AFM Imaging Defect Detection and Classification with Artificial Intelligence and Deep Learning

Atomic force microscopy (AFM) is a powerful tool for investigating the topographical and mechanical properties of a wide range of samples. However, the process of data acquisition and analysis using AFM is time-consuming and labor-intensive, and can be prone to error. Mistaken imaging of the wrong sample, damage to the sample, and tip wear are just a few examples of potential sources of error. Currently, the integration of artificial intelligence (AI) with AFM operations for identifying and classifying sample images is still under investigation. In this paper, we propose a novel deep learning (DL) framework for the classification of AFM images. The key component of the proposed AI framework is a DL neural network model that is trained using the AFM images with various types of blemishes labeled by experts. Then the trained network will be used to analyze future AFM images of the same type of samples. Specifically, the proposed DL model consists of a defect detection layer and a feature extraction and classification layer to ensure the classification accuracy. The Performance of the proposed AI framework was demonstrated in terms of defect detection and classification accuracies in both training and validation. The proposed AI approach will greatly reduce the time and labor cost in AFM quality analysis, and can be potentially extended to other applications, such as manufacturing, material engineering, and biomedical engineering.