Faster-than-light neutrinos: keeping time

There's been a recent announcement of evidence that muon neutrinos may travel faster than light. That would be really weird - in particular it would pose serious problems for Einstein's theories of relativity. But even the discoverers aren't ready to claim that; they just describe their results and say they're puzzling. Personally I think it's unlikely they're right, but figuring out why not may be very interesting.


Faster than Light? par CNRS

Fortunately for us, they have posted a preprint of their paper on arxiv.org. All images below are from that paper.

I can't say much about the particle physics, or the details of instrument calibration, but I can address one possible way people have suggested the result may be wrong: inaccuracies in the time standards at the two endpoints.

The experiment works by making neutrinos at CERN and detecting them 730 km away. It was built to look at flavour oscillations, but by dint of great effort, both installations have good enough time tagging to be able to compute the neutrino time of flight and divide it by the distance. What this report is saying is that they seem to see neutrinos arriving about sixty nanoseconds too early.

There have been comments that if they're deriving their time from GPS, they can't possibly get the few-nanosecond accuracy they need. We use GPS in pulsar astronomy as well, and so I've looked into just what it can do.

First of all, the GPS receiver in my telephone (say) works by comparing the light travel time to several GPS satellites; to get a roughly ten meter accuracy, it must measure these differences to roughly thirty nanoseconds. But if you're serious about measuring time, you don't use that kind of GPS receiver. You use, and they used, a special-purpose high-quality GPS receiver. In fact, they use a pair of receivers set up to keep carefully-synchronized clocks at two different locations, a so-called "time transfer system". They use tricks like relying only on satellites that both receivers can see at the same time. These GPS receivers are used only for long-term stabilization; in fact, it is their one-pulse-per-second output that is used. (Almost all standalone GPS units have one-pulse-per-second outputs, by the way, in case you want microsecond-accurate time for some hobby project. These GPS receivers obviously do much better, the specs claiming 10 nanoseconds.) For short-term stability, they use a cesium atomic clock at each end. This keeps extremely good time — something like a nanosecond drift over the course of an hour, and a nanosecond jitter — and they simply use the GPS receivers to monitor and correct for this drift. So their system is in principle perfectly capable of maintaining time synchronization of the quality they need.

Does the system actually keep time as well as they need? Well, I don't know for sure, though there's nothing obviously wrong with it, but they did have some Swiss metrology people calibrate it, and then some German metrology people come in with a portable time transfer system to test it out, and the metrology people are confident it's good. So I don't think the time synchronization is the place to look for problems with this paper.

Blind neutrino alignment.

I think it's a little more likely that they're being done in by a statistical mess. The problem is, they can't tell when each individual neutrino was generated. All they can tell is which 10.5-microsecond-long packet of collisions it came from. So they took the distribution of neutrino arrival times, and the distribution of collisions within the packet, and lined them up. They did this blindly — that is, without knowing exactly when the speed-of-light arrival time was — and they got the results to line up nicely. Only they were sixty nanoseconds early. I'm rather leery of this packet-alignment stuff, though I shouldn't be, since it's very much the same process we use in pulsar astronomy to line up the arrival time of many-microsecond (radio) pulses to get arrival times good to about a hundred nanoseconds. We even use it for X-ray pulsars, where the data is individual events, just like these folks are using. So really, I guess the problem is more likely to be in some systematic somewhere: some delay in the system is not exactly what they think it is. But I can't spot it.

I'd like to point out, though, just how well these folks are handling this: the paper is very thorough about examining all the ways they can think of that they could be wrong, and providing enough information that the wider community can examine it. The scientists involved are also being, very reasonably, cautious and asking for people to replicate their results, rather than to simply believe them. Most miraculous, this skepticism has survived the press-release machinery pretty well: lots of the general-reader material manages to convey that this is startling but needs further study. I know just how hard it is to hang onto that caution.