Here’s my last story for BBC Future.
Who will find dark matter first? We’re looking everywhere for this elusive stuff: deep underground, out in space, in the tunnels of particle colliders. After the Higgs boson, this is the next Big Hunt for modern physics, and arguably there’s even more at stake, since we think there’s more than four times more dark matter than there is all the stuff we can actually see.
And you can join the hunt. It’s probably not worth turning out your cupboards to see if there’s any dark matter lurking at the back, but there is a different way that all comers – at least, those with mathematical skills – can contribute. A team of astronomers has reported that crowdsourcing has improved the computational methods they will use to map out the dark matter dispersed through distant galaxies – which is where it was discovered in the first place.
The hypothesis of dark matter is needed to explain why galaxies hold together. Without its gravitational effects, rotating galaxies would fly apart, something that has been known since the 1930s. Yet although this stuff is inferred from its gravity, there’s nothing visible to astronomers – it doesn’t seem to absorb or emit light of any sort. That seems to make it a kind of matter different from any of the fundamental particles currently known. There are several theories for what dark matter might be, but they all have to start from negative clues: what we don’t know or what it doesn’t do.
The current favourite invokes a new fundamental particle called a WIMP: a weakly interacting massive particle. “Weakly interacting” means that it barely feels ordinary matter at all, but can just pass straight through it. However, the idea is that those feeble interactions are just enough to make a WIMP occasionally collide with a particle of ordinary matter and generate an observable effect in the form of a little burst of light that has no other discernible cause. Such flashes would be a telltale signature of dark matter.
To see them, it’s necessary to mask out all other possible causes – in particular, to exclude collisions involving cosmic rays, which are ordinary particles such as electrons and protons streaming through space after being generated in violent astrophysical processes such as supernovae. Cosmic rays are eventually soaked up by rock as they penetrate the earth, and so several dark-matter detectors are situated far underground, at the bottom of deep mineshafts. They comprise sensitive light detectors that surround a reservoir of fluid and look for inordinately rare dark-matter flashes.
One such experiment, called LUX and located in a mine in South Dakota, has recently reported the results of the first several months of operation. LUX looks for collisions of WIMPs within a tank of liquid xenon. So far, it hasn’t seen any. That wouldn’t be such a big deal if it wasn’t for the fact that some earlier experiments, have reported a few unexplained events that could possibly have been caused by WIMPs. LUX is one of the most sensitive dark-matter experiments now running, and if those earlier signals were genuinely caused by dark matter, LUX would have been expected to see such things too. So the new results suggest that the earlier, enticing findings were a false alarm.
Another experiment, called the Alpha Magnetic Spectrometer (AMS) and carried on board the International Space Station, looks for signals from the mutual annihilation of colliding WIMPs. And there are hopes that the Large Hadron Collider at CERN in Geneva might, once it resumes operation in 2014, be able to conduct particle smashes at the energies where some theories suggest that WIMPs might actually be produced from scratch, and so put these theories to the test.
In the meantime, the more information we can collect about dark matter in the cosmos, the better placed we are to figure out where and how to look for it. That’s the motivation for making more detailed astronomical observations of galaxies where dark matter is thought to reside. The largest concentrations of the stuff are thought to be in gravitationally attracting groups of galaxies called galaxy clusters, where dark matter can apparently outweigh ordinary matter by a factor of up to a hundredfold. By mapping out where the dark matter sits in these clusters relative to their visible matter, it should be possible to deduce some of the basic properties that its mysterious particles have, such as whether they are ‘cold’ and easy slowed down by gravity, or hot and thus less easily retarded.
One way of doing this mapping is to look for dark matter via its so-called gravitational lensing effect. As Einstein’s theory of general relativity predicted, gravitational fields can bend light. This means that dark matter (and ordinary matter too) can act like a lens: the light coming from distant objects can be distorted when it passes by a dense clump of matter. David Harvey of the University of Edinburgh, Thomas Kitching of University College London, and their coworkers are using this lensing effect to find out how dark matter is distributed in galaxy clusters.
To do that, they need an efficient computational method that can convert observations of gravitational lensing by a cluster into its inferred dark-matter distribution. Such methods exist, but the researchers suspected they could do better. Or rather, someone else could.
Crowd-sourcing as a way of gathering and analysing large bodies of data is already well established in astronomy, most notably in the Zooniverse scheme, in which participants volunteer their services to classify data into different categories: to sort galaxies or lunar craters into their fundamental shape classes, for example. Humans are still often better at making these judgements than automated methods, and Zooniverse provides a platform for distributing and collating their efforts.
What Harvey and colleagues needed was rather more sophisticated than sorting data into boxes. To create an algorithm for actually analysing such data, you need to have some expertise. So they turned to Kaggle, a web platform that (for a time-based fee) connects people with a large data set to data analysts who might be able to crunch it for them. Last year Kitching and his international collaborators used Kaggle to generate the basic gravitational-lensing data for dark-matter mapping. Now he and his colleagues have shown that even the analysis of the data can be effectively ‘outsourced’ this way.
The researchers presented the challenge in the form of a competition called “Observing Dark Worlds”, in which the authors of the three best algorithms would receive cash prizes totalling $20,000 donated by the financial company Winton Capital Management. They found that the three winning entries could improve significantly on the performance of a standard, public algorithm for this problem, pinpointing the dark matter clumps with an accuracy around 30% better. Winton Capital benefitted too: Kitching says that “they managed to find some new recruits from the winners, at a fraction of the ordinary recruiting costs.”
It’s not clear that the ordinary citizen can quite compete at this level – the overall winner of Dark Worlds was Tim Salismans, who this year gained a PhD in analysis of “big data” at the Erasmus University Rotterdam. The other two winners were professionals too. But that is part of the point of the exercise too: crowd-sourcing is not just about soliciting routine, low-level effort from an untrained army of volunteers, but also about connecting skilled individuals to problems that would benefit from their expertise. And the search for dark matter needs all the help it can get.