Posted on 07/12/2006 9:07:26 AM PDT by PatrickHenry
Are the algorithms "embarrassingly parallel" as they say? Do you primarily reduce the equations to matrix mechanics and use SCALAPACK or whatever superceded it to solve?
Ancient minds want to know :-)
Full Disclosure: I clicked on the 'properties' of your animation and surfed over to the website which held it (Dept. of Chem. and Physics, some University in Australia.) The paper looked like a whole mess of tensors and definite integrals, but no mention of basis functions...
Cheers!
Let me spell it out for you: the ability of American scientists to do world-class nuclear physics has kept you free. That isn't Commerce-Clause parsing: it is a crucial part of our ability to make decisive war on hostile nations, and I include in this all of the research that is not directly related to weapons production, because the truths we learn today become the technologies we can't survive without tomorrow.
This is not a matter of mercantilism. Nobody makes money off this. Nuclear physics is an economic loser (thanks to politics), essentially all research is non-corporate, and nuclear physicists are poorly paid. (As a professional programmer I make twice what I ever made as a physicist.)
Cut off nuclear physics research--as we almost have--and in the intermediate term we will lose the ability to defend the country.
Maybe now you can find it in the Constitution.
Now that Wilczek, Gross and Politzer have won the Nobel for asymptomatic freedom, Han, Nobu and Greenberg should win this year for actually coming up with QCD. (at least I think they should).
I honestly know nothing about the computing methods used to do this on a large scale (very few do, and this it's not my research project).
The paper looked like a whole mess of tensors and definite integrals, but no mention of basis functions...
I did do a little bit of lattice gauge QCD in my most advanced grad class - I'd have to dig up the old papers to give you an exact (boring) description, suffice it to say that its an iterative method based on the non-Abelian (i.e. non-commutative) QCD gauge theory Lagrangian (i.e. energy density function). I just recall that it took me hours and hours to manually work through what was a 1x2 grid on the 'lattice' (i.e. 2 pixels in a single time frame of one of these graphics).
Really tough stuff!
I'm embarrassed to say I didn't know they hadn't gotten one already for that! (I haven't been so good on following up on the latest Nobel Laureates.)
I think the prize is a bit of sham any more, anyway; there's so much going on in physics any more, involving such large collaborations of people that singling individuals out for this level of prestige doesn't make much sense.
My apprehension of these things has simply been--for lack of a better term--"propagation of errors".
I once spent a pretty frustrating time on a molecular dynamics code because of non-conservation of energy issues; I finally tracked the problem down to individual sign or transposition errors (in say the 5th decimal place) of four terms out of some 1800 which were hard-wired into the code.
And in a large, very successful commercial code which shall remain nameless, there was a problem in a parallelization module which resulted in a physically impossible temperature jump in one region of the system being studied; to make it worse, that test data set, on that machine, was used as one of the *standards* by which other versions of the code were verified.
All of these things helped contribute greyness to my whiskers. :-)
Cheers!
There are plenty of other physicists who I can think of who deserve the Nobel but haven't yet. Both John A Wheeler and Stephen Hawking are living on borrowed time, needless to say, but Syd Coleman is seriously ill with Parkinson's disease, and his time is running out too. Hopefully, Peter Higgs will still be around to get his Nobel once the LHC goes operational.
Memory, CPU, I/O. Everything. These are like modeling the motion of a car by treating the entire car as being made up of tiny cells. (In QCD, maybe a proton is a few cells across.) It takes a long time. The results are stochastic and thus the computation must be repeated.
It's like doing a elliptic equation but instead of the difference operator, one has non-commuting 3x3 complex (SU3) matrix multiplies and Pauli matrix multiplies.
Bumping this back up due to some thoughts and questions I've been having lately. I've been wondering if it's possible to intepret QCD in terms of wave mechanics. It seems paradoxical to me that as the strong force grows weaker, the gluon wavelengths would become shorter. Or could instead, resonance be a conservation property in quark-gluon interactions, with the quarks gaining mass (and hence, shorter wavelengths) at higher energies, with an equilavent increase in wavelength by the gluons?
A question of very limited interest. The problem, of course, is we don't know enough to formulate the interesting general question.
I'm contemplating your question - it's not an easy one to answer. I can tell you that it has to do with the fact that gluons, unlike photons, actually exert a force on each other (i.e. they "couple" to one another; gluons actually carry a 'color' charge, just like the quarks that exchange them; in contrast, photons carry no electric charge). It's not that individual gluons grow weaker as their wavelength decreases, it's that the net effect of the whole field of gluons grows weaker at short distances due to a sort of 'cancellation' that occurs at close range due to the couplings between gluons. The exact mechanism of how this works is a subject of intense research, and a lot of the details remain a mystery.
This is the best answer I can give without really cracking open the books and putting a lot more time into it - I hope this helps answer your question at least a little.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.