In materials science, performing a synchronous correlative analysis of multiphysics fields at the nanoscale is critical for advancements in material characterization. Atomic force microscopy (AFM) is a fundamental technique for nanoscale characterization; however, multiphysics field noise suppression and feature enhancement remain challenging. Recent advances have demonstrated the exceptional capabilities of deep learning in multiscale feature extraction and high-dimensional data visualization. Accordingly, this paper proposes a multimodal data-fusion-based image enhancement model that captures surface topographical characteristics and facilitates a correlative analysis of surface physical properties. A deep learning enhancer framework is constructed, and a convolutional neural network is employed to extract and enhance features from multiscale AFM data. This approach facilitates the discovery of latent correlations between topographical characterization and the physical property analysis. The proposed model enables precise identification of chromosome-associated regions in super-resolution (SR) topographical images, facilitating the acquisition of accurate surface morphological features. An in-house three-probe AFM system integrated with two conductive probe modules can synchronously acquire the topographical information and physical properties of chromosome surfaces, enabling multidimensional characterization at nanoscale resolution. A systematic correlation analysis of chromosome surface topography in SR images reveals the spatial positions of chromosomal domains (short arms, centromeres, and long arms) and establishes their structural correspondence with their associated physical properties. The proposed method opens up new avenues for exploring the correlation mechanism among material surface topography, mechanical properties, and surface potential distribution, laying the foundation for breakthroughs in materials science.
Tian et al. (Fri,) studied this question.