Computational Electronic Structure Theory (CEST)
We develop density-functional theory (DFT) for the electronic ground state and Green's function methods for excited states. Our favourite Green's function method is the GW approach. We are currently researching the application of GW to core level spectroscopy, a powerful tool to characterize molecules, liquids and adsorption processes at surfaces. We are also going beyond GW by combining it with the configuration interaction method to capture static correlation in strongly correlated systems.
For wide dissemination, we implement most of our developments into the Fritz Haber Institute ab initio molecular simulations package (FHI-aims). If you are interested in using FHI-aims for your own work or if you would like to contribute to FHI-aims, please contact us.
For more information on the GW approach, see our recent review article:
For recent developments see:
Machine learning is a branch of artificial intelligence and is currently revolutionizing research practices in the natural sciences. Machine learning models are trained on materials data already available from experiments or computations by creating statistically optimized relationships between the given data. Once the model is trained sufficiently, it can make predictions for new materials or infer correlations with almost the same accuracy as the data generation method, but in only a fraction of the time and with a fraction of the computational or experimental effort. We currently pursue two main machine learning research lines: BOSS and ARTIST.
BOSS: Bayesian Optimization Structure Search is an active learning technique for global exploration of energy and property phase space, and for accelerated structure determination.
ARTIST: Artificial Intelligence for Spectroscopy is a suite of machine learning methods for excited states and spectral properties. We are exploring kernel ridge regression for individual excitation energies and neural networks for excitation spectra. We are also developing descriptors for atomistic representations. These descriptors are available in the DScribe Phython library.
Chemical diversity in molecular orbital energy predictions with kernel ridge regression, A. Stuke, M. Todorović, M. Rupp, C. Kunkel, K. Ghosh, L. Himanen, and P. Rinke, J. Chem. Phys. 150, 204121 (2019)
DScribe: Library of descriptors for machine learning in materials science, L. Himanen, M. O. J. Jäger, E. V. Morooka, F. F. Canova, Y. S. Ranawat, D. Z. Gao, P. Rinke, and A. S. Foster, Comp. Phys. Commun. 247, 106949 (2020)
Data-driven science is heralded as a new paradigm in materials science. In this field, data is the new resource, and knowledge is extracted from materials datasets that are too big or complex for traditional human reasoning - typically with the intent to discover new or improved materials or materials phenomena. We are currently involved in the development of two materials infrastructures, the Novel Materials Discovery (NOMAD) laboratory, and the Aalto Materials Digitalization Platform.
Biomaterials play a crucial role in our pursuit of a sustainable society. Feedstock from biomass (e.g., wood) processed in biorefineries can provide us with as a renewable source of materials such as chemicals, solvents, and polymers that can subsequently be incorporated into high-value products. Bio materials furthermore offer alternative routes for waste management through biodegradation processes and promote equality in the global economy by decreasing our reliance on scarce raw materials.
To accelerate the development of new technologies for biomaterials, we are researching machine learning-assisted approaches to materials processing and modelling. Our current efforts focus on applying Bayesian optimization, through our in-house developed code BOSS, as a means of planning experiments and predicting their outcome.
For an application of machine learning and BOSS to optimize a novel biorefinery concept for green lignin extraction based on hydrothermal pre-treatment of hardwood followed by aqueous-acetone extraction see:
For an application of machine learning and BOSS to predict the morphology of colloidal, oxidized tannic acid particles see:
Machine Learning as a Tool to Engineer Microstructures: Morphological Prediction of Tannin-Based Colloids Using Bayesian Surrogate Models. S.-A. Jin, T. Kämäräinen, P. Rinke, O. J. Rojas, and M. Todorovic, MRS Bulletin (2022)
Molecules may aggregate into aerosols in the atmosphere. Such cluster formation affects air quality and the climate. We develop and apply artificial intelligence (AI) methods to model molecular processes in the atmosphere and to predict and understand molecular cluster formation. We are also developing digital twins of molecular processes and scientific instruments for a virtual laboratory in atmospheric science. The CEST group is part of the Virtual Laboratory for Molecular Level Atmospheric Transformations (VILMA) Centre of Excellence.
Predicting gas–particle partitioning coefficients of atmospheric molecules with machine learning, E. Lumiaro, M. Todorović, T. Kurten, H. Vehkamäki, and P. Rinke, Atmos. Chem. Phys. 21, 13227 (2021)
Clean energy use and generation is one of society's grandest challenges. We apply our electronic structure theory and machine learning methods to study suitable materials for clean energy solutions. We investigate novel hybrid perovskite solar cell materials, catalysts for hydrogen production and organic light emitters and solar cells.
Activation Energy of Organic Cation Rotation in CH3NH3PbI3 and CD3NH3PbI3: Quasi-Elastic Neutron Scattering Measurements and First-Principles Analysis Including Nuclear Quantum Effects, J. Li, M. Bouchard, P. Reiss, D. Aldakov, S. Pouget, R. Demadrille, C. Aumaitre, B. Frick, D. Djurado, M. Rossi, and P. Rinke, J. Phys. Chem. Lett. 9, 3969 (2018)
Research into future technologies has come to focus on miniature multi-material devices and nanostructures, with the intent of harnessing quantum mechanical phenomena to perform an automated function such as generating signals or separating atoms or charges. Organic and inorganic materials are frequently employed side-by-side to take advantage of their unique capabilities. While the properties of the individual substances or bulk materials are known, it is not always possible to predict or measure what occurs at the boundary between them.
We apply our electronic structure theory and machine learning methods to organic-molecule protected noble metal clusters, DNA-stabilized silver clusters, and organic films on inorganic semiconductors and metals. These systems hold great potential for applications in electronic devices, catalysis, biochemical sensing and medical treatments.