The full paper is on the archive server now.
This blog post interprets the paper.
It is a very nice measurement. The experiment, OPERA, measures the centroid of the "Probablity Density Function" of the time of time of flight from the CERN accelerator to the Gran Sasso Laboratory in Italy. (This in itself is pretty amazing since the neutrinos are detected after travelling through 730 km of solid rock!)
Of couse there are two fundamental systematic uncertainties one must take into account. The first is the measurement of the difference between the time of neutrino creation at CERN and the time of measurement in Italy. The second is the distance between CERN and the detector.
The time difference was obtained by synchronized atomic clocks with greater than 1 ns precision. There is now a massive industry based on precison measurements of distance. This is the world-wide GPS system which allows measurement errors in positions of the order of centimeters. Its extremely hard to see how they could have screwed these up. For example the April 7th, 2009 L’Aquila Earthquake in Italy resulted is an easily identified 7 cm shift in position. (See below.)
The biggest uncertainty is establishing the original time structure of the proton beam that created the neutrinos. The proton pulse is of the order of 10 microseconds in width. The measured effect is a 60 nano-seconds shift from expectations with a 7 nanosecond uncertainty. That means they've established the centroid of their proton pulse width to a precision of better than one part in one thousand. Since they have recorded 16,000 events they have sufficient statistics to do this. The question is whether they've correctly determined the shape in time of the original proton beam.
This is determined from the proton beam intensity which is continuously montored. They compared the this to the measured time profile in their detector. The measured original proton beam intensities as a function of time are shown below. The CERN SPS has two extraction profiles which are easily distinguished.
The observed time structures of the events recored in Italy are shown below. The red lines show the superimposed average time structures of the proton beams with and without the flight time to Italy.
By eye the original time structure of the CERN proton beam appears well reproduced in the detected neutrinos.
After a full blind-analysis they find that their neutrinos travelled (60.7 ± 6.9 (stat.) ± 7.4 (sys.)) ns faster than light over the same distance.
The first uncertainty is the statistical precision of the measurement. The second is their estimate of the systematic uncertainty. Their systematic uncertainty was obtained by a quadratic sum of 12 potentional sources of error. The largest (5.0 ns) was the calibration of the detector that measures the time of the proton beam extraction at CERN.
Overall their measurement gives:
(v-c)/c = (2.48 ± 0.28 (stat.) ± 0.30 (sys.)) ×10-5
This is consistent with an earlier but less precise measurement from an experiment in America with the Fermilab accelerator:
(v-c)/c = 5.1 ± 2.9×10-5
But substantially different from the neutrinos detected from Supernova 1987a which gives a limit on the anti-neutrino velocity difference of:
|v-c|/c < 2×10-9
So what happens now? There are two other high statistics long-baseline experiments in operation in Japan (at the KEK laboratory where I do my CP-violation experiments) and an updated experiment using the Fermilab accelerator in America. You can be certain that both experiments will do their best make the most precise measurements of the neutrino velocities they can. If all three agree what then? Well we will have to wait to see if they do.... As I said this is a most incredible result.
One final point about this. If the neutrinos from SN1987A had had the increase in velvocity over photons measured by the OPERA experiment they would have arived 4 years earlier than the light...
(All images are reproduced from http://arxiv.org/abs/1109.4897v1)