Knowledge extraction for organometallic perovskite solar cells from published data in literature

Abstract

Orgonolead halide perovskite solar cells (PSCs) have been attracted great attention in recent years. This rapid progress is due to excellent light absorption and charge-carrier mobilities of the perovskite materials besides its low-cost and easy processing conditions. In addition to high power conversion efficiency (PCE), reproducibility, hysteresis and longterm stability of perovskite solar cells are major factors to be solved before commercialization of this technology . The objective of this dissertation is to extract useful knowledge from literature to improve the overall performance of this technology for commercialization. The extensive datasets for PCE, reproducibility, hysteresis and longterm stability of PSCs were constructed from the published papers in literature and analyzed using machine-learning tools to determine the effects of materials and perovskite deposition methods employed during cell manufacturing. The evolution of PCE with time was statistically analyzed under different circumstances (i.e. using different materials types or perovskite deposition methods). Then, the databases for PCE, hysteresis and long-term stability were modeled using random forest, association rule mining and decision tree methods to detect the most effective variables and combinations leading to high performance. For reproducibility, pooled variances of different factors were calculated and compared. The mixed cation perovskites, doped mesoporous TiO2 (second electron transfer layer) and LiTFSI+TBP+FK209 (additive to hole transfer materials) were found to promote high efficiency, reproducibility and stability while they lowered the hysteresis; SnO2 (compact ETL), DMF+DMSO (solvent) and diethyl ether (anti-solvent) also had positive effects on these cell characteristics except hysteresis. Hence, it was concluded that the common factors which leaded high PCE, also leaded high reproducibility, low hysteresis and long-term stability. Additionally, our findings were in a reasonable aggrement with the literature showing that the data mining and statistics can be used effectively to derive general results and detect trends, which can not be seen by naked eyes.