Skip to content

Molecular dynamics simulation and experimental research on laser-assisted mechanical polishing of CVD diamond coatings

FieldValue
JournalJournal of Materials Research and Technology
Published2026-08-22
AuthorsHaojie Zhu, Rui Shen, Yi Shen, Weijie Cui, Xinchang Wang
DOI10.1016/j.jmrt.2026.08.197
Citations1

Laser-assisted mechanical polishing produced the most uniform and flat HFCVD microcrystalline diamond coating among the three tested routes. The sequence used about 6 W of actual laser output before mechanical fine polishing with 1 μm diamond abrasive. XPS measured 81.66% sp³ carbon on the deposited coating and 23.19% sp² carbon after laser polishing alone. The authors link the improved finish to laser ablation of sharp grains and a proposed softening step before mechanical removal.

How the two-stage process produced the best finish

Section titled “How the two-stage process produced the best finish”

Laser-only polishing removed sharp {111} protrusions by ablation but left periodic grooves, remelted nanospheres, and micro-undulations. Mechanical-only polishing cut grooves into the hard coating and left large grains and gaps. The combined sequence gave the abrasive a flatter, partly transformed surface, so it removed residual irregularities with fewer deep scratches.

The authors attribute the result to preferential asperity ablation and a local transformation from sp³ bonds to sp² bonds. They propose that the transformed layer lowers binding energy and makes the remaining surface easier to polish.

The simulation linked peak removal to bond disorder

Section titled “The simulation linked peak removal to bond disorder”

Confocal microscopy supplied the 3D topography for the rough surface model. The authors extracted grain height H, equivalent edge length L, and neighboring-grain count n from the measurements. Poisson disk sampling placed the grain centers without overlapping grains. A Boolean cut then shaped a Voronoi polycrystal to match the measured height field.

LAMMPS used a Tersoff potential for carbon bonds. The fixed and thermostat layers were 10 Å and 20 Å thick, and the thermostat stayed at 300 K. After 20 ps of NVT relaxation, the surface received three 350 fs pulses separated by 4.5 ps. Lambert-Beer attenuation and a Gaussian spatial distribution defined the laser energy field. The model captured typical asperity distributions and height trends, but it could not reproduce every grain at the measured scale.

Earlier work had confirmed laser-induced amorphization with Raman spectroscopy and TEM. The cited MD examples focused on single-crystal diamond. This work instead follows the rough polycrystalline surface through repeated pulses.

Sharp peaks lost atoms or clusters after the local temperature reached the sublimation threshold. Residual heat raised the starting temperature of each later pulse, so ablation became stronger across the three-pulse sequence. The coordination number fell from 4 toward 2 and 3. Near the ablated surface, atoms with coordination numbers from 0 through 2 increased.

At 540 fs, the radial distribution function retained only short-range correlations. Because the model had not cooled at 540 fs, this result combined transient thermal disorder with laser-induced change. It was not independent evidence of a permanent amorphous structure. The authors suggest that cooling and carbon recombination leave a disordered sp²-rich surface layer with lower binding energy. That layer provides the proposed reason for easier mechanical removal.

Experimental coatings and polishing conditions

Section titled “Experimental coatings and polishing conditions”

Silicon carbide substrates measured 12 × 12 × 5 mm. Before deposition, the authors abraded and seeded the surface with 7 μm diamond grains in glycerol. They then ultrasonically cleaned the samples in acetone for 10 min. HFCVD used methane and hydrogen with a tantalum filament.

StageConditions used
HFCVD nucleation1500 Pa pressure, 4.2% carbon concentration, 0.5 h duration
HFCVD MCD growth3500 Pa pressure, 2.9% carbon concentration, 20 h duration
HFCVD thermal setupFilament at 2100 to 2300 °C, substrate at 800 to 900 °C, filament separation 25 mm, filament-to-substrate distance 12 mm
Laser rough polishing355 nm wavelength, 12 ps pulse duration, 20 μm focal spot diameter, 800 kHz repetition rate, 800 mm/s scan speed. Actual output was about 6 W, equal to 20% of the 30 W maximum
Mechanical fine polishing1 μm mean diamond abrasive in deionized water, 60 rpm polishing stage speed, 50 N normal axial load

The laser was scanned in raster mode to remove coating peaks before the mechanical stage. Laser scanning confocal microscopy and AFM measured 3D morphology and roughness. SEM examined grain shape and coating thickness, while Raman and XPS measured bonding and surface chemistry.

Surface morphology favored the combined route

Section titled “Surface morphology favored the combined route”

SEM showed sharp triangular and tetrahedral grains consistent with {111} planes in the deposited coating. Laser polishing removed the high protrusions and reduced coating thickness, but small grain edges and gaps remained. Mechanical polishing left conspicuous grooves, polishing marks, residual grain structures, and persistent gaps. The combined route produced a more uniform height distribution and fewer deep scratches or grooves.

