Machine learning based screening of mixed lead-free double-perovskites
Alessio Gagliardi a, Felix Mayr a
a Dept. Electrical and Computer Eng., Technische Universitaet Muenchen, Karlstrasse 45, Munich, 80333, Germany
nanoGe Perovskite Conferences
Proceedings of International Conference on Perovskite Thin Film Photovoltaics and Perovskite Photonics and Optoelectronics (NIPHO20)
Sevilla, Spain, 2020 February 23rd - 25th
Organizer: Hernán Míguez
Oral, Felix Mayr, presentation 037
Publication date: 25th November 2019

Compositional engineering of perovskites enables the precise control of key material properties such as the bandgap [1]. This possibility makes perovskites a promising material for multijunction, “tandem” solar cells, where the combination of two different bandgaps allows to easily break the Shockley-Queisser limit and thus improve efficiency [2]. The remaining challenge is to find structures with the target bandgap which are both stable in the environment as well as non-toxic (i.e. lead-free). To this end, computer simulations allow rapid screening of a large array of compositions for a given structure and subsequent data modeling. However typical hightroughput calculation fall short in capturing the effect of different geometries in modeling thus severely constraining the applicability of the result in predicting “new” structures. This becomes especially problematic for larger systems, such as the mixed, lead-free double perovskites, where different compositions might have varying relaxed geometries and sampling the whole feature space becomes infeasible. Incorporating geometrical information into the data-modeling step is typically achieved by variations on the radial-distribution function (RDF) [3], resulting in a large feature vector fed into a machine learning (ML) pipeline. Recently, we published a new RDF based descriptor, merging geometrical description of the structure with compositional information in the “property density distribution function” (PDDF) [4]. Therein we used a LASSO-based scheme to shrink the feature vector to prevent overfitting on a limited set of data. Building upon an improved database of lead-free perovskites, we are currently exploring neural-network based Autoencoders [5] to improve on the current feature-selection process, allowing us to create a good model with a limited amount of datapoints, and eventually extend to the complete compositional space.

We use our own and third party cookies for analysing and measuring usage of our website to improve our services. If you continue browsing, we consider accepting its use. You can check our Cookies Policy in which you will also find how to configure your web browser for the use of cookies. More info