Reconstructing neutrinoless double beta decay event kinematics in a xenon gas detector with vertex tagging M. Martínez-Vara, K. Mistry, F. Pompaa]Non-NEXT author. Now at SUBATECH, France, B.J.P. Jones, J. Martín-Albo, M. Sorel, C. Adams, H. Almazán, V. Álvarez, B. Aparicio, A.I. Aranburu, L. Arazi, I.J. Arnquist, F. Auria-Luna, S. Ayet, C.D.R. Azevedo, K. Bailey, F. Ballester, M. del Barrio-Torregrosa, A. Bayo, J.M. Benlloch-Rodríguez, F.I.G.M. Borges, A. Brodolin, N. Byrnes, S. Cárcel, A. Castillo, E. Church, L. Cid, C.A.N. Condeb]Deceased, T. Contreras, F.P. Cossío, R. Coupe, E. Dey, G. Díaz, C. Echevarria, M. Elorza, J. Escada, R. Esteve, R. Felkaic] Now at Weizmann Institute of Science, Israel, L.M.P. Fernandes, P. Ferrariod]On leave, A.L. Ferreira, F.W. Foss, Z. Freixa, J. García-Barrena, J.J. Gómez-Cadenase]NEXT Spokesperson, J.W.R. Grocott, R. Guenette, J. Hauptman, C.A.O. Henriques, J.A. Hernando Morata, P. Herrero-Gómez, V. Herrero, C. Hervés Carrete, Y. Ifergan, F. Kellerer, L. Larizgoitia, A. Larumbe, P. Lebrun, F. Lopez, N. López-March, R. Madigan, R.D.P. Mano, A.P. Marques, G. Martínez-Lema, R.L. Miller, J. Molina-Canteras, F. Monrabal, C.M.B. Monteiro, F.J. Mora, K.E. Navarro, P. Novella, A. Nuñez, D.R. Nygren, E. Oblak, J. Palacio, B. Palmeiro, A. Para, I. Parmaksiz, A. Pazos, J. Pelegrin, M. Pérez Maneiro, M. Querol, J. Renner, I. Rivilla, C. Rogero, L. Rogers, B. Romeof]Now at University of North Carolina, USA, C. Romo-Luqueg]Now at Los Alamos National Laboratory, USA, V. San Nacienciano, F.P. Santos, J.M.F. dos Santos, M. Seemann, I. Shomroni, P.A.O.C. Silva, A. Simón, S.R. Soleti, J. Soto-Oton, J.M.R. Teixeira, S. Teruel-Pardo, J.F. Toledo, C. Tonnelé, S. Torelli, J. Torrent, A. Trettin, A. Usón, P.R.G. Valle, J.F.C.A. Veloso, J. Waiton, A. Yubero-Navarro Abstract If neutrinoless double beta decay is discovered, the next natural step would be understanding the lepton number violating physics responsible for it. Several alternatives exist beyond the exchange of light neutrinos. Some of these mechanisms can be distinguished by measuring phase-space observables, namely the opening angle cos ⁡ θ \cos\theta among the two decay electrons, and the electron energy spectra, T 1 T_{1} and T 2 T_{2} . In this work, we study the statistical accuracy and precision in measuring these kinematic observables in a future xenon gas detector with the added capability to precisely locate the decay vertex. For realistic detector conditions (a gas pressure of 10 bar and spatial resolution of 4 mm), we find that the average cos ⁡ θ ¯ \overline{\cos\theta} and T 1 ¯ \overline{T_{1}} values can be reconstructed with a precision of 0.19 and 110 keV, respectively, assuming that only 10 neutrinoless double beta decay events are detected. arXiv: 2502.10198 The NEXT Collaboration 1 Introduction Neutrinoless double beta ( 0\nu\beta\beta ) decay is a hypothetical radioactive process in which two neutrons in a nucleus simultaneously transform into two protons, emitting two electrons: \hphantom{{}^{\text{A}}_{\text{Z}}}{\vphantom{\text{X}}}{}^{\mathchoice{\hbox to0.0pt{\hss$\displaystyle\vphantom{\smash[t]{\text{2}}}\text{A}$}}{\hbox to0.0pt{\hss$\textstyle\vphantom{\smash[t]{\text{2}}}\text{A}$}}{\hbox to0.0pt{\hss$\scriptstyle\vphantom{\smash[t]{\text{2}}}\text{A}$}}{\hbox to0.0pt{\hss$\scriptscriptstyle\vphantom{\smash[t]{\text{2}}}\text{A}$}}}_{\mathchoice{\hbox to0.0pt{\hss$\displaystyle\vphantom{\text{2}}\smash[t]{\text{Z}}$}}{\hbox to0.0pt{\hss$\textstyle\vphantom{\text{2}}\smash[t]{\text{Z}}$}}{\hbox to0.0pt{\hss$\scriptstyle\vphantom{\text{2}}\smash[t]{\text{Z}}$}}{\hbox to0.0pt{\hss$\scriptscriptstyle\vphantom{\text{2}}\smash[t]{\text{Z}}$}}}\kern 0.0pt\text{X}\,\text{$\to$}\,\hphantom{{}^{\text{A}}_{\text{Z+2}}}{\vphantom{\text{X}}}^{\mathchoice{\hbox to0.0pt{\hss$\displaystyle\vphantom{\smash[t]{\text{2}}}\text{A}$}}{\hbox to0.0pt{\hss$\textstyle\vphantom{\smash[t]{\text{2}}}\text{A}$}}{\hbox to0.0pt{\hss$\scriptstyle\vphantom{\smash[t]{\text{2}}}\text{A}$}}{\hbox to0.0pt{\hss$\scriptscriptstyle\vphantom{\smash[t]{\text{2}}}\text{A}$}}}_{\mathchoice{\hbox to0.0pt{\hss$\displaystyle\vphantom{\text{2}}\smash[t]{\text{Z+2}}$}}{\hbox to0.0pt{\hss$\textstyle\vphantom{\text{2}}\smash[t]{\text{Z+2}}$}}{\hbox to0.0pt{\hss$\scriptstyle\vphantom{\text{2}}\smash[t]{\text{Z+2}}$}}{\hbox to0.0pt{\hss$\scriptscriptstyle\vphantom{\text{2}}\smash[t]{\text{Z+2}}$}}}\kern 0.0pt\text{X${}^{\prime}$} + \text{2}\,\text{e}{\vphantom{\text{X}}}^{\text{\hskip 0.90417pt--\hskip 0.90417pt}} . Such a transition would violate the conservation of total lepton number by two units ( \Delta L=2 ), and hence it is forbidden in the Standard Model (SM) of particle physics. The search for 0\nu\beta\beta decay is, in fact, the most sensitive known test of lepton number violation (LNV). At present, the best experimental limits on the half-life of the decay exceed 10^{26} years [ 1 , 2 ] . 