As the development of single-junction solar cells reaches a bottleneck, tandem solar cells have emerged as a critical pathway to further enhance power conversion efficiency. Among them, monolithic perovskite/silicon heterojunction tandem solar cells are currently the fastest-growing technology, achieving the highest efficiencies at relatively low costs. The interconnecting layer, which connects the two sub-cells, plays a crucial role in tandem cell performance. It collects electrons and holes from the respective sub-cells and facilitates recombination and tunneling at the interface. Therefore, the properties of the interconnecting layer are pivotal to the overall device performance. In this work, we applied statistical analysis and machine learning algorithms to systematically analyze the interconnecting layer. A comprehensive dataset on interconnecting layer parameters was established, and predictive modeling was performed using Lasso linear regression, random forest, and multilayer perceptron (a type of neural network). The analysis revealed key feature importance for experimental parameters, providing valuable insights into the application of interconnecting layers in perovskite/silicon heterojunction tandem solar cells. The final optimized interconnecting layer can achieve a proof-of-concept efficiency of 38.17%, providing guidance and direction for the development of monolithic perovskite/silicon tandem solar cells.
Just Accepted manuscripts are peer-reviewed and accepted for publication. They are posted online prior to technical editing formatting for publication and author proofing.

To optimize turn on velocity of the SiC LIMS, we proposed a new structure for the LIMS that incorporates an optimized n+ layer and a multi-light triggered electrode design for the anode. The chip size is 5.5 mm × 5.5 mm in dimension. The experiment results indicate that the saturation laser energy required to trigger the prepared SiC LIMS has been decreased from 1.8 mJ to 40 μJ, with the forward blocking voltage of the prepared SiC LIMSs capable of withstanding over 7000 V. The leakage current is about 0.3 μA at room temperature, and the output current density achieves 4.25 kA/cm2 (with di/dt larger than 20 kA/μs).

Avalanche photodiode (APD) is a kind of photodetector with important applications in optical communication, LIDAR and other fields. APDs fabricated using the recently developed AlGaAsSb as the multiplication material exhibit excellent noise performance. In this work, we report a low-noise separate absorption, grading, charge, and multiplication (SAGCM) InGaAs/AlGaAsSb APD operating at 1550 nm. A double-mesa structure was fabricated to reduce the dark current. Numerical simulations were conducted to compare two different mesa-structured APDs. By analyzing the electric field distribution, it was found that the electric field at the edge of the multiplication region in the double-mesa APD is nearly 100 kV/cm lower than that of the single-mesa structure. Experimental results demonstrate that after device punch-through, the double-mesa APD’s dark current can be reduced by up to four times compared to the single-mesa APD. Quantitative analysis of the dark current components in the AlGaAsSb APD further confirms that the low sidewall electric field in the double-mesa structure effectively suppresses the trap-assisted tunneling. Additionally, noise measurements indicate a k-value of approximately 0.014, which is significantly lower than that of traditional multiplication materials. This work provides preliminary validation for further performance improvements in low noise and low dark current AlGaAsSb APDs.