The authors divide the combined removal into two scales. The laser removes micron-scale asperities and lowers the difference between peaks and valleys. The abrasive then removes residual irregularities from the micron to nanoscale scale through micro-cutting and micro-plastic deformation. The final morphology had better flatness and finish than either route alone. Laser removal of the sharp peaks also reduced local stress concentration before abrasive contact.

Raman spectroscopy detected the initial diamond peak at 1335.6 cm⁻¹, above the stress-free reference of 1332 cm⁻¹. The paper interpreted this shift as compressive stress in the deposited coating. Laser polishing introduced a graphite G peak at 1578.4 cm⁻¹ and broadened the diamond peak. Mechanical-only polishing produced a faint feature near 1600 cm⁻¹, consistent with much less sp² carbon than laser polishing.

The combined route retained disordered sp² carbon. Its spectrum included a D peak at 1350.6 cm⁻¹ and broadened G features at 1567.1 and 1609.3 cm⁻¹. XPS measured 23.19% sp² carbon after laser polishing, while the combined route reduced the sp² fraction without eliminating it. Measured C-O content fell to 1.62% after the combined route.

The Raman and XPS results support partial removal of the laser-formed graphitic and amorphous layer. They do not show that the combined route restored the original sp³ surface.

The molecular dynamics model covered only three pulses to reduce computation time. It also could not reproduce every measured grain at the same geometric scale. The 540 fs radial distribution function was measured before cooling, so it did not establish permanent amorphization by itself.

The experimental narrative compares surface morphology but does not report final Ra or Rq values. It also does not report polishing time, removal rate, replicate counts, or measurement uncertainty. One laser parameter set and one HFCVD parameter set were evaluated, so the paper does not map a process window.

Long-term wear and mechanical-seal performance were outside the reported tests. Some sp²-rich disordered carbon remained after the final mechanical step. That residue must be considered when specifying the finished surface.

Implications for diamond component engineering

Section titled “Implications for diamond component engineering”

Mechanical seals provide a direct design case because the paper links surface roughness to tribological behavior. It also states that poor flatness can raise temperature and cause seal failure. For an HFCVD diamond coating on SiC, laser rough polishing should precede mechanical fine polishing when protruding grains would otherwise concentrate contact stress.

The study’s 6 W output was tied to a 355 nm, 12 ps laser with an 800 kHz repetition rate and an 800 mm/s scan speed. Process transfer must account for pulse energy, focus, scanning speed, coating structure, and abrasive load. Surface acceptance should include morphology, roughness, Raman disorder, and XPS bonding state. A polished finish alone does not establish that the surface chemistry returned to the as-deposited state.

For a mechanical-seal component that calls for a finished polycrystalline diamond surface, the relevant requirements are controlled flatness and low roughness after laser roughing and mechanical fine polishing. 6CCVD’s diamond materials catalog lists MPCVD-grown polycrystalline diamond plates and wafers up to 125 mm, with polishing below Ra 5 nm for inch-size polycrystalline material. engineering support can help compare that specification with the laser-assisted sequence described here.

The study used HFCVD microcrystalline diamond on SiC, so its results do not directly qualify an MPCVD plate.

View Original Abstract

The high hardness and chemical stability of diamond coatings pose a challenge for coating surface polishing. Laser-assisted mechanical polishing is an effective method for polishing diamond coatings, which involves first performing laser rough polishing and then mechanical fine polishing. Laser coarse polishing can significantly improve the polishability of diamond coatings, but its polishing mechanism still needs further clarification. In this paper, molecular dynamics (MD) methods are employed to construct models of the laser energy field and CVD polycrystalline diamond, and to simulate the dynamic evolution of laser polishing polycrystalline diamond. The simulation results show that the extremely high temperature generated by the laser causes the sharp diamond asperities to reach the sublimation threshold and undergo ablation removal in the form of individual atoms or atomic clusters, thereby reducing the surface height differences of the polycrystalline diamond. Furthermore, the diamond surface gradually undergoes a transformation from a long-range ordered polycrystalline state to a short-range ordered amorphous state during laser polishing, which is an important microscopic mechanism for improving the polishability of diamond surface. Subsequently, the micrometer diamond coatings are fabricated using the hot filament chemical vapor deposition (HFCVD) method and the surface were subjected to both laser polishing and mechanical polishing. The surface flatness and finish of the diamond coating achieved through laser-assisted mechanical polishing are superior to those obtained using either polishing method alone. Finally, the mechanisms of material removal and structural transformation in diamond coatings during laser-assisted mechanical polishing are elucidated based on simulation and experimental results.