Berni Alder made Monte Carlo a sure bet

For an 84-year-old, Berni Alder is quite busy. The last few weeks have been especially rushed, as Alder was in Washington D.C. to receive the National Medal of Science from President Barack Obama on Oct. 7.

Alder was among nine recipients of the medal this year, along with another five presented with the National Medal for Science and Technology Innovation.

The computational pioneer was honored for his six decades of work using computers to simulate the properties of materials. Alder invented molecular dynamics and helped develop Monte Carlo methods, which use computers to reproduce the behavior of atoms and molecules through a large number of random steps. The medal citation also recognized his work using shockwaves to study materials under extreme pressure and temperature.

Alder is a professor emeritus in the UC Davis Department of Applied Science, which he helped to set up in 1963; he has worked at the Lawrence Livermore since 1952.

I spoke to Alder after he returned from Washington D.C. 

I understand you're still working.

My working habits are "never before lunch …." I go to work in the afternoons. Three days a week, I'm at Livermore, and two days a week at UC Berkeley.


When you started your scientific career, Harry Truman was president — in your career, you've seen 12 presidents. How has the influence of science changed over those years?

Well, anyone who's lived long enough has seen a few presidents … one every four or eight years, or six years on average, I suppose.

I think for physical sciences, there was a peak after the war  — up til about 20 years ago. Now the biological sciences are much more prominent. But the physical sciences are still very much needed for biology — for example, to build the machines [scientists] need for molecular dynamics, crystallography, microscopes, and so on.

For studying the structure of biological molecules, for example?

Exactly. And a lot of physicists are making the transition. For example, Steven Chu, who's now U.S. Secretary of Energy, went from physics to studying DNA structure. And he approached it in a way only a physicist's mind could do.

Obama seems to me to be one of the most positive presidents about science in my memory. The speech he gave at the medal ceremony  — he gave a very similar speech at the National Academy earlier this year, in April. He's clearly very interested in science, and that's nice.

Your late brother, Henry, was a professor of mathematics here at UC Davis. Did you come from a mathematical or scientific background?

My father was a Ph.D. chemist. Henry of course was a mathematician, in number theory. My other brother, Charles, was a dentist — he was good with his hands.

And you moved to the U.S. in the 1940s? Where did you grow up?

I was born in Germany, but a Swiss citizen. In 1932, we moved to Zurich, just before Hitler came to power. In 1941, we emigrated to the United States; it was just before America entered the war — we had to take a sealed train from Switzerland to Lisbon, and then an American ship.

What are you working on now?

There are two problems I'm working on now, in hydrodynamics and in quantum mechanics.

Now we can deal with a trillion particles at a time in molecular dynamics — that's big enough to study fluid flow, so we can check the classical equations for fluid dynamics, the Navier-Stokes equations. And it turns out, there are discrepancies between what we can find by molecular dynamics and Navier-Stokes equations, because Navier-Stokes makes approximations…particularly ignores fluctuations.

Aren't Navier-Stokes equations what engineers use for modeling aerodynamics, for example?

Yes…the big problem is that they ignore fluctuations in density and temperature and these fluctuations can lead you to different outcomes than you would find from the classical equations, particularly at an instability.

In quantum mechanics, I'm returning to a problem I worked on 30 years ago, which was looking for a new way to solve the Schroedinger equation (which describes how atoms and electrons change over time). This would mean we can predict the behavior of molecules with much higher precision — what I call "chemical accuracy" — and we can study this with a stochastic [random], probabilistic method.

What are the implications of that?

This has implications for the whole of materials science…we could predict the properties of materials from first principles.

For example, with water, we could take two hydrogen nuclei and one oxygen nucleus and eight electrons, and predict the properties of a single molecule of water. And eventually do the same for other materials as well.

How did you become interested in this area in the first place?

The initial problem we were trying to solve was, would a collection of hard spheres go through a phase transition? (The spheres represent atoms and a phase transition would mean going from a solid to a liquid, for example). This is a generic problem because it involves many particles interacting simultaneously.

At the time [1950s] at meetings of applied mathematicians, they would regularly take a vote and about half thought there would be a phase transition and half not. So it was a controversial area, but the problem could not be solved with mathematics.

