Analysis of High Efficient Perovskite Solar Cells Using Machine Learning

Naman Shukla^a*, K. Anil Kumar^a, Madhu Allalla^a, Sanjay Tiwari^a

^aPhotonics Research Laboratory, S.O.S. in Electronics and Photonics, Pt. Ravishankar Shukla University Raipur, Chhattisgarh- 492010

*Corresponding author: naman.shukla43@gmail.com

            [Received: 10 December 2021; Revised: 10 February 2022; Accepted: 15 March 2022]

Abstract. Affordable manufacturing along with high efficiency perovskite solar cell in photovoltaic technology has everyone's attention. Perovskite, which is in the lead role in solar cells, is full of characteristics such as high absorption coefficient, low exciton binding energy, charge carrier capable of having better mobility as well as more diffusion length and availability in suitable energy band. The application of machine learning technology is proving to be a boon to ensure optimum implementation with different properties in photovoltaic device, design, simple construction process and low-cost price. Machine learning is a branch of artificial intelligence which includes large data aggregation, precise structure property installation, demonstration and final model after model validation. The most of the source of database is the simulation and experimental results, calculations and related literature surveys which have a comprehensive compilation of the performance of hybrid perovskite device, collection of structures and properties of elements. Structure-property relationship installation comes under feature engineering which establishes a clear relationship between structure and the properties. In other demonstration process, proper algorithms are selected, data is generated and tested as well as pure estimated values are taken. This article contains a detailed discussion on the involvement of machine learning technology to build high-performance Perovskite solar cells. Proper selection as well as designing of active perovskite absorbent layer by machine learning successfully establishes results by including other parts such as non-toxic (lead free) and stability. Mature machine learning technology becomes a very essential method in determining the solvent combination of hybrid perovskite and in estimating design of the entire solar cell to ensure optimum implementation in the sector of perovskite solar technology. Finally, a phased concept has been briefly discussed to meet the challenges of machine learning and potential future compatibilities related to the prevalence.

Keywords: Perovskite solar cells, machine learning, artificial intelligence, algorithms.

Introduction

Perovskite solar cell is considered as a revolution in photovoltaic solar cell technology. Its features such as low cost, easy fabrication, low temperature chemical synthesis and manufacturing process which is not commonly found in popular silicon-based solar cells (Green et al., 2019; Saliba et al., 2019). Currently, Perovskite solar cell reports a high-power conversion efficiency (PCE) of about 25.2 % (Sahli et al., 2018) Perovskite solar cells have high efficiency due to effective optical absorption coefficient, greater carrier diffusion distance capacity, low defects and increased properties like negligible recombination (recombination). However, continuous corrosion of perovskite stability in the presence of water and oxygen is a matter of research (Koocher et al., 2015; Müller et al., 2015). With this problem focused, many research works are being carried out everywhere in the country and abroad to increase the stability of perovskite compound. The main objectives of machine learning are selection of new perovskite substances under various interfacial technologies.

Figure 1. Features of machine learning

Mature machine learning technology becomes a very essential method in determining the liquid combination of hybrid Perovskite and the estimate design of the entire solar cell to ensure optimum implementation in the Perovskite based solar technology sector. This article discusses a phase concept to meet the challenges of machine learning and potential future compatibilities related to the prevalence.

Review of the machine learning’s applications in the research field of perovskite solar cells

