Saturday, June 1, 2013

The Lone Genius Hypothesis

When I was a little kid, I knew I wanted to be a cosmologist, like Stephen Hawking. I would tell people that my dream was to have a little office somewhere with a giant blackboard, and I would fill it with equations and solve the mysteries of the Universe. Sometimes I imagined that instead of working in my office, maybe for a change of scene I would sit under an apple tree staring off into the sky, contemplating the nature of reality.

Thinking about doing physics. Photo credit: Demelza Kooij

That's not exactly how things turned out. True, there are times when I sit alone in my office and scribble equations. There are times when I sit outside and stare and think. But, to be honest, those times are usually not especially productive. When I really make progress, when I really have breakthroughs -- those are always times when I'm talking to other physicists and astronomers, chewing through new ideas and checking that I'm on the right track. And even more often, the most important work we do is what grows organically from our conversations or e-mails or paper perusals. Sometimes it's hard even to know who should get the credit.

Actually doing physics. Photo credit: CAASTRO.

So I was wrong. But I think adolescent-Katie could be forgiven for imagining a future career of solitary contemplation. When we're presented with images of great theoretical physicists, the picture is almost always of a lone genius, hidden away with a blackboard, making leaps no one else could have seen, using nothing but pure, unadulterated mind-power. (That those lone geniuses are almost always depicted as male is the subject for another discussion entirely.) This image has been brought to the forefront once again in the last couple of weeks by the media attention heaped on the mathematical physicist-turned-hedge-fund-consultant Eric Weinstein, whose name happens to be only one letter off from Einstein, which just adds fuel to the media-hype fire. He has a homemade theory of everything, developed over many years by himself, in his spare time, which he is just now announcing to the public. I haven't attended Weinstein's lectures and I haven't seen his work (very few people have so far), so I'm not going to comment on its genius or lack thereof. I also won't comment on the media attention per se, as others have done plenty of that. What I will say is that the W/Einstein lone-genius model of theoretical physics is, nearly always, in stark contrast to how theoretical physics is actually done.

But... Einstein!

One of the reasons Einstein carries such a hefty cultural weight is that he, like Newton a few centuries before him, appears to have basically single-handedly invented a fundamentally new view of the Universe. Newton did it over the course of 18 months, starting in 1665 while isolated to avoid the Plague, revolutionizing optics and gravity, and inventing calculus along the way. Einstein's turn came in his "annus mirabilus" in 1905 when he published four groundbreaking papers and a PhD thesis. These touched on optics, the size and motions of atoms, and, as you might have heard, the theory of special relativity.

This approach doesn't usually work out. Photo: Associated Press, found here.

Einstein is frequently depicted as having been completely cut off from the academic establishment during this time, being "just a patent clerk." But although he was certainly not a working academic physicist, he still had connections with the community, people to bounce ideas off of, and a (stalled) PhD-in-progress at the University of Zurich. He had also published several papers, though they didn't receive much attention. And working as a patent clerk was actually a fairly technical job, involving evaluating new ideas and requiring a deep understanding of science and engineering. He is, however, I will grant, probably the best example in the modern era of a theoretical physicist revolutionizing science from outside "the establishment." In fact, he's the only one I can think of.

"Hey, who invented quantum mechanics?"

I asked this question of a colleague of mine while writing up this post, not because I thought he'd have a single answer, but because I was curious what the list might look like. There are a few people who should probably get some credit: Maxwell, who first formulated the basic equations of electromagnetism; Hertz, an experimentalist who helped demonstrate the photoelectric effect; Planck, who was so important to quantum theory that its most fundamental constant is named after him; Einstein, who first explained the photoelectric effect from a theoretical point of view; or Pauli, or Heisenberg, or Bohr, or Nernst, or Schroedinger... there was kind of a lot going on around that time. The point is that quantum mechanics is a great illustration of the fact that it doesn't take a lone iconoclast to revolutionize our understanding of the Universe. Even huge breakthroughs that fundamentally change how we see and do physics can come about through a series of incremental steps. Experimentalists see something odd in their experiments, theorists propose possible explanations, experimentalists go back and test the consequences of that theory and the cycle begins again.

