Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
×
Science

IBM and LLNL Scientists Show How Stuff Breaks 15

Maxim writes: "An unprecedented billion-atom calculation has enabled a team of IBM and Lawrence Livermore National Laboratory (LLNL) scientists to demonstrate a major advance in using supercomputers to simulate the strength of materials. Check out the video."
This discussion has been archived. No new comments can be posted.

IBM and LLNL Scientists Show How Stuff Breaks

Comments Filter:
  • by tps12 ( 105590 ) on Tuesday April 30, 2002 @09:59AM (#3435606) Homepage Journal
    IANALLNLS (I am not an LLNL Scientist), but I see some cool applications of this technology on the horizon.

    First, whenever someone says "simulation," I immediately think "games." Call it instinct or hypnosis (or maybe both!), but it's just a gut reaction for me. I mean, billions of atoms. We're talking physics so realistic that the holodeck might not be so far off! Of course, it may be some time until graphics catch up with this (LOL), but I think we're all in for a treat.

    Next of all is Science. Imagine the cost savings when full physics experiments can be carried out with computers! Plus you avoid all of the danger normally associated with Science (i.e., the Frankenstein complex). The implications for astronomy, anatomy, chemistry, geneology, engineering, geography, and anthropology boggle the mind.

    Finally, I think we will finally start to get some answers. One of the biggest barriers to creating the Grand Unifying Theory of Everything (commonly believed to be a combo of Relativity and Quantum Mechanics...not an easy feat) has been the limitations of human reason. On the other hand, computers (which are very good at logical reasoning) lacked the firepower to handle these huge models. If this signifies the breaking down of these barriers (and I think it does), then we might finally get things figured out.

    • Atomic level physics for games would be extremely cool but I think that's still pretty far off. Consider the computing power needed to model an entire world atom by atom, in real time...

      I haven't even seen any good fluid dynamics in games (rivers and waterfalls are always flat and premodeled), or even really convincing rigid body dynamics for that matter.
      • I haven't even seen any good fluid dynamics in games (rivers and waterfalls are always flat and premodeled), or even really convincing rigid body dynamics for that matter.

        This is the beauty of modelling at the atomic level. Once you can accurately model the components of an atom, you just assemble them in different amounts, and the system takes care of itself from there: phase changes and macroscopic behaviors fall out of it.

        It's just like agent-based simulations. Instead of modelling a whole huge-ass thing, you just model the smallest components, and let the computer assemble the macro-simulation.

        IIRC, this is how technologies such as flight sims, voting systems, lottery generators, and letter-sorting machines function. It is also the philosophy on which Linux was based (if you look at it right)! Remember: "if you build it, they will come."

    • by 4of12 ( 97621 )

      Imagine the cost savings when full physics experiments can be carried out with computers!

      Yes, in many cases.

      But I think generally it's over-hyped.

      In the good old days, before computers were available for these kinds of calculations, scientists had to scrimp and scratch their way using classical tools of mathematical analysis.

      These days, everyone runs to the computer before they have the problem properly defined.

      Not to belittle this accomplishment. I think it's still possible to milk a great deal of insight from computational simulations. It's just that most of the time what's milked is a glossy viewgraph to get the next phase of funding...

    • We're talking physics so realistic that the holodeck might not be so far off!

      Reality check calculation:
      Alvogadro's constant:6x10^23
      My body:80kg
      Average grams per mol in my body: 16.5 (assuming 75% water, 25% carbon)
      mol in my body:80kg/16.5g ~= 5000
      Atoms in my body:5000*6x10^23 = 3x10^27

      Atoms in the computer simulation:1x10^9
      Difference in orders of magnitude:18

      Years till simulation:92 (assuming doubling of processing power every 18 months.

      Yup, I'd say holodeck's are just around the corner.
    • Beowu....what?. Oh. Yes, I was just leaving...I'll get my coat.
  • When ... (Score:4, Funny)

    by QuodEratDemonstratum ( 569501 ) on Tuesday April 30, 2002 @10:07AM (#3435665) Homepage

    "Stuff" breaks when you put, on Slashdot, a link to a 63Mb movie.
  • by rtstyk ( 545241 ) on Tuesday April 30, 2002 @01:01PM (#3437046) Homepage
    I'm not a physicist by any stretch of the imagination so take my ramblings with a grain of salt.

    It seems to me that there is a danger in relying too much on simulations. The particle physics has not been fully understood. Simulating how atoms are interacting is great but does the simulation take into account that atoms are made up of other things? There are other forces at work there (I heard :). How accurate can this be? Or is it because any forces within the atoms have no effect on the end result of the breaking of the cube?

    Taking this further, can the simulation show something that isn't expected? In other words can the simulation show an effect such that it was not conceivable? I think not. It can only slow down and reveal a known process right? Otherwise it would theoretically be possible to build an atom level model of the world and speed it up to see what happens. So there is the problem of the computers being unable to come up with anything new.

    The second problem is that the computers in the end depend on a binary decision: 0 or 1. Can atoms on their lowest level be reduced to components of 0 or 1? Or are they infinitely reducable?

    Anyway, going back to the article I think it's great. I can wait for the new SuperGlue (concrete in less than a second!)

    d.
  • by QuantumFTL ( 197300 ) on Tuesday April 30, 2002 @01:19PM (#3437199)
    I am an undergraduate at Cornell University, and I use the Velocity Cluster, the largest cluster of Windows boxes in the world. It's been mentioned on slashdot before, and the system there is actually pretty cool. Anyways, I had remembered seeing that Cornell, in partnership with several other universities, was doing something very similar to this.

    The page about the Cornell Computational Materials Institute can be found here [cornell.edu].

    There's some pretty nifty stuff going on with that... crack propagation is an especially important problem in materials science, and it's nice to see that these large computers can be used to help solve it.

    • "crack propagation is an especially important problem in materials science, and it's nice to see that these large computers can be used to help solve it."

      Yes I hear you material scientists rely far too heavily on smoking crack to get you through those tedious simulation runs..:-S

  • You're all forgetting a little something called the Heisenberg Uncertainty Principle.

    There will always be non-determinism on the atomic/quantum level involved because of it, and makes it impossible to fully simulate and predict effects with *100% certainty( since you don't know how the atoms are going to behave, wrt their position, and momentum.

"Our vision is to speed up time, eventually eliminating it." -- Alex Schure

Working...