This code uses data (or simulation) to evaluate the performance of a likelihood ratio-based cut of 208Tl beta decays occuring in the bulk AV acrylic. The gammas from the 208Tl decay will produce light more spread out in time, providing a handle for discrimination.
It does this in two steps:
- Use one data set to build 1D (PMT hit time residual) PDFs for signal and background hypotheses
- Loop through events of known type in a second data set and compute the ratio of likelihoods for the hypotheses
Likelihood ratio distributions for cuts at several radii may then be analyzed to choose a fiducial volume cut that balances the leaked backgrounds with the irreducible 8B neutrino scatters.
To perform the full analysis, run:
$ python tl208_analysis.py [ncpus]
You will need to edit the data file paths at the top of tl208_analysis.py to
point to your data sets.
Any useful analysis will require very large event samples for PDF generation
and for fake data, on the order of 5 million 208Tl decays each. The analysis
may take a long time (several hours). It can be accelerated by running parallel
processes with the ncpus argument (defaults to 2). However, parallel
processes quickly become I/O bound.
If appropriate PDFs are found in pdfs/, they will be loaded. Otherwise PDFs
will be generated from data.
The code is highly modular, and individual parts of the analysis may be used
in other code or within the Python interpreter. main in tl208_analysis.py
provides a usage example.
The scripts output various files in a few places:
pdfs: PDFs generated from data. These take a very long time to produce, and so are cached for later reanalysis.tl208_analysis.pywill search this location for appropriate PDFs before computing them.figures: Portable Document Format PDF plots of probability distribution PDFs, PDF overlays, likelihood ratios, etc.lratios: Lists of per-event likelihood ratios for different event classes. These are also expensive to compute, so are saved for later re-plotting, evaluating sacrifice, etc.