Computational Simulations and a Phyiscs Phd

[Quaoar]

I will be applying to graduate school this coming winter, and I'm having difficulty deciding what field I should be directing my effort towards.

My background: I have a bachelor's a physics, and my senior project was to write an N-body program that simulated galaxy collisions with the inclusion of dark matter. I've also tackled several other physical problems at my job, such as simulations of sensors, simulations of magnetic fields, etc.

My main interest: Computational simulation of physical systems. The main issue I have is that I feel I would be content designing simulations of almost any physical system. I could focus on galaxies, traffic flow, hurricanes, molecular interations, field calculations, etc, anything where I get to model reality in a virtual environment.

I don't think that computer science is the way I wish to go. I am interested in the theoretic side of the actual problems, not the most efficient way to code. I feel that coding efficiency is something that I can pick up by myself.

So I guess my question is: How should I apply? Should I apply very specifically to a certain field that is deeply dependent on computational simulation? Or should I apply broadly, and not really show preference to any particular sub-field?

 

[Dr Transport]

Most physics today is done via computer simulation... If you are interseted in Computational Physics, there are plenty of places to go. Oregon State has a decent program

http://www.physics.oregonstate.edu/

From my experience, most graduate departments have limited instruction in computational physics per se, mostly just computer methods in physics. I learned what I know about computatinal physics th ahard way by doing and making many mistakes.

 

[Gokul43201]

I will be applying to graduate school this coming winter, and I'm having difficulty deciding what field I should be directing my effort towards.

I would recommend (hard or soft) Condensed Matter. There's fabulously interesting physics in both areas that lend to computational modeling.

 

[Quaoar]

Thanks Dr Transport...any other recommendations? I scored fairly well on the Physics GRE, so I don't believe the top programs are beyond my reach...

 

[Dr Transport]

It depends on whqat you want to do, where you want to go etc... My alma mater had a computatinal physics undergrad program in the 80's and I could have had a masters thru the deans office as an indisiplinary (sic) student in computers and physics but I had absolutly no clue about programming until I went back for a PhD and took a Fortran course. Hind-sight is 20-20 and I would have looked for a way to get into programming much earlier in my career. I'm trying to get back into it full-time thru my employer to little avail.

Soft-mater is a way to go, lots of work in computational fluid dynamics if you prefer. I tend to not push the envelope and poke around some of the more mundane problems and squeak little more out of someone elses theories.

SUNY at Buffalo http://www.physics.buffalo.edu/, has a computational physics program which isn't too bad either. I got to go looking for others, but as I said before, most physics toay is done via compiler and simulation.

 

[robphy]

http://www.math.uchicago.edu/camp/
http://www.cims.nyu.edu/programs/
http://www-sccm.stanford.edu/wrap/program.html
http://www.pacm.princeton.edu/index.html
http://www.acm.caltech.edu/
http://www.cct.lsu.edu/
might be worth looking into.

 

[neurocomp2003]

Guillochon: curious to know what language for the sims you use or what types of packages? or do you code most things from scratch?

 

[Quaoar]

Guillochon: curious to know what language for the sims you use or what types of packages? or do you code most things from scratch?

I have been doing everything from scratch, Fortran/C/C++...occasionally Matlab to prototype.

 

[Clausius2]

I could focus on galaxies, traffic flow, hurricanes...,

Hurricanes, that's kind of cool. You should know there is a lot of science on Computational Fluid Dynamics. The turbulent systems are almost an enigma for nowadays scientists. Try to take a look at UCSD. We have here the SD Super Computer Center, and people working on turbulence uses to simulate with DNS techniques and run it on this center. The dynamics of turbulent structures is something of interest today. We scarcely understand it, and the applications in engineering and science are huge.

 

[neurocomp2003]

Guillochen: so you don't use a numerical packages like Numerical Recipres/ CLAPack/Netlib etc? Do you use a Rendering Package? btw what you do is cool...hopefully i'll get there within the next 2-3 years.

 

[Quaoar]

Guillochen: so you don't use a numerical packages like Numerical Recipres/ CLAPack/Netlib etc? Do you use a Rendering Package? btw what you do is cool...hopefully i'll get there within the next 2-3 years.

Nope. Everything from scratch, even output images.

Anyhow, so it looks like we have some advocates of soft condensed matter...I'll look into that, but could anyone give a brief run-down on the problems the field tackles?

 

[phun]

nowadays every subfield of physics (including experimental) uses computer models to some degree at least, but I think theoretical astrophysics is a field where people rely on modeling almost exclusively.

 

[Gokul43201]

Nope. Everything from scratch, even output images.

Anyhow, so it looks like we have some advocates of soft condensed matter...I'll look into that, but could anyone give a brief run-down on the problems the field tackles?

Here's a couple that I'm aware of:

1. Complex fluids: Simple fluids are those that can be modeled by Navier-Stokes behavior. Properties like viscosity are essentially independent of the flow. A lot is understtod about these fluids. Complex fluids, however, do not even remotely obey NS. They include polymers, various multiphase suspensions (foams, colloids,...) and other such stuff, and exhibit quite bizarre properties (eg:Weissenberg effect, non-linear diffusion) that are very hard to model.

2. Percolation theory: This started off as a soft CM problem involving "percolation" through a network of interconnected channels, but soon found application in other areas like hard CM, high energy physics and many branches of engineering (phase transitions, conductivity of two-phase solids, electrical network analysis).

3. Pattern formation and propagation: All kinds of interesting phase behavior is modeled here - from understanding shape effects in liquid crystals to phase transitions in anything from highly disordered 2D solids to highly correlated electron systems.

