News & Events

Embracing Machine Learning as an Experimental Chemist: Unravelling the Design Principles in Hybrid Organic-Inorganic Perovskites

Date: 
Wednesday, December 11, 2019 - 15:00 to 16:00
Speaker: 
Dr. Shijing Sun
Affiliation: 
Massachusetts Institute of Technology, Photovoltaics Research Laboratory, Department of Mechanical Engineering
Event Category: 
Seminar - Seminar
Location: 
Chemistry D215

Abstract:

Traditionally chemistry is divided into the organic and inorganic fields. Interdisciplinary research on hybrid organic-inorganic materials has opened new avenues for functional materials design.1 For example, the discovery of hybrid lead-halide perovskites that can achieve > 25% solar-to-electricity efficiency has revolutionized the thin-film solar technology, and attracted tremendous interest from both academia and the energy industry.2 The enormous structural and chemical diversity of hybrid perovskites leads to technical applications beyond their purely organic or inorganic counterparts. However, engineering this complex composition space also posed new challenges to chemists.

In this seminar I will present the latest advances in materials discovery resolving two critical challenges hindering the widespread commercial adoption of state-of-the-art perovskite solar cells: 1) their rapid degradation under environmental stimuli and 2) the toxicity of lead.3 Exemplified by two experimental campaigns, I will explain how data-driven “accelerated search” and “smart search” tools effectively guide materials design, and how synchrotron-based advanced characterization can reveal the fundamental crystal chemistry of the newly discovered materials.4

In the first part of my talk, I will introduce a new closed-loop workflow combining high-performance computing and automated experiment in the same machine-learning framework, which rapidly identifed the most stable compositions achieving over 40× improvement in environmental stability compared to the state-of-the-art methylammonium lead iodide. In the second part of the talk, I will discuss the streamlined synthesis and structural-property relationships of novel lead-free perovskites, where a neural network accelerated the classification of X-ray diffraction patterns by 10×, enhancing our understanding of the composition-dependent bangap bowing behaviour that I discovered in a new semiconductor alloy series.5,6 These findings demonstrate the power of “science + data” integrated approaches for materials discovery, and shed light on the design principles of hybrid organic-inorganic perovskites, pushing next-generation solar cells closer to mass manufacturing in fighting the global energy and environmental crises.

 

1.  Cheetham, A. K. & Rao, C. N. R. There’s room in the middle. Science 318, 58–59 (2007).

2.  National Renewable Energy Laboratory, Best research cell efficiencies. Available at: http://www.nrel.gov/ncpv/images/efficiency_chart.jpg. (Accessed: 5th Dec. 2019)

3.  Correa-Baena, J. P. et al. Promises and challenges of perovskite solar cells. Science 358, 739–744 (2017).

4.  Sun, S. et al. Variable temperature and high-pressure crystal chemistry of perovskite formamidinium lead iodide: a single crystal X-ray diffraction and computational Study. Chem. Commun. 53, 7537–7540 (2017).

5.  Sun, S. et al. Accelerated development of perovskite-Inspired materials via high-throughput synthesis and machine-learning diagnosis. Joule 3, 1437–1451 (2019).

6.  Oviedo, F. et al. Fast and interpretable classification of small X-ray diffraction datasets using data augmentation and deep neural networks. npj Comput. Mater. 5, 60 (2019).