From:	Eugene Leitl <eugene.leitl()at()lrz.uni-muenchen.de>
Date:	Mon, 6 Dec 1999 17:37:32 -0800 (PST)
Subject:	Simulating Chemistry: Q&A with John Pople

http://www.rdmag.com/features/0699rd/06pople.htm

Simulating Chemistry

As computers speed up, computational chemistry will become even more
useful, allowing scientists to accomplish much more in less time.

Q&A with John Pople, 1998 Nobel Laureate in chemistry

Computational methods are gradually evolving, bringing new,
easier-to-use, and more sophisticated programs to researchers, says
John Pople, a Nobel Prize-winning professor of chemistry at
Northwestern Univ., Evanston, Ill. These technological changes will
continue in the next century, making chemical theory accessible to
more scientists and engineers.

He predicts that computational methods will become more available to
experimental scientists, who will rely on increasingly sophisticated
software packages. With these, they will be able to predict the
structures and spectroscopic properties of molecules they can't make,
identify unknown compounds, and study unstable molecules and reaction
intermediates.

In an exclusive interview, Pople outlined his views on the future of
computational chemistry and its applications.

R&D: What do you think are some of the most significant changes in
computational chemistry in the last ten years?

Pople: The techniques used to predict the properties of molecules
require sufficient accuracy to reproduce the results of an
experiment. This has only become possible as computers have become
more powerful. At the same time, the corresponding algorithms and
mathematical methods have developed.

R&D: What kinds of improvements do you expect as computer technology
gets better?

Pople: There's a continual search for improved approximation methods,
and there's a continual search to use existing approximations more
efficiently, with fewer mathematical steps.

It's interesting how science has moved from a university environment
out to the commercial area through the software companies that now
sell the quantum-chemistry programs. As a matter of fact, four such
companies are derivatives from my research group.

R&D: Do you think that the changes in computational chemistry will
affect how experimentalists work?

Pople: The chemical community will be more educated as to what can be
done. The methods will become easier to use. They are analogous to
instrumental methods, like looking at the vibrational spectrum of a
molecule or its magentic-resonance spectrum.  These are techniques
that chemists learned to use. They will be able to use a theoretical
quantum-chemistry computer program to learn the properties of a
molecule in the same way.

It's now common for an experimentalist to investigate his problem with
a computational method-perhaps before he does the experiment-to get an
indication of what the results might be. Whether a bond will be a
strong one or a weak one, whether a reaction will proceed easily or
with difficulty-these things can be studied by theory before the
experiment is attempted.

This sort of approach is used, for example, by drug companies that
look at some theoretical studies of a potential drug molecule before
going to the trouble of synthesizing it and testing it in practice.

R&D: What do you think are some of the more important experimental
applications besides drug research?

Pople: Computational chemistry is used for studying the mechanisms of
reactions, for example, in atmospheric physics. There are processes
that can go on with pollutants in the atmosphere, reactions that they
may undertake, and these can be studied by theory.

Another application is in space. People study molecules in
interstellar space that may exist for extended periods but can't be
synthesized on the earth.

R&D: They just aren't stable with other species around?

Pople: That's right. You may have a molecule in interstellar space
that doesn't undergo a collision with anything else or react with
anything for long periods, because there's nothing there. This may be
a molecule that you couldn't study in a terrestrial laboratory.
People hypothesize that such a molecule may exist, then predict its
structure and vibrational frequencies. Then you can go and look for it
using that knowledge to help you.

I mentioned before that you could study possible reactions. When one
molecule undergoes a reaction, and forms some new molecules, it's
often important to know if there's an intermediate or some catalyst
that's affecting the rate of the reaction. You can explore that by
theoretical methods, to see whether you have a reasonable hypothesis.

R&D: How would you go about doing something like that?

Pople: You test out your suggestions. You may think that a certain
species takes part in the reaction, not directly but as a
spectator. It's just present and interacts with the molecules
undergoing the reaction. You can test them out by studying the complex
of the reacting molecules and the possible spectator to see whether
the spectator would affect the rate of reaction. All that can be
studied by theory.