1 1 1 For an in-depth discussion of both the theoretical and experimental aspects of 0\nu\beta\beta decay, we refer the reader to two recent review papers on the topic [ 3 , 4 ] . Most often, 0\nu\beta\beta decay is described as taking place through the so-called mass mechanism , that is, the exchange of light, active Majorana neutrinos via SM left-handed weak currents. However, other LNV mechanisms --- arising, in general, in extensions of the SM involving new physics at energy scales above the electroweak one --- could also mediate the decay and contribute to its rate. 2 2 2 See, for instance, ref. [ 5 ] for a comprehensive list and discussion of these alternative 0\nu\beta\beta -decay mediating mechanisms. These various contributions would have to be disentangled, in the case of a 0\nu\beta\beta -decay discovery, to conclude anything on the features of the LNV sources and, in particular, on the scale of neutrino masses. There are several ways to do that, including a comparison of the 0\nu\beta\beta -decay rate in two or more isotopes (e.g. 136 Xe and 100 Mo or 76 Ge and 100 Mo) [ 6 ] or the decay to excited states [ 7 ] , the measurement of the angular distribution of the 0\nu\beta\beta electrons [ 8 , 9 , 10 ] , or correlating the 0\nu\beta\beta -decay rate with other probes such as cosmological observations [ 11 ] or experiments searching for heavy neutral leptons [ 12 ] . Here, we evaluate the ability of the NEXT technology to discriminate among different LNV mechanisms contributing to the decay amplitude via the reconstruction of the electron tracks in the detector. The NEXT detector is a high-pressure xenon gas time projection chamber that searches for 0\nu\beta\beta -decay: {}^{136}\textnormal{Xe}\to^{136}\textnormal{Ba}^{2+}+2\textnormal{e}^{-} . The detector can produce high-resolution images of large tracks down to mm resolution with excellent energy resolution. The distributions of the kinematic observables of opening angle and single-electron kinetic energy can be reconstructed if the decay vertex location within the detector volume is determined. This can be achieved either through the identification (or tagging ) of the daughter barium isotope, a technique currently under development in NEXT, or through advanced image analysis methods. Figure 1: Average cosine of the opening angle between decay electrons versus the average kinetic energy of the most energetic electron in the decay, for {}^{136}\mathrm{Xe} and some benchmark 0\nu\beta\beta -decay models: standard neutrino mass mechanism ( m_{\beta\beta} ), four different realizations of the minimal left-right symmetric Standard Model (mLRSM models in [ 13 ] ), and four different leptoquark contributions (leptoquark models in [ 14 ] ). Figure 1 shows example particle physics models [ 13 , 14 ] readily available within the \nu DoBe package [ 13 ] , the model generator we consider in this work. The models are shown in terms of the average cosine of the opening angle between decay electrons ( \overline{\cos\theta} ) and the average kinetic energy of the most energetic electron ( \overline{T_{1}} ) 3 3 3 To avoid any ambiguities in the 1 \leftrightarrow 2 electron labeling, throughout this work we assume that T_{1}>T_{2} in each decay, implying T_{1}>Q/2 . considering an ideal experiment detecting with perfect event reconstruction capabilities a sufficiently large number of {}^{136}\mathrm{Xe} 0\nu\beta\beta events to result in negligible statistical uncertainties, and adopting realistic nuclear and atomic physics inputs to compute the phase-space observables. For the standard mass mechanism, we expect \overline{T_{1}}=1.74 MeV and \overline{\cos\theta}=-0.28 in {}^{136}\mathrm{Xe} . Despite the fact that all models span the entire phase-space available ( Q/2T_{2,\mathrm{gen}} . Figure 2 shows the distributions of the single-electron kinetic energy ( T_{1,\mathrm{gen}} ) and opening angle between the emitted electrons ( \cos\theta_{\mathrm{gen}} ) for {}^{136}\mathrm{Xe} 0\nu\beta\beta decay driven by the \langle m_{\beta\beta}\rangle