More details on the “faster than the speed of light” neutrinos

Last night, in response to a worldwide surge in interest, the OPERA experiment released a paper that describes the experiments that appear to show neutrinos traveling faster than the speed of light. And today, CERN broadcast a live seminar in which one of the work's authors described the content of the paper. Both of those emphasized the point of our initial coverage: figuring out whether anything is traveling beyond the speed of light requires incredibly accurate measurements of time and distance, and the OPERA team has made an extensive effort to make its work as accurate as possible.

As a spokesperson for the MINOS neutrino experiment told Ars yesterday, there are three potential sources of error in the timing measurements: distance errors, time-of-flight errors, and errors in the timing of neutrino production. The vast majority of both the paper and the lecture were dedicated to discussing how these errors were reduced (the actual detection of the neutrinos was only a small portion of the paper).

Neutrinos are produced using a proton beam from one of the accelerators that feeds them into the LHC. The protons hit a fixed target and produce unstable particles that decay, releasing a neutrino. The protons move close to, but not at the speed of light, as do the unstable pions; both of these effects were accounted for. The timing of the protons and structure of the two bunches of them used in these experiments is not even, either, so the researchers created a profile of the proton bunch. They also compensated for the timing of the kicker magnet that pushes the bunch out of the accelerator and added detectors that registered them passing through the hardware to get a clearer sense of their timing.

Similar work went into the detector side, where the time between an actual neutrino event and the signal propagating through the hardware and to a field programmable gate array (FPGA) where it was processed was estimated at about 50ns (the neutrinos only arrived 60ns early, so that 50ns is a substantial fraction of the total). But the error in their estimate was only ±2.3ns, as measured by shining a picosecond UV laser on the detector.

Distance travelled created its own problems. The positions of the hardware were measured via GPS, which normally doesn't provide the sort of precision needed for this work. But the labs did multiple samples of the GPS signals, threw out bad ones, compensated for the effect of the Earth's iononsphere, and more. Then, just to check their work, they had an outside team from a German standards institute come in and perform an independent analysis. The end result was a measurement sensitive enough to register both the steady change due to continental drift, as well as a 7cm jump triggered by an earthquake.

Then, the timing of all the events had to be synchronized. At each site, the group put a cesium-based atomic clock, and synchronized it with the GPS signal. Then, they sent a portable atomic clock between the facilities to check. They then ran photons through a fiber optic cable between them, just to make sure.

The end result is that the OPERA team doesn't see any obvious problems in its measurements. All of the errors, when added up, shouldn't be able to account for anything close to the 60ns gap between the neutrinos' arrival and the speed of light. The difference between their speed and that of light is very statistically significant, and the neutrino data itself looks excellent. The team has recorded over 16,000 events now, and the profile of events over time very closely matches the structure of the proton bunches that created them.

But that doesn't mean that this presentation is the last word on the topic. There are a lot of potential sources of error they know about—the paper's table lists a dozen of them. Small errors in each of these could add up to something more significant than their total error. Then there are the classic unknown unknowns. The authors have tried to think of everything, but it's not clear that they can.