Skip to content

First-principles–based reaction kinetics from reactive molecular dynamics simulations: Application to hydrogen peroxide decomposition

Summary

The paper introduces RMD2Kin / ReaxMD2Kin: extract dominant reactions and temperature-dependent rates from condensed-phase ReaxFF RMD, fit Eyring parameters (\(\Delta H^\ddagger\), \(\Delta S^\ddagger\)) from ln(k/T) vs 1/T, then analytically propagate species trajectories—demonstrated for hydrogen peroxide liquid decomposition (1000–2000 K windows in the main example) with QM spot checks on barriers. The opening perspective argues that combustion, CVD, fracking, and related technologies need kinetics tied to actual intermediates, but experiments rarely resolve full networks, so practitioners often rely on oversimplified schemes. QM is accurate yet too costly for \(\sim\)10 nm and \(\sim\)1 ns averaging, whereas ReaxFF RMD is positioned as the workhorse that can be mined automatically for mechanisms and rates without hand-built reaction lists.

Methods

  • Reactive MD: ReaxFF trained to QM (prior work) for H/O chemistry in dense HOOH systems; trajectories long enough to sample bimolecular encounters in liquid (not dilute gas kinetics).
  • Mechanism mining: Identify dominant species (HOOH, H\(_2\)O, O\(_2\), HOO, OH, …) and seven key reactions; count events per interval → instantaneous rates → k(T).
  • Fitting: Eyring analysis yields \(\Delta H^\ddagger\) and \(\Delta S^\ddagger\); recomputed k feeds ODE-style concentration evolution compared against raw RMD (Fig. 1).
  • QM benchmarks: Two-body barrier scans for select steps (values quoted for reactions 1–7, including cases with no simple TS in gas-phase dimer calculations). The authors label the overall pipeline RMD2Kin generically and ReaxMD2Kin when the reactive model is QM-fitted ReaxFF.

Condensed-phase ReaxFF RMD setup. Reactive molecular dynamics on dense hydrogen peroxide systems uses ReaxFF trained to QM (prior publications) in three-dimensional periodic supercells sized for ~10 nm sampling (exact atom counts and densities in pdf_path). Ensemble: NVT thermal sweeps across 1000–2000 K as in the PNAS article; timestep in fs and production/equilibration spans in ps/ns appear in Methods. Thermostat: specified there for high-temperature decomposition kinetics. Barostat / pressure: N/A — no NPT barostat or GPa targets in the showcased liquid H₂O₂ MD. External electric field: N/A. Enhanced sampling: N/A for the RMD2Kin extraction trajectories (direct MD mining).

Findings

  • The seven-reaction mechanism reproduces RMD species populations across 1000–2000 K when propagated with fitted Eyring rates.
  • Negative activation entropy is common; several reactions show negative \(\Delta H^\ddagger\) when radical TS lie below separated reactants—highlighting condensed-phase effects vs gas-phase intuition.
  • Reaction 7 forms a short-lived HO–OH complex that reshapes OH availability; reaction 2 has multiple HOO + HOOH geometries (productive vs exchanging H).
  • Supplemental CFD-scale discussion (Landau–Darrieus) uses HOOH/OH interface tests without observing sustained LDI plumes under sampled conditions. The ReaxFF capability list in the article enumerates prior applications from hydrocarbons and nitramines to electrochemical interfaces, framing HOOH as a tractable exemplar for the kinetics-extraction workflow. The closing discussion positions ReaxMD2Kin as a bridge from atomistic trajectories to continuum reactor or engine models once Eyring parameters are exported.

Limitations

Automation for million-atom, hundreds-of-reaction systems remains future work (manual selection of dominant channels in the showcase).

Relevance to group

Methodological PNAS perspective complementary to Penn State ReaxFF applications; cites ReaxFF ecosystem breadth.

Citations and evidence anchors

  • papers/ReaxFF_others/ilyin-et-al-2018-first-principles-based-reaction-kinetics-from-reactive-molecular-dynamics-simulations-application-to.pdf