Slashdot Mirror
Simulating the Whole Universe
Roland Piquepaille writes "An international group of cosmologists, the Virgo Consortium, has realized the first simulation of the entire universe, starting 380,000 years after the Big Bang and going up to now. In 'Computing the Cosmos,' IEEE Spectrum writes that the scientists used a 4.2 teraflops system at the Max Planck Society's Computing Center in Garching, Germany, to do the computations. The whole universe was simulated by ten billion particles, each having a mass a billion times that of our sun. As it was necessary to compute the gravitational interactions between each of the ten billion mass points and all the others, a task that needed 60,000 years, the computer scientists devised a couple of tricks to reduce the amount of computations. And in June 2004, the first simulation of our universe was completed. The resulting data, which represents about 20 terabytes, will be available to everyone in the months to come, at least to people with a high-bandwidth connection. Read more here about the computing aspects of the simulation, but if you're interested by cosmology, the long original article is a must-read."
Does the simulation include simulated scientists simulating the universe?
"I always wanted to be God." said Dr. Johnson. "When they announced this project, the first words out of my mouth were 'Dibs on God!' I even have plans to introduce a son in a few billion simulated years. This is going to be exciting."
I can search it to find out where I left my cell phone last night.
The only acceptable defense of scientific results is to say that they were the product of the Scientific Method.
The answer is 42. Just google it
Do you have a 1:1 scale map of the world I can use?
Uh yes, but it's being used right now.
... the intergalactic branch of the RIAA has filed a "Cease and Desist" order against the scientists, citing Copyright law; that anyone giving out free copies of the universe without first seeking permission from the copyright holder is a violation of intergalactic intellectual property rights.
Vintage computer adverts: http://www.vintageadbrowser.com/computers-and-software-ads
The same way you simulate anything else. You simplify the problem down to a manageable number of particles that represent larger units of whatever you are simulating. Since in looks like they are interested in mass and gravity at the galactic supercluster scale, they can use particles that weigh much more than any individual star.
The article indicates that the "tricks" these researchers used were the octree and multipole expansion--both of which have been used in gravity and potential theory for many years. They reduce the N^2 interaction problem to N or N Log(N), depending upon implementation. The story makes it sound like these researchers invented the technique; I assume the writer misunderstood the scientists, because it certainly predates them.
Apparently this requires more stack space than exists in this universe.
They learn how large-scale structures formed in the universe
It is, if you don't care about how the universe as we see it came to be. If you do, it isn't.
People have been developing and enhancing codes like this for decades. They're already extremely well-tuned.
Pulling a figure out of your ass
As far as structure formation goes, all you need to know is that there's a chunk of mass.
Largely irrelevant to the effects they're trying to model. Obviously, you aren't a topcoder.
To you and all the other (-1, Redundant) posts on how the system can't simulate every single detail in the Universe: it's a *simulation*, not the real thing, OK?
The first thing you need to do when you plan a simulation is to determine exactly what's significant or not. In this case, they decided that a set of particles with a billion times the mass of our sun would be appropriate. That's because what they are studying is mostly the long range effects of gravitation, where "long range" is defined as a sphere that contains a mass of ten billion suns.
When and if someone wants to study the workings of the Universe at a smaller scale than that, then they will have to simulate at a smaller scale. Phew, people are so dense! Next thing they will say that because a photograph didn't capture every single hair in a person's head or every single pore in their skin, that photo doesn't represent that person at all...
The Read more here link leads to a few pity sentences framing lengthy excerpts from the IEEE article.
BTW, the machine in question, the Max-Planck-Gesellschaft MPI/IPP, is currently ranked 66th. It looks to be a fairly ordinary cluster with none of the exoticism that Cray says we so desperately need
No. Of the four known forces in the universe, only gravity is important in the long range, which defines the overall structure of the universe.
The other three forces are electrical, and two nuclear forces. The nuclear ones are *very* short range, acting only in the atom nucleus. The electrical force is long range, but because there are two different electrical charges, which balance out, there isn't any perceptible electrical attraction in the long range.
