Skip to content

Mechanism of surface quality enhancement in MHz burst femtosecond laser machining of single-crystal diamond revealed by in-situ observation

FieldValue
JournalApplied Physics A
Published2026-04-13
AuthorsYuta Teshima, Reina Yoshizaki, Yuya Kuroda, Naohiko Sugita
DOI10.1007/s00339-026-09581-4
Citations1

The study’s lowest cleaned-groove roughness was Sa = 55 nm at a burst energy of 40 µJ with five sub-pulses. In-situ imaging also showed fewer visible ejecta and less internal darkening as the sub-pulse count increased. The mechanism remains partly interpretive because air ionization and plume chemistry were not measured directly.

Samples were CVD single-crystal diamond measuring 6 mm × 6 mm × 0.5 mm. Each had a mechanically polished {100} surface, and laser irradiation was tangential to that surface.

The laser produced 190 fs pulses at 1030 nm with circular polarization. A 5× objective placed the focus 200 µm below the surface. The calculated beam diameter was 15.7 µm at focus and 25.5 µm at the surface. The Gaussian calculation assumed M² = 1.2 and gave a Rayleigh length of 157.4 µm.

Burst energy took six values: 8, 16, 20, 32, 40, and 80 µJ. The sub-pulse count N ranged from 1 to 10. Sub-pulses were spaced 15.8 ns apart, giving a 63.3 MHz intra-burst repetition rate. Each sub-pulse carried Ep = E/N.

Burst repetition was 10 kHz. The stage moved at 1.0 mm/s over 200 µm and completed five scans. Each condition was repeated four times.

Samples then received 4 min of ultrasonic cleaning. A laser scanning microscope measured Sa over the central 99.25 µm of the groove. The start and end were excluded because of stage acceleration and deceleration. No spatial filtering was applied.

High-speed imaging ran at 1 kfps, so ten bursts occurred between frames.

The main metric was arithmetic mean height Sa on the cleaned groove bottom. The table summarizes the roughness range over N = 1 to 10 and its variability across four repeats.

Burst energyMinimum SaMaximum SaSa standard deviation range
8 µJ169 nm at N = 2277 nm at N = 124 to 61 nm
16 µJ122 nm at N = 4278 nm at N = 114 to 58 nm
20 µJ76 nm at N = 5158 nm at N = 110 to 113 nm
32 µJ110 nm at N = 4170 nm at N = 111 to 68 nm
40 µJ55 nm at N = 580 nm at N = 25 to 11 nm
80 µJ59 nm at N = 278 nm at N = 16 to 20 nm

At 40 and 80 µJ, every Sa value stayed below 100 nm. The standard deviation of Sa ranged from 5 to 20 nm across these conditions. Their groove bottoms were therefore both smoother and more repeatable.

Across 8 to 32 µJ, the standard deviation ranged from 10 to 113 nm. The 20 µJ case was an exception, with an Sa of 76 nm at N = 5 and a standard deviation of 10 nm. Single-sub-pulse runs produced 277 nm at 8 µJ and 278 nm at 16 µJ.

The best count was not constant. It ranged from N = 2 at 8 and 80 µJ to N = 5 at 20 and 40 µJ. More sub-pulses did not always produce a better finish because Ep also decreased.

Michael et al. previously reported the best surface conditions at five sub-pulses for cylindrical single-crystal diamonds. The lowest point here is consistent with that result, although the geometry and conditions differ.

Particle scattering and redeposited debris

Section titled “Particle scattering and redeposited debris”

High-speed images resolved only micrometer-scale ejecta, about 1 µm to several micrometers. Finer debris was outside the camera’s sensitivity.

At 20 µJ and N = 1, about 1 to 4 new particles appeared 9 to 11 ms after machining began. At 80 µJ and N = 1, more than 10 particles appeared at 14 ms.

At 20 µJ and N = 10, almost no scattered particles were visible. Some scattering remained at 80 µJ and N = 10, but it was weaker than under N = 1.

Before cleaning, black particles were common near the groove at N = 1 and N = 2. They were fewer at N = 5 and N = 10. Ultrasonic cleaning removed them completely, linking the in-situ scatterers to redeposited debris.

The debris had aligned and scattered forms. Aligned debris formed a continuous line near the lower groove edge. Scattered debris lay randomly farther from the machined area.

A debris-free zone next to the groove became more distinct at N = 10 than N = 1. The authors associated this clear zone with stronger plasma shielding. At higher N, the deposited debris also appeared finer and more clearly confined to the aligned regions.

Differential images separated surface-connected darkening from isolated internal darkening. At 40 µJ, N = 1 produced repeated internal features. These weakened at N = 2 and N = 5, then became mostly surface-localized at N = 10.

