Research

Dral’s group works on the development and application of AI-enhanced computational chemistry methods.

The goals of this research is to:

  • develop methods breaking through the limitations of the traditional quantum mechanical and dynamics methods
  • provide tools such as software and hardware platforms
  • educate and promote

Below we describe how we achieve these goals.

Method development

Our group develops AI methods to accelerate and improve the accuracy of quantum mechanical and dynamics methods.

Our AIQM1 method approaches beyond-DFT accuracy for organic compounds with only a fraction of the DFT computational cost and is based on the concept of Δ-learning introduced by the group’s PI and his collaborators.

We also develop accurate machine learning potentials that can be used as surrogate models instead of costly quantum mechanical methods.

In addition, we have developed methods for speeding up simulations of various types of spectra including precise UV/vis absorption spectra and two-photon absorption cross sections as well as infrared spectra.

While we provide implementations for traditional molecular dynamics simulations with both quantum mechanical and machine learning methods, we also develop unique approaches offering a radically new way of predicting molecular dynamics and quantum dissipative dynamics directly as a function of time (without doing stepwise propagation).

Software and hardware tools

Almost all of our research is done with the MLatom program and Python library for AI-enhanced computational chemistry. It is an open-source program under a permissive MIT license. The program allows to use of both quantum mechanical and machine learning methods and their combinations for performing a wide range of typical and unique simulation tasks ranging from geometry optimizations to dynamics and spectra simulations. MLatom implements the majority of the methods developed in our group. See the introductory video:

Education and promotion

We are also committed to the education and promotion of AI-enhanced computational chemistry. The group’s PI is an Editor of the book about this topic and many group members are its authors. In addition, we organize regular workshops and prepare tutorials for AI-enhanced simulations.

In addition, we provide the XACS cloud computing platform for performing AI-enhanced simulations in the web browser.