contribution. To avoid any confusion in our notation, the subscript ‘gen’ has been added here to indicate that these are generator-level observables, which do not account for any detector or reconstruction effects. Concerning T_{1,\mathrm{gen}} , the distribution peaks at the lowest possible value of Q/2 , inducing a relatively low value of \overline{T_{1}} for the \langle m_{\beta\beta}\rangle contribution in fig. 1 , and a relatively small asymmetry in the energy spectra of the most energetic ( e_{1} ) and the least energetic ( e_{2} ) electron in the decay. For what concerns the angular correlation, the two electrons are preferentially emitted back to back ( \cos\theta_{\mathrm{gen}}=-1 ) for the \langle m_{\beta\beta}\rangle model. 3.2 Detector simulation The propagation of the 0\nu\beta\beta electrons in gaseous xenon is simulated via nexus [ 54 ] , the Geant4-based simulation framework [ 55 , 56 , 57 ] of the NEXT Collaboration, and Geant4 release 11.2.1. For each event, the initial kinematics of the primary electrons is provided by \nu DoBe . Xenon gas at either 1, 5, 10 or 15 bar is used as sensitive volume, to assess how our results depend on gas pressure. A total of 6,000 simulated events are processed for each detector pressure setting. For simplicity, no detector border effects are considered; a sufficiently large active volume around the 0\nu\beta\beta -decay vertex positions is simulated to ensure full event containment. All relevant electromagnetic physics processes involving electrons, positrons and photons in the keV–MeV energy range are considered. Of particular importance is the simulation of elastic scattering affecting the primary electrons. The Geant4 physics list employed, G4EmStandardPhysics_option4 , described in the documentation as the most accurate set of models of electromagnetic physics, combines a Goudsmit-Sounderson multiple Coulomb scattering (MCS) model [ 58 ] to describe continuous, small-angle scattering with a single Coulomb scattering model applied for large angle scattering. The simulation is performed with a maximum step size among Geant4 hits of 4 mm. A 1 mm maximum step size was also studied but gave negligible differences. Figure 3 presents the simulated mean scattering angle ( \alpha ) of an electron after traversing a layer of 1 mm Xe gas at four different pressures (1, 5, 10 and 15 bar). The shaded band indicates the standard deviation of the computed \alpha at 10 bar, providing a visual representation of the variability and uncertainty associated to the scattering angle. The scattering probability of electrons is higher at lower energies, approaching zero as the energy of the particle increases. This scattering probability also increases as a function of the gas density. The nexus code is run in the so-called fast-simulation mode , meaning no scintillation photons nor ionization electrons are produced, transported, and detected. In this simulation mode, the output files only contain information about the tree of Geant4 particles (primary or secondary) in the event, and the energy deposits they leave in xenon gas for each Geant4 step. Figure 3: Mean scattering angle \alpha for electrons after going through a layer of 1 mm Xe gas for different gas densities: 0.56\times 10^{-3}~\mathrm{g/cm^{3}} (1 bar), 2.8\times 10^{-2}~\mathrm{g/cm^{3}} (5 bar), 5.6\times 10^{-2}~\mathrm{g/cm^{3}} (10 bar) and 8.4\times 10^{-2}~\mathrm{g/cm^{3}} (15 bar) at 291 K. The shaded band represents the standard deviation of \alpha at 10 bar. In the final simulation step, we break each energy deposit into individual ionization electrons and uniformly smear along the length of the Geant4 step. The ionization electrons are then divided into voxels of edge length 1, 2, 4, or 10–mm to capture the impact of varying detector spatial resolutions. Within each voxel, the mean position of the ionization electron cloud is used. This level of spatial separation among hits studied is realistic for a xenon gas detector with a tracking readout plane with sufficient granularity and coverage, and with reconstruction algorithms that are capable of unfolding the image blurring effects caused by diffusion and by the optical smearing produced in the electroluminescence gap. In particular, the existing track reconstruction in the NEXT-White detector, based on the Richardson-Lucy deconvolution algorithm [ 59 , 60 ] , has already demonstrated a FWHM of the reconstructed Gaussian-like charge distribution of order 5 mm for both simulated and real data, independent of the drift distance, hence resolving nearby point-like energy deposits on a similar spatial separation scale [ 20 ] . 