I just hope it has a "You are Here-->" indicator so we all know where we are.
1.Netcraft confirms:In Soviet Russia all your base welcomes a beowolf cluster of CowboyNeal overlords. 2.? 3.Profit!!1!
Are they modeling any of the physical (star formation, etc) interactions of matter or just the gravitational interaction. It seemed like the latter, but the article did mention the apparent non-interaction of dark matter.
Bleh!
Last I heard there was some question as to the speed of gravitational attraction. IE if the effect of gravity is only as fast as that of light then the earth is being acted on by the gravity from the point the sun was at 8 minutes ago or some such while the sun is similarly being affected by the earths poistion from 8 minutes ago.
s /NCOR.11.16.D/display.html
As these mass points get further and further apart this would have a huge effect on the results. Unless of course Gravity is instentaneous across any distance opening the door to some interesting possibilities. Namely the ability to communicate across large distances without delay. Perhaps even FTL travel.
While I find this excercise interesting I also find it a tad ridiculose. So many simplifications have to be made to even attempt it and the whole thing is based on some assumptions that are not necesarrily cold hard fact... such as the mass of the universe. Theory says one thing, observation says another. Dark matter was invented to close the gap. Don't get me wrong, there are a lot of smart people that have come up with an awful lot of observation which seems to confirm its existence, but it could be that our point of veiw is insufficient. After all by all observations the Ptolemaic model of the movement of the heavens was accurate and it had all sorts of added rules for handling what was observed.
Also there is the issue of the N body problem where N is greater than 2. Did you know we cannot accurately model our solar system just using keplers laws ? We have to create stabilising factors in the system to keep the planets paths from becoming unstable in their orbits. And yet here they are attempting to simulate an N body problem where N = 10 billion.
http://www.lactamme.polytechnique.fr/Mosaic/image
That link shows what happens with a pure Keplerian system of equations for 9 bodies.
Thus introducing such things as mass simplification for objects farther away ( creating groupings etc ) and the tree approach for close objects all creates an introduction of error into the equation. Further more they have to use some means of stabilizing the equations similar to solar system models which is a value based on observation but with no understanding for what really controls it ( if they don't do this then the system of equations can't model our own solar system much less 10 billion mass points expanding since 380k years after the big bang ). This is all chance for more error to creep into the equation. Then with all of this they run a simulation for a simplified mass points using simplified interactions with an unkown stabilizing force over the course of billions of years and then expect people to believe that what they wind up with has any significant correlation to reality.
Do not be decieved by impressive things like 4 teraflops and 20 terabytes of information. To me this seems an interesting intellectual excercise, but the chances of the results being meaningful are pretty slim.
I don't ask you to be me. I only ask you not expect me to be you.
(disclaimer: I Am An N-Body Modeller, and although I'm not part of the Virgo collaboration, a large fraction of what I do is study cosmological models like the one described)
It doesn't quite come out in the article, but what's really groundbreaking about this work is the number of particles they're using. When you make models like these, you always have to prioritize how large a volume you want to simulate (the more volume you have, the more representative a fraction of the universe you have and the larger number of structures you can analyze) vs how massive the particles are (the smaller the particles, the smaller structures you can analyze).
The more total particles you have, the less you need to compromise your volume or particle mass. Until now, simulating such a large a fraction of the universe (NOTE: unlike what the submitter said, this is not the full universe; as the article itself says, it's about 0.003 of the Hubble volume) required such large particles that it was impossible to say anything about individual galaxies.
However, with 10^10 particles, the mass of their particles is only about 10^9 solar masses, so they can reliably resolve structures of 10^11 solar masses. For reference, the mass of the Milky Way is roughly 10^12 solar masses. This is a fantastic leap forward - most other modern simulations have 10^8 - 10^9 particles, and so either can only simulate a much smaller fraction of the universe (like the simulations I study), or cannot say anything about galaxies, only massive galaxy clusters.
[TMB]