An official website of the United States government
This paper presents a new approach for predicting thermodynamic properties of perovskites that harnesses deep learning and crystal structure fingerprinting based on Hirshfeld surface analysis. It is demonstrated that convolutional neural network methods capture critical features embedded in two-dimensional Hirshfeld surface fingerprints that enable a quantitative assessment of the formation energy of perovskites. Building on our recent work on lattice parameter prediction from Hirshfeld surface calculations, we show how transfer learning can be used to speed up the training of the neural network, allowing multiple properties to be trained using the same feature extraction layers. We also predict formation energies for various perovskite polymorphs, and our predictions are found to give generally improved performance over a well-established graph network method, but with the methods better suited to different types of datasets. Analysis of the structure types within the dataset reveals the Hirshfeld surface-based method to excel for the less symmetric and similar structures, while the graph network performs better for very symmetric and similar structures.
more »
« less
Machine-learning structural and electronic properties of metal halide perovskites using a hierarchical convolutional neural network
Abstract The development of statistical tools based on machine learning (ML) and deep networks is actively sought for materials design problems. While structure-property relationships can be accurately determined using quantum mechanical methods, these first-principles calculations are computationally demanding, limiting their use in screening a large set of candidate structures. Herein, we use convolutional neural networks to develop a predictive model for the electronic properties of metal halide perovskites (MHPs) that have a billions-range materials design space. We show that a well-designed hierarchical ML approach has a higher fidelity in predicting properties of the MHPs compared to straight-forward methods. In this architecture, each neural network element has a designated role in the estimation process from predicting complex features of the perovskites such as lattice constant and octahedral till angle to narrowing down possible ranges for the values of interest. Using the hierarchical ML scheme, the obtained root-mean-square errors for the lattice constants, octahedral angle and bandgap for the MHPs are 0.01 Å, 5°, and 0.02 eV, respectively. Our study underscores the importance of a careful network design and a hierarchical approach to alleviate issues associated with imbalanced dataset distributions, which is invariably common in materials datasets.
more »
« less
- Award ID(s):
- 1809085
- PAR ID:
- 10418331
- Date Published:
- Journal Name:
- npj Computational Materials
- Volume:
- 6
- Issue:
- 1
- ISSN:
- 2057-3960
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Third-generation photovoltaic materials, including metal halide perovskites (MHPs), colloidal quantum dots (QDs), copper zinc tin sulfide (CZTS), and organic semiconductors, among others, have become attractive in the past two decades. Unlike their first- and second-generation counterparts, these advanced materials boast properties beyond mere photovoltaic performance, such as mechanical flexibility, light weight, and cost-effectiveness. Meanwhile, these materials possess more intricate crystalline structures that aid in understanding and predicting their transport properties. In particular, the distinctive phonon dispersions in MHPs, the layered architecture in quasi-two-dimensional (2D) perovskites, the strong quantum confinement in QDs, and the complex crystal structures interspersed with abundant disorders in quaternary CZTS result in unique and sometimes anomalous thermal transport behaviors. Concurrently, the criticality of thermal management in applications such as photovoltaics, thermoelectrics, light emitting diodes, and photodetection devices has received increased recognition, considering that many of these third-generation photovoltaic materials are not good thermal conductors. Effective thermal management necessitates precise measurement, advanced modeling, and a profound understanding and interpretation of thermal transport properties in these novel materials. In this review, we provide a comprehensive summary of various techniques for measuring thermal transport properties of these materials and discuss the ultralow thermal conductivities of three-dimensional (3D) MHPs, superlattice-like thermal transport in 2D perovskites, and novel thermal transport characteristics inherent in QDs and CZTS. By collecting and comparing the literature-reported results, we offer a thorough discussion on the thermal transport phenomenon in these materials. The collective understanding from the literature in this area, as reviewed in this article, can provide guidance for improving thermal management across a wide spectrum of applications extending beyond photovoltaics.more » « less
-
Hybrid organic–inorganic perovskites (HOIPs) have emerged as a promising class of materials for optoelectronic and spintronic applications. Layered two-dimensional (2D) HOIP variants have received considerable attention, primarily due to their unique properties that can be modulated through the tailored selection of both organic and inorganic components. The spin splitting in the band structure due to the strong spin–orbit coupling is one of the most intriguing properties of such 2D HOIPs materials for their potential utility in spintronics. In addition to observing the spin splitting in equilibrium due to the non-centrosymmetric structure, the possibility of having dynamic spin splitting at room temperature of the otherwise centrosymmetric systems has become a topic of great debate. While modern first-principles molecular dynamics (FPMD) simulation is able to address such a question in principle by taking into account the lattice anharmonicity in electronic structure calculation, the finite-size error poses a great challenge in practice. In this work, we employ a machine learning (ML) model to overcome this practical limitation to investigate the dynamic spin splitting in phenylethyl ammonium lead iodide 2D HOIP. Specifically, we use the deep potential molecular dynamics approach [Zeng et al., J. Chem. Phys. 159(5), 054801 (2023)] for ML FPMD simulation, and we also develop a surrogate model for predicting the spin splitting based on the recent finding that relates the spin splitting to structural descriptors in 2D HOIPs. Our work shows that even in globally centrosymmetric structures, the inclusion of lattice anharmonicity can induce dynamic spin splitting at room temperature.more » « less
-
Abstract Hybrid metal‐halide perovskites (MHPs) have shown remarkable optoelectronic properties as well as facile and cost‐effective processability. With the success of MHP solar cells and light‐emitting diodes, MHPs have also exhibited great potential as gain media for on‐chip lasers. However, to date, stable operation of optically pumped MHP lasers and electrically driven MHP lasers—an essential requirement for MHP laser's insertion into chip‐scale photonic integrated circuits—is not yet demonstrated. The main obstacles include the instability of MHPs in the atmosphere, rudimentary MHP laser cavity patterning methods, and insufficient understanding of emission mechanisms in MHP materials and cavities. This review aims to provide a detailed overview of different strategies to improve the intrinsic properties of MHPs in the atmosphere and to establish an optimal MHP cavity patterning method. In addition, this review discusses different emission mechanisms in MHP materials and cavities and how to distinguish them.more » « less
-
Abstract Modern machine learning (ML) and deep learning (DL) techniques using high-dimensional data representations have helped accelerate the materials discovery process by efficiently detecting hidden patterns in existing datasets and linking input representations to output properties for a better understanding of the scientific phenomenon. While a deep neural network comprised of fully connected layers has been widely used for materials property prediction, simply creating a deeper model with a large number of layers often faces with vanishing gradient problem, causing a degradation in the performance, thereby limiting usage. In this paper, we study and propose architectural principles to address the question of improving the performance of model training and inference under fixed parametric constraints. Here, we present a general deep-learning framework based on branched residual learning (BRNet) with fully connected layers that can work with any numerical vector-based representation as input to build accurate models to predict materials properties. We perform model training for materials properties using numerical vectors representing different composition-based attributes of the respective materials and compare the performance of the proposed models against traditional ML and existing DL architectures. We find that the proposed models are significantly more accurate than the ML/DL models for all data sizes by using different composition-based attributes as input. Further, branched learning requires fewer parameters and results in faster model training due to better convergence during the training phase than existing neural networks, thereby efficiently building accurate models for predicting materials properties.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1038/s41524-020-0307-8
