Skip to content

Silicon‐Driven Facet Regulation Enables Tunable Micro‐Diamond Architectures in Liquid Ga–In

FieldValue
JournalSmall
Published2026-06-22
AuthorsZhi Jiang, Xueying Zhang, António José Silva Fernandes, M. Peres, Gil Gonçalves
DOI10.1002/smll.73876

Faceted single-crystal microdiamonds larger than 50 μm were grown at 900 °C and 1 atm in a plasma-free Ga-In liquid-metal CVD process. Nanosilicon regulated crystal habit, while lowering hydrogen flow increased carbon retention and size. The study produced cubic, truncated-tetrahedral, and octahedral particles.

A horizontal ceramic tube CVD furnace operated at ambient pressure. The growth zone contained a Ga-In eutectic alloy with a 95:5 weight ratio and 99.99% purity. Ferrocene with 98% purity served as the carbon source and sat in a graphite boat at the furnace center.

Defect-rich nanodiamond seeds were 210 nm in size and had greater than 99% purity. Nanosilicon was 50 nm. Both materials were used as received and dispersed through the melt before heating.

After a 15 min purge, the furnace ramped to 900 °C at 10 °C/min. It held for 3 h under N₂:H₂ at 30:100 sccm. The furnace then cooled naturally under flowing nitrogen. Ferrocene decomposition delivered carbon species to the liquid metal interface.

Raman used 532 nm excitation, covered 500 to 2000 cm⁻¹, and had a spectral resolution of about 10 cm⁻¹. SEM imaging used 15 kV. EDS mapped surface elements, while XRD and XPS assessed phase and surface chemistry.

For morphology comparisons, the authors cleaned the crystals in 50 °C HNO₃:H₂SO₄ at 1:1, followed by deionized-water ultrasonication. The general characterization workflow also used ethanol and 10% HCl before drying at 60 °C.

Time-resolved SEM and EDS showed a porous, carbon-rich network on the Ga-In droplets after 0.5 h. At 1.5 h, carbon enrichment increased while Ga, Fe, and Si signals declined. Rough diamond nuclei then appeared.

By 2.5 h, the porous network had reorganized into layered diamond domains with only trace metal. At 3 h, these domains had developed clear facets.

The authors propose that the porous layer is an sp²-rich precursor. Hydrogen etching, Fe-mediated reconstruction, and nanodiamond-seeded stabilization then produce localized sp³ nuclei. Ferrocene supplies carbon and Fe, while the Ga-In melt dissolves and redistributes carbon. The time-resolved measurements support this sequence, but the paper does not directly observe the atomic-scale conversion.

The nanosilicon to nanodiamond ratio set the dominant habit. Without nanosilicon, cubic particles dominated at about 10 μm. At 1:10, truncated tetrahedra became dominant, although a few octahedral grains remained. At 1:1, well-defined octahedra formed most often. A post-treatment comparison still found truncated tetrahedra and octahedra in roughly comparable numbers.

The authors attribute the change to silicon-mediated interfacial kinetics. Transient Si-C and Si-O species alter surface energies and the relative attachment rates of (111) and (100) facets. More silicon improves carbon mobility and stabilizes (111) termination. The evidence shows that nanosilicon changes growth behavior, while its detailed atomic role remains a proposed mechanism.

The main prior liquid-metal reference produced 5 to 10 μm diamonds at approximately 1025 °C. This study used 900 °C and produced several tens of micrometers, with some crystals beyond 50 μm.

Cubic diamonds grew with nanodiamond and ferrocene but no nanosilicon. Their square subunits were stacked along the (100) direction. Raman at 1 h showed a pronounced sp²-rich G band at about 1590 cm⁻¹. After 3 h, the spectrum showed a sharp diamond peak near 1327 cm⁻¹ and an apparent NV⁰ shift near 1430 cm⁻¹.

XRD Fe-C reflections at 43.9° and 50.3° in 2θ disappeared by 3 h. XPS showed dominant sp³ carbon near 283.0 eV. Fe 2p signals appeared at 707.4 eV for Fe-C and 706.6 eV for Fe. The authors interpret Fe as a transient catalyst rather than a stable carbide phase.

Truncated tetrahedra dominated at an NSi:NDD ratio of 1:10. Some crystals carried small pyramidal protrusions on larger facets. Raman shifted from 1319 cm⁻¹ at 1 h to 1331 cm⁻¹ at 3 h. The sharpening was consistent with stress release and improved lattice ordering. The NV⁰ signal appeared at an apparent shift of 1421 cm⁻¹.

