Learn Computational Physics: An Engineering Degree Guide

[vijiraghs]

Hello all,

I am currently doing an engineering degree. I am interested in doing physics related computing. Can anyone tell me what exactly is simulation? I don't have a thorough background in physics but I am willing to learn. Is that okay to work in the field of "computational physics"?

Thanks in advace,
vijiraghs.

 

[swim3051989]

computational physics is often considered as a branch of theoritical physics. Because it is a tool to adapt theories. But, someone think that it belong to experimental physics.
I want to attend graduated program about it, but I don't know what university trains it best.

 

[bromden]

You should be aware that the term "computational physics" is rather broad. It just means applying computational techniques to problems in physics. The fields are diverse: people study biophysics, condensed matter, lattice field theory, chaos, etc.

That being said, you need to figure out what type of physics you want to apply computing to. If you are primarily interested in the methods used, perhaps research in mathematics/computation would be more appropriate. UT Austin, for example, has a http://www.ices.utexas.edu/graduate-studies/admissions/" in computational science.

 

[Astronuc]

Hello all,

I am currently doing an engineering degree. I am interested in doing physics related computing. Can anyone tell me what exactly is simulation? I don't have a thorough background in physics but I am willing to learn. Is that okay to work in the field of "computational physics"?

Thanks in advace,
vijiraghs.

Basically computational physics involves modeling and simulation of the physical processes of some physical system. The computation involves numerical solutions to systems of coupled (and sometimes/often nonlinear) partial differential equations over varying time and spatial scales, from nuclei to atoms to grains to components to large systems such as nuclear reactors, combustion systems, rockets, planets, stars, . . . .

I know some folks from the UT program.

 

[Pythagorean]

Expanding on Astronuc's post, it depends on what aspect of computational physics you're working on. It can be real-world analysis, modeling (which you still have to analyze), or both.

real-world analysis can be, for example, classifying and studying real-world signals with Fourier transforms and other spectral analysis tools. A physics PI I once worked for designed many kinds of acoustic detection systems, using this information collected by undergrads and data technicians.

modeling is generally using differential equations (describing the time evolution of some observable), but then also analyzing the results and making hypotheses.

Then of course, you can integrate both into comparison tests (comparing models to real-world data with coincidence testing or convolution-type algorithms).

In all these cases, it is my not-so-humble opinion that one of the neglected, but very important aspects of computational sciences, is the ability to transform your data into a visual representation that best expresses the answers to your scientific questions. This is the only way you can transform a bunch of numbers into something meaningful to human beings (such as the ones sitting on grant committees and direction boards).

 

[swim3051989]

@bromden: do you know other universities in USA(except UT Austin)? Because I like a little warm climate :)

 

[Astronuc]

Some background on computational physics/engineering (the engineering aspect becomes engineering with multiphysics).

http://www.scidac.gov/

A related program at INL is MOOSE (Multiphysics Object Oriented Simulation Environment): A Parallel Computational Framework for Coupled Systems of Nonlinear Equations
http://www.inl.gov/technicalpublications/Documents/4336141.pdf

http://web.engr.oregonstate.edu/~palmerts/csne_seminar/CSNE_Presentations/Dana-Knoll-OSU.pdf


PETSc = Portable, Extensible Toolkit for Scientific Computation
http://www.mcs.anl.gov/petsc/petsc-as/index.html

and

LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) is a classical molecular dynamics code.
http://lammps.sandia.gov/


Basically one needs to determine the particular partial differential equations that define the mass, momentum, energy distributions and flows in a system (includes solids, fluids and gases). In addition there are constitutive relationships, thermophysical properties, and behavioral models that must be developed. In the case of nuclear systems, one would need to overlay the neutron physics and fissions in a nuclear reactor/fuel, and include dozens or hundreds of radionuclides representing fission products and transuranic elements, as well as the radiation effects on materials. Radiation effects are the transmutations and lattice defects (dislocations, intersitials, holes, voids) in the materials, which may be correlated with fluence, or displacements per atom.

Magnetically confined fusion plasma simulations must track nuclei and particles, their interactions, including fusion, and their interactions with the magnetic field.

 

[daschaich]

@bromden: do you know other universities in USA(except UT Austin)? Because I like a little warm climate :)

To reiterate one of bromden's points, to a first approximation pretty much any research university in the US (or other countries) has some physicists working in part or entirely on computational projects. However, different places will have different specialties, such as particle physics (with subfields ranging from collider experiment simulation and data analysis to lattice gauge theory), condensed matter (lots of subfields), gravity/numerical relativity, plasma physics, and many more. In some places computational physics can be found in applied math, electrical and computer engineering, mechanical engineering, chemistry, astronomy, maybe even computer science. (In my limited experience, computer science research tends to be more abstract and less applied to specific physical problems.)

