Predicting properties from a material’s composition or structure is of great interest for materials design. Deep learning has recently garnered considerable interest in materials predictive tasks with low model errors when dealing with large materials data. However, deep learning models suffer in the small data regime that is common in materials science. Here we develop the AtomSets framework, which utilizes universal compositional and structural descriptors extracted from pre-trained graph network deep learning models with standard multi-layer perceptrons to achieve consistently high model accuracy for both small compositional data (<400) and large structural data (>130,000). The AtomSets models show lower errors than the graph network models at small data limits and other non-deep-learning models at large data limits. They also transfer better in a simulated materials discovery process where the targeted materials have property values out of the training data limits. The models require minimal domain knowledge inputs and are free from feature engineering. The presented AtomSets model framework can potentially accelerate machine learning-assisted materials design and discovery with less data restriction.
Structure–property maps play a key role in accelerated materials discovery. The current norm for developing these maps includes computationally expensive physics‐based simulations. Here, the capabilities of deep learning agents are explored such as convolutional neural networks (CNNs) and multilayer perceptrons (MLPs) to predict structure–property relations and reduce dependence on simulations. This study contains simulated hexagonal boron nitride (h‐BN) microstructures damaged by various levels of radiation and temperature, with the objective to predict their residual strengths from the final atomic positions. By developing low dimensional physical descriptors to statistically describe the defects, these results show that purpose‐specific microstructure representation can help in achieving a good prediction accuracy at low computational cost. Furthermore, the adaptability of the trained deep learning agents is explored to predict structure–property maps of other 2D materials using transfer learning. It is shown that in order to achieve good predictions accuracy (≈95%
- NSF-PAR ID:
- 10461173
- Publisher / Repository:
- Wiley Blackwell (John Wiley & Sons)
- Date Published:
- Journal Name:
- Small
- Volume:
- 15
- Issue:
- 19
- ISSN:
- 1613-6810
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
Abstract -
more » « less
Abstract It is well-known that the atomic-scale and nano-scale configuration of dopants can play a crucial role in determining the electronic properties of materials. However, predicting such effects is challenging due to the large range of atomic configurations that are possible. Here, we present a case study of how deep learning algorithms can enable bandgap prediction in hybridized boron–nitrogen graphene with arbitrary supercell configurations. A material descriptor that enables correlation of structure and bandgap was developed for convolutional neural networks. Bandgaps calculated by ab initio calculations, and corresponding structures, were used as training datasets. The trained networks were then used to predict bandgaps of systems with various configurations. For 4 × 4 and 5 × 5 supercells they accurately predict bandgaps, with a
R 2of >90% and root-mean-square error of ~0.1 eV. The transfer learning was performed by leveraging data generated from small supercells to improve the prediction accuracy for 6 × 6 supercells. This work will pave a route to future investigation of configurationally hybridized graphene and other 2D materials. Moreover, given the ubiquitous existence of configurations in materials, this work may stimulate interest in applying deep learning algorithms for the configurational design of materials across different length scales. -
more » « less
Abstract Understanding and controlling the self-assembly of vertically oriented carbon nanotube (CNT) forests is essential for realizing their potential in myriad applications. The governing process–structure–property mechanisms are poorly understood, and the processing parameter space is far too vast to exhaustively explore experimentally. We overcome these limitations by using a physics-based simulation as a high-throughput virtual laboratory and image-based machine learning to relate CNT forest synthesis attributes to their mechanical performance. Using CNTNet, our image-based deep learning classifier module trained with synthetic imagery, combinations of CNT diameter, density, and population growth rate classes were labeled with an accuracy of >91%. The CNTNet regression module predicted CNT forest stiffness and buckling load properties with a lower root-mean-square error than that of a regression predictor based on CNT physical parameters. These results demonstrate that image-based machine learning trained using only simulated imagery can distinguish subtle CNT forest morphological features to predict physical material properties with high accuracy. CNTNet paves the way to incorporate scanning electron microscope imagery for high-throughput material discovery.
-
more » « less
Abstract Oxidation states (OS) are the charges on atoms due to electrons gained or lost upon applying an ionic approximation to their bonds. As a fundamental property, OS has been widely used in charge‐neutrality verification, crystal structure determination, and reaction estimation. Currently, only heuristic rules exist for guessing the oxidation states of a given compound with many exceptions. Recent work has developed machine learning models based on heuristic structural features for predicting the oxidation states of metal ions. However, composition‐based oxidation state prediction still remains elusive so far, which has significant implications for the discovery of new materials for which the structures have not been determined. This work proposes a novel deep learning‐based BERT transformer language model BERTOS for predicting the oxidation states for all elements of inorganic compounds given only their chemical composition. This model achieves 96.82% accuracy for all‐element oxidation states prediction benchmarked on the cleaned ICSD dataset and achieves 97.61% accuracy for oxide materials. It is also demonstrated how it can be used to conduct large‐scale screening of hypothetical material compositions for materials discovery.
-
more » « less
Abstract Materials discovery from the infinite earth repository is a major bottleneck for revolutionary technological progress. This labor‐intensive and time‐consuming process hinders the discovery of new materials. Although machine learning techniques show an excellent capability for speeding up materials discovery, obtaining effective material feature representations is still challenging, and making a precise prediction of the material properties is still tricky. This work focuses on developing an automatic material design and discovery framework enabled by data‐driven artificial intelligence (AI) models. Multiple types of material descriptors are first developed to promote the representation and encoding of the materials’ uniqueness, resulting in improved performance for different molecular properties predictions. The material's thermoelectric (TE) properties prediction is then utilized as a baseline to demonstrate the investigation logistic. The proposed framework achieves more than 90% accuracy for predicting materials' TE properties. Furthermore, the developed AI models identify 6 promising p‐type TE materials and 8 promising n‐type TE materials. The prediction results are evaluated by density functional theory calculations and agree with the material's TE property provided by experimental results. The proposed framework is expected to accelerate the design and discovery of the new functional materials.
