Abstract: Context. At GeV energies, the sky is dominated by the interstellar emission from the Galaxy. With limited statistics and spatial resolution, accurately separating point sources is therefore challenging. Aims. Here we present the first application of deep learning based algorithms to automatically detect and classify point sources from gamma-ray data. For concreteness we refer to this approach as AutoSourceID. Methods. To detect point sources, we utilized U-shaped convolutional networks for image segmentation and k-means for source clustering and localization. We also explored the Centroid-Net algorithm, which is designed to find and count objects. Using two algorithms allows for a cross check of the results, while a combination of their results can be used to improve performance. The training data are based on 9.5 years of exposure from The Fermi Large Area Telescope (Fermi-LAT) and we used source properties of active galactic nuclei (AGNs) and pulsars (PSRs) from the fourth Fermi-LAT source catalog in addition to several models of background interstellar emission. The results of the localization algorithm are fed into a classification neural network that is trained to separate the three general source classes (AGNs, PSRs, and FAKE sources). Results. We compared our localization algorithms qualitatively with traditional methods and find them to have similar detection thresholds. We also demonstrate the robustness of our source localization algorithms to modifications in the interstellar emission models, which presents a clear advantage over traditional methods. The classification network is able to discriminate between the three classes with typical accuracy of similar to 70%, as long as balanced data sets are used in classification training. We published online our training data sets and analysis scripts and invite the community to join the data challenge aimed to improve the localization and classification of gamma-ray point sources.

Abstract: Context. The existence of a maximum correlation angle (theta(max) & 60 greater than or similar to degrees) in the two-point angular temperature correlations of cosmic microwave background (CMB) radiation, measured by WMAP and Planck, stands in sharp contrast to the prediction of standard inflationary cosmology, in which the correlations should extend across the full sky (i.e., 180 degrees). The introduction of a hard lower cuto ff (k(min)) in the primordial power spectrum, however, leads naturally to the existence of theta(max). Among other cosmological anomalies detected in these data, an apparent dominance of odd-over-even parity multipoles has been seen in the angular power spectrum of the CMB. This feature, however, may simply be due to observational contamination in certain regions of the sky. Aims. In attempting to provide a more detailed assessment of whether this odd-over-even asymmetry is intrinsic to the CMB, we therefore proceed in this paper, first, to examine whether this odd-even parity imbalance also manifests itself in the angular correlation function and, second, to examine in detail the interplay between the presence of theta(max) and this observed anomaly. Methods. We employed several parity statistics and recalculated the angular correlation function for di fferent values of the cuto ff kmin in order to optimize the fit to the di fferent Planck 2018 data. Results. We find a phenomenological connection between these features in the data, concluding that both must be considered together in order to optimize the theoretical fit to the Planck 2018 data. Conclusions. This outcome is independent of whether the parity imbalance is intrinsic to the CMB, but if it is, the odd-over-even asymmetry would clearly point to the emergence of new physics.

Abstract: Continuous crystals can potentially obtain better intrinsic detector spatial resolution compared to pixelated crystals, additionally providing depth of interaction (DoI) information from the light distribution. To achieve high performance sophisticated interaction position estimation algorithms are required. There are a number of algorithms in the literature applied to different crystal dimensions and different photodetectors. However, the different crystal properties and photodetector array geometries have an impact on the algorithm performance. In this work we analysed, through Monte Carlo simulations, different combinations of realistic crystals and photodetector parameters to better understand their influence on the interaction position estimation accuracy, with special emphasis on the DoI. We used an interaction position estimation based on an analytical model for the present work. Different photodetector granulation schemes were investigated. The impact of the number of crystal faces readout by photodetectors was studied by simulating scenarios with one and two photodetectors. In addition, crystals with different levels of reflection and aspect ratios (AR) were analysed. Results showed that the impact of photodetector granularity is mainly shown near the edges and specially in the corners of the crystal. The resulting intrinsic spatial resolution near the centre with a 12 x 12 x 10 mm(3) LYSO crystal was 0.7-0.9 mm, while the average spatial resolution calculated on the entire crystal was 0.77 +/- 0.18 mm for all the simulated geometries with one and two photodetectors. Having front and back photodetectors reduced the DoI bias (Euclidean distance between estimated DoI and real DoI) and improved the transversal resolution near the corners. In scenarios with one photodetector, small AR resulted in DoI inaccuracies for absorbed events at the entrance of the crystal. These inaccuracies were slightly reduced either by increasing the AR or reducing the amount of reflected light, and highly mitigated using two photodetectors. Using one photodetector, we obtained a piecewise DoI error model with a DoI resolution of 0.4-0.9 mm for a 1.2 AR crystal, and we observed that including a second photodetector or reducing the amount of reflections reduced the DoI bias but did not significantly improve the DoI resolution. Translating the piecewise DoI error model obtained in this study to image reconstruction we obtained a spatial resolution variability of 0.39 mm using 85% of the FoV, compared to 2.59 mm and 1.87 mm without DoI correction or with a dual layer system, respectively.

Abstract: Contrary to the see-saw models, extended Higgs sectors leading to radiatively-induced neutrino masses do require the extra particles to be at the TeV scale. However, these new states have often exotic decays, to which experimental LHC searches performed so far, focused on scalars decaying into pairs of same-sign leptons, are not sensitive. In this paper we show that their experimental signatures can start to be tested with current LHC data if dedicated multi-region analyses correlating different observables are used. We also provide high-accuracy estimations of the complicated Standard Model backgrounds involved. For the case of the Zee-Babu model, we show that regions not yet constrained by neutrino data and low-energy experiments can be already probed, while most of the parameter space could be excluded at the 95% C.L. in a high-luminosity phase of the LHC.

Abstract: Convolutional neural networks (CNNs) are widely used state-of-the-art computer vision tools that are becoming increasingly popular in high-energy physics. In this paper, we attempt to understand the potential of CNNs for event classification in the NEXT experiment, which will search for neutrinoless double-beta decay in Xe-136. To do so, we demonstrate the usage of CNNs for the identification of electron-positron pair production events, which exhibit a topology similar to that of a neutrinoless double-beta decay event. These events were produced in the NEXT-White high-pressure xenon TPC using 2.6 MeV gamma rays from a Th-228 calibration source. We train a network on Monte Carlo-simulated events and show that, by applying on-the-fly data augmentation, the network can be made robust against differences between simulation and data. The use of CNNs offers significant improvement in signal efficiency and background rejection when compared to previous non-CNN-based analyses.