An autonomous closed loop found a reproducible perovskite solar cell fabrication method and passivation molecule

AI editorPublished 2026.10.02Posted 2026.10.08 12:01 KST

Nature reported that an autonomous closed-loop framework combining active learning, quantum modelling, Bayesian optimization, and symbolic regression explored passivation molecules together with fabrication processes and identified 5ANI, achieving PCE 27.22% in 0.05 cm² cells and 23.49% in 21.4 cm² mini-modules. [1]

1/ AI usage methods: active learning, quantum modelling, Bayesian optimization, symbolic regression
2/ Applied task: combining passivation molecule exploration with fabrication process optimization
3/ Reported results: 5ANI was identified, achieving PCE 27.22% in small-area cells and 23.49% in mini-modules, with long-term stability confirmed
4/ Nature reported that a machine learning-based active learning, quantum modelling, Bayesian optimization, and symbolic regression combined autonomous closed-loop framework was used to jointly perform passivation molecule exploration and fabrication process optimization for perovskite solar cells.
5/ Using this approach, 5ANI passivation molecule was found and PCE 27.22% was achieved in 0.05 cm² cells and 23.49% in 21.4 cm² mini-modules, with long-term operational stability also confirmed.
6/ The important point for materials researchers is that molecule exploration and process optimization were not separated, but were bound into a single closed loop, creating a flow that repeatedly screened candidates and validated performance.
7/ This approach can be especially useful for solar cell problems in which passivation chemistry and device processing are strongly intertwined.
8/ However, the reported performance was obtained in a specific materials system and process window, so the same numbers will not automatically carry over to other compositions or other scales.
9/ In addition, it is safer to view the interpretations suggested by symbolic regression and quantum modelling as hypothesis-generation tools that still need to be verified experimentally, rather than as complete replacements for the actual cause. [1]

How to apply this

Use closed-loop active learning to jointly optimize passivation molecules and process conditions to improve solar cell performance.

1. Define passivation candidates and process variables simultaneously, and build an initial experimental set.
2. Use the experimental results to update the active learning and Bayesian optimization models so they can propose the next candidates and conditions.
3. Verify the top candidates first in small-area cells, then apply the same conditions to larger-area devices.
4. Evaluate not only performance, but also long-term operational stability and batch-to-batch reproducibility.

Data needed: Passivation candidate molecules, precursor solution compositions, process temperature and speed, small-area and module-level device performance data.
Checking the result: Check whether performance and stability are repeatedly maintained in independent batches and at different device areas, and compare the effect of the proposed molecule against controls.
What can go wrong: If the model is overfitted to a specific process window, a candidate that looks good may lose performance in actual scaling or long-term operation.

이 가이드는 원문의 방법을 바탕으로 AI가 제안한 적용 방향입니다. 원문에 적힌 내용이 아니며 같은 결과가 재현된다는 보장도 아닙니다.

웹 자료를 바탕으로 AI가 자동 요약한 글입니다. 연구의 세부 조건과 결과는 연결된 원문을 확인해주세요.

An autonomous closed loop found a reproducible perovskite solar cell fabrication method and passivation molecule

How to apply this

Related concepts and practice

Sources