Northwestern University researchers have developed a new computational approach to speed up the design of materials that exhibit metal insulating transition (MIT), a rare type of electronic material. shows the potential to launch design in the future and deliver quantum and microelectronic information systems faster ̵1; the technology behind the Internet of Things devices and large-scale data centers provide power. quality for how people work and interact with others.
The new strategy, a collaboration between Professors James Rondinelli and Wei Chen, integrates techniques from statistical reasoning, optimization theory, and computational material physics. This approach combines multi-purpose optimization of Bayes with potentially variable Gaussian processes to optimize ideal features in a family of MIT materials called complex lacunar spinels.
When researchers look for new materials, they often look where existing data is available on similar materials. The design of multiple material property layers has been accelerated in existing works with data-oriented methods supported by high-throughput data generation along with methods such as machine learning.
Such approaches are not available for MIT materials, however, which are classified according to the reversible transition between the conductive state and the insulating state. Most MIT models are built to describe a single material, making it difficult to create models often. At the same time, conventional machine learning methods have shown limited predictability due to a lack of available data, making the design of new MIT materials difficult.
“Researchers understand how to extract information from material data sets,” said Rondinelli, professor of materials science and engineering and Morris E. Fine Professor of Materials and Manufacturing at the McCormick School of Technology. great where it exists and when the right features are available. corresponding author of the study. “But what do you do when you don’t have the big data set or the features you need? Our work breaks down this status quo by building predictive and exploratory models that don’t require large datasets.” or features starting from a small data set. “
An article describing the work, titled “Adaptive optimization without accelerating functional electronic material design,” was published on November 6 in the journal. Reviews on Applied Physics.
The team’s methodology, called Advanced Optimization Engine (AOE), ignores traditional machine learning-based discovery models using latent variable Gaussian process modeling methodology, Only the chemical composition of materials is required to distinguish their optimal nature. This allows AOE based Bayes optimization to efficiently search for materials with the optimum band gap (resistivity / conductivity) and thermal stability (aggregation capacity) – two Identification feature for useful materials.
To validate their approach, the team analyzed hundreds of chemical combinations using simulations based on density function theory and found 12 previously undetermined components of these The complex lacunar spinel shows optimal function and synthesis. These MIT materials are known to host unique rotating structures, a feature needed to power the future Internet of Things and other resource-intensive technologies.
Chen, Professor Wilson-Cook of Engineering Design and professor and chair of mechanical engineering and co-author of the study, said: “This progress transcends the traditional limitations of Material design is based on imposing chemical intuition. “By rearranging functional material design as an optimization problem, we not only found a solution to the challenge of working with limited data, but also demonstrated the ability to explore effectively. New materials optimized for electronics in the future. “
While the researchers tested their method on inorganic materials, they believe that the method can also be applied to organic materials, such as the design of protein sequences in biomaterials. study or sequence monomers in polymeric materials. The model also provides guidance on making better decisions about the optimal design of the material by selecting the ideal candidate compounds for simulation.
“Our approach paves the way for optimizing multiple properties and co-designing complex multi-functional materials where data and prior knowledge are limited,” says Rondinelli.
This research was created from a project that explores the optimization of Bayes in material discovery in an interdisciplinary cluster program in Engineering Prediction and Design (PSED), sponsored by The Graduate School at Northwestern. . It is supported by a grant from the National Science Foundation and the Advanced Research Projects Agency – Energy Difference Program (ARPA-E), which seeks to use emerging AI technologies to solve address major energy and environmental challenges.
“This work highlights the impact of the PSED collaborative interdisciplinary design cluster,” said Chen. “It also highlights important advances occurring in AI and machine learning at Northwestern in design and optimization.”
Novel materials transition between conductive and insulating states
Yiqun Wang et al., Adaptive Optimization does not accelerate functional electronic material design, Reviews on Applied Physics (Year 2020). DOI: 10.1063 / 5.0018811
Provided by Northwestern University
Quote: New approach to determine optimal material design with minimum data (2020, November 6) retrieved November 7, 2020 from https://phys.org/news/2020- 11-approach-optimal-materials-minimal.html
This material is the subject for the fake rights. Apart from any fair dealings for private study or research purposes, no part may be reproduced without written permission. The content provided is for informational purposes only.