Introduction to the Professions
Biology, Chemistry, and Physics 100
lecture notes for Tuesday-Thursday, 18-20 November 2003
Simulation Science
Many months ago we defined science as the application of the experimental
method to the understanding of natural phenomena.
This definition does not encompass all activities that people
traditionally think of as scientific, but it does embrace:
- laboratory work--that's directly covered; and
- theoretical studies--because they illuminate past data and allow
us to make predictions about future experiments.
We now introduce another mechanism for doing science:
computer-based modeling of scientific systems.
We'll use the term "simulation science" to describe this type of modeling.
Modelers or simulators are doing science in that they manipulate the parameters
that define their computer model to help them understand the affects of various
variables on the real system being modeled.
What can we study with a computer model of a physical system?
- The appearance and structure of a collection of objects.
molecular models; CAD drawings
- The energetic stability of a system.
Finite-element analysis; energy minimizations on an atomic scale; Monte Carlo
- The behavior of the system over time.
FEA again; molecular dynamics
How do we construct one?
- Define scale (macroscopic; telescoped; microscopic; atomic)
- Define objects to be simulated
(blocks of steel; planets; bacterial cells; atoms / molecules)
- Define how these are to be understood by the computer
(do they have substructure? how do they interact?
- Set up capability to add new items to or subtract old items
from the list of understood objects.
- Define interactions among objects and with the outside.
... often through potentials
- Incorporate way of tracking things that change
Position, velocity, rotational properties; mobility
What can we learn?
- What a complex system really looks like.
hidden lines handled correctly; rotations are easy; rescaling
- Whether the system is energetically plausible.
- Whether the system will remain stable over time.
- What kinds of lab experiments will enhance our understanding.
- Whether the assumptions that went into the model are reasonable
- look at the verifiable cases and at some extremes.
- Garbage in, garbage out: bad potentials -> bad results
unless you force-fit the parameters, in which case:
bad potentials -> bad parameters.
How can we improve the model once we've started?
- Correspondence with experiment:
Can we reproduce the conditions of an experimental system on the computer
and see if it behaves like the real system.
- Modify as few parameters as possible to improve agreement.
- Stare at it and think of ways to better represent the system.
- Simplify or complicate the model to match known behavior.
What are some applications of simulation science?
- FEA: heat loads, mechanical stesses, shear.
- molecular modeling.
- molecular dynamics.
- astrophysical systems.
What is the relationship between simulations
and laboratory experiments?
- See above: we can make things match.
- Avoid over-fitting the data.
- Use the simulation to guide further experiments.
- Use experimental information as restraints on parameters in the simulation.