ChemTraYzer - Reaction Models from Molecular Dynamics Simulations
Information on molecular properties and reaction kinetics are key components in the design and optimization of internal combustion engines. Models for describing the combustion chemistry of a fuel often comprise hundreds of species and thousands of reactions. Reaction kinetics and thermodynamic properties can be derived from experiments or from ab-initio quantum chemistry methods (QM). The high effort required for both of these approaches makes their use challenging if the underlying reactions and intermediate species of the combustion process are not known a-priorly.
Classical molecular dynamics trajectories (MD), simulated with a reactive force field, allow for studying the chemistry of multiple species at the same time and track their evolution to final products (for example a simulation of a fuel/air mixture during ignition). Force fields are many orders of magnitudes faster in calculating energies and forces compared to ab-initio QM, but their use comes at the cost of accuracy of thermodynamics and kinetics predictions. The low computational cost of MD simulations does not only allow the investigation of combustion processes but also the application to much larger systems like polymerization, catalysis, bio processes, etc.Copyright: Lehrstuhl fuer Technische Thermodynamik der RWTH Aachen
Our software package “Chemical Trajectory AnalYzer” (ChemTraYzer) [1,2,3] is an approach that combines the exploration of chemical space, i.e. species connected by reaction networks, via MD simulations with the semi-empirical force-field ReaxFF [4,5] with an accuracy close to that of quantum mechanical methods for the quantification of thermodynamical properties and reaction kinetics (cf. Figure 2). The code is accessible as open-source software under the MIT license. A guide to installation and usage is given in the 'readme' file. The ChemTraYzer is also available within in the commercial software Amsterdam Modeling Suite.
Currently, we continue the development of the ChemTraYzer in three Projects:
- In an DFG-funded project [DFG Grant No. LE 2221/8-1] we refine the ChemTraYzer methodology to a tool for the highly automatized generation of reliable reaction models of unconventional fuels. Our efforts have already resulted in the extension of the ChemTraYzer approach with automatically reoptimized quantum-mechanical transition-state theory rate constants (see fig. 1).  We intend to further improve the accuracy of the rate constants using non-TST information from the MD trajectories. As some important ignition processes such as the low-temperature ignition are too slow to be uncovered with standard MD we develop acceleration methods to extend feasible simulation times.
- So far, post-processing of MD trajectories lacks a concise representation of results for very large reaction networks. Also, the generation of a suitable set of empirical parameters for ReaxFF is time and resource consuming, as it requires large but accurate training sets containing a variety of stable and unstable species. With the combination of MD and QM refinement, the ChemTraYzer provides a framework for the automatization of the parametrization process.
Both the representation of results and the automatization of the force-field parametrization are addressed in the EU-funded project Automatic generation of Chemical Models (AutoCheMo) in collaboration with the Center for Molecular Modeling at Ghent University and the industrial partner Software for Chemistry & Materials in Amsterdam. Another ambitious goal in the AutoCheMo project is going beyond the commonly used harmonic oscillator approximation in QM calculations by considering coupled anharmonic motions to improve the accuracy of thermodynamic properties and reaction kinetics predictions.
- In the Cluster of Excellence Fuel Science Center we study the production and combustion of unconventional biohybrid fuels on scales ranging from atomistic to the device-level. We employ the ChemTraYzer to discover novel reaction paths necessary to understand ignition behavior and soot formation of novel fuels and their mixtures in order to design fuels with improved properties in the long run.
 Döntgen et al., J. Chem. Theory Comput. 11 (2015), 2517-2524
 Döntgen et al., J. Chem. Inf. Model. 58 (2018), 1343-1355
 Kröger et al., Proc Comb Inst 37 (2018), 275-282
 van Duin et al., J. Phys. Chem. A 105 (2001), 9396-9409
 Chenoweth et al. J. Phys. Chem. A 112 (2008), 1040-1053
DFG, LE 2221/8-1
EU, MSCA-ITN-EID 814143. “AutoCheMo”
Cluster of Excellence, „Fuel Science Center“