Skip to content

Development of a ReaxFF reactive force field for Si/Ge/H systems and application to atomic hydrogen bombardment of Si, Ge, and SiGe (100) surfaces

Summary

Psofogiannakis and van Duin extend an established Si/H ReaxFF parameterization with germanium chemistry—Ge–Ge, Si–Ge, and Ge–H interactions—by training against quantum chemistry data that emphasize relative differences between Ge and Si analogs (Surface Science, DOI 10.1016/j.susc.2015.08.019). The scientific target is plasma-processing and surface hydrogenation of group-IV semiconductors: atomic hydrogen bombardment of (100) diamond-cubic Si, Ge, and SiGe surfaces at technologically relevant kinetic energies. By preserving prior Si training while adding Ge terms through difference fitting, the authors aim to keep compatibility with existing Si ReaxFF datasets while enabling alloy and Ge-specific etching pathways that differ in bond strength and lattice openness.

The introduction also places this work in a process-engineering context: H-plasma treatment is used for passivation and cleaning, but can induce near-surface damage through hydrogenation-driven etching and subsurface defect formation. The ReaxFF extension is therefore motivated as a mesoscale bridge between expensive ab initio approaches and nonreactive empirical models, enabling trajectory-level comparisons of composition and incident-energy effects that matter for semiconductor process windows.

Methods

Force-field training

Parent field: established Si/H ReaxFF augmented with Ge–Ge, Si–Ge, and Ge–H interactions. QM reference: Jaguar DFT with B3LYP and LAV3P** on clusters and related models. Training set: dissociation curves, angle/torsion scans, and geometries benchmarked in Table 1-style comparisons emphasizing Ge–X versus Si–X energetics. Optimization: parameters adjusted to match those QM targets while not degrading prior Si accuracy (difference-fitting philosophy described in Surface Science).

MD application (atomistic dynamics)

Reactive MD with ReaxFF simulates monoenergetic atomic H bombardment of diamond-cubic Si(100), Ge(100), and SiGe(100) slabs in NVT cells with in-plane periodic boundary conditions and vacuum along the surface normal (slab geometry, Surface Science). Engine, timestep, thermostat coupling, substrate temperature, slab sizes, equilibration versus bombardment staging, and the incident-energy grid are specified in the article’s Computational Methods; they are not present on the short local extract used for curation.

Barostat / hydrostatic pressure: N/A — constant-volume slab bombardment.

Electric field / enhanced sampling: N/A.

Findings

Composition trends: Si(100) shows more facile surface and subsurface H uptake, H\(_2\) formation, and etching-related responses than Ge(100) under comparable atomic H bombardment, consistent with bond-strength and lattice openness differences highlighted in the abstract. SiGe is intermediate, with alloy-specific surface disordering that changes how H interacts with the surface. Sensitivity: reaction propensity varies with incident kinetic energy and composition along the scanned conditions. Limitations: monoenergetic atomic H simplifies real H-plasma mixtures (ions, molecules, broad energy distributions); ReaxFF remains approximate relative to DFT for rare channels where the article provides cross-checks.

  • Comparative outcome framing: The paper uses one force field across Si, Ge, and SiGe to isolate composition-driven differences without changing the reactive model family between cases.
  • Mechanistic emphasis from abstract/introduction: Surface adsorption, subsurface incorporation, molecular hydrogen release, and etching are treated as coupled outcomes of bombardment chemistry rather than independent observables.
  • Corpus honesty: The local p1-2 extract establishes the training philosophy and application scope, while full numeric bombardment protocol values and trajectory statistics should be read from the complete article text.

Relevance to group

van Duin-group ReaxFF extension to Ge semiconductors with explicit H-plasma-relevant dynamics.

Citations and evidence anchors

DOI 10.1016/j.susc.2015.08.019.