What Is Computational Physics?

Rhett Allain has published his take on what "computational physics" is.

Many of us practicing physicists do work in computational physics. Some very often, some now and then. At some point, many of us have to either analyze data, do numerical modeling, or solve intractable equations. We either use pre-made codes, modify some other computer codes, write our own code, or use commercial software.

But certainly, this is less involved than someone who specializes in computational physics. But many of us do have the need to know how to do some of these things as part of our job. People who have to simulate particle beam dynamics, and those design accelerating structures are often accelerator physicists rather than computational physicists.

Hum... now I seem to be rambling on and can't remember the point I was trying to make. Ugh! Old age sucks!

Zz.

Unfortunate Superfish

I hope this doesn't taint the name "Superfish".

In case you missed it, this week came the news that Lenovo, the Chinese computer company (to whom IBM sold their ThinkPad laptop series) had been installing on some of their computers a rather nasty tracking software called Superfish Visual Discovery.

I would have paid that much attention to stuff like this weren't it for two reasons: (i) I own a rather new Lenovo laptop and (ii) I am familiar with the name "Superfish" but under a different context.

Luckily, after doing a thorough check of my system, I found no sign of this intrusive software. As for the second reason, there is a rather popular software called "Superfish" out of Los Alamos National Lab that we use quite often. It is a Poisson EM solver often used to solve for EM fields in complex geometry/boundary conditions. I'm guessing that they gave it that name because "poisson" is French for "fish", and it really is a super software! :)

It is unfortunate that in the context of computer technology, the name Superfish is "poison".

Zz.

Sorry, Capt. Kirk. Teleporting You Will Just Take Too Long

... in fact, longer than the age of the universe!

This fun, amusing, and back-of-the-envelope calculations done by a group of students at the University of Leicester in England shows that to sample and transport all the information about a human to another location (as in the Star Trek teleportation device) will just take too long for it to feasible right now.

Our universe has been around for 13.8 billion years. But the seriously tongue-in-cheek paper shows that at a beaming speed of 30 gigahertz, transmitting all the data within a single human would take 4,850,000,000,000,000 (4.85 quadrillion) years. The human dataset includes not only the person's genetic code but also all the memories and knowledge stored in his/her brain. While DNA would take up about 10 billion bits, the brain's information would bring the data total up to 2.6 x 10⁴² (26 followed by 41 zeroes) in bits.

Why not simply increase the bandwidth to speed up the process? That would require an impossible increase in power consumption beyond earth's capabilities.

I think there's something else here that should be pointed out, and something that many non-scientists should be aware of. When we think "outside of the box", we must (i) first know where the box is and where its boundary is (which means one must have the established knowledge first), and (ii) be able to make quick, back-of-the-envelope estimation to tell us if the quantitative aspect of what we wish to do or study produce a reasonable, achievable goal.

The latter is quite important, because it tells us if there's any reasonable chance that we can verify, or accomplish our goal. This means that a device, a detector, an instrument, etc. of certain parameter can either produce, detect, or accomplish that goal. When we seek support to do such work, we must show this because, as is very obvious, no one is going to fund something based on some handwaving argument, and they are especially not going to fund something if the numbers tell us that there's no REASONABLE chance for success!

So if someone wants funding to build such a teleportation device above, would you fund it knowing that it will take 4.85 quadrillion years to transport a human being? If you would, I have a bridge to sell to you.

Zz.

Possibility of LHC Data Pile-Up

Too  much of a good thing is not necessarily a good thing.

So they want a lot of collisions, now they might get too many to handle. The detectors at LHC are preparing for the possibility of data pile-up, now that the LHC is operating at a higher energy and higher luminosity than what the detectors are designed to handle.

Every time two tightly packed bunches of protons cross, they generate not one collision, but on average 27, Lamont says. But within a few weeks, that number is expected to rise into the mid-30s, peaking at around 40 collisions per crossing. The two main detectors at the LHC were designed to handle only around two dozen collisions at once. But they have managed to cope so far.

While this is a good problem to have, it is still a problem, because you simply don't want to have to discard something simply because you can't handle it. It appears that they can, so far, but I can easily see that this number will continue to increase. Considering the daunting size and what they are trying to do, the LHC continues to be an astounding machine that is performing incredibly better than expected.

