Skip to content

The Formation Mechanism and Suppression Methods of Dark Features in Optical Polycrystalline Diamond

FieldValue
JournalFunctional Diamond
Published2026-06-21
AuthorsDongyang Yan, Shuai Xu, Manyu Xia, Yipu Qu, Yi Pang
DOI10.1080/26941112.2026.2689119

Reducing silicon wafer scratching from 10 min to 1 min while adding 2.5 sccm O2 during growth reduced the polished dark-feature area fraction from 4.44% to 0.22% relative to the no-O2 baseline. The resulting optical polycrystalline diamond transmitted 70.8% at 10.6 μm and 66.6% at 1.064 μm. Its absorption coefficients were 0.081 cm⁻¹ at 10.6 μm and 0.246 cm⁻¹ at 1.064 μm.

The best growth condition used 500 sccm H2, 30 sccm CH4, 2.5 sccm O2, about 970 °C, and 16 kPa chamber pressure.

Four free-standing polycrystalline diamond samples tested whether oxygen during growth, silicon scratching time, or intermittent hydrogen-oxygen etching reduced dark features. All films grew on the polished side of a (100)-oriented single-crystal silicon wafer. The substrate measured 50.8 mm in diameter and 3 mm in thickness.

The wafer was scratched with 1 μm diamond powder. Ultrasonic cleaning in deionized water, acetone, and alcohol then lasted 5 min for each liquid. After deposition, the silicon was dissolved in a mixed HNO3 and HF solution to release the diamond wafers.

SampleO2 during growthScratch timeIntermittent H2/O2 etchGrowth rate and thickness
1None10 minNo3.2 μm/h and 912 μm
22.5 sccm10 minNo2.3 μm/h and 864.8 μm
32.5 sccm1 minNo3.4 μm/h and 843.2 μm
42.5 sccm10 minYes2.1 μm/h and 882 μm

The microwave frequency was 2.45 GHz, and the microwave power was 6 kW. Growth temperature was measured with an infrared pyrometer. Intermittent etching paused growth every 72 h. The etch lasted 1 h under 700 sccm H2 and 2.5 sccm O2 at constant chamber pressure and power.

Optical microscopy and image-processing software quantified dark-feature area in each sample’s central region. The fractions ranked from highest to lowest were 5.94% for Sample 1, 2.95% for Sample 2, about 0% for Sample 3, and 1.50% for Sample 4.

The samples were cut at the same location to produce 1-inch wafers for polishing. The reported polished-surface fractions were 4.44%, 4.0%, 0.22%, and 3.3% for Samples 1 through 4, respectively.

Sample 3 had no observable dark features at the microscopic scale, apart from light scattering and refraction at grain boundaries. Sample 4 cracked during polishing, which prevented further polishing and left polishing marks at its edge.

Growth structure and suppression mechanism

Section titled “Growth structure and suppression mechanism”

The scratch count per unit area fell from 5964 after 10 min of scratching to 1404 after 1 min. SEM showed that the 1 min treatment produced lower nucleation density and visible voids from incomplete bonding at grain boundaries. The authors state that the lower density reduced competition during initial grain growth.

Cross-sectional SEM showed continuous columnar growth and dense crystals in Sample 3. Sample 1 instead had discontinuous growth, lateral growth, and a rough nucleation surface. The authors explain that lateral growth obstructs neighboring grains and creates voids that support dark-feature formation.

Sample 1 also had more dark features near its edge. The authors associate this region with lower plasma density and insufficient carbon-atom surface migration. They also report that lateral crystal dimensions increased with thickness in all four samples. This created recessed grain-boundary regions that were difficult to etch.

Earlier work cited by the authors linked dark features to competitive growth, twins, and crystallographic orientation. The present comparison shows that preferred orientation alone cannot explain the results, because all four samples had the same dominant orientation.

Oxygen etching alone lowered the central dark-feature fraction from 5.94% to 2.95%, although it reduced growth from 3.2 μm/h to 2.3 μm/h. The authors attribute further suppression to oxygen removing amorphous carbon and graphite during growth.

XRD showed that all four samples had more than 90% (110) orientation at their centers. Sample 4 reached 99%, while Sample 2 reached 92%. Because the orientation fractions were similar while dark-feature concentrations differed, the authors rejected preferred orientation as the sole cause.

Surface EBSD gave a different view of orientation distribution. Sample 1 had 60.4% (110) orientation, compared with 87.4% for Sample 3. Their high-angle grain-boundary fractions were similar at 60.5% and 56%.

Grain size was more uniform in Sample 3. The area fraction of grains larger than 60 μm was 44% in Sample 1 and 29% in Sample 3. Average grain size was 61.45 ± 35 μm for Sample 1 and 47.9 ± 28 μm for Sample 3.

