Workflow of Characterization for Superconductors

Here is presented how a variety of characterization methods are used to locate and extract key samples for precise analysis using advanced characterization tools, such as Transmission Electron Microscopy. The objective of this work is to pre-select the sample needs from the exact location where materials present the properties and characteristics of interest. This is usually not a problem for homogeneous samples, which is rarely the case in real-life manufacturing. Today Machine Learning (ML) algorithms are allowing rapid interpretation of data, in our case we discuss Raman, SEM, etc. maps where we collect up to 10⁵ spectra to determine areas of predominance with key features that are associated to performance. Those can be areas with defects or pristine conditions. The workhorse for our characterization is Raman because of its simplicity, affordability, accuracy and micrometric resolution. Raman requires minimal sample preparation and provides global evaluation at the meso- and macro or bulk levels. Unfortunately, the amount of data for a map for industrial-ready set ups can reach meters. To understand and analyze the data in an effective manner the results are analyzed using ML. This unique approach helps to study the exact location where in-depth (e.g., atomic resolution) analysis is critical, allowing the identification of areas that will reveal paradigms related to performance. Here we will focus to present our results for superconductors. This simplified approach minimizes the labor intensity of other techniques such as XPS, FIB/pFIB, CF, ToF-SIMS, but particularly transmission and scanning transmission electron microscopy including EDS (2D and 3D), HAADF and EELS.