Zz.

Proof That Physics Is Hard!

Oh, I didn't think such a "proof" exists, but it seems that it does!

Students and researchers alike have long understood that physics is challenging. But only now have scientists managed to prove it. It turns out that one of the most common goals in physics—finding an equation that describes how a system changes over time—is defined as "hard" by computer theory. That's bad news for physics students who hope that a machine can solve all their homework problems, but at least their future jobs in the field are safe from automation.

So there you are! :)

I'll update this when I have the exact reference to the PRL paper.

Edit (3/22/2012) Here is a link to a review of this paper and the exact reference to it in PRL:

http://physics.aps.org/synopsis-for/10.1103/PhysRevLett.108.120503

Zz.

LHC@Home Attracts Large Support

I mentioned this a while back. It appears that a lot of people are eager to participate in the distributive computing effort to do a lot of simulations for the LHC. Since its announcement, the LHC@Home has received a lot of support.

The application Test4Theory, which runs Monte Carlo simulations of events in the LHC, was announced in a CERN press release on 8 August. Within three days, the number of registered volunteers swelled from a few hundred to nearly 8000.

Certainly nothing to sneeze at. Still, they're expecting 40,000 participants in this project, and that will result in an interesting consequence:

According to CERN's Peter Skands, the physicist leading the simulation effort, when the number of active volunteers passes 40,000 – which could happen later this year – the system will become equivalent to a true "virtual collider", producing as many collisions per second as the real LHC.

Fire away!

Zz.

LHC@Home

It seems that the LHC has also gotten into this distributed computing stuff. They are looking for idle computers (such as yours when you're not using it) to do some distributed computing. Can't guarantee that your computer will be hunting for the Higgs, but maybe it might run some code to do some Monte Carlo background simulation! :)

Zz.

The Humongous Computing Task of the LHC

We should not forget that besides the scientific achievement of the LHC, the task of handling the humongous amount of data coming it at such a rapid rate in itself pushes the computing knowledge and ability to a new level. By itself, it forces new innovation in data handling and computing techniques that many commercial companies simply can't do. Yet, this is something we need to know if our technological needs are to continue to expand.

This news article (link may be open for free only for a limited time) examines the astounding effort in piping out data gathered from the ATLAS detector, and how many groups around the world are working together to handle such volume and analysis.

Even after rejecting 199,999 of every 200,000 collisions, the detector churns out 19 gigabytes of data in the first minute. In total, ATLAS and the three other main detectors at the LHC produced 13 petabytes (13 × 1015 bytes) of data in 2010, which would fill a stack of CDs around 14 kilometres high. That rate outstrips any other scientific effort going on today, even in data-rich fields such as genomics and climate science (see Nature 455, 16–21; 2008). And the analyses are more complex too. Particle physicists must study millions of collisions at once to find the signals buried in them — information on dark matter, extra dimensions and new particles that could plug holes in current models of the Universe. Their primary quarry is the Higgs boson, a particle thought to have a central role in determining the mass of all other known particles.

Zz.

Computers Have A Fundamental Speed Limit?

That's what two physicists have argued in a recent PRL paper[1].

Abstract: How fast a quantum state can evolve has attracted considerable attention in connection with quantum measurement and information processing. A lower bound on the orthogonalization time, based on the energy spread DeltaE, was found by Mandelstam and Tamm. Another bound, based on the average energy E, was established by Margolus and Levitin. The bounds coincide and can be attained by certain initial states if DeltaE=E. Yet, the problem remained open when DeltaE[not-equal]E. We consider the unified bound that involves both DeltaE and E. We prove that there exist no initial states that saturate the bound if DeltaE[not-equal]E. However, the bound remains tight: for any values of DeltaE and E, there exists a one-parameter family of initial states that can approach the bound arbitrarily close when the parameter approaches its limit. These results establish the fundamental limit of the operation rate of any information processing system.

In fact, if we go by with Moore's law, the prediction comes to roughly another 75 years before this speed limit is reached.

If components are to continue shrinking, physicists must eventually code bits of information onto ever smaller particles. Smaller means faster in the microelectronic world, but physicists Lev Levitin and Tommaso Toffoli at Boston University in Massachusetts, have slapped a speed limit on computing, no matter how small the components get.

