First Measurement of Neutrino and Antineutrino Coherent Charged Pion Production On Argon
First Measurement of Neutrino and Antineutrino Coherent Charged Pion Production On Argon
First Measurement of Neutrino and Antineutrino Coherent Charged Pion Production On Argon
Neutrinos can produce single pion final states by coherently scattering from the entire nucleus. Both neutral current (NC) and charged current (CC) processes
are possible. In these interactions, the squared fourmomentum transfer to the target nucleus, |t|, is small
so the nucleus remains unchanged. In this Letter, we
focus on the CC coherent pion production from muon
neutrinos and antineutrinos on argon:
+ Ar + + + Ar;
+
+ Ar + + Ar;
(1)
(2)
where the low |t| condition entails that the pions and
muons are forward going with respect to the incoming
neutrino direction.
There are several models from which one can extract
cross sections and kinematical predictions for this interaction. The Rein-Seghal [1] model has been used to successfully describe high energy data within experimental
uncertainties since the first observation of coherent pion
production at the Aachen-Padova spark chamber [2] in
1983. This approach is based on Adlers Partially Conserved Axial Current (PCAC) theorem [3], which relates
the pion production cross section to the cross section for
the pion-nucleus scattering. This model is still the standard for neutrino generators today, such as genie [4],
nuwro [5], and neut [6], with continued updates to the
formalism and the pion-nucleus scattering data that is
used. With recent interest in coherent pion production in the theoretical community, other PCAC models
have been proposed [7, 8]. Microscopic models [911]
2
run is 1.2 1020 and the estimated integrated fluxes are
6.6 1011 muon neutrinos per cm2 and 3.0 1012 muon
antineutrinos per cm2 . The differential flux can be
found in reference [16]. Neutrino interactions comprise
almost 60% of all the neutrino/antineutrino-induced
events in the detector [16]. During this run, the MINOS
near detector [17] placed downstream of ArgoNeuT is
also operational. The muons that exit ArgoNeuTs
TPC volume are matched to MINOS, in which the momentum and charge are reconstructed.
Using the LArSoft software [18], (anti)neutrino interactions are reconstructed, rendering a full characterization of the charged particles emerging in the ArgoNeuT detector. The software also provides the framework for a Monte Carlo (MC) simulation of the experiment. This is achieved by employing genie [4] as the
neutrino event generator and geant4 [19] for the simulation of the propagation of products in the detector.
The complete ArgoNeuT geometry is simulated along
with the signal formation processes and taking into account electronic noise. The simulated events are fully
reconstructed in the same way as for data. The propagation of particles in the MINOS near detector is simulated with geant3 [20]. A standalone version of MINOS simulation and reconstruction is used to characterize the matching of tracks passing from ArgoNeuT
into MINOS.
The search for CC coherent pion production starts
with an event selection which is used to find the two
track topology of Eqs. (1) and (2). Each of the selection criteria described below is chosen
in order to
maximise the significance, defined as s/ s + b, where s
and b are the numbers of signal and background events
which pass the selection in the MC simulation. The MC
used assumes the signal as modeled by Rein-Seghal. We
start by requiring that two tracks are reconstructed in
the event, originating from the same vertex. One track,
identified as the muon, must be reconstructed in both
ArgoNeuT and MINOS and matched between the two
detectors. The unmatched track is the pion candidate.
ArgoNeuTs precise calorimetry is used to discriminate
pions from protons by defining an acceptance window
for the mean dE/dx of the unmatched track. While the
dE/dx of a pion will correspond to a Minimum Ionizing
Particle (1 MIP), a proton track will leave an energy deposition several times higher (> 2 MIP). By applying
a selection criteria on the dE/dx of the pion-candidate
track, the CC quasi-elastic background is almost fully
removed. The calorimetry capabilities of the detector
are further exploited by investigating the Analog-toDigital (ADC) readout at the wires at the vertex. Low
energy protons emerging at the vertex induce high ADC
readouts at the first wire hits which are used to exclude
the event. This selection reduces the background of
interactions where multiple low-energy protons are produced, added either by nuclear effects or a result of deep
inelastic scattering.
The lack of any particles other than the muon and
the pion emerging from the vertex is further reinforced
by another selection criteria. For each event, the charge
readout inside a 20 cm20 cm box defined in the wire
number versus drift time view of the collection plane is
counted; the fraction of this charge that is associated
with the two outgoing tracks must amount to at least
86%(84%) for antineutrino(neutrino) events. The collection plane is used because its response is better calibrated compared with the induction plane. This verification is crucial since it removes background events
with activities around the interaction vertex that are
not originated from the muon and the pion.
