Computational Insights into Tunable Reversible Network Materials: Accelerated ReaxFF Kinetics of Furan-Maleimide Diels–Alder Reactions for Self-Healing and Recyclability
Summary¶
The paper develops a ReaxFF description for furan / N-methylmaleimide Diels–Alder (DA) and retro-DA chemistry relevant to covalent adaptable networks (CANs), trained against M06-2X-D3/cc-pVDZ barriers and paths, then uses bond-boost accelerated ReaxFF MD to study gas-phase retro-DA stereoselectivity, temperature, bond-boost parameters, and a polymer-backbone model system—reported as a first reactive-MD-oriented step toward CAN kinetics with DA linkers.
Methods¶
- QM reference: Gaussian 16; M06-2X with Grimme D3 dispersion and cc-pVDZ; TS characterization by one imaginary mode and IRC paths; Eyring–Polanyi estimates of endo/exo ratios at 300 K from QM barriers.
- ReaxFF training: Starting from Kowalik et al. 2019 parameters, retraining against the QM forward/retro barriers, reaction energies, and geometries along the path (per Methods and Results).
- ReaxFF MD + bond boost: Boxes of 40 endo and 40 exo products (and later 10 products with polymer backbone); bond boost on the four C atoms involved in cycloaddition; target distance 3.0 Å; bond-break detection at bond order 0.3 (~2.45 Å); F2 = 0.25; F1 scanned 90–130 (and higher F1 noted to remove stereoselectivity); 5000 steps boost duration as stated.
- Stereoselectivity readout: Concentrations of endo/exo reagents after successful boosts; comparison to Eyring expectations and literature retro barrier splitting for furan–maleimide-class systems.
1 — MD (bond-boosted ReaxFF). LAMMPS-style ReaxFF dynamics with bond boost (see Computational section: F1/F2, target 3.0 Å, 5000-step boost; NVT-like gas-phase product boxes; N/A for NPT /barostat: not part of the narratives summarized. PBC where used for the simulation cells as in the PDF. Timestep, thermostat, total trajectory length, pressure: N/A if not in this page—see article. Electric field: N/A. Enhanced sampling: bond boost (not umbrella in this summary). 2 — ReaxFF training (relative to Kowalik et al.) and 3 — static QM (Gaussian M06-2X-D3/cc-pVDZ): in the first bullets under Methods; Kowalik parent is the starting parameter set.
Findings¶
- The retrained ReaxFF reproduces QM energy profiles with small mean errors (barriers/reaction energies cited in the paper; endo retro barrier and endo reaction energy show slightly larger deviations than other entries in their Table 1).
- From DFT (endo vs exo retro barrier difference ~2.2 kcal/mol) and ReaxFF (~2.7 kcal/mol), endo/exo ratios at 298.15 K are estimated as ~41 (DFT) vs ~95 (ReaxFF) by Eyring, bracketing an experimental ratio ~69 cited in the article—used as a consistency check for the reactive MD stereoselectivity analysis.
- Bond-boost MD on 40-molecule endo and exo sets shows endo adducts decompose more readily than exo, matching the retro barrier ordering; F1 tunes how often boosts succeed and must stay below the regime where all events succeed (loss of selectivity at overly high F1).
- Higher temperature simulations (reported relative to 300 K reference) and the polymer-backbone model illustrate how environment and acceleration parameters shift endo/exo outcomes, supporting the use of the parametrization for more complex CAN models in follow-on work. Comparisons to DFT/IRC, Eyring ratios, and a cited experimental endo/exo ratio (paper’s Table 1 narrative) are in the bullets. VOR PDF is authoritative for F1/T values.
Limitations¶
Training and dynamics are focused on gas-phase and simplified network fragments; extrapolation to full polymer CAN condensed-phase kinetics requires additional validation.
Relevance to group¶
van Duin-group ReaxFF development and reactive MD methodology for polymerizable / reversible networks.