MiniBooNE Neutrino Result – Guest Blog from Heather Ray

By John Conway | April 11, 2007 4:15 pm
Today at Fermilab, the MiniBooNE experiment announced to a packed auditorium their long-awaited results looking for neutrino oscillations. Below is a guest post from Dr. Heather Ray, a scientist at Los Alamos National Lab, who has been working on the experiment for several years. I have known Heather since she was a graduate student on the CDF experiment at Fermilab, when she was at the University of Michigan. To the right is a photo of her with her significant other, Ivan Furic.

MiniBooNE Neutrino Experiment Results

by Dr. Heather Ray, Los Alamos National Lab

Neutrinos, a fundamental particle of nature, are believed to oscillate, or change from one type to another. In the long list of experiments which have claimed an observation of neutrino oscillations, one stands apart : LSND. The LSND result doesn’t fit in with our picture of oscillations from other experiments, and as such is highly controversial. The MiniBooNE experiment was designed to explore the LSND result, to conclusively prove or disprove the claimed oscillations. MiniBooNE announced it’s first results today (April 11th, 2007). The illustrious rulers of the Cosmic Variance blog have asked me to write a bit about this result. So, let the amazing neutrino story begin!

  • Neutrinos and Oscillations : A Quick Introduction
  • In the Standard Model of physics there are three main categories of fundamental particles: quarks, leptons, and gauge bosons, or force carriers. The leptons are the electron, muon, and tau, as well as their partner neutrinos : &#957e, &#957&#956, and &#957&#964. Neutrinos in the Standard Model have no charge and are massless. Imagine for a minute that neutrinos do have mass. If they have mass then they are able to oscillate, or change type. Neutrinos have definitively been observed changing from one type into another, yet the Standard Model of physics says that neutrinos do not have mass. The simplest solution to this conundrum is to allow the neutrinos to have mass.

    In more technical terms we say that the weak eigenstates ( &#957e, &#957&#956, and &#957&#964) are made up of a combination of mass eigenstates. For example, in a two-neutrino scenario, at the time of creation the muon neutrino is a combination of the two mass eigenstates :

    |&#957&#956(0)> = -sin &#952 |&#9571> + cos &#952 |&#9572>

    where the probability for two-neutrino oscillations is given by :

    Posc = sin2(2&#952) * sin2[ (1.27 * &#916m2 * L) / E ]

    The probability has two terms which are constrained by the design of the experiment (L, the distance from the neutrino source to the detector, and E, the energy of the neutrino beam), and two terms which are fit for when performing a two-neutrino oscillation analysis (&#916m2 and sin2(2&#952), where &#952 is the mixing angle between the two neutrino states and &#916m2ab = m2a – m2b).

    Neutrino physicists illustrate the current status of neutrino oscillations using a two dimensional plot that is the function of the two fit parameters. Oscillation results from the solar and atmospheric sectors (the tiny red and blue dots) have been observed and confirmed by several experiments. The set of several independent measurements allows us to constrain the range of fit parameters for those oscillations. The LSND result, which spans the upper third of this plot, has a large allowed region in parameter space.

    In the Standard Model there are only three neutrinos, all of which interact with matter. The &#916m2 is the mass squared difference between the two neutrino states. These three results represent three differences, or splittings, between the mass states. If the Standard Model of physics is correct and there are 3 and only 3 neutrinos, a summation law should exist : &#916m213 = &#916m212 + &#916m223. You can see that even at LSND’s lowest allowed &#916m2 point the summation law does not hold.

    If LSND’s observation is found to be a true fact of nature, the Standard Model of physics cannot fully accommodate/explain neutrino interactions! This “breaking” of the Standard Model is very exciting to physicists, and indicates there is new physics that we haven’t previously thought possible. Many things could be true – there could be new allowed interactions for neutrinos (Lorentz Violation, CP/CPT violation, the list goes on!), or there could be additional particles – sterile neutrinos, which don’t interact with other matter but only can been seen through mixing with other neutrinos.

    To properly explore the LSND signal we need an experiment that has the same experimental constraints (L/E, from the oscillation probability formula), so that the entire allowed region of LSND can be explored. The follow-up experiment also should have more events (smaller statistical errors), and a different signal signature, backgrounds, and sources of systematic errors. This experiment is MiniBooNE.