The event selection defined makes the most of the
precise calorimetry and the high imaging resolution the
ArgoNeuT detector is capable of, which are a characteristic of LArTPCs. We estimate the selection efficiencies to be (18.4 1.8)% for neutrino and (21.8 0.8)%
for antineutrino events. The inefficiency is dominated
by the track reconstruction inefficiency for overlapping
tracks or complex topologies when the pion interacts
with the argon nucleus. The systematic uncertainties
associated to the assumptions on the kinematics of the
signal events are accessed by estimating the efficiency
using a different generator (nuwro). The difference between the efficiencies obtained with the two generators
is kept as the systematic uncertainty.
A total of 167 antineutrino and 150 neutrino events
have the two-track topology in the TPC with one
track matched to a reconstructed track in MINOS. After applying the event selection described, 30 antineutrino and 24 neutrino candidate events remain. This
event sample contains a background fraction, predominantly resonant and deep inelastic interactions, that
ideally would be reduced by selecting events with low
|t| = (q p )2 , where q represents the momentum
transfer from the neutrino and p is the momentum
carried by the pion. This approach is not feasible because most pions are not contained in the ArgoNeuT
TPC so their momentum cant be estimated. Instead,
we achieve signal from background separation by applying a multivariate method which exploits the topological and calorimetric information reconstructed in
each event. The ROOT Toolkit for Multivariate Analysis [21] was used to create a Boosted Decision Tree
(BDT) which is trained using genie signal and background samples. The classification is based on the angles of the pion and muon tracks, the visible energy
loss of the pion from the TPCs calorimetry, the reconstructed muon momentum from MINOS and the mean
stopping power of the first third of the muon track. The
last of these parameters was added to help distinguish
events where the start of the muon and pion tracks is
overlapping. The angular parameters have the highest discrimination power. An example of a neutrino
24
22
Data
Signal
Background
20
18
16
Event Counts
high charge
Event Counts
low charge
16
Data
Signal
Background
14
12
10
Time (ticks)
14
8
12
10
8
6
-1
-0.5
0.5
BDT Classification
-1
-0.5
0.5
BDT Classification
Time (ticks)
+0.6
39
= 5.5+2.6
cm2
(3)
2.1 (stat)0.7 (syst) 10
+1.2
+0.3
38
2
= 2.61.0 (stat)0.4 (syst)) 10 cm
(4)
per argon nuclei at E = 3.6 1.5 GeV and E =
9.6 6.5 GeV, where the 1.5(6.5) GeV represents the
range that contains 68% of the flux. A comparison between these measurements, existing data, and the ReinSeghal model are shown in Figure 3. The antineutrino
measurement agrees well with the Rein-Seghal model
while the neutrino one deviates by 1.2.
In this Letter, we have presented the first cross section measurement of CC coherent pion production on
argon. This is also the first time that machine learning
techniques have been applied to LArTPC data analysis. The large uncertainties on the final cross section
Flux normalization
+10.0
12.0
+10.0
12.0
4.1
4.3
1.6
2.7
CC QE background
+0.3
0.4
+1.2
0.6
CC RES background
+0.2
0.5
+0.4
0.3
CC DIS background
0.1
0.3
Nuclear Effects
0.3
0.7
POT
0.1
0.1
2.2
2.2
Efficiency
0.8
1.8
Signal modeling
0.8
5.7
Total systematics
+11.3
13.1
+12.9
14.5
[9] L. Alvarez-Ruso, L. S. Geng, S. Hirenzaki and M. J. ViCC-Coh [10-38 cm2 per Argon Nuclei]
ArgoNeuT
ArgoNeuT
Syst. Errors
GENIE (2.8.2)
NuWro (11m)
MINERvA
MINERvA
SKAT
SKAT
CHARM II
4
3.5
3
2.5
2
1.5
1
0.5
0
0
10
12
14
16
18
20
FIG. 3: ArgoNeuTs CC coherent pion cross section measurements (open circle and filled circle) compared to ReinSeghals model as implemented in genie and nuwro [5].
The statistical error is dominant (the systematic uncertainty
is shown alone for comparison). Data from other experiments in the same energy range is also shown. These consist in measurements made by SKAT (filled square, open
square), CHARM II (diamond) and MINERvA (triangle, inverted triangle) [2527]. These measurements are scaled to
Argon assuming the A1/3 dependance from the Rein-Seghal
model.