The process of producing new goods for a wide range of sectors can be reduced and accelerated if the qualities of materials can be correctly predicted.
Researchers from The University of Tokyo Institute of Industrial Science employed core-loss spectroscopy and machine learning to assess the characteristics of organic compounds in a work published in Advanced Intelligent Systems.
Artificial intelligence (AI) is a multidisciplinary discipline with a variety of methodologies, but advances in machine learning and deep learning are causing a paradigm shift in nearly every sector of the tech industry. It is a broad topic of computer science focused with creating intelligent machines that can accomplish activities that would normally require human intelligence.
Energy loss near-edge structure (ELNES) and X-ray near-edge structure (XANES) spectroscopic techniques are used to determine information about electrons, and thus atoms, in materials. They have been used to explore a variety of materials, from electrical devices to medication delivery systems, and have great sensitivity and resolution.
However, the relationship between spectral data and material qualities such as optical properties, electron conductivity, density, and stability is unclear. To extract information from enormous complicated amounts of data, machine learning (ML) algorithms have been deployed. Artificial neural networks, which are based on how our brains work, are used in these approaches to constantly learn how to solve issues.
Deep learning is a sort of machine learning that processes data through a neural network design inspired by biology. The data is processed through a number of hidden layers in the neural networks, which allows the machine to go “deep” in its learning, creating connections and weighing input for the best outcomes.
Although the group has previously employed ELNES/XANES spectra and ML to discover information about materials, what they discovered had nothing to do with the material’s characteristics. As a result, the information was difficult to interpret into advancements.
This descriptor is something that can be directly measured in experiments and can therefore be determined with high sensitivity and resolution. This method is highly beneficial for materials development because it has the potential to reveal where, when, and how certain material properties arise.Kakeru Kikumasa
The researchers have now employed machine learning to uncover information concealed in simulated ELNES/XANES spectra of 22,155 organic compounds. “The system was then fed the ELNES/XANES spectra of the molecules, or their ‘descriptors’ in this case,” says lead author Kakeru Kikumasa.
“This descriptor is something that can be directly measured in experiments and can therefore be determined with high sensitivity and resolution. This method is highly beneficial for materials development because it has the potential to reveal where, when, and how certain material properties arise.”
A model built entirely from spectra proved successful in predicting what are known as intense features. It was, however, unable to anticipate broad features that are influenced by molecule size.
To improve the prediction, the new model was built by including the ratios of three elements in relation to carbon (which is present in all organic molecules) as added factors, allowing for the accurate prediction of extensive features such as molecular weight.
“Our ML learning treatment of core-loss spectra provides an accurate prediction of extensive material properties, such as internal energy and molecular weight. The link between core-loss spectra and extensive properties has previously never been made; however, artificial intelligence was able to unveil the hidden connections. Our approach might also be applied to predict the properties of new materials and functions,” says senior author Teruyasu Mizoguchi.
“We believe that our model will be a very useful tool for the high-throughput development of materials in a wide range of industries.”
The study, “Quantification of the Properties of Organic Molecules Using Core-Loss Spectra as Neural Network Descriptors,” was published in Advanced Intelligent Systems.