The goal of the project is to build a machine learning solution to extract information about molecular structure with sub-femtosecond resolution from high-order harmonic emission spectra and electron diffraction spectra, thereby reconstructing the molecular dynamics processes in real time. In this project we study common but poorly understood biochemical processes such as dissociation and tautomer transfer between DNA bases. It should be emphasized that current research on ultrafast molecular dynamics reconstruction is mainly in a proof-of-principle state and has not yet been applied to complex molecules. The biggest difficulty is the huge amount of data that traditional optimization algorithms cannot handle. Therefore, we apply machine learning to build a solution to reproduce the dynamics of these biochemical processes.
Main tasks of the project
- Simulation of HHG/LIED data of complex molecules, taking into account macroscopic effects such as lateral distribution. This data set is simulated for each position on the chemical reaction path and is considered “experimental” data used to validate the montage program. When used for a specific case, the input data is the HHG/LIED spectrum obtained from experiments, when the molecular structure dynamic parameters are not known.
- Use of machine learning algorithm built in problem III of the previous content to extract single-molecule data (ideal orientation) from experimental HHG/LIED spectra (imperfectly oriented molecules).
- From here, use the solution from problem II of the previous content to reconstruct the structural information from the HHG/LIED of the single molecule. These data are obtained over measurement time. The time on the chemical reaction path is the delay between the laser pulse that excites the molecule into an excited state and the laser pulse that produces ionization and leads to HHG/LIED. The result is the instantaneous molecular configuration during dissociation or isomerization.
- From the above set of figures, using the program to create a movie is a dynamic process. HHG/LIED measurements have a time resolution of several femtoseconds or shorter depending on the experimental laser parameters.
Project impact
- This research is a bridge between theoretical research and application, specifically, providing solutions and software for experimentalists to create molecular videos depicting the process of dissociation and isomerism of complex molecules with high spatial and temporal resolution.
- With foresight, the results of the project contribute to building the foundation for the manufacture of new generation four-dimensional microscopes, serving research in science, especially in physics, chemistry and biology.
- The project helps train young scientists and access current research issues around the world. The project involves doctoral, master students and undergraduate students.
- This project will also help our team develop relationships with experimental research groups in advanced universities, learning new techniques and technologies.