Raman FWHM on the growth side was 5.84 cm⁻¹ for Sample 1 and 3.61 cm⁻¹ for Sample 3. Nucleation-side values across the four samples ranged from 4.07 to 4.76 cm⁻¹. This indicates a decline in growth-side quality for Sample 1 and improvement for Sample 3.

The Raman diamond peak shifted only from 1331.6 to 1333.2 cm⁻¹, which the authors interpret as an absence of significant surface stress. No sp2 peak near 1350 cm⁻¹ was detected. They suggest that dark features may lie mainly inside the sample and may include voids or cracks rather than only amorphous carbon.

Fourier-transform infrared spectroscopy and a UV-Vis-NIR spectrophotometer measured transmission. Relative to Sample 1, Sample 3 increased transmission from 66.7% to 70.8% at 10.6 μm. At 1.064 μm, transmission increased from 47.67% to 66.6%.

Laser calorimetry measured the absorption coefficients of Sample 3. The calculation used a diamond specific heat capacity of 515 J·kg⁻¹·K⁻¹. The measured coefficients were 0.081 cm⁻¹ at 10.6 μm and 0.246 cm⁻¹ at 1.064 μm.

The experiment covered four growth conditions without reported replicate counts or measurement uncertainty. Central-region and polished-surface dark-feature fractions were separate measurements, and the polished comparison used 1-inch pieces cut from 50.8 mm substrates. Detailed absorption coefficients were reported only for Sample 3.

The oxygen mechanism remains partly inferential because Raman did not detect an sp2 phase. The authors attribute the suppression to removal of amorphous carbon and graphite, but their measurements also show incomplete bonding, voids, and cracks. The Sample 4 crack further limits comparison of the intermittent-etch process.

For optical-window selection, the authors’ results support controlling nucleation density, columnar continuity, grain uniformity, and edge growth rather than relying on crystallographic orientation alone. Oxygen alone reduced dark features but lowered deposition rate. Intermittent etching reduced the central fraction to 1.50%, yet Sample 4 grew at only 2.1 μm/h and cracked during polishing.

The 1 min scratch treatment produced the best combined result. Sample 3 had the highest growth rate, the lowest polished dark-feature fraction, and the highest reported transmission at both measured wavelengths. Thickness should still be included in optical-window specifications because grain-boundary recesses increased as the films grew thicker.

The paper’s optical-window application requires polished MPCVD-grown polycrystalline diamond. 6CCVD’s diamond materials catalog lists polycrystalline plates and wafers up to 125 mm, with thickness from 0.1 to 500 μm and inch-size material polished to Ra below 5 nm. The paper’s 843.2 μm Sample 3 is thicker than that listed range, so custom geometry and process translation should be discussed through engineering support.

View Original Abstract

Suppressing the formation of dark features during diamond growth has been a key research focus in the chemical vapor deposition (CVD) of high-quality optical polycrystalline diamond. In this study, the formation mechanism and suppression effect of dark features in diamond under different silicon wafer processing and etching conditions were analyzed. Characterization results from X-ray diffraction (XRD), scanning electron microscopy (SEM), and Raman spectroscopy indicated that the formation of dark features was primarily associated with the quality of crystals, uniformity of grain size, and whether the crystals exhibiting columnar growth. Reducing the silicon substrate scratching duration from 10 min to 1 min, combined with oxygen etching during growth, significantly reduced the concentration of dark features. The shortened scratching duration promoted uniform columnar diamond growth, therefore suppressing the formation of dark features. Meanwhile, the etching of amorphous carbon and graphite by oxygen during the growth process further suppressed the formation of dark features. As a result, by reducing the substrate scratching duration to 1 min and employing growth conditions with gas flow rates of 500 sccm H2 + 30 sccm CH4 + 2.5 sccm O2 at a temperature of approximately 970 °C, the area fraction of dark features on the surface of the diamond optical window was reduced from 4.44% to 0.22%. Its optical transmittance reached 70.8% at a wavelength of 10.6 μm with an absorption coefficient of 0.081 cm−1, and 66.6% at 1.064 μm with an absorption coefficient of 0.246 cm−1.

  1. 2022 - A review on preparation and application of large area polycrystalline diamond films by chemical vapor deposition
  2. 2017 - Effect of boron concentration on the preferred orientation of diamond films deposited using microwave plasma enhanced chemical vapor deposition
  3. 2002 - Deposition and high-oriented growth of diamond films
  4. 2017 - Influence of Ar on nano diamond thin films prepared by MPCVD
  5. 2012 - Effects of process parameters on the preferred orientation of diamond films deposited by MPCVD