Alex Mironenko
Catalysts for renewable energy and chemical production
Catalysis is an indispensable tool for solving energy and environmental problems of the near future. Advances in computational power and quantum mechanical algorithms over the last 15-20 years enabled discoveries of novel catalytic materials on a computer, providing an alternative to a traditional “experimental trial-and-error” approach. Despite successful examples, computational prediction of the optimal catalytic site structure and composition to promote a specific chemical reaction remains an unsolved problem in chemistry, due to the inherent complexity of catalytic phenomena and materials involved.
The long-term focus of the Mironenko Lab is on developing comprehensive theoretical models of a catalytic site and a surrounding environment using novel methods, based on principles of quantum mechanics. We expect that such models will (1) shed light on dynamics of catalytic reactions and time evolution of a catalyst by extending time and length scales amenable to simulations, (2) facilitate high-fidelity in silico predictions of reaction rates, reaction mechanisms, and materials, and (3) help formulate a unifying framework for diverse catalytic phenomena on metals, semiconductors, zeolites, and organometallic complexes. The ultimate goal is to apply such models to chemistries implicated in renewable energy production and environmental remediation, in order to reveal optimal atomic configurations of heterogeneous catalysts that maximize their activity, selectivity, and stability, and which can be synthesized and tested experimentally.
Research Group Website Faculty Profile
Reactive Force Fields Derived from First Principles
Reliable computational description of dynamic catalytic phenomena hinges upon accurate knowledge of the catalytic system energy as a function of all atomic coordinates Ε = ƒ (r1, … , rN) (the potential energy surface, or PES). The density functional theory (DFT) with the semilocal treatment of exchange-correlation effects often represents an optimal PES model in terms of accuracy and computational cost. However, the current routinely used DFT methods are too expensive to study all but the simplest of molecules, surfaces, and solids having < 100-200 atoms, necessitating the development of more computationally efficient, albeit approximate, methods.
Approximate PES models that allow for chemical bond breaking and formation are referred to as the reactive force fields. Although a plethora of reactive force fields has been introduced over the years, such models are often empirical, complex, require large datasets for parameterization, and frequently suffer from parameter transferability issues. We aim to mitigate those drawbacks by the development of a physics-based, minimally empirical reactive force field, obtained by the systematic approximation of the Kohn-Sham equations, which are the basis of the density functional theory. We are interested in determining accuracy and efficiency limits of a fully theoretical PES description, and whether the limits, when found, can be circumvented by a symbiosis of theory and machine learning techniques. Ultimately, an accurate and transferable reactive force field will enable efficient sampling of the system’s configuration space and screening of the chemical compound space at a fraction of computational cost, facilitating in silico catalyst discovery.
Dynamics of Low-Dimensional Metal Oxide Catalysts
Low-dimensional metal oxides, such as oxidized “single atom catalysts”, supported MOx clusters, and 2-dimensional metal-supported surface oxides (“inverse catalysts”), are implicated in diverse catalytic transformations ranging from CO oxidation and CO2 reduction to biomass hydrogenolysis and electrochemical oxygen evolution.
Currently, their atomic structure is either unknown or can only be observed on flat surfaces at ultrahigh vacuum conditions using scanning tunneling microscopy, representing a major obstacle in understanding their catalytic properties and developing structure-property relationships. Quantum-chemical methods together with global geometry optimization tools can, in principle, circumvent this limitation. However, the flexible nature of low-dimensional structures renders their description by commonly employed “static” density functional theory calculations incomplete, necessitating full ab initio molecular dynamics treatment.
We aim to study the dynamics of low-dimensional oxides, using the effective Hamiltonian methods being developed in the group, in order to answer the following questions:
(1) What are the structures and compositions of dominant MOx species as a function of reaction conditions?
(2) What factors affect relative kinetic and thermodynamic stability of 0D, 1D, and 2D oxides, and what are their interconversion pathways?
(3) What are the implications of oxide structural flexibility for catalytic reactivity?
“Bottom-Up” Coarse-Graining for Biomimetic Catalysis
Enzymes are biological catalysts exhibiting extremely high activities and substrate specificities, with current and proposed applications ranging from biofuels production to wastewater treatment. However, enzymes often suffer from the limited stability even at room temperature, difficult recycling, and high cost. Heterogeneous catalysts, such as metal nanoparticles, overcome limitations of the enzymatic catalysis, but at the expense of much lower activity and selectivity.