Various properties have been established to study and analyze the perovskite absorbent layer structure and various characteristics of solar cells. It is through these qualities that we also explain the characteristics of the ideal perovskite. The common molecular formula of perovskite compound is ABX₃ (example: calcium titanium oxide CaTiO₃ etc.). This photovoltaic perovskite contains organic positive ions (e.g. methyl ammonium CH₃NH₃⁺etc.) and inorganic cation (e.g. lead Pb⁺², tin Sn⁺² etc.) respectively in A and B.X displays a halide resin e.g. I^-, Cl^-, Br^-etc.  As a result, perovskite compounds are formed MAPbI₃, MAPbBr₃ and MAPbI_3-xCl_x, etc. The formability and durability of perovskite is explained by goldsmith tolerance coefficient T_f, octahedral coefficient μ and other new versions (Goldschmidt et al., 1926; Allam et al., 2018; Bartel et al., 2019). The probability of building a permanent halide perovskite structure increases when the experimental value of T_f is between 0.81 and 1.11 and the quantity of octahedral coefficient μ is between 0.44 and 0.90 and the position of the T_f being 1 is the ideal cubic-3D perovskite (Kim et al., 2014; Johnsson et al., 2007).] ∆H, ∆L, HOMO, LUMO and E_g determine the main characteristics i.e. J_sc, V , FF and PCE (Li et al., 2019). The results of the formation energy, electronic dynamics and light absorption are determined by the halide composition, electronic structures, unified figure of merit respectively (Yang et al., 2020; Wang et al., 2020). Machine learning has been used for experimental and other device characteristic analysis based on research articles published over the last decade. These works provide a model for development of new permanent halide perovskite as well as suitable ETL, HTL, selection of solvents and process for fixed thickness for highly performance device (Odabas et al., 2019; Odabas et al., 2020; Ren et al., 2019). The machine learning method ensures search and optimum performance of appropriate conditions for achieving Reap-Rest-Recovery Cycle (3-R Cycle) i.e. dynamic performance (Howard et al., 2019). Perovskite solar cell has shown very low recombination with high PCE when it has good quality and low defect density. Therefore, good crystallization of perovskite is one of the essential needs. The perovskite crystallization process has been guided by analysis of crystal pictures and basic factors in harmony with convolution neural network with machine learning (Kirman et al., 2020). Machine learning techniques have been developed to study the imitation of phase transition hybrid perovskite substances (Jinnouchi et al., 2019).

Methodology

Machine learning can understand by dividing it into the main five parts. In the above figure 2, the main five parts of machine learning are description, clustering, classification, estimation and association (Irwin et al., 2005). Simple statistical analysis generally falls under description. The large data describes graphically in it (Tukey et al., 1977). The process of doing large data in small homogenous datasets based on their same properties is done by clustering.  There are two types of clustering that are prevalent - the first type is k-means clustering technology which is often used everywhere. There are other types hierarchical clustering commonly used in the field of material research. Decision trees (DT), logistic regression (LOR), k-nearest neighbor (KNN) algorithm, support vector machines (SVM), Bayesian classification (BC), Artificial neural network (ANN) have been used as common classification techniques. Classification divides data into classes. These classes have been categorized by the magnitude and state (cold, dry, hot) of the temperature of the substances. The forecast method establishes the forecast by establishing a relationship between the input and output variables, as indicated by name. For this, a number of technologies such as artificial neural networks (ANN), multiple regression (MR), regression tree (RT), gradient boosting regression (GBR), random forest (RF) and support vector regression (SVR) have been used. Artificial neural networks based on deep learning algorithms the use of substances has proved to be extremely useful in research.  Association rule mining (ARM) ensures the relationship between properties with large databases and identifying variables (variable) through it. In order to achieve the same objective, the prevailing aprior algorithm is used in data analysis (Larose et al., 2014). Other Subgroup discovery methods are also used in material research. This technique creates interesting associations between different variables and their special properties as desired (Helal et al., 2016; Goldsmith, et al., 2017). Figure 3 shows the phased process of modeling by machine learning technique. The purpose of machine learning is to move from large database to the last task of the model (Lei Zhang, et al., 2020). Perovskite Solar Cell (n-i-p type) model is displayed in Figure 4. It proposes selection of latest Perovskite active layer by model, selection of electron and whole transport layer in line with energy level (HOMO and LUMO) of absorbent perovskite layer and selection of proper electrode for charge termination. Through feature engineering, high absorption of light as well as sustainability can be ensured for lead free halide by altering the properties such as tolerance factor, octahedral factor, bandgap energy and other ionic charges. Formability values are indicated the stability of perovskite structure, determined by the calculations. It acts as the output variable after data processing of machine learning. Machine learning plays the key role to investigate the internal dissociation of energy and the structural factors. Association rule mining/ decision trees are helpful to increase in stability and diminish the degradation, employed to the preparation of dataset and analysis of hysteresis data (Choudhary et al., 2019; Odabas et al., 2020).