"If we believe in Moore's laW ... then it would take about 75 to 80 years to achieve this quantum limit," Levitin said.

"No system can overcome that limit. It doesn't depend on the physical nature of the system or how it's implemented, what algorithm you use for computation … any choice of hardware and software," Levitin said. "This bound poses an absolute law of nature, just like the speed of light."

Still, 75 years is a very, very long time as far as technology is concerned. While a fundamental limit is a fundamental limit, I can certainly see new physics popping up in 75 years that will require a re-evaluation of this conclusion.

Zz.

[1] L.B. Levitin and T. Toffoli, Phys. Rev. Lett. v.103, p.160502 (2009).

More Really Is Different

I've mentioned before, many times in fact, on here about the issue regarding reductionism and "The Theory of Everything". I argued that a large fraction of practicing physicists, especially those in condensed matter physics, do not buy into this idea of a theory of everything, simply based on emergent phenomena that we observe in condensed matter physics.

The main "poster child" for such argument is Phil Anderson's "More Is Different" paper published quite a while back. Now comes a paper that strengthen Anderson's assertion that More Really Is Different.

Gu et al.[1] published a paper in which, using the 2D Ising model that simulates a cellular automata of magnetic spins.

Abstract: In 1972, P.W. Anderson suggested that ‘More is Different’, meaning that complex physical systems may exhibit behavior that cannot be understood only in terms of the laws governing their microscopic constituents. We strengthen this claim by proving that many macroscopic observable properties of a simple class of physical systems (the infinite periodic Ising lattice) cannot in general be derived from a microscopic description. This provides evidence that emergent behavior occurs in such systems, and indicates that even if a ‘theory of everything’ governing all microscopic interactions were discovered, the understanding of macroscopic order is likely to require additional insights.

A News and Views in Nature[2] this week article reviewing this paper explains it a little bit clearer for those of us not familiar with these 2D computation and the intricacies of cellular automata.

In their study, Gu et al. mapped the dynamics of a certain CA into the lowest-energy (ground) states of Ising models. In this framework, Figure 1 can now be interpreted as a snapshot of a two-dimensional spatial lattice of spins. They grouped spins into blocks that encode the logic operations needed to produce universal computation in the corresponding CA. They then defined the 'prosperity', p, of two-state systems as "the probability that a randomly chosen cell at a random time step is live" (live meaning state 1).

Using the computational properties of the CA, Gu and colleagues were able to show that p is undecidable for infinite, periodic Ising systems. They argued that, as a consequence, many macroscopic properties of an Ising system, including the system's magnetization and degeneracy (number of independent configurations) at zero temperature, depend on p and hence are also undecidable. Because Ising models have been used to describe not only magnetic materials but also neural activity, protein folding and bird flocking, the consequences of Gu and colleagues' results transcend both computer science and physics.

Nice stuff! This would be another compelling argument against reductionism and the fallacy of the "Theory of Everything".

Zz.

[1] M. Gu et al., Physica D: Nonlinear Phenomena, v.238, p.835 (2009). Also see the arXiv version here.
[2] P.M. Binder Nature v.459, p.332 (2009).

For RHIC and LHC, Data Is King!

While people are very caught up with the powering up of the LHC and anticipating first collision, many people forget that there's a another monumental task ahead for many people - the handling of an believable amount of data that will be coming out of the various detectors at the LHC. No amount of real-world application comes close to matching the data-handling task that has to be carried out once the LHC is in full operation.

This article looks at the daunting task of distributing just the anticipated data coming from the ATLAS detector at the LHC.

As the sole Tier 1 computing facility for ATLAS in the United States, Brookhaven provides a large portion of the overall computing resources for U.S. collaborators and serves as the central hub for storing, processing and distributing ATLAS experimental data among scientists across the country.

This mission is possible because of the ability to build upon and receive support from the Open Science Grid project, Ernst said. Yet, even after ramping up to 8 petabytes of accessible online storage – a capacity ten times greater than existed when ATLAS joined the RACF eight years ago – the computing center’s scientists still have plenty of testing and problem-solving to conduct before the LHC begins operations this fall.

It is not an understatement that many advances in computing that we see today were driven by the needs of scientific projects such as this. What is being done here will eventually trickle down to various parts of society in a few years.

Zz.