Although the computational discovery of new materials can simplify the screening process before experimental synthesis, due to the huge potential combination of material components and structures, this screening process is still difficult and long, and it is also challenging to systematically explore the material composition and structural space.
fig. 1 performance ()of prediction of various test datasets using the ce feature models with different center atom definitions.
It is more difficult if the data of the target material is limited, and the migration machine Xi across crystal structures from large data sets known to other materials has become one of the important strategies in material design.
fig. 2 formation energies predicted by the ml with ce features (dnn-ce) and dft using various datasets.
The team of Prof. Liu Yi and Prof. Feng Lingyan from the Institute of Materials Genome Engineering of Shanghai University proposed a deep migration Xi method based on a large-scale spinel oxide computational dataset for thermodynamically stable perovskite oxides.
fig. 3 heat map of formation energies of 5329 abo3 perovskite oxide structures predicted by the transferred learning model in this work, containing 73 constitution elements at the a and b sites, respectively, sorted by the atom number.
Using the calculated formation energies of 5329 spinel oxide structures, they developed a deep neuronal network (DNN) source domain model with "central environment (CE)" characteristics, and then fine-tuned the DNN model parameters by Xi a small dataset of 855 perovskite oxide structures, and obtained a migration Xi model with good mobility in the perovskite oxides in the target domain.
fig. 4 heat map of tolerance factor of 5329 perovskite oxide structures calculated in this work, containing 73 constitution elements at the a and b sites, sorted by the atom number.
The average absolute error (MAE) of the formation energy of the perovskite structure by the migration Xi model** is only 0106 EV atom, which is better than 0132 ev/atom。Based on the migration Xi model, the authors quickly estimated the formation energies of 5329 potential perovskite structures containing 73 elements.
fig. 5 tolerance factor (t) vs. octahedral factor (μscatter plot ofperovskite oxide structures, where the colormap corresponds tothe transfer learning predicted formation energy of perovskitestructure.
The formation energy of the binding ** and the inclusion of the tolerance factor (0.).7 < t ≤ 1.1) and octahedral factor (045 < 0.7) they identified 1314 potentially thermodynamically stable perovskite oxides. Of the 1314 potential perovskite oxides, 144 have been synthetically confirmed experimentally, 10 have been identified by other calculations, 301 have been recorded in the Materials Project database, and the remaining 859 oxides have not yet been reported in the literature.
fig. 6 statistical distribution of the formation energy of perovskite structures predicted by machine learning and the screening process for stable perovskite structures.
This study combines machine Xi characterization based on structural information and migration Xi methods, and uses abundant known structural data to create new structures at a low additional computational cost, providing a new and effective acceleration strategy for expensive high-throughput computational screening material designs.
fig. 7 crystal structures and constituent elements of spinel oxides and perovskite oxides studied in this work.
*The novel perovskite oxides provide a wealth of candidates for the experimental synthesis and exploration of renewable energy and electronic materials applications. Related**Recently published on NPJ Computational Materials106 (2023)。To read the original text on your mobile phone, please click "Read the original text" in the lower left corner of the bottom of this article, and you can also ** full text pdf file after entering.
fig. 8 general schematic diagram of dnn-ce models and the workflow of transfer learning method in this work.
editorial summary
transfer learning across crystal structures: “center-environment” feature accelerates materials predicting
discovering new materials through computational methods has simplified the screening process before experimental synthesis. however, systematically exploring the material space remains challenging due to the vast potential combinations of material compositions and structures. in cases where data on the target material is limited, cross-crystal structure transfer learning from large-scale known datasets of other materials has become an important strategy in materials design.
this study proposes a deep transfer learning approach based on a large-scale dataset of spinel oxide compounds to predict the thermodynamically stable perovskite oxides. prof. liu and prof. feng’s team at materials genome institute of shanghai university utilized the formation energy of 5,329 spinel oxide structures to develop a deep neural network (dnn) source domain model with “center-environment” (ce) features. the ce-dnn model was then fine-tuned using a small dataset of 855 perovskite oxide structures to achieve a transferable learning model with good performance in the perovskite oxide target domain.
