High-throughput thermophysical and structural characterizations of semiconductors using transient thermoreflectance

Pump-probe thermoreflectance (Pump-probe TR) is a non-contact technique for thermal properties characterization of materials, which is also used to analyze material quality and detect defects. For wafers, uniform thermophysical properties and thickness are critical to avoid localized thermal hotspots and device failure. Traditional spot detection methods for fitting unknown parameters are time-consuming and labor-intensive when handling large datasets. This work proposes a high-throughput method for semiconductor thermophysical and structural characterizations, integrating automatic scanning measurements with deep learning techniques. A multilayer perceptron (MLP) model processes the data, demonstrating reliability through global scanning of Au-sapphire samples, revealing variations in thermal boundary conductance (TBC). The method also detects material defects and is validated in industrial applications with GaN-on-Si samples, offering high accuracy and speed. In addition, we demonstrated the feasibility of the method for structural characterization of materials by scanning measurement of the thickness of Si thin film. This high-throughput method can reduce data processing time to the millisecond level, significantly reducing time and labor compared to traditional methods, making it particularly efficient for high-volume wafer thermophysical and structural characterization.