This has happened a number of times since Einstein's era. In addition to quantum mechanics, we've seen the appearance of the Standard Model of Particle Physics, quantum field theory, the concordance model of cosmology (including dark matter and dark energy), and the as-yet purely theoretical frameworks of supersymmetry and string theory. None of these advances could be attributed to one person, nor did they generally involve people working in isolation on theories of their very own.

So how does it usually work?

Physics is, these days, an immensely collaborative field. There are a lot of conferences. There are institutes and workshops and collaboration visits and endless seminars and dissections of research papers. Newly built physics institutes tend to have hallways lined with blackboards or dry-erase-glass cubicles to get people out of their offices to collaborate. We talk to each other, not because we are inherently very social (though a lot of us are), but because it's a really productive way to proceed. Personally, I find I think better when I'm explaining my ideas to someone. Some people, after staring at the same equations for days, just need to get the math written down and show it to other people to make sure it really makes sense. And, even more importantly, we're not all experts on all areas of physics. One person might have spent four years working on a particular quantum mechanical process in the early universe while another might be an expert on strong-field gravitation, and together they can create a much clearer picture of, say, how gravitational waves might be produced right after the big bang. Or two people might have slightly different perspectives on the same subfield of physics, because they were taught by different people or did projects on different things. For whatever reason, it turns out that talking to other physicists is one of the most productive things a physicist can do, if he or she wants to really make a breakthrough. And here I'm just talking about pure theory -- if you want to actually test any of this stuff, to see if it's on the right track for describing the actual universe in which we live, you have to be in touch with experimentalists and observers and find out what kind of tools they have available too.

Progress through collaboration: CMS at the LHC. Photo credit: CERN.

The way theoretical physics is funded (though keep in mind the funding system has its own problems) is a good clue to what we've found to be successful over time. Unless perhaps you win a MacArthur "Genius Grant," neither grant decisions nor academic hiring are determined solely by how incredibly brilliant you are. They're determined by how much science you produce, how good it is, how much it adds to previous research, and how your advisors and collaborators see your work. "Quality of the Investigator" is only one section of a grant application -- you have to also explain how your work fits in with the work of others at the institute where you'll work, and why it's a good environment for you. I've actually had a fellowship application rejected based entirely on the institute not being "a good fit" -- the assumption being that without anyone to talk to, I just wouldn't be all that productive there.

So, the synergy factor is not to be dismissed. (You would be amazed how many papers out there include in the acknowledgments something like "We thank [conference/workshop] where part of this work was carried out.") Smart people are smarter when they work together.

What about the W/Einsteins of the world?

To clarify again: I have no intention of passing judgment on Weinstein's ideas. It's entirely possible he's onto something incredible, and it's entirely possible the work will lead to nothing at all. It might turn out to solve all the problems of cosmology, or it might already be ruled out by experiment. I haven't seen the paper or heard the talk, so there's really no way for me to hazard an educated guess.

But I will say that if you think you might want to solve the biggest outstanding problems in theoretical physics, I don't recommend the lone-genius approach. Maybe Weinstein had some really good reason not to talk to other physicists about his work before now. Perhaps he was worried it might be wrong and didn't want to embarrass himself, or perhaps he was worried it might be right and he'd be scooped or not get all the credit. Or maybe he just doesn't like to talk to physicists all that much. It's even possible that he thought his ideas were so revolutionary that no one else would understand. But I kind of doubt that. We physicists love finding new ways to think about things. We love stretching our minds and seeing things from another point of view. It's why we do this work at all. And it's why we spend so much time talking to each other about it.

Saturday, April 27, 2013

The Art of Darkness

The Universe is a very dark place.
The contents of the Universe, according to recent results from the Planck Satellite. Image copyright ESA.
This post focuses on the blue bit; for more on the pink segment, see my earlier post.

You've probably been hearing a lot about dark matter detection lately. In the past couple of months, there have been announcements of announcements, delays of announcements, press conferences, ambitious claims, cautious optimism, not-so-cautious optimism, and various "hints," "signs" and "clues." But what does it all mean? Have we actually detected dark matter?

Short answer: Um, maybe. Also: It's complicated. Really complicated.

