Optimization of a New Reactive Force Field for Silver-Based Materials
A ReaxFF reparameterization for Ag and Ag–thiolate systems using >120 QM structures, MCFF optimization in ADF, and validation MD in ADF/LAMMPS against prior Au/S thiolate literature behavior.
Summary¶
The authors introduce AgSCH-ReaxFF, fit with Monte Carlo force-field (MCFF) sampling to DFT training data on silver clusters, slabs, and thiolate adsorption. The new parameter set improves average geometries and energetics versus earlier Ag/Au ReaxFF descriptions for many benchmarks, while acknowledging limits at highly under-coordinated edge sites. MD snapshots compare Au vs Ag thiolate SAMs, including staple motifs on gold but not on silver under their silver model. The parameterization targets organosulfur chemistry at Ag interfaces relevant to SERS, self-assembled monolayers, and nanoparticle coatings where prior Au-centric ReaxFF descriptions are a common starting point but Ag-specific bonding errors accumulate.
Methods¶
- Training data: Diverse Ag clusters (sizes and isomers), bulk/surface slabs, and thiol/thiolate reactions; QM references primarily DFT as described (functionals/basis sets in article and SI).
- Optimization: MCFF implementation in ADF to adjust selected ReaxFF parameters against the training set.
- MD validation: Production runs with ADF and LAMMPS ReaxFF modules; timestep 0.25 fs, Berendsen thermostat (5 fs damping), total ~100 ps segments as reported; NVT heating/cooling protocols for SAM studies (e.g., snapshots near 100 K for comparison figures). PBC for slab and SAM supercells; barostat: N/A in the NVT validation windows described here; pressure: N/A for those constant-volume NVT runs. Electric field: N/A; umbrella / metadynamics: N/A.
Findings¶
- New parameters better reproduce average cluster and slab energetics vs older Ag ReaxFF and track Ag\(_{20}\) pyramid thiolate adsorption comparably to gold benchmarks.
- Edge-shortening artifacts for under-coordinated Ag atoms remain imperfect—an intrinsic limitation noted by the authors.
- SAM simulations on Ag(111) do not show Au–S–Au staple motifs that appear in gold SAMs with prior parameterizations—consistent with differing experimental pictures for Ag.
- The article positions MCFF optimization as a practical route to large training-set fits while flagging residual errors where coordination is extremely low—an important caveat for nanocluster applications.
Limitations¶
Residual errors for low-coordination geometries; thiolate chemistry extrapolated beyond training chemistries should be revalidated. SAM comparisons to gold staple motifs are qualitative structural contrasts; quantitative free energies of disorder and domain boundaries on Ag(111) may require longer sampling than the ~100 ps illustrative segments quoted in the article. MCFF fits can be sensitive to weighting of cluster versus surface targets; reproduce optimization settings from the primary text before porting parameters into new workflows.
Relevance to group¶
Core ReaxFF parameterization reference for Ag and organosulfur/metal interfaces, relevant to catalysis and nanocarbon/metal hybrid contexts.