Skip to content

Toward a Mechanistic Understanding of the Formation of 2D-GaNx in Epitaxial Graphene

Summary

The work links MOCVD growth, surface analytics (SEM, Auger mapping, STEM/EDS), and DFT to explain how graphene thickness and oxygen functionalization steer gallium intercalation and subsequent nitridation to 2D-GaN\(_x\) under epitaxial graphene on SiC. Oxygen-rich buffer-layer regions near steps adsorb plasma-generated oxygen and block intercalation, whereas thicker, nonfunctionalized terraces intercalate gallium; immediate NH\(_3\) annealing after TMGa exposure is required to avoid gallium loss during heating. Epitaxial graphene on SiC is a platform for 2D nitride integration because sp\(^2\) carbon can template intercalation chemistry while SiC supplies C-face vs Si-face step structure that localizes oxygen uptake.

Methods

1 — MD application (atomistic dynamics). N/A — the study does not report production reactive or classical MD trajectories; prior ReaxFF work on gallium intercalation is cited in the introduction for context only.

2 — Force-field training. N/A — no ReaxFF or classical FF reparameterization in this work.

3 — Static QM / DFT. The authors use density functional theory to relax and compare adsorption energy (and related interaction energy targets) for epoxy- and hydroxyl- functionalized graphene with gallium and NH\(_3\)-related motifs, contrasting basal (nonoxidized) terraces with oxidized regions. Full functional (e.g. PBE-class and hybrid or dispersion choices), PAW/plane-wave or equivalent basis settings, a documented k-mesh (k-point) grid for the periodic supercells, and the list of relaxed structures and other computed property sets are given in the electronic-structure section of the PDF (this wiki does not restate every numerical cut-off from the VOR; the corpus pdf_path is a galley-style file for this ingest).

4 — Experiments and correlative analysis. MOCVD uses epitaxial graphene on SiC with as-grown and He/O\(_2\)-plasma-exposed conditions; Auger maps (e.g. C/O contrast) and STEM/EDS-style cross-section imaging (per ACS Nano Methods) tie graphene thickness and local oxygen content to TMGa exposure, gallium intercalation vs droplet formation, and NH\(_3\)-driven nitridation to 2D-GaN\(_x\). Correlative readouts link O-rich buffer-layer areas to blocked intercalation and O-functionalized thin regions to 2D-GaN\(_x\) when ammonia annealing follows gallium immediately.

Findings

Outcomes and mechanisms. Thin buffer-layer graphene near SiC steps attains oxygen functionalization under plasma or air-equivalent exposure more readily than thicker, less reactive terraces; that O-rich state inhibits Ga intercalation and can yield surface gallium droplets, whereas thicker, low-oxygen regions allow gallium under the graphene. 2D-GaN\(_x\) after TMGa + NH\(_3\) is observed preferentially in O-functionalized thin regions when ammonia is applied right after gallium exposure, while intercalated Ga under thick, nonfunctionalized graphene may not convert in the same window. DFT supports a picture where hydroxyls and related O groups weaken direct Ga binding to the surface but enhance NH\(_3\) reattivity at the graphene, promoting local nitridation and 2D-GaN\(_x\) at O-rich patches, consistent with the experimental emphasis on O-functionalized buffer graphene for uniform coverage. Co-authors Nayir and van Duin place the theory in the same Penn State line as graphene-related reactive modeling, though ReaxFF is not the main engine here.

Comparisons. The work contrasts O-rich vs low-O EG regions, thin vs thick graphene on SiC, and sequencing of Ga and NH\(_3\) exposure; see the PDF (prefer VOR over the corpus galley file for pagination) for SEM/Auger/STEM correlation and DFT energy tabulations.

Sensitivity and levers. Graphene thickness, step- and plasma-driven oxygen uptake, and MOCVD timing (immediate NH\(_3\) after gallium) are the central process knobs the authors highlight.

Limitations (as framed on this page). Reactive MD in cited work is not driving this study’s new results. Kinetics of full MOCVD cycles and entropic effects in ALD-like precursor coverage exceed the 0 K-style DFT sampling summarized here; wafer nonuniformity in graphene from sublimation growth can outrun lab coupon maps—see also ## Limitations below and the VOR PDF for author discussion.

Limitations

Reactive MD cited from related work is not the primary engine here; kinetics of full MOCVD cycles exceeds the DFT models.

Wafer-scale nonuniformity in graphene thickness and step density from SiC sublimation growth can amplify oxygen plasma heterogeneity beyond the lab coupons analyzed—expect process-window sensitivity when scaling GaN\(_x\) coverage.

Relevance to group

Co-authored by van Duin; couples epitaxial graphene/III–V chemistry with DFT interpretation relevant to 2D nitride synthesis.

Citations and evidence anchors