Skip to content

A reactive molecular dynamics model of thermal decomposition in polymers: I. Poly(methyl methacrylate)

Evidence and attribution

Authority of statements

Prose sections below (Summary, Methods, Findings, etc.) are curated summaries of the publication identified by doi, title, and pdf_path in the front matter above. They are not new primary claims by this wiki.

For definitive numerical values, reaction schemes, and interpretations, use the peer-reviewed article (and optional records under normalized/papers/ when present)—not this page alone.

Summary

Thermal decomposition of polymethyl methacrylate (PMMA) is often summarized as depolymerization to methyl methacrylate, but initiation steps and the relative fate of primary versus tertiary radicals in the condensed phase remain debated when only TGA or TGA/MS data are available. Stoliarov, Westmoreland, Nyden, and Forney present reactive molecular dynamics (RMD): classical MD in which covalent bonds may break and form according to a reactive modification of a valence force field, implemented in their MD_REACT program. They apply the method to bulk PMMA heating so chemistry evolves in situ in the dense polymer, avoiding gas-phase-only mechanistic assumptions.

Methods

Reactive force-field training / QM reference (checklist A)

  • Lineage: Reactive molecular dynamics (RMD) implemented in MD_REACT; valence energy follows a CVFF-type decomposition (bonds, angles, torsions, nonbonded pairs) with bond stretching replaced by Morse functions so bonds can dissociate (Sec. 2, Polymer 44, 883–894).
  • QM training data: CBS-QB3 reaction enthalpies \(\Delta H_0\) for a Table 1 set of model reactions representing PMMA decomposition; D parameters taken from dissociation reactions (with zero-point corrections as stated), \(\pi\)-bond energetics from \(\beta\)-scission reactions; \(r_e\) from B3LYP/CBSB7 optimized geometries (Appendix A lists Morse parameters and provenance).

Molecular dynamics protocol (checklist B)

  • Engine: MD_REACT; Hamilton’s equations integrated with Verlet velocity integrator (Sec. 3.1).
  • System: Bulk PMMA models—primarily one 15-unit chain (~230 atoms); four-chain bundles (15 units each, ~920 atoms) for inter-chain chemistry; periodic boundary conditions; nonbonded interactions via atom-based summation with 16.5 Å cutoff (Sec. 3.1).
  • Equilibration: simulated annealing from 600 K, NPT at 101 kPa, then energy minimization steps; Morse bonds temporarily replaced by harmonics to avoid premature scission; equilibration 0.5–15 ps; reported mass density ~1.02–1.04 g cm\(^{-3}\) (weak T dependence) (Sec. 3.1).
  • Production RMD: NVT after equilibration; timestep 0.2–1 fs (smaller at higher T to avoid divergence); total RMD time 2–100 ps per run; target temperatures 1000, 1200, 1500 K (lowest 1000 K chosen for observable chemistry in feasible time); 10 single-chain and 3 four-chain runs per T (Sec. 3.1–3.2).
  • Thermostat: direct velocity scaling preferred over Nosé–Hoover at these conditions because Nosé–Hoover gave large fluctuations/divergence; velocity scaling kept instantaneous T within ±10 K when used (Sec. 3.1).

DFT / static QM in support of the FF

  • CBS-QB3 thermochemistry for the Table 1 training reactions (as above).

Findings

  • Volatile speciation: methyl methacrylate dominates volatile products (~80–90% mass fraction of volatiles from 1000–1500 K in simulations), with trace H\(_2\), CO, CO\(_2\), light hydrocarbons, and methyl formate—qualitatively consistent with compiled experimental product analyses cited in the paper (Sec. 3.2).
  • Effective rate of monomer production: Arrhenius fit to simulation-derived \(k\) (Eq. (9)) gives \(A \approx (1.9\text{–}2.9)\times 10^{12}\ \mathrm{s}^{-1}\) and \(E \approx 53 \pm 14\ \mathrm{kJ\,mol^{-1}}\) (2\(\sigma\)); authors compare to experimental mass-loss \(E_a\) ranges 60–270 kJ mol\(^{-1}\) and discuss underestimation of overall mass-loss activation energy, noting experiments near 500–700 K vs simulations 1000–1500 K (Sec. 3.2, Fig. 3).
  • Initiation mechanism: backbone and side-group scissions each <~20% of initiation events; a two-step C–C channel (their Reactions (11)–(13)) accounts for ~50–60% (plus ~20–25% for related channels)—contrasting with textbook random-scission or Manring side-group dominance hypotheses (Sec. 3.2).
  • Radical fate / “unzipping”: tertiary radicals mainly \(\beta\)-scission to monomer; primary radicals compete with termination channels (16)–(17)—at 1500 K, <0.5 monomer on average from the primary site before termination vs 2–3 at 1000 K (Sec. 3.2).
  • Authors’ limitations / future work: FF is a first-order approximation to PES; quantum effects and longer time/length scales not captured; concluding remarks call out condensed-phase initiation dynamics differing from gas-phase bond dissociation (Sec. 4). The article appears in Polymer 44, 883–894 (2003).

Limitations

  • Model chemistry and FF errors still bound mechanistic conclusions; extrapolation to additives, flame conditions, or other polymers requires separate validation.

Relevance to group

Reactive MD lineages connect conceptually to ReaxFF applications in polymers and combustion; Westmoreland ties to Penn State-adjacent reaction-engineering networks.

Citations and evidence anchors

  • DOI: https://doi.org/10.1016/S0032-3861(02)00761-9