Skip to content

Reaction path identification and validation from molecular dynamics simulations of hydrocarbon pyrolysis

Summary

The article combines ChemTraYzer (CTY) analysis of ReaxFF CHO2016 NVT RMD trajectories from LAMMPS for iso-octane and n-heptane high-temperature pyrolysis with a validation workflow: NEB transition-state refinement (GFN1-xTB and ReaxFF in AMS, then B3LYP-D3-BJ/6-31G(d′,p′) in Gaussian 16) to test pathways, focusing on benzene formation routes relative to a Langer et al. mechanism. The work argues CHO2016 can recover known routes while proposing many novel elementary steps, but also documents large errors vs DFT for some hydrogen migration barriers and problematic behavior for spin-forbidden channels—motivating validation and possible reparameterization.

Methods

  • RMD setup: LAMMPS with ReaxFF CHO2016 (Ashraf et al. parametrization as cited); 3D PBC gas cells; NVT at 2500 K; 100 molecules iso-octane simulated 5 ns and 100 molecules n-heptane 3 ns; >5000 total atoms in each cell (from cited molecule counts; exact atomic number per frame in Table/Methods in kinetics PDF); initial mass density 0.1 g cm\(^{-3}\) (pressures ~182 bar and ~207 bar respectively); geometry/connectivity every 1 fs for CTY.
  • CTY post-processing: Bond-order-based connectivity changes; recrossing filtering; long-lived species threshold ~10 fs; reaction lists compared to Langer et al. mechanism via SMILES (RDKit); NetworkX pathway search with user-guided intermediates and constraints excluding very small carbon/hydrogen species as described.
  • Path validation: Extract 10 consecutive geometries spanning 50 fs around an event (5 frames before/4 after); NEB in AMS with 10 Ha/Bohr\(^2\) springs and climbing image; CHO2016 NEB uses tapered bond orders option; end-point B3LYP-D3-BJ/6-31G(d′,p′) verification of TS.
  • Compute cost (reported): Order 30 h/ns (iso-octane) and 27 h/ns (n-heptane) on 16 cores for RMD; CTY analysis ~10 h/ns (paper’s Section 2.4).

1 — RMD (production pyrolysis). Engine: LAMMPS + ReaxFF CHO2016. System: 100-molecule iso-octane and n-heptane cells at 0.1 g cm\(^{-3}\) (~180–210 bar internal pressure from ideal-gas-like density—see paper). PBC; NVT at 2500 K; output every 1 fs to CTY; run lengths 5 ns and 3 ns as stated. Timestep, thermostat, barostat, E-field, umbrella/MTD: not expanded here beyond NVT 2500 K—if absent from the indexed summary, use the kinetics PDF §2 for the exact fs step and thermostat parameters. 2 — DFT/TS for validation (Gaussian B3LYP, AMS NEB, GFN1-xTB in places). Treated in the same workflow bullets. 3 — NEB and B3LYP verification are reaction-path follow-ups, not a separate “production DFT” paper block.

Findings

  • CTY yields very large reaction sets (21,495 reactions after 5 ns iso-octane; 7,422 after 3 ns n-heptane) compared to ~10,298 reactions in the reference mechanism, with >90% labeled new by strict SMILES matching—illustrating both discovery potential and the need for screening.
  • A benzene formation pathway from n-heptane pyrolysis is extracted that connects toward 1,3-pentadiene chemistry in the reference network but uses non-mechanism steps; DFT vs ReaxFF energetics along steps such as 1,3-pentadiene → 1,2,3-pentatriene show large endothermicity, so such routes are high-temperature-biased.
  • For pathways tested, CHO2016 can align with QM for some channels but underestimates hydrogen migration barriers by up to ~40 kcal mol\(^{-1}\) vs DFT in cases highlighted, and can lower barriers for spin-forbidden reactions—stressing QM validation for RMD-discovered steps. The discussion compares the Langer reference mechanism, CTY reaction sets, and NEB+DFT barriers. Kinetic claims here reference ~2500 K RMD; low-temperature kinetics extrapolation is caveated in ## Limitations.

Limitations

SMILES matching marks many near-duplicate elementary processes as new; RMD at 2500 K explores chemistry far from low-temperature kinetics; GFN-xTB / ReaxFF NEB barriers carry known error bars.

Relevance to group

Benchmarks ReaxFF hydrocarbon chemistry discovery workflows relevant to combustion/soot contexts that intersect reactive MD practice in the broader community.

Citations and evidence anchors