2. Diagnóstico y/o determinación de necesidades
3.2 Análisis de los resultados
As well as a single bin cross section measurement we examined distributions in other variables between data and Monte Carlo. Systematic uncertainties are less
thoroughly evaluated for these than for the single bin analysis so the data should be interpreted with care. Nonetheless these distributions are useful for two reasons:
• Differences between generators can be resolved by determining which best matches data. In this case the results presented should be taken as strictly preliminary and to be refined in future generations of gas interaction analysis.
• Areas where there are notable differences between generators highlight targets for future development. We have already identified improved reconstruction of low energy protons as a target for next generation TREx software for this reason.
9.2.1 Muon kinematics
The kinematic properties of our muon candidate were evaluated. We expect these measurements to be somewhat reliable given our reliable simulation of muons and propagation of systematic uncertainties relating to muon momentum.
Muon momentum is shown in Figure 9.2. When both statistical or systematic uncertainty are taken into account most of the bins agree within 1σ.
The angles of our muons relative to the forward beam direction are also shown (Figure 9.3). Again most of our bins agree within 1σ between real data and simulation. There does appear to be a slight but consistent deficit in muons in the highly forwards direction. This could hint at a higher ratio of out of fiducial volume background relative to true signal events in data compared with Monte Carlo†. Nonetheless these differences are not hugely significant when taking statistical and systematic uncertainties into account.
9.2.2 Secondary particle kinematics
Little development has so far been done with the specific goal of accurately recon- structing secondary particle kinematics in real data, nor on the systematic uncer- tainties of such measurements in simulation. This is a long term goal for the gas interaction analysis. The raw distributions presented here give an indication of what we are aiming to precisely quantify in the future.
These quantities are particularly interesting because our gas TPC allows a comparison between generators over variables which have not yet been precisely measured in experiment. In particular there is no experimentally verified model
†
Signal events are expected to produce predominantly forwards going muons whilst background features many muons entering from the sides of the TPC.
Muon momentum / MeV 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Events 0 20 40 60 80 100 120 Real data Neut MC Genie MC
Events passing selection by momentum
Figure 9.2: Selected muon momentum in real data and Monte Carlo.
θ cos -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 Events 20 40 60 80 100 120 Real data Neut MC Genie MC
Events passing selection by muon angle relative to beam axis
for MEC final states, and precise measurements of secondary particles and their kinematics present a chance to rectify this. GENIE generally favours events with a larger number of protons than NEUT, each carrying a lower proportion of the event’s total momentum. GENIE also predicts a much larger number of protons with energy between 50 MeV and the Fermi momentum at 250 MeV whilst NEUT predicts a sharp drop off in energies below around 250 MeV. Presently no data exists for tuning the generators in this low proton energy region.
We restrict ourselves to measuring particles of 100 MeV momentum or more, corresponding to ranges starting at around 30 cm. This is due primarily to our currently inability to trust PID algorithms at low energies. It also cuts out short tracks for which systematic uncertainties are presently still uncertain.
We checked the overall number of paths emerging from each candidate vertex (Figure 9.5). In doing this we found notably fewer high multiplicity events in real data. A possible non-physics cause for this is additional noise obscuring the true number of tracks emerging from a vertex. Despite our handling of hairy events it can still be very difficult to both accurately reconstruct the number of tracks emerging from a messy vertex and provide the successful fit needed to evaluate the track’s momentum. This is something which will require further study while developing
TRExfor future analyses.
As shown in Figure 9.5 we also checked the number of selected paths with proton PID (defined in both real data and Monte Carlo as a proton likelihoodLp >
0.9; see Section 7.2.1). Results here are similar to the case of total path multiplicity, with substantially fewer particles identified in data than in Monte Carlo. This supports the hypothesis that unsimulated effects of messy topologies are causing us to lose candidate secondary particles. If a vertex is difficult enough that we cannot reliably reconstruct a track as having over 100 MeV momentum it also makes sense that we cannot reliably assign it a good PID value.
Finally we checked the highest momentum of our candidate protons. This is illustrated in Figure 9.6. The normalisation is much smaller for real data than Monte Carlo simply because so many fewer events were identified as containing even one proton.