MIEM HSE: Machine Learning Method for Quantum Process Modeling

Main article: Quantum computers and networks in Russia

2025: Introduction of a machine learning method to model quantum processes

HSE scientists, together with colleagues from the University of Southern California, have developed an algorithm that quickly and accurately predicts the behavior of quantum systems - from quantum computers to solar panels. The HSE announced this on February 25, 2025. With its help, it was possible to simulate the processes in the MoS₂ semiconductor and find out that the movement of charged particles is influenced not only by the number of defects, but also by their location. These defects can slow or accelerate charge transfer, creating effects that used to be difficult to account for when applying standard methods.

Modern electronics work thanks to quantum effects. Semiconductors, LEDs, solar panels - all these devices depend on how electrons behave in materials. Describing such processes with high accuracy is difficult: modeling requires enormous computing power. To calculate the motion of electrons in a material of a thousand atoms, supercomputers have to perform millions of operations.

Usually, the method of molecular dynamics is used in modeling quantum systems: it allows you to predict how atoms and electrons will move over time. However, if electron states change rapidly, standard modeling methods become too resource-intensive.

Researchers at the Moscow State Institute of Electronics and Mathematics (MIEM) of the Higher School of Economics solved the problem using machine learning. This algorithm analyzes small fragments of material, learning from their local properties, and then builds predictions about the behavior of the entire system. Scientists have studied the two-dimensional semiconductor molybdenum sulfide (MoS₂), a promising material for optoelectronics and photovoltaics. In particular, it can serve as a working layer of solar cells. Ideally, molybdenum (Mo) and sulfur (S) atoms form an ordered lattice, but in real materials the structure is rarely ideal: defects may be present in it.

Defects are irregularities in the arrangement of atoms. In MoS₂, they can manifest as vacancies (lack of sulfur or molybdenum atoms), extra atoms between layers, local displacements, or other deviations from the ideal lattice. Defects change the behavior of electrons: in some cases, they impair conductivity, but sometimes they can give the material new properties, for example, increase its sensitivity to light or make it the best charge conductor.

To understand how defects affect electron motion, we focused on small fragments of material. The algorithm first studied the local properties of the system, and then predicted the behavior of the entire structure. It's like learning a language: first you remember individual words, and then you begin to understand whole sentences, "said Liu Dongyu, associate professor at the Higher School of Economics.

It turned out that it is important not only the number of defects, but also their location. Defects can delay or accelerate the movement of charged particles, creating traps for charge carriers within the semiconductor band gap. Standard methods do not cope well with the calculation of these effects, since the calculations must take into account the interaction of defects with each other and with the atoms of the material, which is difficult to do when using small computational cells. Machine learning allows you to overcome these dimensional limitations and take into account the synergistic effect of multiple defects in the material.

It is important that this method not only speeds up calculations, but also helps to study real quantum systems. The results of our research will be able to narrow the gap between theoretical modeling and experimental research on materials. We have developed a new approach to studying charge motion in complex systems, combining precise calculations, molecular dynamics, and machine learning. This method will help to investigate materials in which electrons transfer energy and information, which is important for electronics and power, - said Andrei Vasenko, professor at the Higher School of Economics.