What is COMPUTATIONAL PHYSICS?
Computation vs Theory vs Experiment in Physics
Both experimental and theoretical physics are incomplete without
the option to compute whenever it is neccessary.
The goal of computational physics is not to replace theory or experiment,
but to enhance our understanding of physical processes.
Four different numerical techniques:
The numerical techniques listed here link into the lectures for
the graduate course ``Computational Physics'', which is also open
to undergraduates in their final semester.
- solution of equations.
Areas of physics where computational techniques are widely implemented
include my own areas of condensed matter/statistical physics, astrophysics,
nuclear physics, particle physics etc etc
In the Physics Department at the Technion ( Italics for professors giving talks in this course)
Condensed matter/statistical physics (J. Adler and students in association
with the experimental groups of E. Polturak, E. Ribak and R. Kalish ).
Students have done many projects with all three, remind me to talk about them later!)
Astrophysics (N. Soker, H. Peretz and A.
Experimental particle physics (Yoram Rozen and Shomit Tarem)
have been among the most active areas using computers in physics research.
In the last years computational physics class there were 3 students who did
projects that potentially lead to a new track of computations in their
research groups. These involve ``big'' codes - EPOCH for plasma physics,
QUANTUM ESPRESSO for energy bands and LAMMPS for Molecular Dynamics in models of quantum dots. More about this later....
All the above and other fields of physics (and all the areas of
Chemistry, Engineering, Economics etc)
can benefit from computational methodology. Using computers
in a specific field like physics is quite distinct from Computer Science,
since in a specific field the emphasis is on solving problems of that field.
Other computational scientists at the Technion include Simon Brandon in
Chemical Engineering and Dan Mordehai in Mechanical Engineering.
The use of computers in physics often leads to advances
in computers, for example:
- John Hammersley writes about the early days of simulation
of models in statistical physics
on computers. These simulations were requested by AT&T Bell to test their
``newest'' machines in 1956! View
extracts. See also pages about the theory of percolation here and of a new interactive percolation site
- Scott Kirkpatrick, formerly vice president of IBM
and now a faculty member at the Hebrew University, introduced
``simulated annealing'' into the design of computer chips. Simulated annealing
mimics the annealing process whereby a real crystal is cooled into
its crystalline groundstate. It was developed to find the groundstate
``spin glasses'' and can be
used to find the lowest ``energy''
a travelling salesman (lowest energy here corresponds to the shortest
no. of miles he has to drive,)
a computer chip (least amount of wires connecting the devices.)
An introduction to simulated annealing suited to
physics undergraduates is
``A Walk in Phase Space''
- The Wold Wide Web protocols originated from the internal communication
needs of CERN in Geneva, Switzerland.
Some general guidelines for Computational Physics are:
- ALWAYS compare with exact solutions for limiting cases, to validate
your programming and algorithms. Do this with each calculation as a powerful
way to find the ``trivial'' mistakes.
- If possible use more than one approach; in particular each of the four
classes of approaches listed above is based on different (implicit) assumptions
and so use of more than one is a powerful way to cross-check.
- There is no intrinsic value in calculating on a computer.
- There is intrinsic value in understanding the physics of complex systems
which usually involves use of computers.
- Visualization is a very important tool for both research and presentation.
This is usually, but not always done today on computers. See the
discussion in my
lecture notes from the
Oporto Summer School, 1998.
Back to the index page