everyone wants to claim the intellectual ownership of idea/research. Besides, grant reviews every 3 years makes it difficult to do long-term research.

Government labs may be able to pick up some slack, their funding is less competitive and may encourage long-term projects, as well as collaborations between large number of people. But government management is, as everything government-related, is highly inefficient and slow to change.

Aaron said...

Government labs certainly appear to be the best available solution to this problem. I was amazed at how many former Bell Lab employees are now employed at Sandia National Labs (my current employer). Unlike university research, collaborations here are encouraged. Also, the funding here seems much more available (if I remember correctly, about 50% is internal funding). I have not been here long enough to comment on incoherent ponderer's assessment of goverment management, but I would not be surprised if things are a little bit better than average here since the lab is run by Lockheed Martin opposed to directly by the goverment.

me, myself, and i said...

From my prospective after 6 years at a government lab, there is neither the will nor the money to pick up the slack on basic research. We do have a very collaborative environment, and lots of equipment (both of which are nice, of course). But we also are short manpower, have very limited funding, and are under huge pressure to do more and more applied stuff, because you know there's a war on.

My DoD lab is in effect a "contract" lab--we need a sponsor for everything we do, and unlike in academia, our salaries are NOT guaranteed. Since we all want to eat and maintain our health insurance, proposing fundable research is more important than proposing interesting research. The riskier and most original ideas are weeded out, because it is just too much of a gamble to risk having nothing at the end of the day.

I know not all of the government labs are like that, but even at the "non-contract" labs funding is a huge issue. Getting external funding is really hard when we can't go to NSF or NIH, and DOE can fund 5 grad students for the cost of a single Fed. Internal funding (for us, at least) is also on a 3 year funding cycle, so long term projects are hard. Without external funding, it is difficult to get enough manpower to get anything really interesting done.

Alison Chaiken said...

The shorter time horizon at what few industrial research labs remain definitely means no more basic research. As has been discussed, universities to some extent may pick up the slack on basic research, but what they can't readily do is develop major new products like the disk drive or the cell phone. Complex products like these are the result of years of work by interdisciplinary research teams that universities and startups can't duplicate. As someone who has worked in an industrial research lab for 10 years, I'm amazed, for example, that Texas Instruments managed to ship the Digital Light Processor. Shorter time horizons and smaller labs may mean no more of these revolutionary innovations. The only solution is for corporations who do R&D (or at least D) to stop issuing quarterly earnings reports that make multi-year projects look like a poor investment.

I do believe that the industrial research situation is better at (for example) Samsung, Philips, Hitachi and Lumileds. The fact that only Lumileds is based in the U.S. says as much about the move away from domestic manufacturing of electronics as it does about the state of R&D.

Doug Natelson said...

Even a good research environment that successfully innovates is no guarantee of business success, if management is clueless. Look at Xerox and the mouse/GUIs, or the fact that AT&T layed off most of their flat-panel display group in the early 1990s.

National labs should be able to fill some of these roles, but they're problematic. The DOE ones sometimes suffer from the lack of clearly defined mission. They have a very tough time attracting research superstars from the outside, particularly at the senior level. There are political decisions (let's have 5 DOE nanocenters and slightly resource-starve each of them, rather than having 2 nanocenters that really have unparalleled capabilities - that way the wealth is spread around to all constituents) that complicate matters.

Aaron said...

Doug, that is good point. The CINT facility I mentioned to you actually has two locations, one here at Sandia and one at LANL. I still wonder why that is a better idea than having one facility (at either location) and focus on making it the premier nanoscale research center. I mean the two labs are a couple hour drive from each other.

David said...

I find the discussion of the "location" of basic research quite interesting, given both my status as a pretty "basic" researcher as well as a former DOE-labs person myself.

I think that the government labs could easily pick up the "slack" from the Bell Labs-type facilities and be able to do the kind of long-term "vision"-stuff. During my personal experience (LANL), I was continually impressed by the people there and the fact that some pretty incredible science goes on there. I think that, if you could privately talk to them, though, there would be concerns about the new management there and the overall direction of lab research. Are the kinds of research proposals that can generate the new basic science going to be favored by the new management there, or will a more applied science model dominate?

When I left, it was not clear, and probably won't be so until at least this July, if not longer. Their first funding cycle that is run by the new management (LANS) is just starting and the funding decisions will be made this summer. It will be interesting to see if the same kinds of proposals that were funded under UC's management get funded under the LANS management. I think that there is some significant worry that it won't be.

A comment on Aaron's comments: I think that the rationale for the split structure of CINT was to exploit capabilities at each place and make an overall stronger bid. For example, Sandia is quite strong in microelectronics and fabrication while LANL has facilities like NHMFL (High Magnetic field) and LANSCE (Neutron Scattering) that area largely unavailable elsewhere. There are many such examples, so don't take this as some kind of complete list, but this is the basic idea. I do not think that the idea is that you have broad duplication of capabilities at Sandia and LANL, but have complimentary capabilities.

