Skip to content

Plasma-assisted atomic layer etching of single-crystal diamond

FieldValue
JournalJournal of Vacuum Science & Technology A Vacuum Surfaces and Films
Published2026-04-15
AuthorsJack S. Draney, A. Z. Panagiotopoulos, D. B. Graves
DOI10.1116/6.0005266

Classical molecular dynamics predicts that 5 eV oxygen ions can remove an ion-damaged amorphous carbon layer from single-crystal diamond (100) without etching the underlying crystal. The simulated selective window extends from about 0.2 eV to nearly 5 eV. A proposed cycle using argon followed by oxygen ends with a nearly atomically smooth, fully oxygen-terminated surface. These results describe a process proposal, not an experimental demonstration. The target application is near-surface nitrogen-vacancy center devices, where surface defects can reduce coherence times and add electromagnetic noise.

The authors ran classical molecular dynamics in LAMMPS with a ReaxFF potential fitted through Jax-ReaxFF to over 16,000 density functional theory calculations. The calculations used the PBE functional with D3 dispersion, while argon was modeled with a purely repulsive ZBL potential. Most simulations used a diamond (100) surface. Impacts were normal to the surface and represented by energetic neutrals. This representation assumes that ions recombine through Auger neutralization before reaching the surface.

Each incoming atom started 6 Å above the cell. The impact and collision cascade ran for 0.5 ps. A stratified Berendsen thermostat then ran for 0.5 ps with a 62.5 fs time constant. One monolayer of fluence represents 10^15 impacts per cm2 and corresponds to 72 impacts in the standard cell. The system generally required at least 10 ML to reach steady state, and statistics were usually collected after 50 ML. Common-neighbor analysis separated cubic diamond from amorphous carbon.

The authors call this variant ion damaged layer etching, or IDLE. Unlike conventional plasma ALE, which modifies the surface chemically before argon removal, the proposed sequence uses argon first. Argon creates the amorphous layer, and low-energy oxygen later removes it.

The first step uses argon ions to modify and clean the surface. At 20 eV, the model predicts an amorphous carbon layer of about 1 ML. A 2 ML layer corresponds to roughly 2 to 3 Å. The physical sputtering threshold for diamond (100) is about 30 to 35 eV, so 20 eV irradiation can damage the surface without removing crystalline carbon by sputtering. Most amorphous carbon forms within the first few ML of exposure.

Higher energy gives deeper damage and more removal. In the combined-cycle example, 50 eV argon creates about 4 ML of amorphous carbon. Earlier work cited by the authors used normal 100 eV argon to smooth diamond with about 1 nm roughness. At 100 eV, argon implantation in the amorphous layer is typically a few percent relative to carbon.

Selective etching used O ions, or hyperthermal oxygen atoms, at normal incidence. The amorphous layer in the yield test was several nanometers thick and formed with high-energy argon impacts. Its yield exceeded that of pristine diamond (100) from about 0.2 eV to nearly 5 eV. The authors describe this as effectively infinite selectivity, which would stop the etch when crystalline diamond is exposed. They later define a broader control range of about 0.2 to 10 eV and target about 5 eV.

The prediction agrees with prior oxygen-beam studies. Thermal oxygen near 0.04 eV caused mild smoothing, while 5 and 7.5 eV hyperthermal oxygen did not significantly etch (100) facets but attacked other facets and terrace edges. Those studies support orientation-dependent removal, although the present paper does not model every diamond face.

The modeled cycle began with 20 ML of 5 eV O ions, which chemisorbed a 1 ML oxygen layer. It then applied argon before returning to oxygen.

StepSpecies and energyFluenceSimulated result
Initial coverageO+ at 5 eV20 ML1 ML oxygen on diamond (100)
Surface modificationAr+ at 50 eV25 MLAbout 4 ML amorphous carbon; about 1 ML oxidized carbon removed
Amorphous removalO+ at 5 eVOver 100 MLAmorphous carbon mostly removed after about 25 ML; 1 ML oxygen maintained

The argon step creates about 4 ML of amorphous carbon, etches about 1 ML of oxidized carbon, and removes most adsorbed oxygen. Oxygen exposure then maintains 1 ML coverage while removing the amorphous layer. After about 25 ML, most amorphous carbon is gone. Further exposure continues until carbon etching stops and oxygen coverage no longer changes.