The Truth is Out There

I'll start by saying the one thing we're really pretty sure of: dark matter is real. We've known for a very long time that the matter we can see in our telescopes -- stars, galaxies, gas, dust -- doesn't have enough gravitational pull to explain the motions of the cosmos. There are some very good explanations of dark matter and its evidence on the web out there already, but in brief, given how fast stars and galaxies are moving (stars moving in galaxies, galaxies moving in clusters), the matter we can see isn't enough to hold them all together. The first evidence of this came out in the 1930s, and since then, astronomers have hypothesized that some mysterious new component of matter that we can't see -- dubbed dark matter -- is pervading and surrounding galaxies and clusters and keeping everything from flying off into space.

Sorry about the afterimage.
Artist's impression (well, mine) of a galaxy embedded within a spherical dark matter "halo." Image of Andromeda Galaxy credit GALEX, JPL-Caltech, NASA, from APOD.

If this was the only evidence, it might be reasonable to suggest that it's not a new form of matter, but rather an altered law of gravity that explains the inconsistency. But it turns out that evidence for dark matter being a fundamentally new kind of matter pops up virtually everywhere we look -- from the way light bends around massive objects, to the history of galaxy formation, to the chemical make-up of the early universe. Some of the strongest evidence for dark matter is found in the aftermath of collisions of galaxy clusters, since these cosmic train wrecks can effectively separate dark matter from stars and gas.

The Bullet Cluster, a.k.a., dark matter's smoking gun.
Composite image of the Bullet Cluster of galaxies, with optical Hubble Space Telescope image, Chandra X-ray image of ionized gas in pink, and dark matter abundance determined from gravitational lensing portrayed in blue. Credit: X-ray: NASA/CXC/CfA/M.Markevitch et al.; Optical: NASA/STScI; Magellan/U.Arizona/D.Clowe et al.; Lensing Map: NASA/STScI; ESO WFI; Magellan/U.Arizona/D.Clowe et al. Annotated image from animation found here.

So we know dark matter is out there, but we don't know what it is. We think it's probably some kind of new elementary particle, and the leading theories all suggest that it should have some interaction with light and/or ordinary matter (i.e., particles contained in the Standard Model of particle physics), so over the last few decades the physics community has put a lot of effort into finding a way to detect those interactions. There are basically three approaches:

  • Direct detection: If dark matter is a new elementary particle that interacts mainly via gravity and only very weakly via any other force, dark matter particles should be passing through the Earth all the time, and, very occasionally, you'd expect one to bump into something. Direct detection experiments look for that collision, called "nuclear recoil" (because you're looking for the movement of the atomic nucleus, not the electrons). Basically they put a box full of some target material (in the case of the CDMS experiment, that's silicon or germanium) in a heavily shielded lab deep underground where virtually no standard model particles can get in. Then, very sensitive detectors watch for one of the target nuclei to be bumped. If the scientists can rule out other explanations for the bump (like radioactive decay of the material around the target sending in neutrons, for instance), and if the recoil energy is what they expect dark matter to produce, then they have a dark matter event candidate.
  • Indirect detection: In many of the models of dark matter, the dark matter particle is its own antiparticle, which means that if two dark matter particles collide precisely enough, they annihilate. In theory, this produces standard model particles that we can see. If that's correct, then one way to find dark matter particles is to look at where dark matter is densely concentrated (like in the Galactic Center) and see if there are gamma rays or high-energy particles being produced in a way that ordinary astrophysics can't explain. There are other ways dark matter particle physics could be probed with direct detection, like if the particle decays or has other (non-annihilating) interactions with itself or other matter, but annihilation is the most common thing to look for. One of the reasons we think annihilation happens is because it leads to a natural way to explain the production of dark matter in the early universe -- the idea being that dark matter was annihilating and being produced all the time in the beginning when the universe was very dense, and it was only when expansion allowed dark matter particles to interact less frequently that they were able to exist in a more or less stable way for long periods of time.
  • Collider production: If two dark matter particles can annihilate to make standard model particles, then you should be able to reverse the process and make dark matter particles by colliding standard model particles at high energies. This is the idea behind the search for dark matter at colliders such as the LHC. A dark matter particle produced in a collider would pass right through the surrounding detectors without leaving a mark, so the way we'd see it would be to look for "missing energy." You add up all the energy of all the particles you do detect in the collision aftermath, compare it to the total energy you put in, and see if the missing energy is consistent with what a dark matter particle would spirit away.
Ways to make, destroy, or detect dark matter particles.
Different ways to detect dark matter (DM) particle interactions with standard model (SM) particles. Image found on the MPIK website, originally produced by Jonathan Feng. "Thermal freeze-out" is what happens when the dark matter is no longer dense enough to annihilate all the time due to the expansion of the early universe.