Figure 3. Machine learning steps to getting a model of perovskite solar cell

Conclusion

Perovskite compound-based research has recorded a historic growth over the years. In the same vein, the use of machine learning in physics research has also increased rapidly. The application of machine learning has given amazing momentum to the journey of bringing perovskite solar cell technology from new born to adult. As a result, a good number of research articles, review articles, magazine reports, and comments have been published. The computing techniques and experimental works for perovskite technology development are set to grow in future. Similarly, development in machine learning and data management technology is also possible. However, going from research to experimental level by machine learning model or from matter to device is a challenging task itself. It is enough to understand the fact that the stability of matter does not guarantee the stability of the device. The difference can be seen only because of the process of manufacturing the device, storage and operational conditions, and dependence on maintenance of different laboratories. The source of data required for this error prevention is linked to a number of various sources, with a good quantity of non-uniformity levels, mismatches and missing data, etc., which will yield good and accurate results.

Acknowledgement

I am thankful for research facilities at Photonics Research Laboratory, School of Studies in Electronics and Photonics, Pt. Ravishankar Shukla University, Raipur, Chhattisgarh, India.

References

A.Y.-T. Wang, R.J. Murdock, S.K. Kauwe, A.O. Oliynyk, A. Gurlo, J. Brgoch, K.A. Persson, T.D. Sparks, Machine Learning for Materials Scientists: anintroductory guide towards best practices, Chem. Mater. 32 (2020) 4954–4965, https://doi.org/10.1021/acs.chemmater.0c01907  

C.J. Bartel, C. Sutton, B.R. Goldsmith, R. Ouyang, C.B. Musgrave, L.M. Ghiringhelli, M. Scheffler, New tolerance factor to predict the stability ofperovskite oxides and halides, Sci. Adv. 5 (2019) 1–10, https://doi.org/10.1126/sciadv.aav0693

C. Müller, T. Glaser, M. Plogmeyer, M. Sendner, S. D¨oring, A.a.Bakulin, C. Brzuska, R. Scheer, M.S. Pshenichnikov, W. Kowalsky, A. Pucci, R. Lovrinˇci´c, Water infiltration in methylammonium lead iodide perovskite: fast and inconspicuous, Chem. Mater. 27 (2015) 7835–7841, https://doi.org/10.1021/acs.chemmater.5b03883.

Ç. Odabas¸ı, R. Yıldırım, Performance analysis of perovskite solar cells in 2013–2018 using machine-learning tools, Nanomater. Energy 56 (2019) 770–791, https://doi.org/10.1016/j.nanoen.2018.11.069.

Ç. Odabas¸ı, R. Yıldırım, Assessment of reproducibility, hysteresis, and stability relations in perovskite solar cells using machine learning, Energy Technol. (2020), 1901449, https://doi.org/10.1002/ente.201901449.

Ç. Odabas, R. Yıldırım, Machine learning analysis on stability of perovskite solarcells, Sol. Energy Mater. Sol. Cells 205 (2020), 110284, https://doi.org/10.1016/j.solmat.2019.110284.

D.T. Larose, C.D Larose, Discovering Knowledge in Data: An Introduction to Data Mining, second ed., John Wiley & Sons, New Jersey, 2014.

F. Sahli, J. Werner, B.A. Kamino, M. Br¨auninger, R. Monnard, B. Paviet-Salomon, L. Barraud, L. Ding, J.J. Diaz Leon, D. Sacchetto, G. Cattaneo, M. Despeisse, M. Boccard, S. Nicolay, Q. Jeangros, B. Niesen, C. Ballif, Fully textured monolithic perovskite/silicon tandem solar cells with 25.2% power conversion efficiency, Nat. Mater. 17 (2018) 820–826, https://doi.org/10.1038/s41563-018-0115-4.

Green, M., Ho-Baillie, A. and Snaith, H. (2014). The emergence of perovskite solar cells. Nature Photon 8, 506–514.

H.S. Kim, S.H. Im, N.G. Park, Organolead halide perovskite: new horizons in solarcell research, J. Phys. Chem. C 118 (11) (2014) 5615–5625, https://doi.org/10.1021/jp409025w.

J.J. Irwin, B.K. Shoichet, ZINC--a free database of commercially available compounds for virtual screening, J. Chem. Inf. Model. 45 (2005) 177-182.

