Fig. 1. Overview of the first version of the R3 B setup in Cave C at GSI [9], used with permission.
Fig. 2. The full 30 dp NeuLAND geometry, as it was considered in our simulations.
Fig. 3. The current state of the construction of the NeuLAND detector (8 dp) [10];
figure used with permission.
Fig. 4. The Different steps in the NeuLAND data analysis. The upper branch illustrates the steps used for simulated data and the lower branch illustrates the steps used for
measured data.
Fig. 5. Illustration of the TDR algorithm for multiplicity determination in the situation for a 30 dp NeuLAND and a 600 MeV neutron energy. The number of detected neutrons
refers to the number of assigned shower heads by the ‘‘Perfect Tracking’’ algorithm. The red numbers in the upper left corners refer to the fraction of events in each histogram
that falls between the respective cuts (black lines) in that histogram. See text for further details.
Fig. 6. Performance of the multiplicity determination in NeuLAND. For different neutron multiplicities fired by the particle gun, the percentage of events where that multiplicity
was established, is plotted against the number of NeuLAND double-planes (dp). The full results are shown as solid lines for the TDR algorithm, the new DNN algorithm and for
the ‘‘Perfect Tracking’’ algorithm, which gives the highest possible performance. The dashed lines correspond to the restricted scenario; the part of events where false-positives are
excluded. The simulations are performed with a neutron energy of 200 MeV. 200,000 events were simulated per multiplicity-figure. See text for further details.
Fig. 7. Same as Fig. 6, but for a neutron energy of 600 MeV.
Fig. 8. Same as Fig. 6, but for a neutron energy of 1000 MeV.
Fig. 9. Invariant Mass spectra for 4 neutrons, each carrying an energy of 200 MeV. The spectra (a, b, c, d and e) were reconstructed from different NeuLAND configurations (in
terms of dp). Four times the mass of the free neutron was subtracted from the total invariant mass, which is why the figures are labeled ‘Invariant Mass Difference’. The results
of the TDR, the DNN and the ‘‘Perfect Tracking’’ algorithms are plotted, together with their statistical uncertainties (around the data) and their systematic uncertainties from the
physics list (separate bands). To obtain the results for each curve, 200,000 4𝑛 events were simulated. See text for further details.
Fig. 10. Same as Fig. 9, but for a neutron energy of 600 MeV.
Fig. 11. Same as Fig. 9, but for a neutron energy of 1000 MeV.
Fig. 12. Performance of the neutron multiplicity determination as a function of 𝛥𝜃max . The full results are shown as solid lines for the TDR, DNN and ‘‘Perfect Tracking’’ algorithms.
The dashed lines correspond to the restricted scenario; the part of events where the false-positives are excluded. The simulations are performed for a neutron energy of 600 MeV,
a NeuLAND configuration of 30 dp and a distance between NeuLAND and the target of 14 m. See text for further details.
Fig. 13. Invariant Mass spectra for 4 neutrons, each carrying an energy of 600 MeV. A NeuLAND configuration of 30 dp at a distance of 14 m was used. Four times the mass
of the free neutron was subtracted from the total invariant mass, which is why the figures are labeled ‘Invariant Mass Difference’. The spectra are plotted for different values of
𝛥𝜃max . To obtain the results for each curve, 200,000 4𝑛 events were simulated. See text for further details.
Fig. 14. This project has received funding from the European Union’s Horizon 2020
research and innovation programme under grant agreement No. 654002.
Table 2
Optimalization of the DNN network structure.