And then there's all the other magic that falls under non-linear dynamics (eddies and vortices, granular systems, traffic flow, neural networks)

 

[desA]

1. Complex fluids: Simple fluids are those that can be modeled by Navier-Stokes behavior. Properties like viscosity are essentially independent of the flow. A lot is understtod about these fluids.

Actually, precious little is really understood about the Navier-Stokes behaviour.

At the slow end of the scale, we have incompressible, viscous flows.

At the high speed end, we have high-speed, compressible, flows with Mach numbers crossing through M=1 & into hypersonics. These are the N-S equations minus the viscosity terms - the so-called Euler equations.

There is a huge no-man's land between these two extremes, with some extrapolation backwards from the high-speed compressible forms.

The simple facts are that the N-S equations are notoriously unstable, requiring all sorts of numeric kludges just to provide numeric stability - but, unfortunately, at the expense of accuracy. Why do N-S solvers 'blow up'? No-one has yet found a suitable theory for this.

120 years ago, Reynolds performed a famous pipe flow experiment, where he observed the onset of turbulence. No-one, to this day, can explain his findings.

At what set of circumstances does the 'onset of instability' occur? Again, lots of research, but no clear-definative answers. What mechanisms are at work?

So, in reality this is a very, very fertile research ground, without moving away from Newtonian-type fluids.

 

[Clausius2]

Actually, precious little is really understood about the Navier-Stokes behaviour.

At the slow end of the scale, we have incompressible, viscous flows.

At the high speed end, we have high-speed, compressible, flows with Mach numbers crossing through M=1 & into hypersonics. These are the N-S equations minus the viscosity terms - the so-called Euler equations.

There is a huge no-man's land between these two extremes, with some extrapolation backwards from the high-speed compressible forms.

The simple facts are that the N-S equations are notoriously unstable, requiring all sorts of numeric kludges just to provide numeric stability - but, unfortunately, at the expense of accuracy. Why do N-S solvers 'blow up'? No-one has yet found a suitable theory for this.

120 years ago, Reynolds performed a famous pipe flow experiment, where he observed the onset of turbulence. No-one, to this day, can explain his findings.

At what set of circumstances does the 'onset of instability' occur? Again, lots of research, but no clear-definative answers. What mechanisms are at work?

So, in reality this is a very, very fertile research ground, without moving away from Newtonian-type fluids.

AMEN!

And just add MagnetoHydrodynamics in which there is a lot to do.

 

[Gokul43201]

desA, I agree with your assessment. Descriptions like "a lot" and "precious little" are naturally, subjective and mean different things coming from different frames of reference. Compared to our understanding of complex fluids, our understanding of simple fluids may appear to be "a lot". But compared to essentially linear systems, "precious little" is really known about highly nonlinear systems like a Navier-Stokes fluid.

 

[asherp]

If you're into magnetohydrodynamics (MHD), consider space physics! It covers different areas of the solar system, with emphasis on the sun-earth environment. E&M fields, solar wind speed, density, and temperature are the focus. Space weather forcasting in particular is a new and growing field, and it can be quite lucrative if you want to get into the satellite industry. There's a good summary of the field on the Center for Integrated Space Weather Modelling website www.bu.edu/cism.

 

[J77]

I went to a talk by David Gross about the future of Physics the other day...

He made a comment that in the future there will be no more analytic work, things will be done as "computer assisted proofs".

Caused a slight ripple in the audience

 

[Quaoar]

So I guess the general consensus is that I should decide what topic interests me most, because there is simulation work in almost every field of physics...

 

[neurocomp2003]

learning about N-body problems and fluid dynamics i think would be a start...Game Physics may be a area you might be interested in or designing Physics-based Virtual Reality. Astrophysics Sims/Geophysics Sims fascinate me th emost.

 

[franznietzsche]

I'm surprised no one has said this yet, but Astrophysics. I'm currently working as a summer student at Los Alamos in Computational Astrophysics. Everything that has been mentioned so far is covered in some of the stuff I've done, or that the people in the same room as me are working on. This is what I love about astrophysics, I don't have to specialize, I can work in any and every part of physics, because almost all are important in astrophysics. Nuclear and particle physics, Plasma Physics, Computational Fluid Dynamics all come together in very intricate ways. The guy next to me is doing supernovae modeling with Smooth particle hydrodynamics. I'm working on asteroseismology and stellar evolution. In plasma physics there is a huge range of problems to work on (magneto hydrodynamics being just a subset) that encompass many of the things discussed already. And of course, this is really just the stellar astrophysics stuff done here. There's also the cosmology, and galactic astrophysics simulation work related to dark matter and dark energy constraints. In astrophysics, and computational astrophysics you can work on almost everything and anything in physics because its all relevant.

 

[Quaoar]

I'm definitely leaning towards computational astrophysics, franz, as my undergraduate work focused on it, and it is the topic that interests me the most. The main concern is that I would basically have to commit myself to academia at that point, which I'm still not 100% comfortable in doing (although, I am gradually getting to that point :) )

 

[desA]

So I guess the general consensus is that I should decide what topic interests me most, because there is simulation work in almost every field of physics...

I'm reminded of a saying:

The more you learn, the more you discover how little you actually know

This may be extended to:

The more you learn, the more you discover how little you actually know, the more you realize how little is really known.

 

[creepypasta13]

I'm definitely leaning towards computational astrophysics, franz, as my undergraduate work focused on it, and it is the topic that interests me the most. The main concern is that I would basically have to commit myself to academia at that point, which I'm still not 100% comfortable in doing (although, I am gradually getting to that point :) )

do you really have to commit to academia at that point? how likely is it that one can find employment as an engineer working on CFD for defense, instead of stuff like antennas?

 

[comp_math]

You can apply to a computational astrophysics program. I am applying to a computational and applied math program but my focus is on mechanics.