  • MiniBooNE
  • MiniBooNE is located at Fermi National Accelerator Laboratory, in Batavia, IL. To produce our neutrino beam we start with an 8 GeV beam of protons from the Booster. The proton beam enters a magnetic focusing horn where it strikes a beryllium target. The interactions of the protons+Be produce positively and negatively charged mesons (pions and kaons). The positively charged mesons will decay to produce a neutrino beam, while the negatively charged mesons will decay to produce an anti-neutrino beam.

    There are still a lot of mysteries surrounding the interactions of neutrinos. We don’t yet know if neutrinos mix with the same probability as anti-neutrinos. Therefore, the LSND result, which claims observation of anti-&#957&#956 &#8594 anti-&#957e oscillations, needs to be explored using both neutrinos and anti-neutrinos. For the first check of LSND we chose to focus the positively charged mesons, which means we’re looking for &#957&#956 &#8594 &#957e oscillations. This choice was solely dictated by physics : the proton + Be interactions produce far more positively charged mesons. This means the rate of collection for our neutrino sample is much higher than our rate of collection for an anti-neutrino sample. We chose to collect the quick data set first, and then proceed with analyzing that data while collecting the anti-neutrino data set.

    The mesons decay in flight into the neutrino beam seen by the detector : K+ / &#960+ &#8594&#956+ + &#957&#956, where the &#957&#956 comprise the neutrino beam seen at MiniBooNE. These mesons decay in flight in our vacuum decay region. Following the decay region is an absorber, put in place to stop any muons and undecayed mesons. The neutrino beam then travels through approximately 450 meters of earth before entering the MiniBooNE detector.

    MiniBooNE is a 12.2 meter diameter sphere. The detector is filled with pure mineral oil and lined with photomultiplier tubes (PMTs). PMTs work like a reverse light bulb – instead of putting in electricity to produce light the PMTs collect light from neutrino interactions in our detector and output an electrical pulse. There are two regions of the MiniBooNE detector : an inner light-tight region and an optically isolated outer region known as the veto region, which aids in vetoing cosmic backgrounds.

  • Detecting Neutrino Interactions
  • Neutrinos interact with material in the detector. It’s the outcome of these interactions that we look for. These neutrino interactions in the MiniBooNE detector leave a distinct mark in the form of Cerenkov and scintillation light. Cerenkov light is produced when a charged particle moves through the detection medium with a velocity greater than the speed of light in the medium (v > c/n). This produces an electro-magnetic shock wave, similar to a sonic boom. The shock wave is conical and produces a ring of light which is detected by the PMTs. We can use Cerenkov light to measure the particle’s direction and to reconstruct the interaction vertex. This effect occurs immediately with the particle’s creation and is known as a prompt light signature.

    Charged particles moving through the detector also may deposit energy in the medium, exciting the surrounding molecules. The de-excitation of these molecules produces scintillation light. This is an isotropic, delayed light source, and provides no information about the track direction. We can however use the PMT timing information to locate the point, or vertex, where the neutrino interaction occurred.

    We can use the patterns of light seen in our PMTs to determine what type of neutrino interacted in our detector. In the charged-current quasi-elastic events, a neutrino interaction in the detector will produce the lepton partner of the neutrino. For example, an electron neutrino interaction will produce an electron, and a muon neutrino interaction will produce a muon. Electrons travel for only a very short time before their velocity falls below the Cerenkov threshold. They multiple scatter along the way, as well. This leaves a fuzzy Cerenkov ring in the detector. Muons tend to travel for a much longer distance. As they travel through the detector they lose energy, and the angle at which the Cerenkov light is being emitted shrinks. Muons also emit scintillation light. The signature of a muon in the detector isn’t one of a ring, as in the case of an electron. It is instead a filled in circle of light. Neutral pions decay into two photons, which then pair produce. The electrons from this pair production each create a ring in the detector.

  • Components of the Oscillation Analysis
  • MiniBooNE is performing a blind analysis. This means that we can either :

    • see some of the information in all of the data : we can check the charge per PMT as a function of time, to verify our detector isn’t failing,
    • see all of the information in some of the data : we are able to select data sets which will have no oscillation events present, if we assume maximal allowed oscillations from LSND. We can use these data sets to then tune and verify our Monte Carlo simulation.

    but we can’t see all of the information in all of the data. Having access to all of the information in all of the data is unblinding. Prior to unblinding we had to have all components of the analysis completely fixed. We aren’t allowed to go back and change event selection cuts or error estimates once we unblind.