If any of these subjects interests you more than others, it will really help narrow down the options.

Personally, when I was applying to grad school, I was very interested in doing computational physics, and I wanted to keep both condensed matter theory and particle theory as options. (Particle experiment computing hadn't really agreed with me.) The main reason I ended up going to Boston University was that they had strong computational groups in both condensed matter theory and particle theory. (Another place I considered with similar strengths was Carnegie Mellon.) I eventually ended up doing lattice gauge theory at BU, supported by their Center for Computational Science,
http://ccs.bu.edu

Now I'm at the University of Colorado, which in addition to lattice gauge theory has a good applied math group working on computational physics. The applied mathematicians here collaborate with others at Tufts, Penn State, Columbia and various national labs on SciDAC projects like those Astronuc linked to (PETSc, TOPS, etc.)
http://www.scalablesolvers.org
http://ccma.math.psu.edu

 

[Astronuc]

One can also browse journals to learn what is being done in various areas, e.g., CFD, computational mechanics, computational materials science, etc, as well as who is doing it and where.

Journal of Computational Physics
http://www.journals.elsevier.com/journal-of-computational-physics/?navopenmenu=-2

Journal of Computational Physics thoroughly treats the computational aspects of physical problems, presenting techniques for the numerical solution of mathematical equations arising in all areas of physics. The journal seeks to emphasize methods that cross disciplinary boundaries.

Multiphysics simulation crosses disciplinary boundaries, as well as specialties within physics.

For example - http://math.stanford.edu/~memoli/papers/hm.pdf

Computational Mechanics - http://www.springer.com/materials/mechanics/journal/466

International Journal for Computational Methods in Engineering Science & Mechanics
http://www.tandf.co.uk/journals/titles/15502287.asp

Computational Materials Science
http://www.journals.elsevier.com/computational-materials-science/

 

[Pythagorean]

You might even consider computational neuroscience as computational physics. There are many computational physicists working in the field:

http://people.brandeis.edu/~pmiller/

and many such studies are published in Physical Review E

http://pre.aps.org/accepted/E/6407fR57T0b1a70bb9463f7099e4c7ae3926dc055

 

[swim3051989]

I really like to study about quantum gravity. So that, I am finding out some graduated program for it although I am learning about IT bachelor:(

 

[daschaich]

I really like to study about quantum gravity. So that, I am finding out some graduated program for it although I am learning about IT bachelor:(

By coincidence, I saw today a recent PhD thesis on computational calculations of causal dynamical triangulation, completed at Jagiellonian University in Kraków. If this sounds like something that may interest you, then you could check out the authors cited in the bibliography to find other places where similar things are being studied.
http://arxiv.org/abs/1111.6938

 

[swim3051989]

Thank you. But this university in Poland while I am planning to come USA :)

 

[Sina]

computational physics is experimental theoretical physics :)

simulation is well like in may aspects games. Racing games are car simulations, there is a lot of classical mechanics going in the background where the game solves equations of motion numerically.

I will given an example from a field that is familiar to me, simulation of molecules. You want to see how molecules behave under certain circumstances (interaction with other molecules, change of enviromental conditions such as temperature, solvent etc). You have some theoretical models that describe how these molecules behave and they all boil down to solving certain linear or nonlinear algebraic, integral or differential equations. There are numerical methods to do these and you employ them. At the end you can understand how the molecule behaves (i.e how its coordinates, momentum or other internal variables such as charge distribution etc changes) and you output them. There fore you have simulated the behaviour of a molecules under some hyphotetical conditions that you have imagined.

Now the experimental part is, all these models require certain parameters or assumptions which can be changed up to certain degree. Experimental part is that you usually simulate your model for different paratmers or assumptions and then you look at the results and compare them with ACTUAL experimental (lab) results to see which parameters and assumptions fit best. Therefore you experiment on these parameters and assumptions rather than rigorously proving that they should produce results up to certain error terms (actually sometimes rigorous calculations may not be possible at all). Thus many times, just like actual experimental physicists, you may also have to statistically study your results too.

 

[swim3051989]

uhm, I think computational physics biases about theory more. It simulates abstracted model while it don't need to practice such as loop quantum gravity...

 

[jk]

@bromden: do you know other universities in USA(except UT Austin)? Because I like a little warm climate :)

Austin has a warm climate (putting it mildly)

 

[SophusLies]

computational physics is experimental theoretical physics :)

Having worked in computational physics previously, that is the best description possible. A lot of the time I was debugging code and checking if the model is correct. Very much analogous to an experimental physicist wrestling with lab equipment.

 

[swim3051989]

Austin has a warm climate (putting it mildly)

I think Austin is very hot. Moreover, it has had fire forest :(