Image-based impedance spectroscopy for printed electronics

Abstract The field of printed electronics has been extensively researched for its versatility and scalability in flexible and large-area applications. Impedance is of great importance for the performance and reliability of electronics. However, its measurement requires electrical contacts, which makes it difficult on complex or bio-interfaces. Although the printing process is accessible, impedance characterization may be cumbersome, which can create a bottleneck during the manufacturing process. This paper reports the first effort at developing a convolutional neural network (CNN) based image regression model to replace impedance spectroscopy (IS). In our study, the CNN model learned the features of inkjet-printed electrode images that are dependent on the printing and sintering of nanomaterials and quantitatively predicted the resistance and capacitance of the equivalent circuit of the inkjet-printed lines. The image-based impedance spectroscopy (IIS) is expected to be the cornerstone as a revolutionary approach to electronics research and development enabled by deep neural networks.