Neutrinos seem to oscillate: they change back and forth between one type and another and, by extension, have a tiny mass. But one experiment that predicted a particularly large mass looks to have been mistaken.
When is a discovery not a discovery? When it can't be reproduced, of course. That scientific ground-rule has plagued the members of the LSND (Liquid Scintillator Neutrino Detector) collaboration since they first saw evidence1 for so-called neutrino oscillations. Had the LSND results been confirmed, they would have rewritten much of what we think we know about the ever-elusive neutrinos (Box 1). But results just announced2 from the MiniBooNE detector at Fermilab, near Chicago, could prove those earlier results' nemesis.
The story begins in 1996, with observations made by LSND of the decay products of a pion particle beam at the Los Alamos Meson Physics Facility (LAMPF) accelerator in New Mexico. The neutrinos came mainly from the decay of positively charged pions into positive muons and muon neutrinos, and the subsequent decay of these positive muons to positrons (positively charged electrons), muon antineutrinos and electron neutrinos (Fig. 1a). What the LSND collaboration found on examining the reaction products was an excess of electron antineutrinos1 — which are produced nowhere in the positive-pion decay chain. The conclusion was that muon antineutrinos were changing into electron antineutrinos while propagating. This is the process known as neutrino oscillation, and is itself by now relatively uncontroversial3. But the oscillations that LSND saw seemed to indicate much larger neutrino mass differences than other experiments had predicted. That was indeed controversial.
Perhaps the LSND results pointed to some new physics even more surprising than neutrino oscillations. Many possibilities have been suggested, including the existence of other, 'sterile' neutrino families beyond the three that participate in the interactions of particle physics' standard model. But the first attempt to test the results — KARMEN, a comparable, but slightly less sensitive experiment at the Rutherford Appleton Laboratory in Didcot, England — failed to see any signal, although it could not rule out the LSND claim. A new experiment with much higher rates, and therefore higher sensitivity, was needed to test the LSND claim conclusively.
Thus was born Fermilab's MiniBooNE, so called because it was a scaled-down version of the proposed Booster Neutrino Experiment. In MiniBooNE, a beam of either positive or negative pions decays to form beams of positive or negative muons, and muon neutrinos or antineutrinos, whose oscillations are looked for (Fig. 1b). But the muons also decay, producing, among other things, electron neutrinos and antineutrinos — a potential source of confusion for the oscillation measurement. This is minimized at MiniBooNE by stopping the muons before significant numbers of them decay, and also — unlike the LSND and KARMEN experiments — using a higher-energy pion beam, which produces a clean differentiation in energy between the different sources of neutrinos.
The muon neutrino beam propagates for around 500 metres before hitting a detector consisting of 1,520 light-sensitive photomultiplier tubes embedded in a 12-m-diameter tank of mineral oil. The charged particles resulting from neutrino interactions travel faster than the speed of light in the oil, resulting in the emission of a characteristic cone of 'Čerenkov' light, rather like the sonic boom produced by a jet breaking the sound barrier. The detection of this light allows the particle's energy and position to be measured, and one type of neutrino to be distinguished from another through its characteristic interactions (Fig. 2).
The analysis of the data2 is complicated, partly owing to physics and partly to psychology. The physics complication lies in the number of background signals that mimic the real one. Added together, these outnumber the postulated 'real' events by two to one. To add to the fun, neutrino cross-sections — the probability that a neutrino will interact in a particular way — are poorly known in the MiniBooNE energy range.
Psychology enters the picture in the attempts to quantify the background signal. The history of science is littered with false results produced by scientists who see how each tuning of an analysis procedure affects their desired signal and its undesired background, and who subconsciously tune their analysis to produce the result they want to see. To prevent this, MiniBooNE adopted two independent forms of blind analysis, hiding away the data that would contain the potential signal until all the analyses had been finalized.
This procedure greatly increases the reliability of the results, but it also adds considerably to the complexity of the analysis. The two separate analyses reveal that the number of events corresponding to electron neutrinos matches expectations from backgrounds. One analysis produces a few more events than expected, the other a few less, but neither deviation is statistically significant. The results thus rule out the simple oscillation explanation of the LSND results at the 98% confidence level.
The work is not yet done, however: 98% is a little less confidence than we like to have in our fundamental laws. MiniBooNE is working to increase its sensitivity still further, and also to understand a discrepancy between the expectation and experiment that occurs mostly at energies below those at which the oscillation signal should occur. In addition, a bolt-hole for the new physics that the LSND result might have signalled remains to be closed: LSND used antineutrinos; MiniBooNE, neutrinos. Such a gross difference in the oscillations of these matter and antimatter analogues would be a truly revolutionary result. But no experimental results rule it out: experimenters must therefore try to collect enough data with an antineutrino beam to discount this possibility.
None of these quibbles seems strong enough to challenge the basic conclusion. It is worth mentioning here that the MiniBooNE collaboration includes many influential members of the LSND collaboration. They would not be human if they didn't have a strong desire to see their signal confirmed and most of neutrino physics rewritten: various rewards await those who rewrite the laws of physics. And yet they helped construct an experiment to produce a conclusive test of their results, intentionally designed it so that it would be almost impossible to bias the results one way or the other, and when it ruled against them, announced it openly to the world. That might not win a Nobel prize, but it is still science at its best.
Aguilar, A. et al. Phys. Rev. D 64, 112007 (2001).
Aguilar-Arevalo, A. A. et al. preprint available at http://arxiv.org/abs/0704.1500v2 (2007).
Giunti, C. J. Phys. G 34, R93–R109 (2007).
Fukuda, Y. et al. Phys. Rev. Lett. 81, 1562–1567 (1998).
Ahmad, Q. R. et al. Phys. Rev. Lett. 89, 011301 (2002).
Eguchi. K. et al. Phys. Rev. Lett. 90, 021802 (2003).
Ahn, M. H. et al. Phys. Rev. D 74, 072003 (2006).
Michael, D. B. et al. Phys. Rev. Lett. 97, 191801 (2006).