DFT Modeling and Development of Electrocatalysts

NAISS 2024/5-213


NAISS Medium Compute

Principal Investigator:

Tore Brinck


Kungliga Tekniska högskolan

Start Date:


End Date:


Primary Classification:

10407: Theoretical Chemistry

Secondary Classification:

10403: Materials Chemistry

Tertiary Classification:

10402: Physical Chemistry



Our focus is currently on the development of heterogenous electrocatalyst for electrosynthesis of important base chemicals, such as ammonia, alcohols and olefins. The research includes both independent computational studies and collaborations with internationally renowned experimental groups. We will continue exploring the unique capacity of boron doped silicon and germanium compounds, e.g. B-silicene and B-germanene, to selectively bind N2 and CO, and function as single atom catalysts. Further doping to produce diatom and effective triatom catalysts that promote C-C cross coupling and results in higher valued products will also be explored. We will continue the development of solid catalysts of Cu and Zn ligated with e.g. NCN units for electrocatalytic reduction of nitrate and nitrite to ammonia. The electrocatalytic modeling is highly challenging and requires resource demanding periodic DFT computations using VASP. To model the reaction mechanism and free energy surface of a single reaction for a single catalyst typically requires the computation of structure, phonon frequencies and free energy of more than 25 reaction intermediates and products. Highly demanding transition state calculations using the nudge elastic band (NEB) approach are needed to determine the kinetics of key chemical and electrochemical steps. NEB calculations parallelize well and the performance is good using 16 nodes or more on Dardel. Electrochemical reactions take place in aqueous solution and implicit solvation using Vaspsol is applied. In some cases resource demanding DFT MD simulations with explicit water molecules are needed. Long MD simulations are used, e.g. to estimate catalyst stability. Furthermore, the development of new catalysts requires structure prediction when crystal structures are not available, e.g. our collaborators are developing catalysts based on solid solutions of ligated transition metals in varying ratio. For structure prediction we use USPEX, which is based on an evolutionary algorithm and is coupled to VASP or CP2K for ranking and structure optimization. Structure prediction is highly demanding as a very large number of structures has to be optimized but the process is well parallelized in USPEX. In additions to the conventional DFT computations we will continue to develop the molecular surface property approach (MSPA). It is computationally efficient as it allows all the catalytic sites and their associated activities to be estimated from local surface descriptors obtained from a single DFT computation of the bare catalyst. The MSPA descriptors will be used in the analysis and development of catalysts, e.g. large chiral nanoparticles with well-defined structures have been studied for selective glucose oxidation, and for the development of machine learning models for use in catalyst design. We have developed python codes for automated computation and analysis of MASP that work with a variety of DFT codes, e.g. GPAW, CP2K and VASP ( In addition, a new Fortran program is being developed that will allow for real time analysis of these properties in terms of local minima and maxima as well as descriptors related to the statistical distribution of surface properties.