3.3 Event reconstruction For the results presented in the next sections, we take into account the most important limiting factors in reconstructing the kinetic energy of the most energetic electron in the event, T_{1,\mathrm{reco}} , and the opening angle between the electrons’ initial momenta, \cos\theta_{\mathrm{reco}} . The non-perfect reconstruction is affected both by simulated detector effects discussed above and from our realistic classification of the various hits into e_{1} and e_{2} reconstructed tracks. Even assuming perfect knowledge of the decay vertex, as we do here, this task is non-trivial, given the complicated event topologies. Our approach is to use a simple reconstruction algorithm over more sophisticated methods, which will be followed up on future work. Our algorithm takes the vertex position as seed hit. We then sort all hits in the event in order of increasing distance from the vertex. The first hit (that is, the nearest hit to the vertex) is assigned to the e_{1} group, while the second hit is assigned to the e_{2} group. For the remaining hits, we proceed iteratively. For each hit in the event, from third to last, its distance to the last hits assigned to the e_{1} and e_{2} groups is computed. The hit is then assigned to the group that minimizes such distance, forming the new last hit in that group. The process ends when the farthest hit from the vertex has been classified, such that every hit is assigned to one of the two hit groups. The kinetic energy of the two decay electrons is simply the sum of the hit energy depositions in the two groups. As for the generator-level quantities, the 1 \leftrightarrow 2 electron labeling is redefined to ensure T_{1,\mathrm{reco}}>T_{2,\mathrm{reco}} for all events. For the calculation of \cos\theta_{\mathrm{reco}} , we found that using the first hit associated with each track instead of fitting the first few nodes gave the best performance. This performance difference is likely due to the multiple Coulomb scattering, which quickly washes out the angular information after the first few hits. In the scenario where the closest hits were part of the same track, the angular reconstruction tends to give a small opening angle ( \cos\theta_{\mathrm{reco}}\sim 1). For these cases, we check if the second node in each reconstructed track produces a larger angle. If it does, we then re-assign the mis-reconstructed node to the other track, accounting for the energy re-assignment. Figure 4: Two-dimensional projection in the XY plane of one simulated 0\nu\beta\beta decay assuming the neutrino mass mechanism and 10 bar xenon gas pressure with a 4 mm bin size. The left panel shows the entire event, while the right panel shows a zoomed-in portion near the event vertex, indicated by a cross. The markers show the smeared and binned energy deposits, with colors indicating the assigned electron. The arrows in the right panel show the directions of the decay electrons’ initial momenta projected onto the XY plane. As an illustration, a two-dimensional projection of a simulated 0\nu\beta\beta decay (for the neutrino mass mechanism) at 10 bar detector pressure and a 4-mm voxelization is shown in figure 4 . The color coding of the hits represents the reconstructed ( e_{1} or e_{2} ) tracks. Clearly, the electron tracks feature fairly complicated topologies, primarily due to the large amount of multiple Coulomb scattering along the tracks in dense xenon gas. Nevertheless, the capability to reconstruct the single electron kinetic energies and initial directions is preserved. We note that our simple reconstruction technique has been developed to address the central question of this paper of how well the energy and angular information is preserved. Future work will include improved reconstruction techniques, such as employing ML or track-building algorithms, when applied to tracks with realistic detector effects such as diffusion and assessing sensitivities to specific models. 