Skip to content

Atomistic Scale Analysis of the Carbonization Process for C/H/O/N-Based Polymers with the ReaxFF Reactive Force Field

CHON-2019 extends CHO-2016 with nitrogen parameters fit to DFT so ReaxFF can follow polyacrylonitrile and PBO carbonization, including N₂ release and ring statistics, in ADF-driven reactive MD.

Summary

The authors develop CHON-2019 by adding nitrogen-containing bond and angle terms to the CHO-2016 ReaxFF baseline using density-functional reference data, with adjustments that remove the valence-angle correction when nitrogen participates (restoring linear HCN) and strengthen the N₂ triple bond relative to CHON-2010. Reactive simulations in the ADF ReaxFF engine study carbonization of ideal ladder polyacrylonitrile (PAN), oxidized PAN, and poly(p-phenylene benzobisoxazole) (PBO) after preequilibration at 300 K, heating at 10 K/ps to 2800 K, and holding 900 ps in NVT to monitor N₂, small gases, and five-, six-, and seven-membered carbon rings as graphitic precursors.

Methods

ReaxFF training and QM targets (A)

CHON-2019 extends CHO-2016 with nitrogen parameters fit to DFT data: N₂ energetics, N-containing bond/angle terms, and interactions of N₂ with radicals formed during carbonization; specific valence adjustments include removing the valence-angle correction when N participates (restoring linear HCN) and strengthening the N₂ triple bond vs older CHON-2010 behavior (see article/SI).

Polymer initial structures and equilibration (B)

Precursors: ideal ladder PAN, oxidized PAN, and PBO models. Cells are compressed to ~1.6 g/cm³, NVT-equilibrated 100 ps at 300 K from independent snapshots; Figure 3 contrasts random vs aligned chain packings where discussed.

Reactive MD carbonization protocol (B)

ADF ReaxFF RMD integration: Berendsen thermostat (100 fs coupling), 0.25 fs timestep, bond-order cutoff 0.3, PBC; heating 10 K/ps to 2800 K, then 900 ps NVT production (shock T exceeds typical furnace ramps to increase reaction events in sub-ns windows). System size (atoms) and cell metrics: PDF Tables / SI. Barostat / servocontrol of pressure: N/A — NVT after heating. Electric field: N/A. Umbrella / metadynamics / replica exchange: N/A — direct RMD only. Visualization in VMD.

Static QM beyond training (C)

DFT details for the training corpus appear in the article; this paper’s result section is trajectory-centric.

Findings

Mechanisms

Oxidized PAN produces more six-membered rings than ladder PAN under identical protocols—oxygen steers early graphenic nucleation. N₂ release traces differ ladder vs oxidized PAN; PBO differs due to its heterocyclic backbone. 5-/7-membered rings plateau near 500 ps while 6-membered rings grow to ~60% of ring motifs by 900 ps.

Limitations

2800 K shock heating accelerates chemistry beyond industrial carbonization schedules. Parameter scope follows the training chemistry emphasized in the article.

Limitations

The 2800 K shock heating schedule accelerates chemistry beyond typical furnace pyrolysis and can overestimate reaction rates relative to experiment. Parameter training emphasizes gas-phase kinetics targets, so condensed-phase quantitative agreement should be checked case by case. The abstract and introduction frame CHON-2019 as extending CHO-2016 so nitrogen release and heterocyclic backbone chemistry in aerospace precursors can be followed without switching force fields mid-simulation.

Relevance to group

van Duin-group CHON ReaxFF for carbon fiber precursor pyrolysis pathways.

Citations and evidence anchors

  • papers/Kowalik_JPCB_2019_CHON_polymer.pdf

Reader notes (navigation)