Research Highlights Archive

Evidence for Neutrino Oscillations from the LSND Experiment


One of the only ways to probe small neutrino masses is to search for neutrino oscillations, where a neutrino of one type (e.g. νμ) spontaneously transforms into a neutrino of another type (e.g. νe) For this phenomenon to occur, neutrinos must be massive and the apparent conservation law of lepton families must be violated. The probability for 2-flavor neutrino oscillations can then be expressed as P=sin2(2θ) sin2(1.27 m2L/E) , where θ is the mixing angle, m2 is the difference in neutrino masses squared in eV2, L is the neutrino distance in meters, and E is the neutrino energy in MeV. In 1995 the LSND experiment1 published data showing candidate events that are consistent with νμ - > νe oscillations. 2 Additional data are reported here that provide stronger evidence for νμ - > νe oscillations 3 as well as evidence for νμ - > νe oscillations. 4 The two oscillation searches have completely different backgrounds and systematics from each other.


The Liquid Scintillator Neutrino Detector Liquid Scintillator Neutrino Detector (LSND) experiment at Los Alamos 5 was designed to search with high sensitivity for νμ - > νe oscillations from μ+ decay at rest. The LANSCE accelerator is a most intense source of low energy neutrinos due to its 1 mA proton intensity and 800 MeV energy. The neutrino source is well understood because almost all neutrinos arise from π+ or μ+ decay; π- and μ- are readily captured in the Fe of the shielding and Cu of the beam stop. 6 The production of kaons and heavier mesons is negligible at these energies. The νe rate is calculated to be only 4*10-4 relative to νμ in the 36 < Eν< 52.8 MeV energy range, so that the observation of a significant νe rate would be evidence for νμ -> νe oscillations.

The LSND detector consists of an approximately cylindrical tank 8.3 m long by 5.7 m in diameter. The center of the detector is 30 m from the neutrino source. On the inside surface of the tank 1220 8-inch Hamamatsu phototubes provide 25% photocathode coverage. The tank is filled with 167 metric tons of liquid scintillator consisting of mineral oil and 0.031 g/l of b-PBD. This low scintillator concentration allows the detection of both Cerenkov light and scintillation light and yields a relatively long attenuation length of more than 20 m for wavelengths greater than 400 nm. 7 A typical 45 MeV electron created in the detector produces a total of approximately 1500 photoelectrons, of which ~ 280 photoelectrons are in the Cerenkov cone. The phototube time and pulse height signals are used to reconstruct the track with an average r.m.s. position resolution of ~ 30 cm, an angular resolution of ~ 12 degrees, and an energy resolution of ~ 7%. The Cerenkov cone for relativistic particles and the time distribution of the light, which is broader for non-relativistic particles, give excellent particle identification. Surrounding the detector is a veto shield 8 which tags cosmic ray muons going through the detector.

νμ - > νe Oscillation Data

The signature for a νe interaction in the detector is the reaction νe p ->e+ n followed by n p -> d Υ (2.2 MeV). A likelihood ratio, R, is employed to determine whether a Υ is a 2.2 MeV photon correlated with a positron or is from an accidental coincidence. R is the likelihood that the Υ is correlated divided by the likelihood that it is accidental and depends on the number of hit phototubes for the gamma, the reconstructed distance between the positron and the Υ, and the relative time between the Υ and e+. After subtracting the neutrino background with a recoil neutron, there is a total excess of 51.0+20.2-19.5 +-8.0 events, which if due to neutrino oscillations corresponds to an oscillation probability of (0.31+-0.12+-0.05)% .

There are 22 events beam on in the 36 with R>30, while the estimated background (beam off plus neutrino-induced background) is 4.6+-0.6 events. The probability that this excess is a statistical fluctuation is <10-7. The observed average value of cos(θb), the angle between the neutrino di"RECT"ion and the reconstructed positron di"RECT"ion, is 0.20+-0.13, in agreement with the expected value of 0.16 for νe p interactions. If the observed excess is due to neutrino oscillations, Fig. 1 shows the allowed region (90% and 99% likelihood regions) of sin2(2θ) vs m2 from a maximum likelihood fit to the L/E distribution of the entire data sample. Some of the allowed region is excluded by the ongoing KARMEN experiment at ISIS, 9 the E776 experiment at BNL, 10 and the Bugey reactor experiment.

Six months of additional data have been collected in 1996 and 1997. For this running period the beam stop was reconfigured with the water target replaced by a close-packed high-z target for testing tritium production. The muon decay-at-rest neutrino flux with this new configuration is only 2/3 of the neutrino flux with the old beam stop; however, the pion decay-in-flight neutrino flux has been reduced to 1/2 of the original flux, so that the 1996 and 1997 data serve as a systematic check. Events with R>30 are referred to as "gold-plated". Preliminary results from 1996 and 1997are given in Tables 1 and 2. Table 1 shows the number of "gold-plated" events from the entire 1993-1997 data sample, while Table 2 shows the total numbers of excess events and the corresponding oscillation probabilities from fits to the R distributions for the running periods 1993-1995, 1996-1997, and 1993-1997. The preliminary oscillation probability for the entire data sample is (0.31+-0.09+-0.05)%.