4 Event-by-Event Reconstruction Performance Figure 5 shows the performance for T_{1,\mathrm{reco}} reconstruction at a fixed voxelization size of 4 mm for each pressure, while figure 6 shows the performance for angular reconstruction at a pressure of 10 bar for each voxelization. For the energy reconstruction, we note little dependence on pressure and voxelization. The event-by-event standard deviations of the difference between reconstructed and generated single electron energies range from 225 to 273 keV for all gas pressures and voxelizations studied. Such resolutions largely exceed the energy resolution on the T_{1}+T_{2} kinetic energy sum due to the NEXT technology itself, approximately 1% FWHM or \sim 10 keV sigma. This justifies our approximation of not introducing any additional energy smearing effect to our analysis. For the \cos\theta_{\mathrm{reco}} reconstruction performance, we note a stronger dependence on both the voxelization size and pressure with event-by-event reconstructed minus generated standard deviations ranging from 0.18 to 0.68. The best performance is achieved at the lowest pressure and the smallest voxelization. This is due to the improved spatial resolution and fewer scatters in a given voxel size. As can be seen in fig. 6 , smaller voxels (at a fixed pressure) tend to improve reconstruction performance. By using a 10 mm voxel size at 10 bar, a significant amount of information has been lost for the \cos\theta_{\mathrm{reco}} variable due to the multiple scattering over that length scale. Figure 5: Left: reconstructed T_{1} for a fixed voxelization of 4 mm for different pressures (area normalized). The black line shows the generated event distribution. Right: the distribution of reconstructed minus generated single electron kinetic energy. At this voxelization, excellent performance is achieved, with standard deviations for this difference ranging from 230 to 240 keV. There is also little variation in the performance with gas pressure. Figure 6: Left: reconstructed \cos\theta for a fixed pressure of 10 bar for different voxelizations (area normalized). The black line shows the generated event distribution. Right: the distribution of reconstructed minus generated for \cos\theta . At this pressure, the performance starts to deviate from the generated one when the voxelization is increased. The 10-mm distribution tends towards flat, indicating that a significant amount of information in \cos\theta is lost. One limitation of our hit classification algorithm is apparent in the left panel of fig. 6 where there is a peak towards small opening angles near \cos\theta_{\mathrm{reco}}=+1 . This peak is responsible for the asymmetric tail toward positive values in the \cos\theta_{\mathrm{reco}}-\cos\theta_{\mathrm{gen}} residuals shown in the right panel of fig. 6 . For a certain fraction of events, our assignment of the nearest hit to the vertex for each of the two electron tracks fails. Despite our efforts to reduce this impact, using the second-nearest node in the reconstructed track (see section 3.3 ), it is challenging to completely remove all mis-reconstructed events. Moreover, the fraction of mis-reconstructed events with small opening angle increases with voxel size and gas pressure. Upon further inspection, many mis-reconstructed events correspond to events where there is a strong asymmetry in the energy of the electrons. The electron that carries a small fraction of the energy does not travel far, and its information can be lost within a single voxel. 5 Statistical Accuracy and Precision in Kinematic Observables For each configuration described in 3.3 , the obtained datasets are then divided into several subsamples, each containing a fixed number n3.8\times 10^{26} yr at 90% CL [ 2 ] . In the following, we study how well we can reconstruct the following average quantities: \overline{X}=\frac{1}{n}\sum_{i=1}^{n}X_{i}\,, (1) where X=\{T_{1},\cos\theta\} and n=\{5,10,...,100\} . In particular, we evaluate two separate performance indicators: accuracy and precision. The accuracy, or bias, refers to how well the energy or angle reconstructed average values reproduce the generated average values, in the limit of very large signal sample statistics ( n\to\infty ). The precision, or resolution, refers instead to the variations from one signal dataset to another, in the realistic finite n case, in reconstructing the average values. We are particularly concerned with precision. While biases can largely be corrected for, although some model dependence may be introduced in the process, resolution effects constitute an irreducible uncertainty in reconstructing the energy and angle average values. Numerically, we estimate the accuracy as the difference between reconstructed and generated average values over the full datasets with N events. For a fixed number of signal events n