Skip to main content
Fig. 13 | Advanced Structural and Chemical Imaging

Fig. 13

From: Deep data analysis via physically constrained linear unmixing: universal framework, domain examples, and a community-wide platform

Fig. 13

Non-negative matrix factorization of X-ray diffraction microscopy. a The low-rank factors and b the sample distributions resultant from applying NMF to the XDM dataset in Fig. 1b, with k = 6 representatives. The low-rank factors are readily interpreted as different classes of spatially resolved XRD scans, with k = 1–3 belonging to SRO and k = 4–6 to PZT. In each U k , a dashed line indicates the Q z position of the SRO or PZT 103 Bragg peak as measured with standard XRD (i.e., spatially averaged across the entire sample). Note that each U k is associated with a single V k , whereby the image intensity in V k indicates the strength of association between sample regions and the structure encapsulated by the diffracted intensity in U k

Back to article page