Optimizing geophysical muon radiography using information theory

Abstract

Cosmic ray muons are highly penetrating, with some reaching several kilometres into solid rock. Consequently, muon detectors have been used to probe the interiors of large geological structures, by observing how the muon flux varies with direction of arrival. There is an increasing need to discriminate between materials differing only slightly in bulk density. A particularly demanding application is in monitoring underground reservoirs used for CO2 capture and storage, where bulk density changes of approximately 1 per cent are anticipated. Muon arrival is a random process, and it is the underlying expectation values, not the actual muon counts, which provide information on the physical parameters of the system. It is therefore necessary to distinguish between differences in muon counts due to real geological features, and those arising from random error. This is crucial in the low-contrast case, where the method can reach the information theoretic limit of what a data source can reveal, even in principle. To this end, methods to analyse information availability in low-contrast muon radiography have been developed, as have means to optimally interpret the available data, both for radiography and for tomography. This includes a method for calculating expectation values of muon flux for a given geological model directly, complementing existing Monte Carlo techniques. A case study, using a model of carbon capture is presented. It is shown that the new data analysis techniques have the potential to approximately double the effective sensitivity of the detectors.