WATOC Award Winners: Katja-Sophia Csizi and Maximilian Mörchen


In our third set of interviews featuring PeerJ Award winners from the 12th Triennial Congress of the World Association of Theoretical and Computational Chemists (WATOC 2020), PeerJ Physical Chemistry recently spoke to Katja-Sophia Csizi and Maximilian Mörchen – both based at ETH Zurich in Switzerland – about their interests.

In total, ten PeerJ Awards were given to the best posters presented at WATOC 2020, as voted for by the attendees. Interviews with all the award winners will be posted over the coming weeks.



Free August at PeerJ Physical Chemistry!

A reminder that throughout August, submissions to PeerJ Physical Chemistry will receive a 100% APC waiver if ultimately published in the journal following peer review. See the announcement blog for details.



Katja-Sophia Csizi Ph.D. candidate at ETH Zurich, Switzerland. 

Can you tell us a bit about yourself and your research interests?

I studied general chemistry in Berlin and Munich. Very early, I was fascinated by the predictive power of computational chemistry and how to couple computational method development to experiments. For my PhD, I moved to ETH Zurich, where I am working on the automated exploration of chemical reactions in nanoscale structures via black-box constructible quantum-classical hybrid models.

What first interested you in this field of research?

The exploration of the chemical reaction space of nanoscale systems, for instance proteins and enzymes, is at the heart of understanding important chemical phenomena in nature. The underlying chemistry is extremely versatile and also challenging to describe with computational methods, which is very exciting to me. The sheer dimensionality of such systems requires the application of models that are accurate and computationally efficient, but can cope with a molecular size of several thousands of atoms. This can be realized by the combination of quantum and classical methodologies. However, such hybrid models are difficult to construct due to the huge amount of tunable parameters, which is why the field is largely dominated by the time-consuming case-by-case construction of models. An efficient and accurate alternative is the development of protocols that automatically construct such models from first-principles criteria only. These procotols can then be applied in the context of chemical reaction network exploration. This is still a new and very exciting field and we hope that application of out-of-the-box models to manifold chemical problems becomes valuable even for non-experts in this field.

Can you briefly explain the research you presented at WATOC?

At WATOC, I presented the automated protocol for the construction of quantum-classical hybrid models from a raw molecular input structure, which has been developed in our group. This setup comprises different components, ranging from automated preprocessing of a raw, 3D molecular input structure, to automated parametrization of a classical force field from a quantum reference and automated selection of a quantum region. Our methodology has also recently been released as Swoose module on GitHub within our Scine software project. To validate our
setup, we build the bridge from theory to experiment by applying our black-box model to a specific enzyme class in a cooperation with EAWAG Dubendorf. In this work, we aim to understand the mechanistic implications of cis-dihydroxylation of aromatic hydrocarbons catalyzed by Rieske Dioxygenases. For this class of dioxygenases, we have both experimental reference data and computational data at hand, which allows us to validate whether our black-box model can describe key mechanistic steps at good accuracy.

How will you continue to build on this research?

In future work, we aim to perform high-throughput QM/MM explorations via our Scine/Chemoton chemical reaction network exploration software infrastructure, especially in an interactive framework. We are confident that our automated QM/MM protocol offers a valuable tool for the community, because it is particularly powerful for systems that include non-standard chemistry (where chemical intuition might fail) due to its formulation in the framework of first-principles rules only.


Maximilian Mörchen Ph.D. candidate at ETH Zurich, Switzerland.. 

Can you tell us a bit about yourself and your research interests?

My research generally focuses on the development of highly accurate electronic structure methods, such as Coupled Cluster (CC) and the Density Matrix Renormalization Group (DMRG). I am particularly interested in the inclusion of static correlation effects in the CC theory as well as the automatic selection of active spaces for Complete Active Space (CAS) methods.

What first interested you in this field of research?

The beauty of developing modern, high-precision ab initio correlation methods is the combination of mathematics, physics and programming. Creating a mathematical framework and exploiting physics are at the heart of developing new theories and have always fascinated me. The high complexity of these types of methods even led to the development of diagrammatic models to reduce the complexity and provide a different perspective on these problems.

The implementation is a story in itself. Designing a program structure, optimizing efficiency and enhancing algorithms is another side of this field that I really enjoy. Besides the development, I remain fascinated by how precisely correlation methods predict reactions without relying on any additional assumptions. Even though the correlation energy accounts for only around 1% of the total energy of a molecule, it is of utmost importance for the accurate and reliable description of chemical problems. This shows how complex even small molecular systems are.

Can you briefly explain the research you presented at WATOC?

At WATOC, I presented our new implementation of the autoCAS program package. autoCAS exploits DMRG and concepts from quantum information theory to automatically determine crucial orbitals for an active spaces, enabling black-box usage of CAS methods. While the previous implementation featured a graphical user interface for interacting with autoCAS, our new implementation is fully controllable over the command line and scriptable via the corresponding Python3 package. This allows users to streamline CAS calculations, customize work ows and implement new interfaces to existing electronic structure codes. In addition, I presented our analysis of the Tailored CC method, where we showed that active spaces determined via autoCAS provide an excellent choice for tailored CC.

What are your next steps?

We are working on a new graphical interface of autoCAS to further investigate the system under study and plan to utilize autoCAS in reaction network explorations. Moreover, we are developing a scheme to refine our method combining DMRG and CC in context of strongly correlated systems.


You may also like...