Portraits of a Universe in Motion

February 2, 2006 1:31 PM

Galaxy DynamicsGRAVITAS is an ongoing project to visualize and animate the dynamics of galaxies using supercomputer simulations.
posted by ZippityBuddha (21 comments total)

I wonder how many variables they have in their formulas, because the resulting animations don't look look like they'd take a supercomputer to compute. Beautiful animations, regardless.
posted by pmbuko at 1:50 PM on February 2, 2006

Incredible. Like visual music. Some of the best vids I've seen on the net. It's going to take a while to download all the HD vids but I must have them. A thousand thanks, ZippityBuddha.
posted by MetaMonkey at 2:02 PM on February 2, 2006

It's also Kiefer Sutherland's favorite word.
posted by JeffK at 2:06 PM on February 2, 2006

It's also Kiefer Sutherland's favorite word.
Yes JeffK, but do you know his least favorite word?
posted by super_not at 2:08 PM on February 2, 2006

I wonder how many variables they have in their formulas, because the resulting animations don't look look like they'd take a supercomputer to compute. Beautiful animations, regardless.

Computing transformations over time in a galaxy is horrifically computationally intensive. Even just a few thousand stars can really bog things down. There's a decent coverage of the various approaches here, but the upshot is that there's not likely ever going to be a way to do it that doesn't take a really big box or cluster.
posted by gurple at 2:09 PM on February 2, 2006

The "N-word".

Oh, and these animations are amazing. Just imagine how many civilizations are destroyed in this collision.
posted by JeffK at 2:15 PM on February 2, 2006

Ultra cool videos!! Thanks!
This stuff goes on in my mind all the time.
posted by Balisong at 2:15 PM on February 2, 2006

"an exquisite ballet of mutual annihilation"

Nice to know there are others out there who appreciate magnificent violence.
posted by Smedleyman at 2:16 PM on February 2, 2006

super_not: I'm guessing, nigger ?
posted by elpapacito at 2:32 PM on February 2, 2006

Really neat stuff.
posted by brundlefly at 2:36 PM on February 2, 2006

These look exactly like fancy versions of the "galaxy" screensaver that's included with xscreensaver on Linux. They're a little more colorful, but computationally intensive? Not to the point that anyone would notice if you cheated and used simpler math.
posted by odinsdream at 2:39 PM on February 2, 2006

Also - I really enjoyed the link. Beautiful stuff.
posted by odinsdream at 2:41 PM on February 2, 2006

Just imagine how many civilizations are destroyed in this collision.

True. But I'd bet that the number of civilizations that will rise and fall during the course of the collision more than makes up for it.
posted by brundlefly at 2:42 PM on February 2, 2006

The cross-section of interaction is extremely low for star collisions. Collisions between galaxies are far more common than collisions between stars, by several orders of magnitude.

I've seen Dubinski present these simulations at several different conferences. One of the coolest things he mentioned is that his rendering is done on the fly. If I recall correctly, he uses a completely home-brewed code, using Smoothed particle hydrodynamics.
posted by headlessagnew at 3:00 PM on February 2, 2006

The Linux galaxy screensaver is definitely an attempt to reproduce some aspect of the beauty of galactic collision. However, it's a heavily simplified model. In the screensaver, a small number of invisible galactic cores interact gravitationally using n-body simulation. The star particles then interact with these cores, rather than each other. This way, the simulation is O(n+m) in the number of stars n and cores m, rather than O(n^2) in the number of stars, as is n-body.

Nice summary link, gurple.
posted by rlk at 3:02 PM on February 2, 2006

Scott Draves has been doing this for years, except, um, groovier. And distributed.

It's fractal flames and it's called Electric Sheep.
posted by Freen at 3:03 PM on February 2, 2006

Not to the point that anyone would notice if you cheated and used simpler math.

We don't get great leaps and bounds in technology/learning when you cheat the math to make it look pretty.

I assume there was not only a light point, but a spacific gravity for each, as well as a mass/gravitational constant, vector calculated by every other point's gravity, I immagine there must be surface tension, and newton's law for colliding bodies...

If you are going to do it to the limit, do it right.
posted by Balisong at 4:09 PM on February 2, 2006

Hmm, watching these while listening to Tarkus = whoa. It's like being inside the 70s, man. And by the 70s, I mean 2001, which was filmed in the late 60s, but you get my drift.
posted by Eideteker at 7:13 PM on February 2, 2006

I downloaded the 30MB AVI of the collision (Milkyway/Andromeda) and it only plays music. Missing codec?

I figured a galactic collision would mostly be about shifting orbits, since even galaxies are mostly empty space.
posted by Goofyy at 1:42 AM on February 3, 2006

odinsdream: if you'll notice, there are a lot of stars in each galaxy. They start pretty small, but they expand to immense size with very dense coverage. And, as others are pointing out, it looks like the effect is calculating every star compared to every other star, not just 'stars against galaxy centers'.

Very lovely stuff. Thanks for posting it, Zippity.
posted by Malor at 1:48 AM on February 3, 2006

headless is right...when galaxies collide, their stars rarely do, as they are very far apart compared to their diameters...a good visualization:
stars are like ping pong balls in major cities (one in new york, one in chicago, one in l.a....) while galaxies are like dinner plates flying aroung your living room...they crash all the time
posted by sexyrobot at 2:36 AM on February 3, 2006