Researchers at the University of Tokyo’s Institute of Industrial Science are using artificial intelligence to help interpret data generated in materials science spectroscopy experiments to help develop new drugs and organic conductors.Provided by: Institute of Industrial Science, University of Tokyo
Scientists have developed a machine learning algorithm that predicts the electronic energy levels of organic molecules. Trained on a database of over 22,000 molecules, this breakthrough technology has the potential to accelerate the design of functional molecules such as pharmaceuticals.
Organic chemistry, the study of carbon-based molecules, is not only fundamental to the science of life, but is also important to many current and future technologies such as organic light-emitting diode (OLED) displays. Understanding the electronic structure of a material’s molecules is key to predicting the material’s chemical properties.
In a recently published study by researchers at the Institute of Industrial Science, the University of Tokyo, a machine-learning algorithm for predicting the density of states in organic molecules, the number of energy levels an electron can occupy in an organic molecule. was developed. The ground state within the molecule of a substance. These predictions based on spectral data are very useful for organic chemists and materials scientists when analyzing carbon-based molecules.
Commonly used experimental techniques for finding density of states can be difficult to interpret. This is especially true for a method known as core-loss spectroscopy, which combines energy-loss near-edge spectroscopy (ELNES) and X-ray absorption near-edge structure (XANES). These methods irradiate the material sample with an electron beam or his X-rays. By scattering the resulting electrons and measuring the energy emitted by the molecules of the material, the density of states of the molecule of interest can be measured. However, the spectrum only contains information about the electron-free (unoccupied) states of the excited molecule.
To address this issue, a team at the Institute of Industrial Science, the University of Tokyo analyzed core loss spectroscopy data and trained a neural network machine learning model to predict the density of electronic states. First, we calculated the density of states and corresponding core loss spectra for over 22,000 molecules to build a database. Also added simulated noise. The algorithm was then trained on the core-loss spectrum and optimized to predict the exact density of states for both occupied and unoccupied states in the ground state.
“We tried to use models trained on smaller molecules to extrapolate the predictions of larger molecules.[{” attribute=””>accuracy can be improved by excluding tiny molecules,” explains lead author Po-Yen Chen.
The team also found that by using smoothing preprocessing and adding specific noise to the data, the predictions of density of state can be improved, which can accelerate adoption of the prediction model for use on real data.
“Our work can help researchers understand the material properties of molecules and accelerate the design of functional molecules,” senior author Teruyasu Mizoguchi says. This can include pharmaceuticals and other exciting compounds.
Reference: “Prediction of the Ground-State Electronic Structure from Core-Loss Spectra of Organic Molecules by Machine Learning” by Po-Yen Chen, Kiyou Shibata, Katsumi Hagita, Tomohiro Miyata and Teruyasu Mizoguchi, 17 May 2023, The Journal of Physical Chemistry Letters.
DOI:10.1021/acs.jpclett.3c00142