The breakthrough was to use simulation — that turned out to be the crucial thing.

In physics this is called the "many body problem," and Monte Carlo and molecular dynamics are considered among the key developments in applied mathematics in the last century. So resolving the existence of the phase transition was a big development, although it was anticipated, at least, by some of us.

What was totally unexpected came out of molecular dynamics in the late 1960s. One could quantitatively show that fluid flow at an extremely small scale — 100 Angstroms — over a very short time, one hundredth of a nanosecond, is the same as at scales of centimeters and minutes. That was a total surprise, and it completely anticipated both theory and experiment.

Another area that I worked on early on at Livermore was on shockwaves — studying materials under very high temperatures and pressures using shockwaves.

Using computer simulations?

No, these were actual experiments with explosives at Site 300 (the Lab's explosives testing station). During the war (at Los Alamos), they had manipulated explosives to create shockwaves that trigger a nuclear weapons, but these controlled explosives could also be used to quantitatively determine the properties of materials.

We wanted to know the properties of materials under very high temperatures and pressures, and we could do the experiments with these explosively generated shockwaves. I worked on that project for about five to seven years, late 1950s to early 1960s.

One of the things we did was to shock graphite into diamond. This overthrew a theory at the time, because it was thought that diamonds could only be made inside a planet. So when diamonds were found in meteorites, the assumption was that the meteorites must have come from the breakup of a planet.

Our shockwave experiments showed that diamond in meteorites most likely formed on impact (when meteorites hit the Earth).

I used to give a talk called "millionaire for a microsecond." Of course you could only recover a miniscule amount of diamond — the crystals were very small, microcrystals…if you wanted to be trendy you would call them nanocrystals.

What's the difference between molecular dynamics and Monte Carlo simulation?

There is overlap between the two. Monte Carlo, which relies on making a lot of random moves, can only get to an equilibrium state, so you can use it for modeling a molecule but not, for example, for how it changes over time. With molecular dynamics you can do all kinds of problems: whether there are time-dependent changes or equilibrium phase transitions, and so on.

You were one of the founding faculty of the UC Davis Department of Applied Science in 1963. What were you trying to achieve?

With Edward Teller and some other Livermore scientists, yes. The original idea was to create a sort of applied physics department. The Livermore Lab had facilities such as electronic computers, magnetic fusion experiments, an explosives station…big apparatus that students couldn't access at universities.

The idea was to get graduate students involved in these big projects. It didn't quite work out, though. There's the considerable distance between Livermore and Davis, and these big machines are not such a good pedagogical environment, I think.

For a Ph.D., to develop your creativity, it's better to work in a small team or as an individual. Although there were some fine people who came up through the program.

The current program is more like a regular university department, and that's largely located on the Davis campus.

Over the years I have had a number of students through Davis — I still have a graduate student and postdocs whom I work with.

I noticed that at the award ceremony, you received the first award. And then the last award the president presented was to IBM  — represented by the CEO, Samuel Palmisano — for developing the BlueGene supercomputer (which is located at Livermore). That seemed like an interesting contrast  — you were in at the beginning of scientific computing, and here is IBM BlueGene, which is the latest in supercomputing.

Yes, that was a nice bracketing. It was a very complementary award, because actually we have used BlueGene for the hydrodynamics work.

BlueGene is ideal for molecular dynamics and Monte Carlo simulations because both depend on a lot of random moves. The algorithms are simple, but it's very intensive computer use. So the computer needs to perform a large number of simple calculations at the same time.

We started working on Monte Carlo methods in 1949 with mechanical computers — we couldn't run a big simulation on them, but we could develop the principles. They ran on punch cards and card sorters and you had to program them with a plug board, like an old telephone exchange.

In the late 1950s, electronic computers became powerful enough to do these calculations.

When you were starting out, 60 years ago, could you have foreseen how powerful computers would become?

Oh no, it was unforeseeable — computing power has increased by 10 (to the power) 12 from the first electronic machines at Livermore.

Also what we did not foresee is how these methods would permeate all of physics, chemistry, biology and materials science.

Andy Fell covers physical and mathematical sciences, biological sciences and engineering for the UC Davis News Service.

Oct. 23, 2009