So, have we found it yet?

There's been a lot of hype. A few weeks ago, the team behind an experiment called AMS-02, a cosmic ray detector that hangs off the side of the International Space Station, made an announcement that they found a signal "consistent" with dark matter, but that it was "not yet sufficiently conclusive to rule out other explanations." They held a press conference and released a short publication summarizing the work. As I pointed out in a blog post for IoP's Physics Focus, the tone of the announcement, and the especially media hype that followed, went far beyond what was really justified by the results. What they actually saw was an excess of positrons over what would be expected from standard astrophysical processes. The excess might have arisen from dark matter annihilation. But it could also have come from something else. Like pulsars, which are known to accelerate particles and which could certainly produce a positron excess like the one seen by AMS. I go into more detail on this in my Physics Focus blog post, but the gist is that while the AMS signal is intriguing, it's really difficult to pin it on dark matter with any degree of certainty.

But that didn't keep the media from running away with the idea. Here's a sampling of the kinds of headlines I saw in response to the AMS result:

"Experiment believed to detect evidence of dark matter" - Boston Globe
"Strong hints of dark matter detected by space station, physicists say" - Guardian
"CERN Scientists Continue to Prove Their Value with First Evidence of Dark Matter" - Atlantic Wire
"Hints of Dark Matter Have NASA Scientists over the Moon" - Space News

Being excited about the prospect of a big discovery is fair, but overhyping it doesn't help anyone. Especially because only a couple of weeks later, another experiment, called CDMS, also claimed a possible detection of dark matter, and news articles said pretty similar things, sometimes without even referencing AMS:

"Researchers May Have Finally Detected a Dark Matter Particle" - Universe Today
"Homing in on Dark Matter" - Sky & Telescope
"Dark matter researchers think they've got a signal" - The Register
"Another dark-matter sign from a Minnesota mine" - Nature News Blog

Actually, the CDMS result got quite a bit less press, which was surprising to me. Pretty much any way you look at it, it's a much more direct result, if (as I'll explain) fantastically confusing.

Deep dark secrets

CDMS (or, specifically, CDMS-II) is an underground dark matter direct detection experiment. It's located in an old iron mine in Minnesota and it consists of super-cooled targets of silicon and germanium surrounded by sensitive detectors that can measure the positions and energies of any movements they see in their target nuclei. They expect to see, as a background, electron recoils from a variety of processes, and they can distinguish these from recoils of nuclei by looking at the way bumped electrons would ionize the target material. There are a number of ways they slice up the data to take out the electron background, but they also expect a tiny number of neutrons to get into the detector (either from space or from radioactive decay more locally) and bump into their target nuclei, and these would look exactly like dark matter collisions. The only way to deal with those is to estimate the number they expect from neutrons, and get excited if they see way more than that.

"Aww, look at the little WIMPy candidates." -@sc_k
The dark matter candidate events found by the CDMS-II experiment. Plot from presentation by Kevin McCarthy at the APS meeting. The full presentation can be found here and the paper is here. I was alerted to this plot by this tweet.

In the end, CDMS found three candidate events. In the plot above, they're labelled Candidate 1, 2 and 3. (I hope the CDMS folk actually named the candidate events, in the style of the IceCube collaboration, who found two extragalactic neutrinos and called them Bert and Ernie.) The collaboration claims that the chance of these events actually being dark matter -- as opposed to misidentified background or random chance -- is 99.81%. That corresponds to what we call a 3-sigma result, which, by particle physics convention, is officially "evidence" but not officially a "detection." For comparison, the Higgs Boson discovery was deemed a true discovery when it reached 5-sigma.

