eReaxFF: a pseudoclassical treatment of explicit electrons within reactive force field simulations
Summary¶
The paper introduces eReaxFF, extending ReaxFF with pseudoclassical explicit electrons (and holes) while retaining the bond-order reactive framework. ACKS2 (atom-condensed Kohn–Sham DFT to second order) is integrated for charge-related quantities. The authors train the extension so that electron affinities (EAs) of saturated, unsaturated, and radical species track experiment and selected DFT checks, then illustrate explicit-electron molecular dynamics on conjugated hydrocarbon radicals and compare to Ehrenfest reference dynamics. The abstract emphasizes orders-of-magnitude cost reduction relative to quantum chemistry dynamics and states that literature ReaxFF parameters largely transfer into eReaxFF with optimization focused on explicit-electron-related terms.
Methods¶
MD application (Section IV.A). Two conjugated hydrocarbon radicals, C₁₂H₁₉• and C₁₄H₂₃•, carry an extra electron on the conjugated segment to mimic an excited state. After NVT relaxation at 1 K, NVT production runs at 400, 500, and 600 K use a Berendsen thermostat (100 fs damping), velocity Verlet integration, and Δt = 0.1 fs, probing temperature-driven electron migration among the conjugated chain, aliphatic linker, and radical site; the same setups are compared to Ehrenfest reference dynamics. Section IV.A and Figures 3–4 discuss multi‑ps evolution of energy and charge localization rather than a single headline production length. Gas-phase radical models do not use NPT barostats, applied electric fields, or bias-based enhanced sampling in the protocol described. Boundary conditions are not spelled out on the opening pages summarized here; the indexed JCTC text does not name a commercial MD package.
Force-field formulation and training. eReaxFF retains the standard ReaxFF bonded and nonbonded decomposition while adding explicit electron–nuclear couplings for pseudoclassical carriers and integrating ACKS2 for charge-related quantities (Sections I–II). Training targets electron affinities across bonding classes; a successive one-parameter search minimizes weighted squared error against literature EA data, adjusting only explicit-electron-related parameters (including Gaussian electron width) while leaving the bulk of literature ReaxFF parameters fixed. M06-2X/aug-cc-pVTZ calculations in Jaguar 7.5 provide spot checks for selected unsaturated species where eReaxFF and experiment disagree, showing DFT also underestimates those EAs in the cases discussed.
Static QM as a standalone campaign. N/A — DFT supports EA training and validation only.
Findings¶
The abstract reports qualitatively correct electron affinities versus experiment for the training set and good agreement between eReaxFF MD and Ehrenfest dynamics for the radical demonstrations, while conventional ReaxFF fails many of the same EA targets (Figure 2). Section IV.A describes how temperature changes the timescale for electron migration from the polyacetylene-like segment toward the radical site, including transient localization on the aliphatic linker at elevated temperature; the 400 / 500 / 600 K comparison for C₁₂H₁₉• illustrates thermally accelerated pathways that are rarer at lower temperature. The authors note that full quantum calculations delocalize the added electron more than the pseudoclassical treatment—an explicit limitation—and position eReaxFF as a large-scale alternative to costly time-dependent DFT nonadiabatic dynamics.
Limitations¶
- Pseudoclassical electrons are not a full quantum electronic solution; accuracy limits near conical intersections and strongly correlated states must be monitored.
- Parameter portability still requires systematic validation per chemically distinct subsystem.
Relevance to group¶
Landmark method paper from the van Duin + Harvard + Ghent collaboration defining eReaxFF, a cornerstone of the group’s next-generation reactive FF roadmap.
Citations and evidence anchors¶
- Abstract and Sec. I in
papers/Islam_JCTC_eReaxFF_2016.pdf; DOI:10.1021/acs.jctc.6b00432.