Dr. Nenad Vukmirović from the Institute of Physics Belgrade and his research group are dedicated to understanding charge transport in semiconducting materials - an essential process underlying the performance of all electronic devices. "To improve these devices", he explains, "it is absolutely necessary to perform simulations of these materials". Given the complexity of such simulations, the group relies heavily on high-performance computing (HPC) resources. "We need high-performance computing resources to perform these simulations as fast as possible", Nenad emphasizes.
The group investigates three broad classes of materials, each requiring a tailored simulation strategy. The first are disordered organic materials, in which electronic states are localized and charge moves by hopping between sites. "For these materials, we use methods that are capable of performing electronic structure calculations for systems containing on the order of ten thousand atoms", says Nenad. These large-scale simulations are followed by multi-scale modeling to bridge length scales from a few nanometers to about 100 nanometers, enabling the prediction of macroscopic transport properties such as temperature-dependent mobility.
The second class includes conventional crystalline materials, like silicon. Here, simulations are grounded in density functional theory (DFT), which provides detailed insight into electronic states and lattice vibrations—phonons. "We calculate the coupling between electrons and phonons, which actually limits the charge transport in the material", he explains. These electron-phonon interactions must be computed on dense momentum-space grids, requiring highly specialized computational methods to ensure accuracy.
Finally, the group also explores materials with intermediate properties, where strong electron-phonon coupling may lead to the formation of polarons—quasiparticles composed of electrons coupled with phonons. "For these materials, we are going to develop methods that will be able to simulate the mobility of charge, which are polarons", Nenad notes. This research is conducted within the framework of PolMoReMa, a project funded by the Science Fund of the Republic of Serbia.
Across all these efforts, HPC remains indispensable. It enables the group to handle large, complex systems, perform advanced multi-scale modeling, and ensure that simulations faithfully capture the physical behavior of materials critical for future electronics. As he explained, "the usage of High-Performance Computing in physics research in general and in particular, for example, in the research of the physics of materials, has become almost an everyday practice in these fields". The increasing complexity of physical models and the demand for precision in theoretical validation have made HPC an indispensable tool in modern scientific workflows.
Nenad illustrated this with two concrete examples. First, he emphasized that the scientific process fundamentally depends on developing and testing theories capable of explaining experimental results. These theories often involve "complicated mathematical equations" that cannot be solved analytically. To obtain numerical solutions, researchers rely on large-scale computational simulations. The speed of these simulations, he noted, is crucial: "to be able to get the actual result of your theory that you can then compare to the experiment", access to powerful computing resources is required. While desktop systems might take "days or months" to complete a calculation, HPC infrastructures can produce the same results "in just a few hours or a day". This acceleration enables scientists to explore many theoretical variations efficiently—an essential factor in refining models to achieve agreement with experimental data.
His second example concerned charge transport in materials, a domain that typifies the dual challenge of accuracy and efficiency. In this field, his research group develops approximate methods that are "computationally efficient so that they can be easily applied to realistic materials". These approaches are validated against "numerical exact methods", which, while more precise, are far less tractable computationally. Both approaches, however, depend heavily on HPC resources, as "for both calculations, the approximate methods and the exact methods, we need high-performance computing resources so that we can perform these calculations as fast as possible". Such computational comparisons are key to ensuring that simplified models retain scientific reliability while remaining practical for complex systems.
Finally, Nenad cautioned that the benefits of HPC do not come without effort. As he put it, "there is no completely free lunch". Leveraging HPC infrastructures effectively requires not only access to hardware but also specialized expertise in software development and optimization. Researchers must adapt or entirely redevelop their computational codes to run efficiently on parallel architectures, since "it is not the same as developing code to use just on single processor machines". This highlights an increasingly important aspect of computational science: the need for continuous investment in both human capital and algorithmic innovation to fully exploit the potential of high-performance computing in advancing modern physics.
