General/complete physics simulation software?

[MrDocat]

Is there a complete physics simulator that I can use to do lots of things while I learn physics? For example, I just learned about barymeters. It would be very interesting to simulate one. It could be done setting up a small atmosphere filled with a gas, the glass model and the liquid. Some time ago I was learning classical mechanics and it would be interesting if I could set some experiments with springs and such.

I'm not talking about focused applets; I mean a general software where you are free to set your small experiences, like you would in a lab, regardless of the topic. It is not something hard to program and would be in my view the mother of the learning tools. Is there something like that?

 

[jedishrfu]

I haven't seen any software product like this. It may appear to be easy to program but in fact it would be quite hard to be as flexible as you say allowing you to setup any kind of experiment and then simulate it in an easy-to-use fashion.

 

[MrDocat]

Not that hard if it is implemented as particles (atoms) interactions only. For example, you code the dynamics of an iron atom (mainly electromagnetic forces, I guess) and the algorithm will do the magic of giving the solid metal all it's properties. Right? Now, if you make your tests small enough (for example, using only few thousands of atoms to represent a solid cube) you will have something that is either accurate, not that much computer expensive and easy to use (scripting example: createCube(pos,dimensions).fill(Iron_Atom)). Am I mistaken? (I think I am, but I would like to know where.)

 

[hadsed]

To do that would take an insane amount of work to be able to model everything. The closest, most general thing I can think of is some type of gaming engine. It's more sensible to write smaller programs to do the more specific things.

You should try writing your own. There are uncountable benefits to writing a program to simulate physics. You'd learn how to program, first, and in doing that you would understand the concept you're trying to model at a much deeper level, or at least more thoroughly since you'll be thinking about it a lot. You'll also have something to show professors for doing research if you decide you're interested in modeling physics for a career...

 

[cgk]

Not that hard if it is implemented as particles (atoms) interactions only. For example, you code the dynamics of an iron atom (mainly electromagnetic forces, I guess) and the algorithm will do the magic of giving the solid metal all it's properties. Right? Now, if you make your tests small enough (for example, using only few thousands of atoms to represent a solid cube) you will have something that is either accurate, not that much computer expensive and easy to use (scripting example: createCube(pos,dimensions).fill(Iron_Atom)). Am I mistaken? (I think I am, but I would like to know where.)

That would be fun, wouldn't it? It is far from possible, however, and unlikely to become possible in the forseeable future. Atoms are quantum objects, and you might have heard (or learned) that the Schroedinger equation is not solvable at all analytically except for trivial cases, and doing it numerically is very very very difficult (some would say it's even more impossible). Depending on the level of accuracy you are aiming for, the current limit on the numbers you can simulate is somewhere between one, ten and a few hundreds. In order to get a metal qualitatively right you'd need millions of them (metals are complicated. If you think a solid is just a big molecule, you are wrong).

And that is not even touching the fact that the phenomena we know from real life are rooted in several different orders of magnitude, which often interact. For example, plasticity and many electronic and other behaviors of material we know from real life are dominated by *defects* in the crystals (dislocations, grain boundaries for plasticity and point defects and material boundaries for many electronic effects). Of course such effects don't need quantum effects at all, but classical mechanics simulations are also limited. That is why there are also continuum mechanical methods around... and using all such kinds of techniques is at this moment expert knowledge (for the respective techniques) which requires a lot of deep insight into both the numerical methods themselves and the physical problems they attempt to describe.

If you still think such simulations are possible, then look up what a Frank-Reed source and a dislocation stack are and do, and then ask yourself how you would want to simulate something like that without knowing what you are looking for in advance.

In short: There is no general simulations software. Researchers are working on this topic... but I would not hold my breath.

 

[fluidistic]

Not that hard if it is implemented as particles (atoms) interactions only. For example, you code the dynamics of an iron atom (mainly electromagnetic forces, I guess) and the algorithm will do the magic of giving the solid metal all it's properties. Right? Now, if you make your tests small enough (for example, using only few thousands of atoms to represent a solid cube) you will have something that is either accurate, not that much computer expensive and easy to use (scripting example: createCube(pos,dimensions).fill(Iron_Atom)). Am I mistaken? (I think I am, but I would like to know where.)

I'm doing a molecular simulation of water melting ice with only about 1500 molecules (over 7000 "atoms" for that model). To simulate 10 ns of real time, it takes me about 10 days for the calculations to be done. I use a dual core E6300 at 1.86 Ghz. As far as the accuracy goes, I'm supposed to find a melting point of around 271 K +/- 1 K.