Event reconstruction covers two parts: Multiplicity and first interaction points of neutrons. There are several methods to determine both individually. For example, the combination of the calorimetric multiplicity method R3BNeulandMultiplictyCalorimetric and cluster selection via R-Value with R3BNeulandNeutronsRValue is equivalent to the classic TDR reconstruction.
Multiplicity methods and neutron methods can be mixed and matched freely, however most neutron methods require multiplicity as input. Most methods require some sort of calibration.
Overview:
- multiplicity/R3BNeulandMultiplicityBayes Simple probability calculation from event properties
- multiplicity/R3BNeulandMultiplicityCalorimetric Classic calorimetric cuts
- multiplicity/R3BNeulandMultiplicityCheat Get number of reacted neutrons from simulation
- multiplicity/R3BNeulandMultiplicityFixed Set a fixed value to each event
- multiplicity/R3BNeulandMultiplicityScikit Use a pre-trained pickled scikit-learn model
- neutrons/R3BNeulandNeutronsCheat Get correct neutrons from simulation
- neutrons/R3BNeulandNeutronsRValue Classic R-Value sorting for clusters
- neutrons/R3BNeulandNeutronsScikit Use a pre-trained pickled scikit-learn model
Multiplicity
Multiplicity is saved in R3BNeulandMultiplicity data structures for each event. This contains an array to store the probability for each possible multiplicity value, e.g., [0, 0, 0.000911, 0.105381, 0.893707, 0, 0]. The provided GetMultiplicity method would return 4 in this case. However, you might want to introduce filters for less clear-cut cases. Note that some algorithms, like the calorimetric one, will provide the multiplicity as one-hot: [0, 0, 0, 0, 1, 0, 0].
Calorimetric method
It was found that plotting the number of clusters vs the total energy is an acceptable way of determining the neutron multiplicities, at least for large numbers of planes.
This method needs to be calibrated. In this process, the cuts applied to the 2D histogram need to be created and adjusted such that the "best" result is obtained. This is handled by the Neuland::Neutron2DCalibr class, which is not a task as specifically set inputs are needed (events have different numbers of primary neutrons). Cuts created by this process are stored in R3BNeulandMultiplicityCalorimetricPar.
At this point, efficiency matrices can already be created for the calibration dataset, e.g.:
| 600 MeV | 1n | 2n | 3n | 4n |
|---|---|---|---|---|
| 0n | 0.06 | 0 | 0 | 0 |
| 1n | 0.90 | 0.16 | 0.01 | 0 |
| 2n | 0.04 | 0.76 | 0.23 | 0.04 |
| 3n | 0 | 0.07 | 0.68 | 0.32 |
| 4n | 0 | 0 | 0.08 | 0.56 |
With the number of generated neutrons on the columns and the detected neutron multiplicity in the rows. Note that in this example, the chance to detect no neutron whatsoever is only 6% in the case of one incident neutron. In any other case, the total detection efficiency is 100%.
The cuts are saved in the parameter file via R3BNeulandMultiplicityCalorimetricPar. Provided with the total energy and number of clusters, this class then can return the neutron multiplicity.