Mixed signals

Obviously, a result at 99.81% confidence, while maybe not quite a detection, is intriguing. And, due to CDMS's ability to distinguish backgrounds, I would say it's far more intriguing than the AMS result as far as dark matter implications are concerned. But there are a number of very good reasons the physics community is staying cautious on this one. The biggest reason is that the simplest model of dark matter that could explain the CDMS result has already been ruled out by other experiments. There are lots of detectors in the direct detection game right now, and at the moment, many of them seem to be giving us very conflicting information. There have been detections -- tentative or otherwise -- claimed by four different experiments now, if CDMS is included. The others are DAMA/LIBRA, CoGeNT, and CRESST -- and, actually, a previous signal was claimed by CDMS but has since been considered more likely to be background). All these results could be signs of a dark matter particle -- specifically, a weakly interacting massive particle (or WIMP), but it's difficult to find a way to make them agree with one other. They all seem to find particles with different masses and interaction rates. Even worse, combining the constraints from other experiments, such as XENON and EDELWEISS, and even previous results from CDMS, seem to rule out all the claimed detections.

"looks like Pollock's painting" -Resonaances Blog
Constraints and hints from direct detection experiments. The horizontal axis is the mass of the dark matter particle and the vertical axis measures its interaction with standard model nuclei. Filled regions indicate signals interpreted as dark matter; lines indicate upper limits. Everything to the upper right of a line is ruled out to 90% confidence by that experiment.  The lines are, roughly from left to right: XENON100 (dark dash-dotted green), XENON10 (light dash-dotted green), CDMS II Ge (dark and light dashed red), EDELWEISS (orange diamonds), and CDMS II Si (dark blue solid and black dotted). The asterisk is the best-fit point for CDMS's candidate events. This plot and more details can be found in arXiv:1304.4279 by the CDMS Collaboration.
AMS wasn't discussed in the CDMS paper, but I should point out that the best candidate dark matter model for the AMS result and the CDMS dark matter candidates do not agree either. It's a little difficult to compare them directly, because one is looking at dark matter annihilation and the other at dark matter interactions with nuclei, but the inferred particle masses are very different. To explain the AMS result, the dark matter particle would need a mass in the TeV (trillion electron-volt) range, whereas CDMS needs a particle with a mass a thousand times lighter. (Even though it's technically energy, an electron-volt is used as a measure of mass for fundamental particles, via E=mc2. GeV is a billion electron-volts and TeV is a trillion. For comparison, a proton is 0.938 GeV.)

AMS? CDMS? Total mess?

In the astro/physics community, the response to the result from CDMS has been mixed.

It's really not clear what we should make of all these conflicting results, and it's even less clear how to reconcile them. It could be that several of the experiments have just made mistakes or been misinterpreted, and with more data and more careful analysis we'll find out which signals were actually background events or random fluctuations. Or, it could be that dark matter is way more complicated than we realized. For instance, maybe it interacts differently with protons than with neutrons, or maybe there's more than one kind of dark matter particle, or maybe we've made an error with our assumptions about how dark matter is distributed in our galaxy, and fixing that will alleviate some of the tensions in the data. A few papers posted recently have also argued that the CDMS analysis of the XENON 100 constraint made it out to be more constraining than it is, so the CDMS result is maybe not entirely ruled out by XENON 100. But even that wouldn't explain all the other signals, and the results still don't easily agree.

As usual in science, the only thing to do now is to get more data. The business of dark matter detection is still in a fairly early stage -- as the detectors take more data and become more sophisticated, hopefully these signals and limits will start to make more sense. And of course we will keep looking in other places. The LHC is starting to place interesting limits on the dark matter parameter space, and even beyond AMS, efforts at indirect detection is also giving us some intriguing signals that may or may not have anything to do with dark matter. Some of us (e.g. me) are also looking toward the early universe to see if we can find hints for dark matter's effects on the first stars and galaxies.

Meanwhile, the preprint archive is happily aglow with new theory papers trying to piece this all together. It really is an exciting time; sometimes it's fun to have no idea what's going on.