XRD showed a weak SiC peak at 35.4° in 2θ during early growth, which faded later. A Si (111) peak at 28.3° persisted. XPS also detected C-O near 286.3 eV and SiO₂ near 103.2 eV. The authors found no evidence that silicon entered the diamond lattice and treated it as an interfacial promoter.

Octahedra dominated at NSi:NDD = 1:1. Raman shifted from 1327 cm⁻¹ at 1 h to 1330 cm⁻¹ at 3 h, with a narrower line. The apparent NV⁰ shift was 1435 cm⁻¹. Weak bands at 938 and 980 cm⁻¹ were assigned to Si-OH groups, probably from oxidation and acid cleaning.

XRD showed a faint Si (111) reflection and diamond peaks at 2θ values of approximately 43°, 75°, 91°, and 119°. XPS showed sp³ carbon near 283 eV and a small C-O component near 286.2 eV. The authors linked complete octahedra to more uniform growth across (111) facets. Mixed octahedral and truncated products show that the ratio does not provide perfect morphology control.

Lowering the hydrogen flow scaled the crystals independently of habit control. With N₂ fixed at 30 sccm, reducing H₂ from 100 to 30 sccm increased net carbon retention at the Ga-In interface. The authors associate this with higher effective supersaturation and faster deposition. Nucleation density increased, and well-faceted crystals grew beyond 50 μm. SEM across the samples generally showed faceted crystallites between 10 and 100 μm.

Control runs without nanodiamond and nanosilicon produced sp²-rich graphene or graphite-like carbon despite the graphite boat. This result assigns the effective carbon supply to ferrocene rather than the boat. Diamonds also formed on both a graphite-boat configuration and a silicon wafer with nanodiamond. The authors report that metal and seed distribution changed droplet formation, nucleation density, and final habit.

Ga-In residue remains a practical measurement issue. Oxide-assisted wetting and hydrophobic trapping can retain metal films or patches on diamond. Such residue can obscure facets and add EDS or XPS signals. Earlier samples grown for 2.5 h or less were not given the mixed-acid cleaning because their porous, hydrophobic surfaces retained little metal. After mixed-acid treatment, EDS found only trace Fe and Si, and carbon dominated the particles. Some untreated samples still carried patch-like Ga-In.

The paper reports no crystal yield, particle size distribution, quantitative defect density, or crystal orientation statistics. Raman sometimes showed a weak, broad band near 1440 cm⁻¹, which the authors associate with minor NV centers possibly inherited from nanodiamond. That signal was not quantified.

For material selection, the measured controls are direct. Use nanodiamond without nanosilicon for cubic particles. A 1:10 ratio favors truncated tetrahedra. A 1:1 ratio favors octahedra, although products remain mixed. Reduce H₂ when size above 50 μm is required. Device engineers should require residue analysis and crystal statistics before transferring the particles to a surface-sensitive application.

View Original Abstract

ABSTRACT We report an ambient‐pressure liquid‐metal‐assisted CVD strategy that enables shape‐programmable growth of microscale diamond by coupling a liquid‐metal Ga-In with ferrocene (Fe(C 5 H 5 ) 2 ) as a carbon precursor, nanodiamond seeds, and nanosilicon. Building on liquid‐metal diamond synthesis, this approach pushes liquid‐metal growth toward lower temperature (900°C, 1 atm) while enabling single‐crystal diamonds to scale from ∼10 µm to several tens of micrometers with well‐developed faceting. Ferrocene decomposition supplies a sustained interfacial carbon flux captured and redistributed by the Ga-In melt toward seed‐rich liquid-solid interfaces. Defect‐rich nanodiamond provides the crystallographic template required for reliable sp 3 nucleation despite low carbon solubility in Ga-In. Nanosilicon plays a complementary role by tuning interfacial kinetics and facet competition, enabling control of crystal habit: cubic (∼10 µm), truncated‐tetrahedral, and fully faceted octahedral diamonds are obtained by adjusting the nanosilicon: nanodiamond ratio, with octahedral crystals reaching ∼50 µm. Crystal size is further scaled by regulating hydrogen flow: lowering the H 2 rate increases carbon retention at the liquid‐metal interface, raises supersaturation, and accelerates diamond deposition. Together, habit control and size scaling establish a practical route for facet regulation and size control under ambient pressure, offering tunable microscale single‐crystal diamonds under mild conditions.

  1. 2021 - Quantum Control of the Tin‐Vacancy Spin Qubit in Diamond