I am a computational chemist with a holistic view on tackling challenging problems in chemistry & engineering. In particular, I enjoy finding new, unexpected applications or synergies of established techniques and theories.
Below, you can find an overview of some of my past & present projects and interests. You're welcome to reach out for specific inquiries.
How do materials behave? Why can we synthesize one material, but not the other? How easily can we turn a pharmaceutical powder into a tablet? Why does a crystal grow in a particular shape? How does a catalyst work? These are all questions I try to answer using computational tools.
At Johnson & Johnson, my postdoc project concerns the development & evaluation of in silico tools to derisk & guide formulation of small-molecule pharmaceuticals. In casu this mostly means property prediction of molecular crystals with density functional theory (DFT). DFT nowadays does a pretty good job in describing energetics of many molecular crystals (lattice energies and relative polymorph stabilities) and if it doesn't, we usually have a pretty good idea why. We have further explored its ability to describe elastic properties and ionization states in acid–base multicomponent crystals.
Both properties are highly sensitive to the employed approximations (i.e., DFT functional), but good practices have been identified to study relative elasticities (elastic anisotropy) and to assist the assignment of experimentally ambiguous salts or cocrystals.
Are elastic properties of molecular crystals within reach of density functional theory? Accuracy, robustness, and reproducibility of current approaches. K. M. Bal, A. Collas. Cryst. Growth Des. 24, 3714–3725 (2024).
Charting the salt–cocrystal continuum of acid–base multicomponent crystals with hybrid density functional theory. K. M. Bal, A. Collas. CrystEngComm 26, 6765–6773 (2024).
I have done some fundamental work on DFT modeling practices in materials science. Specifically, I have worked on charged materials and some of the computational issues you run into when studying adsorption and catalysis in such systems. Many standard practices cannot be used indiscriminately. Be careful when you use periodic boundary conditions and local (non-hybrid) functionals! I also studied some charge–activity relations.
Modelling molecular adsorption on charged or polarized surfaces: a critical flaw in common approaches. K. M. Bal, E. C. Neyts. Phys. Chem. Chem. Phys. 20, 8456–8459 (2018).
Overcoming old scaling relations and establishing new correlations in catalytic surface chemistry: combined effect of charging and doping. K. M. Bal, E. C. Neyts. J. Phys. Chem. C 123, 6141–6147 (2019).
Besides predictive modeling, I have also collaborated with applied researchers to explain experimental observations. One interesting example of such study was a collaboration with imec about the formation (and prevention) of defects in semiconductor materials during manufacturing. About my part of the collaboration I made a poster that I'm quite fond of.
Reactive plasma cleaning and restoration of transition metal dichalcogenide monolayers. D. Marinov et al. npj 2D Mater. Appl. 5, 17 (2021).
My predominant research focus during my postdoc at the University of Antwerp has been on enhanced sampling techniques. Such methods allow us to simulate slow, activated processes within molecular dynamics simulations. From those simulations, we can then extract chemically relevant observables such as reaction rates, free energy differences, and free energy barriers. Input data to reproduce or adapt my work can be found on PLUMED-NEST.
Free energy barriers are commonly used to describe kinetics. We show how to extract this quantity in a consistent manner from enhanced sampling simulations.
Free energy barriers from biased molecular dynamics simulations. K. M. Bal, S. Fukuhara, Y. Shibuta, E. C. Neyts. J. Chem. Phys. 153, 114118 (2020).
A strategy that uses concepts from nonequilibrium thermodynamics and machine learning to accelerate the sampling of high-barrier processes. It can be more efficient than techniques such as metadynamics in certain difficult cases.
Reweighted Jarzynski sampling: Acceleration of rare events and free energy calculation with a bias potential learned from nonequilibrium work. K. M. Bal. J. Chem. Theory Comput. 17, 6766–6774 (2021).
I show that transition state theory—in combination with my barrier calculation approach—is generic enough to be even applicable to nucleation processes. Key here is the use of a recrossing correction, which ought to be equally applicable to any type of diffusive process as well.
Nucleation rates from small scale atomistic simulations and transition state theory. K. M. Bal. J. Chem. Phys. 155, 144111 (2021).
Extending and validating bubble nucleation rate predictions in a Lennard-Jones fluid with enhanced sampling methods and transition state theory. K. M. Bal, E. C. Neyts. J. Chem. Phys. 157, 184113 (2022).
The collective variable-driven hyperdynamics (CVHD) method can accelerate molecular dynamics simulations so that slow chemical reactions can occur within the short MD time scale.
It is a rather generic technique that we have applied to predict the outcome of processes ranging from fuel pyrolysis & combustion, etching in nuclear fusion devices, diffusion, and catalysis. The high flexibility of the method—allowing it to accelerate reactions in very wide time scale ranges within a single simulation—is achieved by using a metadynamics-style bias potential.
CVHD has been implemented in the Amsterdam Modeling Suite (AMS), where it can be combined with ChemTraYzer to analyze reaction networks.
Merging metadynamics into hyperdynamics: Accelerated molecular simulations reaching time scales from microseconds to seconds. K. M. Bal, E. C. Neyts. J. Chem. Theory Comput. 11, 4545-4554 (2015).
Thermodynamically stable greenhouse gasses such as CO2 can be removed with high conversion using plasma technology. But how is this possible? We have shown that we can interpret the plasma as an environment in which the chemical equilibrium is shifted. That is, the plasma as a whole is out of equilibrium, but if we zoom in on specific reaction stoichiometries we see that product concentrations evolve to some equilibrium state. So, a partial chemical equilibrium (PCE) is established due to a background plasma nonequilibrium. Different plasma conditions yield different equilibrium states and, thus, different conversions. That's how we sometimes beat thermodynamics! Analyzing these shifts in conversion also yields mechanistic insights.
On the kinetics and equilibria of plasma-based dry reforming of methane. Y. Uytdenhouwen, K. M. Bal, E. C. Neyts, V. Meynen, P. Cool, A. Bogaerts. Chem. Eng. J. 405, 126630 (2021).
My very first research work was on Monte Carlo simulations. More specifically, I studied force-bias MC methods that are very MD-like from an implementation perspective. In addition, they result in MD-like particle moves. The question than becomes, can you associate a time scale with these moves?
Not really, it turns out. fbMC is, however, quite good at quenching and relaxing amorphous systems, usually more efficiently than MD. I have implemented one such algorithm in LAMMPS. There are also commercial implementations in AMS and QuantumATK.
On the time scale associated with Monte Carlo simulations. K. M. Bal, E. C. Neyts. J. Chem. Phys. 141, 204104 (2014).