Herrero-Garcia, J., Schwetz, T., & Zupan, J. (2012). Astrophysics independent bounds on the annual modulation of dark matter signals. Phys. Rev. Lett., 109(14), 141301–5pp.
Abstract: We show how constraints on the time integrated event rate from a given dark matter (DM) direct detection experiment can be used to bound the amplitude of the annual modulation signal in another experiment. The method requires only mild assumptions about the properties of the local DM distribution: that it is temporally stable on the scale of months and spatially homogeneous on the ecliptic. We apply the method to the annual modulation signal in DAMA/LIBRA, which we compare to the bounds derived from XENON10, XENON100, cryogenic DM search, and SIMPLE data. Assuming a DM mass of 10 GeV, we show that under the above assumptions about the DM halo, a DM interpretation of the DAMA/LIBRA signal is excluded for several classes of models: at 6.3 sigma (4.6 sigma) for elastic isospin conserving (violating) spin-independent interactions, and at 4.9 sigma for elastic spin-dependent interactions on protons.
|
Herrero-Garcia, J., Patrick, R., & Scaffidi, A. (2022). A semi-supervised approach to dark matter searches in direct detection data with machine learning. J. Cosmol. Astropart. Phys., 02, 039–19pp.
Abstract: The dark matter sector remains completely unknown. It is therefore crucial to keep an open mind regarding its nature and possible interactions. Focusing on the case of Weakly Interacting Massive Particles, in this work we make this general philosophy more concrete by applying modern machine learning techniques to dark matter direct detection. We do this by encoding and decoding the graphical representation of background events in the XENONnT experiment with a convolutional variational autoencoder. We describe a methodology that utilizes the `anomaly score' derived from the reconstruction loss of the convolutional variational autoencoder as well as a pre-trained standard convolutional neural network, in a semi-supervised fashion. Indeed, we observe that optimum results are obtained only when both unsupervised and supervised anomaly scores are considered together. A data set that has a higher proportion of anomaly score is deemed anomalous and deserves further investigation. Contrary to classical analyses, in principle all information about the events is used, preventing unnecessary information loss. Lastly, we demonstrate the reach of learning-focused anomaly detection in this context by comparing results with classical inference, observing that, if tuned properly, these techniques have the potential to outperform likelihood-based methods.
|