Cracks still formed at 40 µJ and N = 10. One appeared about 45 µm below the surface 1 ms after the first pulse reached the sample. It reached the surface by 3 ms.

The crack advanced at about 45 degrees. Its alignment with diamond crystallography suggests propagation along {111} cleavage planes.

The observed length was about 65 µm over 2 to 3 ms. This gives an apparent average rate of 21 to 33 mm/s.

That rate was not an instantaneous crack-tip velocity. The camera sampled at 1 kfps while bursts arrived at 10 kHz. The growth event covered about 20 to 30 bursts. The measured rate was therefore orders of magnitude below the intrinsic dynamic crack speeds reported for diamond.

Crack growth also differed by direction. Growth in the machining direction reached the surface continuously. Growth against it advanced in two stages. The authors expect crack arrest after irradiation stops, but no stop-restart test was performed.

Ep = E/N separated the observed regimes. The paper estimated Ip as 2.25 × 10^12 times Ep, with Ep in µJ and Ip in W/cm².

At 8 to 20 µJ, roughness varied non-monotonically with per-sub-pulse peak intensity. The optimum near 9 × 10^12 W/cm² lay just below the 10^13 to 10^14 W/cm² scale for pronounced air ionization.

Reducing Ip by increasing N was therefore interpreted as suppressing air ionization and plasma defocusing. Plasma-mediated focusing fluctuations can change the fluence reaching the diamond and produce depth variations in the ablated surface.

Raising N too far reduced Ep toward the cited diamond damage threshold of about 1 to 4 J/cm². Removal then became intermittent, and roughness could rise. This behavior produced the near-threshold processing window.

At 40 and 80 µJ, Sa stayed low across the tested intensity range. The authors linked this weak N dependence to stable removal well above threshold.

Reduced visible scattering also had a possible plume explanation. Later sub-pulses may reheat the first plume and increase ionization. That could promote gas-phase cluster nucleation and finer ejecta.

Plume overlap or direct emission of small clusters could produce the same trend. The camera could not resolve clusters below its micrometer-scale sensitivity, so the plume mechanism was not uniquely identified.

Process limits and engineering implications

Section titled “Process limits and engineering implications”

The evidence supports tuning E/N rather than choosing N alone. At 40 and 80 µJ, the groove bottoms were repeatable across the tested sub-pulse counts. Lower-energy runs required an optimum near threshold.

At low energy, an operating window should include both per-sub-pulse intensity and the repeatability of Sa. A roughness value alone can hide large run-to-run variation.

Burst mode also did not eliminate subsurface damage. The crack at 40 µJ and N = 10 matters for any application that cannot tolerate modified material below the machined surface.

Sa was measured only after ultrasonic cleaning and within the constant-speed groove center. It does not describe debris during processing or edge start-and-stop regions.

The study did not directly measure plasma density or air ionization. Its mechanism for reduced scattering therefore remains a plausible interpretation, not a confirmed cause. Most findings apply to the tested CVD {100} surface, tangential geometry, and laser conditions. Other crystal orientations were not tested.

The study used a 0.5 mm CVD single-crystal diamond plate with a mechanically polished {100} face for femtosecond-laser-machined grooves. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal plates and wafers with thickness from 0.1 to 500 µm and polishing below Ra 1 nm. The engineering support page can be used to specify the required plate geometry and starting surface finish.

View Original Abstract

Diamonds are expected to have widespread applicability across various fields; however, owing to their extreme hardness, precise processing using conventional methods remains challenging. Among the different approaches, laser processing has attracted considerable attention owing to its non-contact nature, which enables machining without inducing subsurface damage, and its flexibility in creating complex, non-planar geometries. Burst pulse laser irradiation has been reported to improve surface quality; however, the underlying physical mechanisms remain unclear. In this study, the machining process of single-crystal diamond {100} surfaces was investigated through in-situ high-speed imaging combined with post-processing microscopic evaluation, to elucidate these mechanisms. The experimental results revealed that increasing the number of sub-pulses per burst generally enhanced the surface quality, particularly under low-energy irradiation, which may be associated with reduced air ionization as the energy delivered by each sub-pulse decreases. In the near-threshold regime, however, the surface quality exhibited a non-monotonic dependence, indicating an optimal processing window; an excessively large sub-pulse count per burst can destabilize material removal. In addition, the amount of optically visible scattered particles decreased and the deposited debris became finer with increasing number of sub-pulses per burst. This tendency may be associated with burst-induced plume interaction, including possible gas-phase cluster nucleation driven by subsequent sub-pulses within each burst. These tendencies are particularly pronounced under low-energy irradiation. The insights obtained from this study significantly advance the understanding of the material removal mechanisms involved in the ultrashort pulse laser machining of single-crystal diamonds. Ultimately, these findings can contribute to establishing robust and precise laser-based processing methods for highly brittle and difficult-to-machine materials.