Nanozymes are an emerging class of catalysts that take the best from both worlds and combine enzymatic ultrahigh activity with nanoparticle stability and separability. Nanozymes include metal oxide nanoparticles, such as Fe3O4 with peroxidase-like activity, or metal nanoparticle cores decorated with a mix of ligands that either modulate catalytic activity of the core or behave as catalytic sites themselves, acting cooperatively.
We aim to obtain atomic-level understanding of the catalytic action of representative nanozymes and the origin of their similarities to enzymes using a combination of first-principles calculations, molecular simulation, and microkinetic modeling. Due to intertwined solvent, ligand, and active site dynamics, coupled with chemical reactivity, it is essential to reduce the complexity of systems studied by integrating out inessential dynamics while retaining computational accuracy, in order to make nanozymes computationally tractable. We aim to achieve this by developing “bottom-up”, mixed-resolution coarse-grained models that reproduce underlying all-atom simulations as closely as possible while incorporating bond breaking/formation, described at the DFT level of theory. Ultimately, we will formulate the design principles for active and stable nanozymes.
Multiscale Modeling of Selective C-C Coupling
There is a consensus among the 97% of climate science experts that excessive CO2 emissions is the main contributor to the ongoing climate change and its associated economical and human life risks. There is an urgent need to develop an economically competitive carbon recycling technology that would enable efficient and selective conversion of CO2, water, and solar/wind energy to high-energy-density C2+ liquid fuels, which would be compatible with the existing petroleum infrastructure. While the selective electrochemical synthesis of C1 compounds, such as CO, HCHO, and HCOOH, has been reported, the transformation of C1 species to C2+ fuels and value-added chemicals remains challenging due to associated high capital costs or selectivity issues owing to fundamental constraints in the metal catalysis.
Despite those challenges, there have been reports of ultraselective conversions of C1 platform chemicals to C2-C3 species, catalyzed by homogeneous catalysts. However, the mechanistic details of those reactions often remain poorly understood. We aim to reveal elementary reaction sequences that lead to C3 formation and elucidate the origins of their high selectivity. Once the reaction mechanisms and states of catalysts under experimental conditions are known with certainty, we will focus on formulating the design principles for the effective homogeneous C-C coupling catalyst, which will be subsequently adopted to perform an in silico design of a corresponding heterogeneous catalyst. To reveal the fundamental chemistry behind the selective C-C coupling, we will employ the multiscale modeling framework, which includes density functional theory, microkinetic modeling, electronic structure analysis, and which will yield accurate estimates of observables (reaction rates, reaction orders, rate-limiting steps, and selectivity trends) that will be compared to corresponding experimental values.
Several of selective C-C coupling reactions occur in non-aqueous solvents. To model the reaction energetics accurately, we will employ a state-of-the-art, implicit solvent-based computational protocol that allows pKa predictions for charged intermediates within 1-2 units specifically in nonaqueous solvents.
Improving the Accuracy of Density Functional Theory
Density functional theory (DFT) is essential for reliable predictions of reaction enthalpies, entropies, and activation barriers in catalysis. Accuracy of DFT computations is dependent on the choice of the exchange-correlation energy functional Εxc, [ρ] where ρ is the electron density. While the most accurate Exc (“hybrids” and “double hybrids”) approach chemical accuracy (<1 kcal/mol errors) in predictions of enthalpies of formation for certain systems, these functionals are inherently non-local (Exc = ∫∫ ƒ (r1,r2) dr1 dr2) and too computationally expensive to be used in heterogeneous catalysis. For this reason, a more computationally efficient “local approximation” (Εxc = ∫ g (r)dr) has been adopted, which, however, lacks quantitative accuracy for transition metal compounds and reaction barrier predictions in general.
In the group, we investigate the “divide-and-conquer” approach to the development of the exchange-correlation functional. We identify the most problematic parts of the chemical system systematically and apply non-empirical, non-local corrections only to those parts while describing the rest of the system with local Exc. By doing so, we combine the accuracy of nonlocal and efficiency of local Exc functionals at a minimal extra computational cost. The new methodology demonstrates promising results in describing dissociation energetics of molecules, cohesive energies of solids, lattice constants, and atomic and molecular ionization energies.