Organic semiconductors are exotic, carbon-based optoelectronic materials that are utilized in a broad spectrum of optoelectronic applications, stretching from organic light-emitting diodes to perovskite solar cells. They exhibit unique solid-state physical properties owing to the soft, van der Waals bonding between individual molecules. The presence of strong coupling between the electronic and structural dynamics gives rise to unique and fascinating phenomena in these molecular solids, such as a transient localization of the electronic states, but is also a performance-limiting factor for their charge transport and optoelectronic properties, such as carrier mobilities or excited-state lifetimes. Using a multiscale computational workflow, we aim to explore bulk charge transport parameters in organic semiconductors and procure molecule-specific quantitative predictions which would empower the prescreening of a diverse set of materials for the design of highly efficient optoelectronic devices.

Organic light-emitting diodes (OLEDs) are at the forefront of display and lighting innovation, yet the development of efficient and stable blue emitters remains a key challenge. Our research centers on the design of advanced blue emitters for OLEDs, with a special focus on thermally activated delayed fluorescence (TADF) materials. By leveraging the TADF phenomenon—which enables the harvesting of both singlet and triplet excitons for near-unity internal quantum efficiency—we investigate cutting-edge classes such as multi-resonance TADF (MR-TADF), inverted singlet–triplet gap (INVEST) emitters, and hybrid long-range/short-range charge transfer (LR/SR-CT) compounds. Using multiscale simulations, we aim to optimize these materials for superior efficiency, stability, and color purity, accelerating the discovery of next-generation blue OLED technologies.

Harnessing the power of machine learning and inverse design, our research accelerates the discovery of high-performance materials for optoelectronic applications, including thermally activated delayed fluorescence (TADF) emitters and organic solar cells. We enable highly accurate prediction and targeted optimization of both ground- and excited-state molecular properties by integrating graph neural networks, data-driven modeling, and physics-informed machine learning frameworks. Our approach significantly streamlines virtual screening, enhances the inverse design of molecules with tailored optoelectronic characteristics, and facilitates the optimization of critical device parameters such as efficiency, spectral quality, and operational stability. By combining quantum mechanical descriptors with interpretable, robust models, we rapidly navigate vast chemical spaces and extract meaningful structure–property relationships—ultimately driving the intelligent design of next-generation OLED and photovoltaic materials.

The accuracy of computational predictions is often governed by the employed atomistic model representing the underlying atomic interactions in the target compound. However, in many instances, the bottleneck has been the availability of sufficiently efficient interatomic potentials providing reliable energies and forces. Electronic structure methods can often accurately predict atomic interactions but are too computationally demanding to be applicable for large-scale systems. In contrast, empirical force fields are much faster to evaluate and can be used for large systems, but they often lack the accuracy needed for truly predictive simulations. The omnipresent issue of compromise between accuracy and speed can be solved by interatomic potentials developed by employing machine learning techniques. These potentials aim to provide the accuracy of electronic structure calculations at approximately the computational cost of force fields. Our objective is to build such machine-learning potential to perform material and reaction simulations which typically deal with slow chemical bond evolution in complex environments.

As global concerns about fossil-fuel depletion and greenhouse gas emissions continue to escalate, the demand for alternative energy from renewable and clean sources is rapidly increasing. The development of fuel cells and batteries may provide practical clean energy alternatives that can replace internal combustion engines in automobiles and also power personal electronics. However, despite decades of research, electrochemical energy conversion and storage technologies continue to lag behind fossil fuels in performance and cost. Fundamental problems that hinder the advancement in this area are primarily the lack of understanding of transport and catalytic mechanisms as well as the complexity of modeling chemical processes and dynamics at the interfaces between each component of electrochemical devices. Our goal is to develop computational tools based on multiscale theory to investigate and predict the behavior of complex systems that are relevant to renewable energy technologies. A key area of research is the study of proton and ion conduction mechanisms in electrolytes for batteries and fuel cells.

Our research investigates cutting-edge photocatalytic materials—including gold nanoparticles, metal–organic and covalent organic frameworks (MOFs/COFs), and halide perovskites—for sustainable energy conversion and environmental remediation. We focus on enhancing light-driven processes such as hydrogen evolution, carbon dioxide reduction, and pollutant degradation by engineering charge separation and transport at the nanoscale. Key efforts include tuning ligand–surface interactions, optimizing nanostructures for improved stability and reactivity, and understanding structure–function relationships through a combined experimental–theoretical approach. We integrate electronic structure modeling and surface interaction simulations to guide material design and interpret photocatalytic mechanisms. This computational insight complements synthetic strategies and accelerates the development of rationally engineered systems.