    Our oscillation analysis can be boiled down to this simple algorithm : determine a set of event selection cuts which will isolate the electron neutrino events but remove the majority of all other events. There are a certain amount of electron neutrino events inherent in the beam, which come from kaon decays. There are also a small amount of other types of events (delta decays, pi0 events) which will pass the electron neutrino cuts, but which are not from true electron neutrino events. The sum of the estimated intrinsic electron neutrino events plus the fake events is the total number of events we expect to see, if no oscillations are present. We compare the number of events observed in data to the number expected, as a function of the reconstructed neutrino energy in these events. If we observe oscillations we should see an excess of data events over the expectation, whose shape will change as a function of the oscillation parameters.

  • Awesome-o Results!
  • This plot shows the MiniBooNE final sensitivity, compared to the prediction from our 2003 Run Plan. Curves are shown overlaid on the allowed LSND region.

    This plot shows the final result from the likelihood analysis. Data points are the black dots. The expected event spectrum is shown in red, and is broken down into the intrinsic electron neutrino and fake event shapes in the green and blue.

    We have two separate analyses that are used in the oscillation search : one which depends on likelihood variables, and one which depends on a boosted decision tree. These two analyses have a small overlap in event composition, and provide a good check of each other. We have complete confidence in our analysis if these two analyses find the same result. Before we unblinded our data we had to decide which of the two analyses we would call our primary analysis, for the purpose of quoting numbers in publications. We made this decision based on the expected sensitivity found using Monte Carlo. The sensitivity is the amount of parameter space allowed by the LSND result that we expect to be able to probe. Our sensitivity studies showed that with the likelihood analysis we were able to achieve a sensitivity which agrees quite well with the sensitivity MiniBooNE was designed for. This is something to be quite proud of! We also agreed that the final result quoted for the two neutrino oscillation search would be from 475 MeV to 3 GeV, based on the LSND best fit region.

    MiniBooNE unblinded on Monday, March 26th, 2007. When we opened the box we found no evidence for an excess of events over the background prediction. The MiniBooNE neutrino data set agrees with the no neutrino oscillation hypothesis, in the range of reconstructed neutrino energy from 475 MeV to 3 GeV. The probability that MiniBooNE and LSND both are due to two-neutrino oscillations is only 2%. The likelihood and the boosting analyses also agree quite well in measured excess events.

    I’m sure by now you’ve noticed that MiniBooNE observes an excess of events in the low energy region. We first saw this excess two weeks ago, when we unblinded. At that point we began working like madmen to determine what those events could possibly be. We’re rechecking our detailed understanding of various low energy background events: interactions in the dirt surrounding the detector, radiative delta decays, low energy neutral current events, you name it, we’re exploring it. Obviously, we don’t want to make any additional comments about this excess until we’re certain that we’ve performed all possible checks on our event predictions in that region. We’re hoping to have this excess mystery resolved in the next few months.

  • Now What?
  • MiniBooNE’s neutrino oscillation analysis was the first of two analyses needed to conclusively explore the LSND result. An anti-neutrino oscillation analysis will also need to be performed. At MiniBooNE it would take many more years to accumulate the data set needed to perform this anti-neutrino analysis. Instead, I’m hoping that we can continue this exploration at the Spallation Neutron Source, at Oak Ridge Laboratory in TN. The SNS is designed to be a world-class neutron facility. One of the side-effects of the process which produces the neutrons is that you get an amazing neutrino beam for free! Funding permitting I’m hoping we can begin taking data at the SNS within 3 to 4 years. At the SNS we could perform neutrino and anti-neutrino measurements simultaneously (no need to switch the horn polarity!), look for oscillations, sterile neutrinos, and look for CP/CPT violation in the neutrino sector.


    • S. Ahmed et al. [SNO Collaboration], Phys. Rev. Lett. 92, 181301 (2004),
    • G. Fogli et al., Phys. Rev. D 67, 093006 (2003),
    • A. Aguilar et al. [LSND Collaboration], Phys. Rev. D 64, 112007 (2001),
    • D. Smith, “Calculating the Probability for Neutrino Oscillations”,
    • The HARP Collaboration, CERN-SPSC/2003-027, SPSC-P-325
    • S. Kopp, “The NuMI Neutrino Beam at Fermilab”,
    CATEGORIZED UNDER: Guest Post, Science

    Discover's Newsletter

    Sign up to get the latest science news delivered weekly right to your inbox!

    Cosmic Variance

    Random samplings from a universe of ideas.

    See More

    Collapse bottom bar