Fig. 4: The UML-PSD workflow for crystal assemblies. | Nature Communications

Fig. 4: The UML-PSD workflow for crystal assemblies.

From: Machine learning for microscopy data analytics targeting real-time optical characterization of semiconductor nanocrystals

Fig. 4: The UML-PSD workflow for crystal assemblies.

a The time-averaged image of a PL movie (.tiff), recorded for three self-assembled MAPbI3 crystals, with background pixels set to zero. b Illustration of the V&L interface that extracts and normalizes the unlabeled blinking traces corresponding to relevant pixels. Calculation of the <SS> and <%Mscl> profiles for the normalized ensemble of PL trajectories determines the Kopt at cluster 3 and estimates the optimum binning window at 40 frames. Metric profiles for the raw traces (in green) and optimally binned trajectories (in violet) are shown in bold. c Schematic of the segregation (UML-PSD) module, where the UML section applies the K-means algorithm to cluster the pixels exhibiting analogous blinking behavior, generating a cluster map of the crystal assembly based on the ensemble of binned PL trajectories. d The initial stage of the PSD module retrieves the raw PL trajectories depending on the cluster indexes. e The subsequent part computes the PSDs for the labeled blinking traces, extracting the pixel-wise power-law exponents (β) to construct the corresponding β-maps for the clusters.

Back to article page