Organic chemistry, the study of molecules made of carbon, not only supports the science of living things, but is important for many current and future technologies, such as organic diode displays light-emitting diode (OLED), Machine learning allows researchers to see beyond the spectrum.
Understanding the electronic structure of molecules is important to predict the chemistry of the substance.
In a recently published study by researchers from the University of Tokyo‘s Institute of Industrial Sciences, a machine learning algorithm was developed to predict the density of states in organic molecules, namely the number the energy of electrons can penetrate into the ground, say infractions.
These predictions, based on spectral data, can be of great help to organic chemists and material scientists when analyzing carbon-containing molecules.
This experimental procedure is often used to determine the density of states that can be difficult to define. This is especially true for the so-called key loss spectroscopy method, which combines energy loss proximity spectroscopy (ELNES) and X-ray absorption proximity structure (XANES).
These methods cause electron beams or X-rays to penetrate a sample of material; the resulting scattering of electrons and the measurement of the energy generated by the molecules make it possible to measure the density of states of the molecule of interest.
However, the information in the spectrum relates only to the unoccupied (unoccupied) electron states of the excited molecules.
To solve this problem, a team from the University of Tokyo’s Institute of Industrial Sciences trained neural network machine learning techniques to analyze mass spectroscopy data and predict the density of electronic states.
First, a database was built by calculating the density of states and the model fit for more than 22,000 molecules. They also added simulated noise. Next, an algorithm is trained on the missing point bias and improved to predict the correct state density of occupied and unoccupied states in the ground state.
“We tried to make predictions for larger particles using samples of smaller particles. “We found that the accuracy can be improved by removing small particles,” says lead author Po-Yen Chen.
The team also found that by using smoothing before processing and adding specific noise to the data, the state density prediction can be improved, which can increase the acceptance of the prediction for use with real data.
“Our work can help researchers understand the properties of molecules and accelerate the design of functional molecules,” says lead author Teruyasu Mizoguchi. This can include drugs and other stimulant compounds.
Source: University of Tokyo
