Abstract
In recent years, Artificial Intelligence techniques have proved to be verysuccessful when applied to problems in physical sciences. Here we apply anunsupervised Machine Learning (ML) algorithm called Principal ComponentAnalysis (PCA) as a tool to analyse the data from muon spectroscopyexperiments. Specifically, we apply the ML technique to detect phasetransitions in various materials. The measured quantity in muon spectroscopy isan asymmetry function, which may hold information about the distribution of theintrinsic magnetic field in combination with the dynamics of the sample. Sharpchanges of shape of asymmetry functions - measured at different temperatures -might indicate a phase transition. Existing methods of processing the muonspectroscopy data are based on regression analysis, but choosing the rightfitting function requires knowledge about the underlying physics of the probedmaterial. Conversely, Principal Component Analysis focuses on small differencesin the asymmetry curves and works without any prior assumptions about thestudied samples. We discovered that the PCA method works well in detectingphase transitions in muon spectroscopy experiments and can serve as analternative to current analysis, especially if the physics of the studiedmaterial are not entirely known. Additionally, we found out that our MLtechnique seems to work best with large numbers of measurements, regardless ofwhether the algorithm takes data only for a single material or whether theanalysis is performed simultaneously for many materials with different physicalproperties.