the mean absolute error (mae) of the perovskite structure formation energy predicted by the transfer learning model is 0.106 ev/atom, which is better than the mae of 0.132 ev/atom of the model trained solely using small perovskite data. based on the transfer learning model, the formation energy of 5,329 potential perovskite structures containing 73 different elements was further predicted. combining the predicted formation energies with structural factor criteria, including tolerance factor (0.7 < t ≤ 1.1) and octahedral factor (0.45 < 0.7), a total of 1,314 potentially thermodynamically stable perovskite oxides were predicted.
among 1,314 predicted potential perovskite oxides, 144 were experimentally synthesized, 10 were predicted by other computational works, and 301 are documented in the materials project database, while the remaining 859 oxides h**e not been reported in literatures. the combination of structural information features and transfer learning methods in this study enables the low-cost prediction of new structures using existing big data, providing an effective acceleration strategy for expensive high-throughput computational material design. the predicted stable novel perovskite oxides offer a rich platform for exploring novel perovskite experimental synthesis, renewable energy and electronic materials applications.this article was recently published in npj computational materials
Original text abstract and its translation
Center-Environment Deep Transfer Machine Learning across Crystal Structures: from Spinel Oxides to Perovskite Oxide Xi s
yihang li, ruijie zhu, yuanqing wang, lingyan feng & yi liu
abstractin data-driven materials design where the target materials h**e limited data, the transfer machine learning from large known source materials, becomes a demanding strategy especially across different crystal structures. in this work, we proposed a deep transfer learning approach to predict thermodynamically stable perovskite oxides based on a large computational dataset of spinel oxides. the deep neural network (dnn) source domain model with “center-environment” (ce) features was first developed using the formation energy of 5329 spinel oxide structures and then was fine-tuned by learning a small dataset of 855 perovskite oxide structures, leading to a transfer learning model with good transferability in the target domain of perovskite oxides. based on the transferred model, we further predicted the formation energy of potential 5329 perovskite structures with combination of 73 elements. combining the criteria of formation energy and structure factors including tolerance factor (0.7 < t ≤ 1.1) and octahedron factor (0.45 < 0.7), we predicted 1314 thermodynamically stable perovskite oxides, among which 144 oxides were reported to be synthesized experimentally, 10 oxides were predicted computationally by other literatures, 301 oxides were recorded in the materials project database, and 859 oxides h**e been first reported. combing with the structure-informed features the transfer machine learning approach in this work takes the advantage of existing data to predict new structures at a lower cost, providing an effective acceleration strategy for the expensive high-throughput computational screening in materials design. the predicted stable novel perovskite oxides serve as a rich platform for exploring potential renewable energy and electronic materials applications.
Summary:
In data-driven material design, when the data of the target material is limited, the transfer machine Xi based on a large number of known source material data, especially across different crystal structures, has become a research strategy with practical needs and application scenarios. This paper proposes a deep migration Xi method, based on the known large-scale spinel oxide calculation data and the newly added small amount of perovskite oxide calculation data, and the new thermodynamically stable new perovskite oxide across crystal structures.
Firstly, a deep neural network (DNN) source domain model based on the "central environment" (CE) features containing structural information was developed by using the calculated formation energy of 5329 spinel oxide structures, and then the DNN model parameters were fine-tuned by Xi learning a small dataset of 855 perovskite oxide structures to obtain a migration Xi model for perovskite oxides in the target domain with good mobility. Based on the CE-DNN migration Xi model, we further estimated the formation energies of 5329 perovskite structures containing 73 element combinations. Binding to the formation energy of ** and structural factors (including tolerance factor (0.).7 < t ≤ 1.1) and octahedral factor (045 < 0.7)), we identified a total of 1314 potentially thermodynamically stable perovskite oxides, of which 144 oxides have been experimentally synthesized, 10 oxides have been recorded in other computational literature, 301 oxides are recorded in the materials project database, and the other 859 oxides are reported for the first time in this paper.
Based on feature engineering containing structural information, the migration machine Xi method in this study leverages the abundant data available to provide an effective acceleration strategy for expensive high-throughput computational screening of new crystal structure properties at a low additional computational cost. The novel perovskite oxides in this work provide a rich platform for material candidates and improvements for exploring renewable energy and electronic materials applications.