Novel event classification based on spectral analysis of scintillation waveforms in Double Chooz
T. Abrahão, H. Almazán, J. C. dos Anjos, S. Appel, J. C. Barrière, T. J. C. Bezerra, L. Bezrukov, Timothée Goubault de Brugière, C. Buck, J. Busenitz, A. Cabrera, L. Camilleri, M. Cerrada, E. Chauveau, P. Chimenti, J. I. Crespo-Anadón, J. Dawson, Z. Djurcic, A. Etenko, M. Fallot, D. Franco, H. Furuta, I. Gil‐Botella, A. Givaudan, H. Gómez, L. F. G. Gonzalez, M. C. Goodman, T. Hara, J. Haser, D. Hellwig, A. Hourlier, M. Ishitsuka, J. Jochum, C. Jollet, K. Kale, Philipp Kampmann, M. Kaneda, T. Kawasaki, E. Kemp, H. de Kerret, D. Kryn, M. Kuze, T. Lachenmaier, C. Lane, T. Lasserre, C. Lastoria, D. Lhuillier, H.p. Lima, M. Lindner, J. M. López-Castaño, J. M. LoSecco, Б. Лубсандоржиев, J. Maeda, C. Mariani, J. Maricic, T. Matsubara, G. Mention, A. Meregaglia, T. Miletic, A. Minotti, D. Navas-Nicolás, P. Novella, L. Oberauer, M. Obolensky, A. Onillón, A. Oralbaev, C. Palomares, Iuri Muniz Pepe, G. Pronost, B. Reinhold, R. Santorelli, S. Schönert, S. Schoppmann, M. Settimo, R. Sharankova, V. V. Sinev, M. Skorokhvatov, P. Soldin, A. Stahl, I. Stancu, L.f.f. Stokes, F. Suekane, S. Sukhotin, T. Sumiyoshi, Yujing Sun, A. Tonazzo, C. Veyssière, B. Viaud, M. Vivier, S. Wagner, C. H. Wiebusch, G. Yang, F. YermiaJournal of Instrumentation2018 · DOI · arXiv
Liquid scintillators are a common choice for neutrino physics experiments, but their capabilities to perform background rejection by scintillation pulse shape discrimination is generally limited in large detectors. This paper describes a novel approach for a pulse shape based event classification developed in the context of the Double Chooz reactor antineutrino experiment. Unlike previous implementations, this method uses the Fourier power spectra of the scintillation pulse shapes to obtain event-wise information. A classification variable built from spectral information was able to achieve an unprecedented performance, despite the lack of optimization at the detector design level. Several examples of event classification are provided, ranging from differentiation between the detector volumes and an efficient rejection of instrumental light noise, to some sensitivity to the particle type, such as stopping muons, ortho-positronium formation, alpha particles as well as electrons and positrons. In combination with other techniques the method is expected to allow for a versatile and more efficient background rejection in the future, especially if detector optimization is taken into account at the design level.
Introduction
In the field of neutrino physics, liquid scintillators serve as primary detection media due to their high light yield upon particle interaction. However, a significant challenge in large-scale detectors is the effective implementation of Pulse Shape Discrimination (PSD). While different particles, such as electrons and alpha particles, produce distinct light-pulse shapes, these differences are often obscured in massive detectors. As pulses travel through the scintillator, scattering and reflections "smear" the timing information, making traditional time-domain analysis—focused on rise and decay times—highly difficult. This limitation impacts the ability to distinguish neutrino signals from instrumental noise and background radiation.
Methodology: Spectral Shape Discrimination (SSD)
To overcome the limitations of time-domain analysis, we propose a shift to the frequency domain using Spectral Shape Discrimination (SSD). The workflow begins with pulse reconstruction, where signals from 390 photomultiplier tubes (PMTs) are aggregated and corrected for time-of-flight based on the event vertex.
Example scintillation pulses in the time domain for low (left) and high (right) amplitude signals.
Following normalization to ensure energy independence, a Discrete Fourier Transform (DFT) is applied to the reconstructed pulse. A key advantage of the Fourier power spectrum is its inherent robustness against timing jitters, as it is independent of the pulse's starting position within the readout window. Finally, an optimized discriminator variable ($\Omega$) is constructed as a weighted sum of Fourier coefficients, with weights mathematically tuned to maximize the separation between distinct event classes.
