Deep Learning-Assisted Graphene Coverage Estimation for Transfer Quality Optimization

Monolayer graphene is a suitable active material for several electronics and sensor applications, for which large-area individual sheets are necessary on insulating substrates. Chemical vapor deposition (CVD) enables the synthesis of large-area graphene of high quality on metallic supports. However, graphene grown in this manner has to be transferred to suitable target substrates, for which there are still several challenges involved to obtain intact and defect-free sheets on a large scale. Here, we present a deep-learning assisted workflow for quantifying the coverage of transferred graphene based on high resolution optical image data, offering an accessible and scalable method for quality assessment. The workflow is based on training a neural network-based classifier (U-Net classifier) using an initial set of images to automatically identify areas covered by graphene. The trained model is subsequently used to quantify graphene coverage on samples obtained by different transfer methods. Utilizing this methodology, we first show that appropriate substrate pre-treatment is essential to maximize the coverage of transferred graphene. Moreover, we demonstrate that a post-transfer annealing treatment significantly improves the stability of graphene sensors in aqueous environments, a critical factor for sensing applications.