Motion of randomly generated stars

dgroth

Hello, I am new to this forum - and impressed with the posts. I have recently developed an addiction to astrophysics. I wish to do the following:

- define a spherical "sandbox" with a radius of let's say 1, which I think of as "universe"
- within that sphere I generate N number of points. Each point have the following attributes:
1) a coordinate: p(x, y, z)
2) a velocity: v(x, y, z)
3) a mass
- define time as t

Then, kick-off a program (that I'm hoping to write) which will represent all the points (stars) as I move to t+1, t+2 etc... Many points should collide, others turn into orbits, others will fly outside my "universe". I wish my model to be "accurate" according to gravity fields of each star.

Can anyone share with me how I could go about achieve this goal ? (let's start with 3 stars)

In advance, thank you.
Regards, Dan.

granpa

n stars requires n^2 calculations. you'll run out of processing power rather quickly I'm afraid.

dgroth

I apperciate the calculation involved. But thanks for drawing my attention to this. However, I have access to a rather massive computer all for myself: IBM p6 with 64CPUs and 256GB of RAM. Anyway, I have already started working on a grid solution to distribute the calculations, if necessary (which I assume it will).

Regards, Dan.

dgroth

Hmmm, I see what you mean granpa...now with a few steps back and some thinking - I might have been a bit ambitious. Just for the laugh: my target was/is 1 million stars.

But... I'm not going abandon this easily. Nothing easy is worth doing. Since I don't know the first thing about celestial motion, I'm going back to the basics: Newton. I'll move onto GR later. First I building I'm building a gridable physics architecture.

I have one question that may help me:
- is there a distance limit for which gravitational contribution of other stars can be ignored (stars extremly far away) ?
- if I have a cluster (I mean just a group) of stars, does all the members of the group behave in similar way, ie. can get away with applying a "pull" to the group rather consider each start individually? This will help me understand if I can grid clusters across the network.

Regards, Dan.

Oberst Villa

Welcome to the forum, Dan. What you are planning to do falls in the general field of "N-Body Simulations", googling for this might yield useful info.

Also this recent review paper http://arxiv.org/abs/0806.3950 might be interesting, section 3.2 Tree Codes seems to be related to what you wrote in #4 (grouping stars to clusters).

A website from one of the authors of this paper, www.artcompsci.org contains (among other stuff related to scientific computation) some further info about N-Body Simulations.

granpa

you might consider trying to form a spiral arm of a spiral galaxy. you could limit it to 2 dimensions and you might not even require ANY interaction between the stars. you could probably do a million stars quite easily. just put them in circular orbits of all different radius's around a central mass (but with all the dark matter I'm not sure they follow an inverse square law though) then somehow put a slight mass concentration in one area (not a concentration of stars) and see if their orbits shift into that area thereby creating an even greater mass concentration.

http://abyss.uoregon.edu/~js/ast122/lectures/lec26.html