Skip to content

Sub-melt nanosecond pulsed-laser induced densification and strain-field relaxation in single-crystal diamond

FieldValue
JournalCarbon Trends
Published2026-06-05
AuthorsA. Z. Khan, Tae Jeong Kim, Gabe Guss, Ted A. Laurence, Sonny Ly
DOI10.1016/j.cartre.2026.100664

Sub-melt nanosecond pulsed-laser annealing reduced developed surface area by 60 to 90% in responsive regions. It also produced an effective compaction of about 4 to 6.5 nm in single-crystal CVD diamond. Raman profiles detected strain-related linewidth recovery to a depth of about 2 to 3 µm. STEM and geometric phase analysis did not observe edge dislocations, stacking faults, or grain boundaries in the examined region. The 532 nm process used 8 ns pulses and fluences stepped from approximately 17 to 32 J/cm². Raman and GPA changes are strain measurements, while a lower dislocation density remains an inference.

Two single-crystal CVD coupons, each nominally 8 mm × 8 mm × 0.7 mm, were used. Their as-received surfaces had been mechanically polished and laser diced without CMP. They carried dense polishing striations, edge chatter, micro-step features, and residual particulates. Each coupon was ultrasonically cleaned for 10 min in acetone, then 10 min in isopropyl alcohol, and dried in N2 air.

PLA used a frequency-doubled Nd:YAG source at 532 nm. It delivered 8 ns pulses at 10 Hz to an approximately 1.100 mm 1/e² beam. Sites were placed on 1 mm centers in a 7 × 7 grid for SCD-1 and a 7 × 5 grid for SCD-2. The timing was about 620 ms between shots and 3 s between sites. Fluence was stepped across approximately 17 to 32 J/cm². SCD-1 received one pulse per site, while SCD-2 received two. A fused silica wedge and a profiler at an equivalent plane provided a fluence measurement for each shot.

Surface metrology quantified nanometer-scale compaction

Section titled “Surface metrology quantified nanometer-scale compaction”

Pre- and post-PLA white-light interferometry used the same stage coordinates. The Sensofar VSI setup covered about 1.75 mm × 1.32 mm with a 50 µm Z range and a 10× objective. The authors removed a least-squares plane, masked dust and dropouts, and applied no extra smoothing. They calculated ISO 25178 areal parameters and ISO 25178-2 volumes from the leveled maps. All 30 SCD-1 ROIs were measured before and after treatment.

Sdq measures RMS surface slope, while Sdr is the percentage increase in true area over projected area.

MeasurementReported changeCondition
Surface slopeSdq decreased 45 to 65%Responsive regions
Height spreadSp and Sz decreased 30 to 65%Responsive regions
Void volumeVv decreased 57 to 60%Fluences above the response threshold
Roughness amplitudeSa and Sq decreased 48 to 57%Representative responsive ROIs

The full SCD-1 grid showed a broad trend but some spatial variation. Sa fell at 29 of 30 ROIs, with a median reduction of about 44% and an interquartile range of 34 to 51%. Sq fell at 26 of 30 ROIs, with a median of 37% and an interquartile range of 24 to 51%. Sdq fell at 22 of 30 ROIs, and Sdr fell at 23 of 30. Ten ROIs had at least 40% reductions in Sa, Sq, and Sdq.

The reported effective compaction magnitudes were 4.6 nm at ROI 25 and 5.1 nm at ROI 28 after one pulse. At ROI 25, this change corresponded to 10,626 µm³ over an area of about 2.31 mm². The simultaneous loss of peak, valley, and void volume was consistent with compaction rather than lateral feature movement.

The authors used a Mullins-type surface diffusion relation to interpret the slope and area metrics. Their thermodynamic model treats reduced true surface area and defect-mediated strain energy as driving forces for smoothing. They interpret the volume collapse as compaction of surface-connected free volume.

Raman measured strain recovery below the surface

Section titled “Raman measured strain recovery below the surface”

Pre- and post-PLA Raman spectra were acquired at the same ROIs. Depth scans used 532 nm excitation, a 100× oil-immersion objective with NA 1.3, and a 25 µm confocal pinhole. The depth axis was corrected for refractive-index mismatch and registered with the surface at z = 0. A Gaussian fit to the diamond line near 1332 cm⁻¹ gave peak position, FWHM, and integrated area. A cleaning-only tile served as the depth-matched control.

After PLA, an FWHM below 2.70 cm⁻¹ was reached about 0.5 to 1.5 µm earlier than in the control. At matched depths, treated areas had an FWHM about 0.07 to 0.10 cm⁻¹ narrower. The improvement became indistinguishable from the control at approximately 2 to 3 µm. Peak-position excursions near the surface were shallower and returned to the bulk value sooner.

No G peak was observed, which the authors treat as evidence against graphitization in this processing window. The linewidth change is a direct measurement. Its assignment to dislocation-associated strain remains an interpretation supported by the independent GPA observations.

A cross-section from responsive ROI 28 was prepared by FIB for STEM and geometric phase analysis. The lamella was thinned to about 60 to 100 nm and examined at 200 kV. GPA used atomic-resolution STEM images, one or two non-collinear Bragg reflections, and a fixed strain scale of ±0.02.

