The NEXT experiment, a search for neutrinoless double beta decay, has developed new and innovative techniques for addressing a common challenge in applying AI to experimental data: how does one train a model on imperfect simulation and successfully use it on data, with an understanding of how well the AI will work on the data? 

The techniques developed by Argonne, in a collaboration with the University of Valencia, have shown that symmetry-preserving augmentation techniques, in combination with statistical measurements of the intermediate state of the model, can greatly enhance the reliability of AI techniques for experimental data.

The search for neutrinoless double beta decay is a top priority for the DOE Office of Science, Nuclear Physics, as highlighted in the 2015 long range plan.  Experimental discovery of neutrinoless double beta decay would provide insights into the matter/antimatter asymmetry of the Universe and probe some of the most critical open questions in fundamental physics.

The NEXT series of experiments involves competitive neutrinoless double beta decay searches in Canfranc, Spain, with Argonne leading the design of the next-generation time projection chamber.  The AI techniques developed in this project are applicable to current and future searches for neutrinoless double beta decay in NEXT.  With an improvement of more than 50% in the figure of merit, these techniques extend the scientific reach and impact of the NEXT experimental program.