Approaching Novel Perovskites Photovoltaic Devices through Machine Learning and Interfacial Engineering
Name
zhang-rqzhang-sm-eecs-2023-thesis.pdf
Description
Thesis PDF
Size
13.79 MB
Format
Adobe PDF
Checksum (MD5)
183ecc8a925fc110dfdd095209b141c6
Author(s)
Zhang, Ruiqi
Advisor(s)
Bulović, Vladimir
Date Issued
September 2023
Publisher
Massachusetts Institute of Technology
Abstract
Organic metal halide perovskites have shown plenty of extraordinary optoelectronic properties which make them good candidates for various photovoltaic applications [1-5]. The fascinating optoelectronic properties of perovskite largely take credit to their low exciton binding energy, strong light absorption coefficient, relatively long carrier diffusion length, and carrier recombination lifetime [6-9]. However, even with an increasing number of studies carried out, perovskite solar cell is still facing plenty of challenges towards commercialization. Two main challenges towards large-area commercialization include first the harsh fabrication environment and cost of large-area coating; and second the redundant fabrication process with a huge labor force impelled. In this thesis study, an intermedia thin film layer tris(4-carbazoyl-9ylphenyl)amine (TcTa) with a thickness of 3 nm is discovered in a large-area compatible perovskite solar cell structure ITO/SnO2/(MAFACs)1Pb(IBrCl)3/PV2000/TcTa/Au that reaches a power conversion efficiency above 14%. The TcTa intermediate film is compatible with substituting gold top electrodes and preventing sputter damage while maintaining a similar solar cell performance (etc. sputtered Ni). In addition, a machine learning algorithm is developed to predict the solar cell current-voltage properties only based on the film stack optical properties before the solar cell is fabricated. The algorithm is developed and tested based on the 3D/2D perovskite solar cell structure [10] with resulting in an average prediction regression loss below 5% and a best prediction accuracy above 99%. Multiple different machine learning algorithm is also carried out to analyze the prediction results and learning weights for the model.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
