Coronavirus (Covid-19): Latest updates and information
Skip to main content Skip to navigation

Work at Warwick

This page describes some of the contributions made by physicists at the University of Warwick to the Deep Underground Neutrino Experiment (DUNE).

Event reconstruction using Pandora pattern-recognition software

Liquid-Argon Time-Projection Chambers, or LArTPCs, provide high spatial and calorimetric resolution, providing fine-granularity images of particle tracks and electromagnetic showers. The three wire planes give three two-dimensional images of the event which acts as input images and are reconstructed using algorithms developed using the Pandora Software Development Kit (SDK).

Pandora employs a multi-algorithm approach where each algorithm performs a specific task in a particular event topology. Chains of many tens of algorithms gradually build up a picture of events and collectively provide a robust reconstruction. The aim is to assign input "hit" objects to clusters in each wire plane, to match clusters representing the same particle between the different wire planes, then to organise the reconstructed particles into hierarchies representing the charged particles in the detector and their subsequent interactions or decays.

Pandora offers two main chains of algorithms for use at LArTPCs: PandoraCosmic, which is optimised for the reconstruction of cosmic-ray muons and delta rays, and PandoraNu/PandoraTestBeam, which is optimised for the reconstruction of neutrino interactions or test beam interactions, where multiple charged particles emerge from an identified interaction vertex position. These algorithm chains can be combined intelligently to deliver a "consolidated" event reconstruction for surface-based or deep-underground LArTPCs with one or more drift volumes.

raw and reco hits using pandora

(a). One of the three two-dimensional images of an event inside the detector as seen from the W-view which acts as an input image for Pandora.

(b). Various algorithms operate on these images , gradually building it up, to reconstruct the final event.

One of the most important outputs of Pandora reconstruction is the PFParticle (PF stands for "Particle Flow") which corresponds to a clear track or a clear shower and is associated with a list of 2D clusters and 2D hits. It is also associated to a list of reconstructed 3D space points and a vertex. The PFParticles are placed in a hierarchy.

Warwick physicists are writing algorithms using Pandora and its outputs to tackle various pattern recognition problems such as finding primary and secondary vertices of a neutrino interaction and discriminating between track-like and shower-like topologies. Recent work has also included optimisation of the algorithms used to reconstruct cosmic-ray muons, the addition of new algorithms to support the single-phase and dual-phase ProtoDUNE detectors at CERN, and the use of convolutional neural networks to provide additional input information to aid pattern-recognition decisions.

Data acquisition for ProtoDUNE

Two prototype liquid argon time projection chambers, known as ProtoDUNEs, were constructed at CERN, Geneva in 2017-2018. One of these prototypes took data from a dedicated particle beam in the second half of 2018. Both prototypes will also take data from cosmic rays in 2019. Physicists at Warwick supplied software that does the data acquisition from the photon detectors in one of the ProtoDUNEs. They are also involved in monitoring the performance of the photon detectors using the online monitor.

Resolution of deltaCP
Data acquisition for ProtoDUNE

Optimisation of target

The beam of muon neutrinos or muon antineutrinos is produced by colliding a beam of protons with a solid target, as part of the Long-Baseline Neutrino Facility (LBNF) accelerator complex at Fermilab. These collisions produce charged pions, which decay quickly to either antimuons and muon neutrinos (from positively charged pions) or to muons and muon antineutrinos (from negatively charged pions). It is possible to select either muon neutrinos or muon antineutrinos for the beam by deflecting one sign of charged pions away from the beamline using magnetic horns. For example, deflecting negatively charged pions removes most of the muons and muon antineutrinos from the beamline. The antimuons are stopped by a layer of graphite but the muon neutrinos pass through this layer and travel on to the DUNE detectors.

Physicists at Warwick have made extensive simulations of different types of neutrino target in order to optimise the flux of neutrinos at the far detector. These simulations have shown that the best type of target is a graphite cylinder of length 2 metres. Optimising the neutrino flux in this way will improve the measurements of δCP and the neutrino mass hierarchy that DUNE will make when it starts taking data in 2026.

Resolution of deltaCP
Graphite cylinder neutrino target of length 2 metres

Particle identification

It is not possible to detect neutrinos or antineutrinos directly. They can only be detected indirectly when they interact in a neutrino detector, with the detection being made using the particles that are produced in the interaction. One common type of interaction is between a muon neutrino and a neutron in a detector which produces a muon and a proton (other particles can also sometimes be produced in these interactions). The equivalent interaction for an electron neutrino produces an electron and a proton (and sometimes other particles).

Physicists at Warwick are working on particle identification in the DUNE far detector. Initially this was focussed on distinguishing electrons from muons in that detector since that enables interactions of electron neutrinos to be distinguished from those of muon neutrinos. The discrimination between electrons and muons is based on the different shapes of their energy depositions in the detector and the different rates of their energy depositions per unit length. This work has been extended to include identification of other particles, e.g. protons and pions.

Warwick physicists are also investigating the use of machine learning to identify particles in the DUNE detectors.

Neutrino energy reconstruction

In the DUNE experiment, the distance between the neutrino target and the far detector is fixed at 1300 km. Under these circumstances, oscillations from one type of neutrino to another depend on the energy of the initial neutrino. For this reason, it is essential to know the energies of neutrinos that interact in the DUNE detectors. However the neutrino or antineutrino beam has a broad range of energies, which means that the energy of the neutrino in a given interaction cannot be known in advance. The only way to determine the energy of a given neutrino is to make a measurement in the detector of the products of the interaction; this is known as "reconstructing" the neutrino energy.

Physicists at Warwick have developed software to reconstruct the energies of muon neutrinos or antineutrinos and electron neutrinos or antineutrinos in the DUNE far detector. Several different algorithms are used in this reconstruction. For example, a muon neutrino interacts in the far detector to produce a muon that is contained within the detector and a proton. In this example, the momentum of the muon is estimated from the length of its track, and the energy of the proton is estimated from the charge deposited in the detector that is not in the muon track. The reconstructed energy of the incoming neutrino is then estimated as the sum of the muon momentum and proton energy.