The final surface resembles the crystal surface before bombardment. It has few defects, is nearly atomically smooth, and is completely covered by oxygen. The simulations predict 60% to 80% of the termination in the desired C-O-C ether form. At steady state, about 35% of surface carbon is ketone terminated with a C=O bond. The study reports 95% confidence intervals from three simulation runs.

The authors modeled isolated terraces to test how oxygen repairs the rough amorphous-diamond interface. After 30 and 60 ML of 5 eV oxygen exposure, terrace carbon was removed from the step edge along the underlying C-C bond direction. About 100 ML removed the full initial terrace. The remaining terrace was not perfectly ether terminated.

A separate surface with one missing carbon atom lost its entire row in the [110] direction. After 20 ML, no carbon outside that row had been removed. These results support edge-selective layer removal, but the mechanism remains unresolved. Other crystal faces may etch at different rates and could roughen the surface rather than smooth it.

Plasma delivery is the main experimental challenge identified by the authors. For argon, the DC floating potential is about 5.2 Te. With an electron temperature of 2 to 5 eV, an unbiased surface would receive argon ions at roughly 10 to 25 eV. That range is sufficient for the proposed argon step, and RF bias can add energy when needed.

Controlling oxygen is harder because its energy distribution must remain narrow. A substantial population above 10 eV could etch the underlying crystal and remove self-limiting behavior. The authors suggest source pulsing and electron-beam-sustained plasmas. Such sources can produce densities of 10^11 to 10^12 cm^-3 with surface ion energies of 1 to 5 eV. Fast neutral oxygen atoms also need to form within a few mean free paths of the surface because collisions slow them.

Several limits remain. The process is predicted only by molecular dynamics, and experiments are under way. The simulations cover impacts lasting about 1 ps, so the oxygen termination may change over longer experimental times. The authors also note that current best practices use many hours of vacuum annealing followed by hot acid cleaning, yet shallow NV coherence remains below bulk values. Plasma results therefore require reporting of ion flux, radical flux, and energy distributions.

The proposed sequence offers separate controls for cleaning and atomic layer resolution. Higher argon energy or fluence can remove more material, smooth a rough surface, and sputter clean contamination. Lower argon energy limits substrate removal when near-single-layer control is the priority. Oxygen energy then sets whether the amorphous layer is removed selectively.

For near-surface NV centers, usually placed less than 10 nm from the surface, the target is a smooth (100) crystal with one oxygen layer and minimal processing defects. Ether termination is preferred for these quantum devices. The model suggests that energy-controlled argon and oxygen could create a controlled damage-and-removal sequence, but no NV coherence, experimental etch rate, or long-term surface stability has yet been measured.

The paper targets single-crystal diamond (100) for near-surface NV-center device development and needs a controlled, oxygen-terminated surface. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal plates and wafers up to 125 mm, with thickness from 0.1 to 500 micrometers and polishing to Ra below 1 nm. The linked engineering support page provides custom fabrication support for specifying the starting material and surface plan.

View Original Abstract

Applications of near-surface nitrogen-vacancy (NV) centers in diamond are often limited by surface defects created during processing. Understanding and controlling plasma-induced surface damage is important for preserving the optical and spin properties of diamond NV centers. We report molecular dynamics simulations of a novel form of plasma-aided atomic layer etching of diamond. In this proposed scheme, the initial surface modification step consists of Ar+ ion bombardment. This creates an amorphous layer at the surface, the thickness of which is controlled by the ion energy. Amorphization is known to help smooth the surface, at least locally, and would also serve to sputter clean an initially contaminated surface. A second step impacts the amorphous carbon layer with O+ between about 1 and 5 eV. Simulations show that this energy range will remove the amorphous carbon but will not etch the underlying diamond. Though this energy range is not trivial to achieve in a conventional low-temperature plasma, potential methods for creating these low-energy O+ ions are discussed. In addition to a-C etching, the O+ impacts are predicted to remove any isolated (100) diamond terraces by selectively attacking the edges of the terraces. The proposed atomic layer etching (ALE) approach reverses the conventional ALE sequence in which the surface modification step is usually chemical modification (oxidation in this case), followed by a removal step using Ar+ impacts. This plasma ALE procedure is predicted to create a diamond surface that is atomically flat and defect free.