Table 1: Preliminary numbers of "gold-plated" events with R>30 from the entire 1993-1997 data sample

Positron Energy Events Beam On Events Beam Off Neutrino Background Total Excess
20"<"E<60 MEV 61 15.6+-1.0 11.5+-1.5 33.9+-8.0
36"<"E<60 MEV 29 5.2+-0.6 3.0+-0.6 20.8+-5.4

Table 2: Preliminary number of excess events and the corresponding oscillation probabilities from fits to the R distributions for the running periods 1993-1995, 1996-1997, and 1993-1997

Data Sample Fitted Excess Neutrino Background Total Excess Oscillation Probability
1993-1995 63.5+20.0 12.5+-2.9 51.0+-20.2 (0.31+-0.12+-0.05%)
1996-1997 35.1+-14.7 4.8+-1.1 30.3+-14.8 (0.32+-0.15+-0.05%)
1993-1997 100.1+-23.4 17.3+-4.0 82.8+-23.7 (0.31+-0.009+-0.05%)

νμ -> νe Oscillation Data

The signature for νμ-> νe oscillations is an electron from the reaction νeC -> e-X in the energy range 60< Ee <200 MeV. Using two independent analyses,4 a total of 40 beam-related events and 175 beam-unrelated events are observed, corresponding to a beam on-off excess of 27.7+-6.4 events. The neutrino-induced backgrounds are dominated by μ+ -> e+ νμνe and π+ -> e+ νe decays-in-flight in the beam-stop and are estimated to be 9.6+-1.9 events. Therefore, a total excess of 18.1+-6.6+-3.5 events is observed above background.

The excess events are consistent with νμ-> νe oscillations with an oscillation probability of (0.26+-0.10+-0.05)%. A fit to the event distributions yields the allowed region in the (sin2(2θ), m2 parameter space shown in Fig. 2, which is consistent with the allowed region from the νμ-> νe search. These two searches have completely different backgrounds and systematic errors, and together they provide strong evidence that the observed event excesses are indeed due to neutrino oscillations.


In summary, the LSND experiment observes excesses of events for both the νμ - > νe and νμ-> νe oscillation searches, corresponding to oscillation probabilities of (0.31+-0.12+-0.05)% and (0.26+-0.10+-0.05)%, respectively. These two searches have completely different backgrounds and systematics and together provide strong evidence for neutrino oscillations in the range 0.2 < m2 < 2.0 eV2. This implies that at least one neutrino has a mass greater than 0.4 eV.

Neutrino Oscillation Experiments


  1. The LSND Collaboration currently consists of the following people and institutions: E. D. Church, I. Stancu, W. Strossman, G.J. VanDalen (Univ. of California, Riverside); W. Vernon (Univ. of California, San Diego); D.O. Caldwell, S. Yellin (Univ. of California, Santa Barbara); D. Smith,(Embry-Riddle Aeronautical Univ.); R.L. Burman, J.B. Donahue, F.J. Federspiel, G.T. Garvey, W.C. Louis, G.B. Mills, V. Sandberg, B. Sapp, R. Tayloe, D.H. White (Los Alamos National Laboratory);R. Imlay, H.J. Kim, A. Malik, W. Metcalf, M. Sung, N. Wadia, (Louisiana State Univ.): K. Johnston (Louisiana Tech Univ.); A. Fazely (Southern Univ); L.B. Auerbach, R. Majkic (Temple Univ.).
  2. C. Athanassopoulos et al., Phys. Rev. Lett. 75, 2650 (1995)
  3. C. Athanassopoulos et al., Phys. Rev. C 54, 2685 (1996); C. Athanassopoulos et al., Phys. Rev. C 77, 3082 (1996)
  4. C. Athanassopoulos et al., LA-UR-97-1998, submitted to Phys. Rev. C
  5. C. Athanassopoulos et al., NUCL. Instum. Methods A 388, 149 (1997).
  6. R.L. Burman, M.E. Potter, and E.S. Smith, Nucl. Instrum. Methods in Phys. Res. A 368, 416 (1996).
  7. R.A. Reeder et al., Nucl. Instum. Methods A 334,353 (1993).
  8. J.J. Napolitano it et al. Nucl. Instrum. Methods A 274, 152 (1989).
  9. B. Bodmann et al., Phys. Lett. B 267, 321 (1991), B. Bodmann et al., Phys. Lett. B 280, 198 (1992), B. Zeitnitz et al., Prog. Part. Nucl. Phys. 32 351 (1994).
  10. L. Borodovsky et al., Phys. Rev. Lett. 68, 274 (1992).
  11. B. Achkar et al., Nucl. Phys. B434, 503 (1995).

Neutrino Oscillations

A collaboration of the University of California at Riverside, the University of California at San Diego, the University of California at Santa Barbara, Embry-Riddle Aeronautical University, Los Alamos National Laboratory, Louisiana State University, Louisiana Tech University, Southern University, and Temple University.

Figure 1

Figure 1: Plot of the LSND (m2) vs sin2(2θ) favored regions. They correspond to 90% and 99% likelihood regions after the inclusion of the effects of systematic errors. Also shown are 90% C.L. limits from KARMEN at ISIS (dashed curve), E776 at BNL (dotted curve), and the Bugey reactor experiment (dot-dashed curve).

Figure 2

Figure 2: The 95% confidence region for νμ-> νe oscillations (solid curve) along with the favored regions for νμ -> νe oscillations(dotted curve).