Machine learning-guided antireflection coatings architectures and interface modification for synergistically optimizing efficient and stable perovskite solar cells

Abstract

In recent years, single-junction perovskite solar cells (PSCs) have experienced unprecedented development, approaching the Shockley-Queisser (S-Q) theoretical efficiency limit, due to versatile optimization strategies targeting functional layers to minimize energy loss. The antireflection coating (ARC), as part of the light-management strategy, plays a critical role in reducing optical loss to achieve higher efficiency. The development of multifunctional ARC that can simultaneously enhance visible light transmittance while suppressing ultraviolet (UV) light transmission, along with excellent adhesion and wear resistance on glass substrates, remains a significant challenge in current research. Herein, we propose ultra-thin ARC made of multilayer dioxides, SiO₂–TiO₂–SiO₂ (STS) films, optimized using a machine learning approach with a Bayesian optimization algorithm. This process involved parameterized modeling of multilayer dioxide ARC, physical simulations using the Transfer Matrix Method (TMM), and evaluation of antireflective performance. The optimal configuration of STS ARC consists of 100 ​nm SiO₂, 10 ​nm TiO₂, and 10 ​nm SiO₂, increasing the transmittance of FTO glass by 9.2% in the 400–800 ​nm wavelength range. The ARC effectively enhances external quantum efficiency, achieving 96.94%, thereby increasing the short-circuit current density (J_SC) and power conversion efficiency (PCE) by 4%. PSCs with STS ARC retain 81.2% of their initial efficiency after continuous UV illumination for 300 ​h, while control devices degrade to approximately 69%, demonstrating effective UV filtration and improved operational stability. This ARC exhibit hardness exceeding 9H on the pencil hardness scale and achieve ISO class 0/ASTM class 5B in adhesion tests, meeting the outdoor durability requirements for PSCs. In addition to optical energy loss, the accumulation of defects on the surface of the perovskite layer induces non-radiative recombination energy loss and serves as initiation sites for lattice degradation. To address this, we use 3-amidinopyridinium iodide (3-PyADI) to passivate interface defects, further improving the PCE to 24.44%. The stability of the device remains at 93% of the initial PCE after 1000 ​h under atmospheric conditions. The proposed ARC and PSCs structure are expected to enhance optoelectronic performance and environmental stability, providing a promising and practical path for the development of PSCs.