Researchers from Jinan University, Guangdong Mellow Energy and Shanghai Jiao Tong University have reported a flexible perovskite solar cell architecture that combines data‑driven machine‑learning optimization with a targeted passivation scheme based on amorphous grain‑boundary engineering to simultaneously address power conversion efficiency, operational stability and mechanical reliability. The perovskite microstructure is tailored to suppress nonradiative recombination, mitigate ion‑migration pathways and enhance fracture tolerance under repeated bending, enabling flexible devices with high efficiency, extended durability under environmental stressors and robustness suitable for lightweight, large‑area modules for wearable, vehicular and building‑integrated photovoltaic applications.
The optimized architecture delivers flexible perovskite solar cells with a power conversion efficiency of 24.52%, while maintaining 92.5% of the initial efficiency after 10,000 bending cycles, 95% after 300 days in ambient conditions and 80% after 650 h of continuous maximum power point tracking. Certified flexible modules reach 21.09% efficiency at an aperture area of 21.07 cm² and 17.38% at 0.5 m² (86.9 W output), while a larger 1.4725 m² module delivers 226 W with a specific power of 558 W kg⁻¹, underscoring the scalability of the amorphous grain‑boundary engineering approach. These metrics collectively show that the strategy both advances the state of the art in flexible perovskite device efficiency and mitigates key reliability bottlenecks that have historically impeded the transition to large‑area, application‑relevant modules.
The work leverages machine learning to navigate the complex parameter space that governs flexible perovskite performance, including composition, interface chemistry and processing conditions. Instead of relying on trial‑and‑error optimization, the model identifies high‑value regions in this multidimensional space, accelerating convergence toward device architectures that maximize power conversion efficiency while remaining compatible with low‑temperature, flexible‑substrate fabrication. This data‑centric framework is particularly well suited to perovskite photovoltaics, where subtle changes in precursor stoichiometry or post‑treatment can dramatically influence defect density and film morphology.
A key innovation lies in engineering amorphous grain boundaries within the perovskite layer to passivate defects that normally act as nonradiative recombination centers and trigger degradation. By introducing tailored passivation species that form an amorphous intergranular phase, the team effectively decouples adjacent grains electronically and mechanically, reducing trap‑assisted recombination while improving strain accommodation during flexing. This amorphous boundary network also disrupts continuous ion‑migration pathways, thereby enhancing operational stability under electrical bias and illumination, a longstanding bottleneck for flexible perovskite modules.
Flexible perovskite devices must tolerate repeated bending without catastrophic cracking or delamination of the active layers. The amorphous grain‑boundary framework acts as a mechanically compliant buffer that redistributes local stress, increasing the fracture toughness of the perovskite film and delaying crack initiation and propagation under cyclic strain. Coupled with mechanically optimized charge‑transport and encapsulation layers, this microstructural design supports stable performance over extensive bending cycles, a prerequisite for wearable and conformable photovoltaic applications.
Because the demonstrated strategy relies on passivation chemistry and process optimization rather than exotic substrates or high‑temperature treatments, it is compatible with roll‑to‑roll and other scalable coating techniques used in flexible electronics manufacturing. The integration of machine‑learning‑guided process windows with amorphous grain‑boundary passivation provides a blueprint for transitioning flexible perovskite solar cells from small‑area laboratory devices to large‑area modules without sacrificing efficiency or stability, positioning flexible perovskites as strong candidates for next‑generation photovoltaic technologies in sectors where lightweight, conformable and high‑power‑density energy harvesting is essential.