The outermost amorphous rim was attributed to FIB preparation. The underlying diamond remained continuous and crystalline. Contrast decayed gradually with depth rather than ending at a sharp interface. The examined region contained no edge dislocations, stacking faults, or grain boundaries. GPA maps showed low-amplitude, smoothly varying strain without prominent dipole-like singularities.

The authors therefore infer that strong near-surface strain concentrators were reduced or redistributed. The limited field of view does not establish that dislocations were absent from the whole coupon. The maps also do not provide a direct dislocation count or a point-defect concentration.

Comparison with melt and ultrafast regimes

Section titled “Comparison with melt and ultrafast regimes”

Prior laser studies discussed by the authors focused on high-fluence melting and ablation or ultrafast pulses for color-center formation. High-fluence nanosecond and excimer processing passes through a molten state and can produce cracks, graphitized rims, or new extended defects if poorly controlled. Those prior regimes differ from the continuous crystalline structure and absent Raman G peak observed after sub-melt PLA here.

The two-pulse case reached 6.5 nm of effective compaction. This compares with 4.6 and 5.1 nm at the one-pulse ROIs. The authors describe the gain as sublinear and attribute it to diminished defect-mediated absorption after the first pulse.

Evidence remains limited to two single-crystal coupons. The full grid statistics apply to the 30 SCD-1 ROIs, while Raman and STEM used selected sites. The study did not directly count dislocations or point defects. It also did not test fracture strength, thermal cycling, charge stability, or spin coherence. The paper does not report specimen temperature or ambient gas during PLA. Fluence windows for other defect states remain unresolved.

Heteroepitaxial and polycrystalline diamond were proposed for future study rather than tested. Wafer-scale treatment and transfer into a CVD workflow were also suggested rather than demonstrated.

For material selection, the results support PLA as a near-surface finishing step for industrial-grade CVD diamond with polishing or laser-dicing damage. The paper reports reduced void volume, smoother topography, and lower strain proxies. It does not establish a reduction in bulk dislocation density.

For power and RF electronics, the authors expect lower stress concentrators to help fracture resistance and thermal-cycle reliability. Direct mechanical testing is still required. For shallow color-center devices, the measured strain relief is relevant to charge stability and spin coherence. Those device metrics were not measured in this study.

The device applications discussed in the paper use single-crystal CVD diamond with a polished, characterized surface for power, photonics, and quantum processing. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal diamond plates and wafers up to 125 mm, with polishing to Ra below 1 nm. The paper’s laser exposure and co-registered characterization requirements can be reviewed through engineering support for the selected plate or wafer.

View Original Abstract

Dislocations and polishing-induced defect networks in synthetic diamond introduce local strain fields and broaden Raman features, limiting performance in optical, thermal, and electronic applications. Laser annealing is emerging as a promising approach to repair surface and near-surface defects in diamond without entering the melt regime, yet surface densification, defect-state modification, and associated structural changes have not been well quantified. In this work, we show that sub-melt nanosecond pulsed-laser annealing (PLA) induces near-surface densification and defect-mediated strain relaxation in single-crystal Chemical Vapor Deposition (CVD) diamond. Single- and two-pulse PLA were applied, and structural evolution was quantified using co-registered ISO 25178 white-light interferometry, depth-resolved Raman spectroscopy, and cross-sectional STEM with geometric phase analysis (GPA). Across a 5 × 6 grid (n = 30), responsive regions exhibit large reductions in local slope (Sdq 45-65%), developed area (Sdr 60-90%), height spread (Sp, Sz 30-65%), void volume (Vv 57-60%), and roughness amplitude (Sa, Sq 48-57%), consistent with densification of ∼4-6.5 nm. Raman profiling shows narrowing of the diamond line and improved spectral uniformity to depths of ∼2-3 μm. Given that the Raman probing depth significantly exceeds the densified layer thickness, this response is interpreted as consistent with long-range strain-field redistribution originating from the near-surface region. STEM-GPA strain maps further support this interpretation, showing smoother strain fields, suppressed hotspots, and redistribution of localized strain concentrations following PLA. These results are consistent with defect-mediated strain relaxation and densification-driven modification of the near-surface energy state. The approach provides a scalable pathway for improving near-surface structural quality in diamond relevant to electronic, photonic, and quantum applications.

  1. 2020 - Diamond semiconductor performances in power electronics applications [Crossref]
  2. 2021 - Materials challenges for quantum technologies based on color centers in diamond [Crossref]
  3. 2024 - Multi-microscopy characterization of threading dislocations in CVD-grown diamond films [Crossref]
  4. 2023 - Reduction of dislocation density in single crystal diamond by Ni-assisted selective etching and CVD regrowth [Crossref]
  5. 2019 - Imaging crystal stress in diamond using ensembles of nitrogen-vacancy centers [Crossref]
  6. 2017 - Characteristic luminescence correlated with leaky diamond Schottky barrier diodes [Crossref]
  7. 2023 - Dislocations as natural quantum wires in diamond
  8. 2020 - Designing ultrahard nanostructured diamond through internal defects and interface engineering at different length scales [Crossref]
  9. 2020 - Direct observation of room-temperature dislocation plasticity in diamond [Crossref]
  10. 2017 - Mode-I fracture toughness prediction of diamond at the nanoscale [Crossref]