R&D: Do you think they could use theoretical programs to identify
unknowns?

Pople: If you have a spectrum and don't know what the molecule is, but
have a suspicion, you could work out the spectroscopy theoretically
and then compare that with the spectrum that you see. That may help
you identify the molecule. That's an important area of application.

R&D: Do you think that there will be changes that will affect applied
R&D?

Pople: In so far as such applications are determined by underlying
chemical processes, an understanding of chemical processes is
important to the development. The whole area of combustion, in which
there are many intermediate short-lived species, can be affected by
studies of such intermediates-how stable they are, what their nature
might be.

R&D: Do you think that advances in computer technology or improvements
in computational methods are required to do calculations on bigger
molecules?

Pople: The two go together-the methods and the algorithms as well as
the computer technology.

R&D: Can you describe some of approximations that people are trying to
do?

Pople: Some methods require integration of a wave function or wave
functions over space, and you have to do this by forming a
mathematical grid-a large grid of points in space. And one question is
what's the most effective grid to use. A large grid will make it
accurate, but, on the other hand, you don't want to make it too large
so that the calculations don't get out of hand. You have to decide on
the best compromise between accuracy and efficiency.

R&D: How do you compromise on something like that?

Pople: There are other factors, like where you put the grid
points. What part of the molecule do you put most of them in to be
most effective? Do you need many grid points far away from the center
of the molecule? How does the density of the grid points affect the
accuracy of your answer? All these are theoretical questions that you
would have to study.

Carrying out a computation is another challenge. Even after you've
designed the computation, you may modify the computer algorithm to get
the same answer using fewer multiplications, for example by changing
the loop structure of the program.  Modern techniques in computer
science are quite important in making computations effective.

R&D: Do you think that computer companies will design CPU chips for
specific purposes like quantum chemistry?

Pople: That does not seem to be the way to go. That has been suggested
many times, and some people have actually experimented with special
chips. However, it's usually found that you can do better with a
general-purpose chip that will do additions and multiplications and
move things about from one location to another. It's generally better
to use what extra power you have, rather than to design specialist
chips.

R&D: So you don't think that's going to be a trend.

Pople: It hasn't proved so up till now, although it may be.  Some
computers have chips designed to do square roots, for example. I don't
know to what extent that's true at the present time, but certainly
they have been designed to do that. A special circuit put on a chip to
do a square root-that's the kind of thing that [computer engineers]
would work on.

That doesn't really affect the way I do things, or at least I'm not
conscious of it. You just call for a square root and you don't know
how it handles it inside the processor.

Another thing is that computers can do several calculations at
once. This happens in parallel computers. You have many processors all
doing multiplications simultaneously.  That can be used to make
computations more effective-to get a bigger computation done in a
given amount of time.

If you have many processors going in parallel, the calculation has to
be such that you don't need the results from one to do the computation
in the next one. Avoiding that kind of conflict would be another
important feature of computer programming.

R&D: Do you think there will continue to be problems that experimental
chemists won't be able to solve, where they will call in a
computational specialist?

Pople: Probably so. They must be educated as to the limitations of the
programs.  Sometimes they need to be educated as to the meanings of
the answers. It may be that only a professional person would
understand or recognize that answers might be unreliable. So there's a
market for that.

R&D: Now that experimentalists can do more things themselves with
prepackaged software, how do you think what computational chemists do
will change?

Pople: There are continual developments. And the program packages that
are available commercially will improve. Computational chemists will
be working on the next versions, as they get better all the time.

Similar Messages
06/03/1993:  summary: shortage of compuational chemists
06/08/1993:  undergrad computational chem
04/16/1997:  computational chemistry
05/03/1995:  undergrad computational chem
09/26/1994:  Re: CCL:BS/MS Computational Chemists?
08/01/1996:  Re: CCL:M:Heat of formation calculation using MOPAC.
03/11/1996:  Law of conservation of difficulty: violations.
05/03/1995:  summary (II): teaching material for computational chemistry
08/09/1994:  *summary* education in comp. chem.
10/01/1993:  torsion of conjugated systems -- summary


Raw Message Text