J. Kirman, A. Johnston, D.A. Kuntz, M. Askerka, Y. Gao, P. Todorovi´c, D. Ma, G. G. Priv´e, E.H. Sargent, Machine-learning-accelerated perovskite crystallization, Matter 2 (2020) 938–947, https://doi.org/10.1016/j.matt.2020.02.012.

J. Li, B. Pradhan, S. Gaur, J. Thomas, Predictions and strategies learned from machine learning to develop high-performing perovskite solar cells, Adv. Energy Mater. 9 (2019) 1–10, https://doi.org/10.1002/aenm.201901891.

J.M. Howard, E.M. Tennyson, B.R.A. Neves, M.S. Leite, Machine learning for perovskites’ reap-rest-recovery Cycle, Joule 3 (2019) 325–337, https://doi.org/ 10.1016/j.joule.2018.11.010.

J.W. Tukey, Exploratory Data Analysis, Addison-Wesley Publishing Company, 1977.

J. Yang, Y. Wang, T. Wu, S. Li, Correlating the composition-dependent structural and electronic dynamics of inorganic mixed halide perovskites, Chem. Mater. 32 (2020) 2470–2481, https://doi.org/10.1021/acs.chemmater.9b04995.

K. Choudhary, M. Bercx, J. Jiang, R. Pachter, D. Lamoen, F. Tavazza, Accelerated discovery of efficient solar cell materials using quantum and machine-learning methods, Chem. Mater. 31 (2019) 5900–5908, https://doi.org/10.1021/acs.chemmater.9b02166.

Lei Zhang, Mu, He, Shaofeng Shao, Machine learning for halide perovskite materials, Nano Energy 78 (2020) 105380, https://doi.org/10.1016/j.nanoen.2020.105380

M. Johnsson, P. Lemmens, Crystallography and chemistry of perovskites, handbookof magnetism and advanced magnetic materials, in: H. Kronmüller, S. Parkin (Eds.),Volume 4: Novel Materials, John Wiley & Sons Ltd, Chichester, Uk, 2007, pp. 2098–2106 https://doi.org/10.1002/9780470022184.hmm411.

M. Saliba, Polyelemental, multicomponent perovskite semiconductor libraries through combinatorial screening, Adv. Energy Mater. 9 (2019), 1803754, https://doi.org/10.1002/aenm.201803754.

N.Z. Koocher, D. Saldana-Greco, F. Wang, S. Liu, A.M. Rappe, Polarization dependence of water adsorption to CH3NH3PbI3 (001) surfaces, J. Phys. Chem. Lett. 6 (2015) 4371–4378, https://doi.org/10.1021/acs.jpclett.5b01797.

O. Allam, C. Holmes, Z. Greenberg, K.C. Kim, S.S. Jang, Density functional theory - machine learning approach to analyze the bandgap of elemental halideperovskites and ruddlesden-popper phases, ChemPhysChem 19 (2018) 2559–2565, https://doi.org/10.1002/cphc.201800382.

R. Jinnouchi, J. Lahnsteiner, F. Karsai, G. Kresse, M. Bokdam, Phase transitions of hybrid perovskites simulated by machine-learning force fields trained on the fly with bayesian inference, Phys. Rev. Lett. 122 (2019), 225701, https://doi.org/10.1103/PhysRevLett.122.225701.

S. Helal, Subgroup Discovery Algorithms: A Survey and Empirical Evaluation, J. Comp. Sci. and Tech. 31 (2016) 561-576.

V.M. Goldschmidt, Die gesetze der Krystallochemie, Naturwissenschaften 14 (1926) 477–485, https://doi.org/10.1007/BF01507527.

Z. Ren, S. Tian, T. Heumueller, E. Birgersson, F. Lin, A. Aberle, S. Sun, I.M. Peters, R. Stangl, C.J. Brabec, T. Buonassisi, F. Oviedo, H. Xue, M. Thway, K. Zhang, N. Li, J.D. Perea, M. Layurova, Y. Wang, Physics-guided characterization and optimization of solar cells using surrogate machine learning model, in: 2019 IEEE 46th Photovolt. Spec. Conf., IEEE, 2019, pp. 3054–3058, https://doi.org/10.1109/PVSC40753.2019.8980715.