Scientists Solve Mystery of Star Formation Near Black Holes 88
eonlabs writes "A new paper has been published on the formation of stars in close proximity to a supermassive black hole. Their formation has not been well understood until now, but with the help of a year of supercomputer time, scientists have been able to model the interstellar processes needed to produce them. The results not only match up well with earlier observations, but provide clues as to how their formation is remotely possible. It also helps clear up previous research in this area. 'The simulations...followed the evolution of two separate giant gas clouds up to 100,000 times the mass of the Sun, as they fell towards the supermassive black hole. ...The disrupted clouds form into spiral patterns as they orbit the black hole... In these conditions, only high mass stars are able to form and these stars inherit the eccentric orbits from the elliptical disc.'"
The paper itself was published in Science, but you'll need a subscription to read more than the abstract.
Also on the BBC (Score:5, Informative)
It's also on the BBC News site:
http://news.bbc.co.uk/1/hi/sci/tech/7574255.stm [bbc.co.uk]
complexity analysis, you imposter (Score:2, Informative)
Yes and we're...
Oh, you are, are you?
...ecstatic that it was possible to do so without O(e^n) and taking more than our lifetime.
This type of simulation is very common in computational astrophysics from stellar scale to large-scale structure, and everything in between. The two common computational techniques are the particle mesh methods. It's easy to think about the complexity of the particle problem: every particle interacts with every other particle, giving O(n^2), which is unworkable for even moderate values (astrophysically speaking) of n.
Judiciously sacrificing a little accuracy for a lot of efficiency gets you down from O(n*n) to O(n*log n) calculations. Mesh and hybrid methods used in MHD and FD simulations are similarly tractable; that is, only just, if you have a lot of processors for the (*moderately*) large n in what is now very attainable memory.
No part of the complexity analysis of any of these physical processes is super-polynomial. So I don't think you know what you're talking about, and I certainly don't think you're Ian Bonnell or Ken Rice, nor apparently even a student with basic experience with either complexity analysis or the basic physical concepts being modelled (on a computationally-large scale).
I haven't read the paper yet, but this isn't exactly a bolt from the blue for the researchers; they've published other papers on this subject matter which are typical for the field. They're working on the same problems as lots of people, using the same methods as lots of people, and only "ecstatic" about the whole affair when they actually get supercomputer time, again like other people.
This is both useful and impressive.
It is not at all impressive, which should be obvious to anyone actually doing this research. It would take a very naive and conceited individual to trumpet this research as such if it were his own.
That is all.
Your modesty and lack of impersonation are noted by all.
I know there are at least a few Slashdot readers who use this sort of thing in their work; maybe one will chime in here if the record isn't already straight enough.