Results and Discussion
The SSD technique demonstrated significant improvements over traditional methods across several classification tasks. In distinguishing events between the Neutrino Target and the Gamma Catcher volumes, SSD achieved a figure of merit ($Q$) of 3.56, more than doubling the performance of the best time-domain method ($Q = 1.42$).
Spectral coefficients for two event classes shown on linear (left) and logarithmic (right) scales, illustrating the basis for SSD.
The method also proved highly effective at identifying instrumental "light noise" that had bypassed standard filters. Furthermore, SSD showed sensitivity to subtle physical effects, such as the pulse broadening caused by ortho-positronium formation (with a lifetime of approximately 3.7 ns). Particle identification capabilities were also observed, with statistical separation achieved between alpha particles and electrons, as well as a detectable shift between electron and positron signals.
Conclusion
The development of SSD represents a significant advancement in event classification for large liquid scintillator detectors. By leveraging the stability of frequency components over time-domain features, the method achieves unprecedented background rejection and particle identification. Future improvements may include the use of non-linear classifiers like Artificial Neural Networks and the optimization of detector hardware to further enhance spectral resolution.
Summary updated
Abstract
Liquid scintillators are a common choice for neutrino physics experiments, but their capabilities to perform background rejection by scintillation pulse shape discrimination is generally limited in large detectors. This paper describes a novel approach for a pulse shape based event classification developed in the context of the Double Chooz reactor antineutrino experiment. Unlike previous implementations, this method uses the Fourier power spectra of the scintillation pulse shapes to obtain event-wise information. A classification variable built from spectral information was able to achieve an unprecedented performance, despite the lack of optimization at the detector design level. Several examples of event classification are provided, ranging from differentiation between the detector volumes and an efficient rejection of instrumental light noise, to some sensitivity to the particle type, such as stopping muons, ortho-positronium formation, alpha particles as well as electrons and positrons. In combination with other techniques the method is expected to allow for a versatile and more efficient background rejection in the future, especially if detector optimization is taken into account at the design level.
BibTeX
@article{abrahao2018novel,
author = {T. Abrahão and H. Almazán and J. C. dos Anjos and S. Appel and J. C. Barrière and T. J. C. Bezerra and L. Bezrukov and L. Bezrukov and Timothée Goubault de Brugière and C. Buck and J. Busenitz and A. Cabrera and L. Camilleri and M. Cerrada and E. Chauveau and P. Chimenti and P. Chimenti and J. I. Crespo-Anadón and J. Dawson and Z. Djurcic and A. Etenko and M. Fallot and D. Franco and H. Furuta and I. Gil‐Botella and A. Givaudan and H. Gómez and L. F. G. Gonzalez and M. C. Goodman and T. Hara and J. Haser and D. Hellwig and A. Hourlier and M. Ishitsuka and J. Jochum and C. Jollet and K. Kale and Philipp Kampmann and M. Kaneda and T. Kawasaki and E. Kemp and H. de Kerret and D. Kryn and M. Kuze and T. Lachenmaier and C. Lane and T. Lasserre and C. Lastoria and D. Lhuillier and H.P. Lima and M. Lindner and J. M. López-Castaño and J. M. LoSecco and Б. Лубсандоржиев and J. Maeda and C. Mariani and J. Maricic and T. Matsubara and G. Mention and A. Meregaglia and T. Miletic and T. Miletic and A. Minotti and D. Navas-Nicolás and P. Novella and L. Oberauer and M. Obolensky and A. Onillón and A. Oralbaev and C. Palomares and Iuri Muniz Pepe and G. Pronost and B. Reinhold and R. Santorelli and S. Schönert and S. Schoppmann and M. Settimo and R. Sharankova and V. V. Sinev and V. V. Sinev and M. Skorokhvatov and P. Soldin and A. Stahl and I. Stancu and L.F.F. Stokes and F. Suekane and S. Sukhotin and T. Sumiyoshi and Yujing Sun and A. Tonazzo and C. Veyssière and B. Viaud and M. Vivier and S. Wagner and C. H. Wiebusch and G. Yang and F. Yermia},
title = {{Novel event classification based on spectral analysis of scintillation waveforms in Double Chooz}},
journal = {Journal of Instrumentation},
year = {2018},
doi = {10.1088/1748-0221/13/01/p01031},
}