There also has been a broad push for several years to increase connections between those two labs (heck, Sandia started out its existence as LANL's Z-division right after WWII, so I guess this is really to re-connect the labs more). If I recall correctly, this was a particularly strong sentiment at DOE during the formation of the CINT proposal. The decision to have CINT/LANL and CINT/Sandia has to be viewed in the context of what DOE said they wanted at the time from LANL and Sandia, which was more of this kind of collaboration.

Doug's comment: "let's have 5 DOE nanocenters and slightly resource-starve each of them, rather than having 2 nanocenters that really have unparalleled capabilities - that way the wealth is spread around to all constituents".

I personally think that one way you could look at the DOE nanoscale lab structure is that, at a five-year review or some other appropriate point, some of the facilities will lose any further funding, and DOE will continue to fund the "successful" ones. To some degree, this is like a venture capital funding scenario. (Whether this is appropriate for our tax money is obviously an open question.) It remains to be seen what the future of all five centers is.

I guess my personal feeling is that the national labs could pick up the slack from the Bell Labs-type labs, if they are allowed to, and that there is clearly room at the University level to do this as well. The fundamental question is if the political will exists at DOE/DOD and that Congress _wants_ to see this. We, as scientists, need to do a better job of making our voices heard, in my opinion.

Doug Natelson said...

David - The problem I see with revisiting the nanocenters at the 5-year mark is that then you've spent $90M each on buildings and facilities, let alone support staff. Is there really going to be political will to shut one of these down after that kind of initial investment? I went to the first planning workshop for CINT, and at the time it sounded like they wanted to make something like the ITP but for experimentalists. A place where there would be special in-house expertise and technical resources that you just couldn't get at a university. That vision seems to have shrunk a bit. I visited ORNL last year, and saw their new nanocenter getting off the ground. It's nice and all, with some good electron microscopes and a cleanroom, and of course some good in-house staff. Did it really outclass what I have access to at Rice and, if I'm willing to drive a little, UT in Austin? Not so much. An example of something that would really be special: a state-of-the-art (30nm compensated mask) photolithography setup like one in industry, but with in-house experts who can do mask design for you. No university has that. That would be unique. It looks like all the centers were given enough resources to build buildings and fill them with some toys, but not enough resources to go out and steal in-house experts (like ITP did).

I know some excellent people at the DOE labs, and I know that they can do great work. As you say, the question is what are the expectations. I still think that some of the labs have poorly defined missions. What is Livermore really supposed to be doing? (And don't tell me that the NIF is actually a prototype fusion reactor. That's just silly.) What is Brookhaven really supposed to be doing? I'm not saying they don't have some great people; I just don't know what the lab missions are. LANL and Sandia seem to be more clearly set, though "stockpile stewardship" is pretty unexciting. ORNL has the Spallation Neutron Source. Argonne has their synchrotron, as does LBNL, but what is Argonne supposed to be doing besides running a user facility? I don't mean that as a knock - I genuinely don't know. The NREL has a mission, though it's one that doesn't seem to have the support of the administration....

Anonymous said...

Doug - EXACTLY!
I happen to have done my postdoc research at one of the DOE nanocenter facilities, and it's precisely the question I would ask: how are you trying to be different from anything that is available at universities?

I think the answer is large projects that most smaller university groups/grants cannot support. Even then I am not convinced that it cannot be done at universities.

Ideally you also want to fund not just the facilities at a much higher rate, but also staff. Having a photolithography facility and have two or three staff members is not going to accomplish much. Senior staff are managers, and its junior people who do lion's share of the work. In DOE labs there are no grad students, just postdocs, who don't last long enough to develop something long-term. So can someone guarantee that DOE group with a great facility, and one or two junior staff members who actually do science (as opposed to senior people who just travel giving manager/mission talks) can outperform a university group lead by an ambitious faculty with 10 grad students and 3 postdocs? How about 20 grad students and 8 postdocs? I am not so sure.

The real success of DOE labs were large facilities that universities cannot afford - neutron reactors, synchrotrons, etc. This is where Oak Ridge and Argonne have the edge - their nanocenters are tied to those facilities. Yes, they become user facilities, but it's not a bad model because it leads to unique and cutting-edge science. Whereas buying a whole bunch of electron microscopes and understaffing the facility will not lead to anything university cannot do.

But another way to look at it - all five centers are tied with some of the (arguably) best DOE labs. So having five instead of two will help to foster some interdisciplinary research with other departments - chemistry, materials, physics, etc.