Skip to content

Die to wafer direct bonding of (100) single-crystal diamond thin films for quantum optoelectronics

FieldValue
JournalDiamond and Related Materials
Published2026-09-29
AuthorsDominic Lepage, Amin Yaghoobi, Heidi Tremblay, Dominique Drouin
DOI10.1016/j.diamond.2026.114218

Direct bonding of (100) single-crystal diamond films to a 100 mm fused-silica wafer reached 45.1 ± 0.6 MPa dry shear strength. Each bonded element was a 1 × 1 mm × 20 µm nitrogen-doped platelet. The result exceeded the paper’s previous (100) diamond benchmark of 14 MPa. The authors attribute the interface to van der Waals attraction rather than covalent molecular bonding.

The paper’s comparison table places the result above every listed attempt. Earlier (100) direct bonds achieved 0, 2, and 14 MPa. A (111) direct bond reached 35 MPa, and 25 µm polycrystalline fusion bonding reached 32 MPa. Although (111) diamond bonded more readily, most commercial CVD diamond is grown in the (100) direction. Direct bonding was preferred over solder and fusion because it can retain mechanical and optoelectronic performance without restricting device integration. The authors therefore describe 45.1 ± 0.6 MPa as a record for (100) single-crystal diamond.

CVD-grown high-purity single-crystal diamond remains limited to areas of a few mm². Polycrystalline diamond cannot reproduce the optical and quantum properties required for these devices. Nitrogen-doped, quantum-grade CVD single-crystal diamond began as 3 × 3 mm × 500 µm pieces. The pieces were diced and polished into 1 × 1 mm × 20 µm platelets. One crystal yielded 27 usable 1 mm² areas, reducing material use and later thinning. Nitrogen doping served the intended commercial quantum magnetometry application. The authors selected it for high photoluminescence intensity and long spin coherence times.

Laser dicing leaves glue, particles, and graphitic or pyrolytic carbon. The usual 210 °C boiling nitric, perchloric, and sulfuric mixture creates corrosive fumes and fire risk. It can also leave salt residues and complicates platelet retrieval. The replacement sequence began with 30 minutes of sonication at 37 kHz in warm acetone. Platelets then spent 50 minutes in a 50 nm alumina polishing suspension at room temperature. A 1% anionic detergent bath at 70 °C lasted 25 minutes, followed by thorough 70 °C deionized-water rinsing. The 2.8 L tank held eight filters. The authors report clean surfaces suitable for bonding and reproducible nanofabrication.

Fused silica provided low refractive index, high purity without luminescent or magnetic contaminants, and transparency on all sides for photonic addressing. AFM measured RMS roughness of 0.61 ± 0.08 nm for diamond and 0.37 ± 0.10 nm for silica. Contact angles tracked hydroxyl coverage. Deactivated silica reached 73 ± 1°, compared with 4 ± 1° after indirect plasma and 0 ± 1° after UV ozone. Cleaned diamond measured 23 ± 2°, then 16 ± 4° after indirect plasma and 6 ± 1° after UV ozone. The original hydrogen-terminated diamond measured 58 ± 5°.

Deactivation used 350 °C for 2 hours. UV ozone exposure lasted 10 minutes at 65 °C, while indirect plasma ran for 60 minutes. Water vapor was injected at 10 mbar for 5 minutes before the platens closed. The methodology specifies 4 MPa platen pressure. Temperature then ramped at 1 °C/min from 21 °C to 100 °C and held for 1 hour. Samples then stayed in a desiccator for 72 hours. Final annealing used 1 °C/min to 100 °C, 2 °C/min to 250 °C, a 24-hour hold, and a 24-hour cooldown. A PTFE gasket limited pressure between the platelets and top wafer and reduced cleavage risk.

Testing used a Nordson Dage 4000Plus with a 50 kg load cartridge and a modified silicon-carbide-coated blade. No normal compressive load was applied during the shear test. Dry shear reached 45.1 ± 0.6 MPa after indirect plasma and 34.3 ± 0.4 MPa after UV ozone. Deactivated silica produced 30.8 ± 0.4 MPa. The interface was dragged without a force-curve break, rather than separated at a molecular fracture point. One activated platelet cleaved during processing. Its two halves gave 43.1 MPa.

Isopropanol reduced the corresponding shear values to 12.7 ± 0.4, 17.7 ± 0.4, and 18.0 ± 0.4 MPa. Platelets also underwent 15-minute sonication in oxylene, methyl isobutyl ketone, ZED-N50, ZEDN50 developer, and 1% detergent in water. No platelet moved during these liquid exposures. Strong mechanical stress, such as a high-powered water jet, could still move the platelet across the wafer.

The authors separate direct observations from their proposed mechanism. The force trace lacked a fracture break, and deactivated and activated surfaces both carried load. Liquid immersion weakened adhesion. These observations favor van der Waals attraction rather than a network of covalent Si-O-C bonds.

A Lifshitz calculation based on the measured AFM roughness predicted about 40 MPa. This was close to the 30.8 to 45.1 MPa experimental range. The model used a Hamaker constant of 130 × 10^-21 J in air or vacuum. It used 28 × 10^-21 J in water and 16 × 10^-21 J in isopropanol. Breaking bonds over one molecular length produces much larger forces in the molecular model. That model predicts about 17 GPa for C-O and 19 GPa for Si-O.

The authors attribute the difference to mismatched protonation. Diamond hydroxyls are mainly in-plane and have a reported pKa near 10. Silanols include associated groups with pKa 5.6 and isolated groups with pKa 8.5. Their proposed explanation is that C-OH groups do not substitute for Si-OH during water-mediated polymerization. The measured data support this mechanism as the authors’ interpretation, not a direct bond observation.

The demonstrated process covers multiple 20 µm films on 100 mm double-side-polished fused silica. The authors also report using the method on silicon and sapphire, but give no shear data for those carriers. Surface cleanliness and roughness appear more influential than activation chemistry. AFM roughness and the distance dependence of van der Waals forces therefore deserve process control.

The interfaces survived standard nanofabrication steps, but liquid exposure lowers shear strength. One platelet also cleaved during bonding. The paper gives 50 kPa as the general platen pressure but 4 MPa in its methodology. The text does not reconcile the values. It does not report bonded-area yield, long-term cycling, thermal reliability, or residual stress.

For quantum magnetometry, nitrogen-doped single-crystal film remains the material requirement. For broader devices, this work supports direct parallel integration of thin single-crystal diamond on compatible carriers. The measured interface remains vulnerable to liquids and cannot be treated as a covalent bond.

The process needs a polished 20 µm nitrogen-doped (100) single-crystal diamond film for quantum magnetometry. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal material and polishing to Ra below 1 nm. The paper’s 1 × 1 mm geometry and nitrogen doping need confirmation through engineering support.

View Original Abstract

This work unlocks the manufacturing of nanophotonic quantum systems that exploit the unique material properties of single-crystal diamond (SCD). We achieve this by introducing a semiconductor-compatible process for the direct bonding of multiple high-quality, ultrathin diamond films onto a carrier wafer, enabling the subsequent parallel nanofabrication of optoelectronic integrated circuits. Central to this approach is a new diamond surface-preparation method that avoids boiling tri-acid mixtures while producing exceptionally clean 15 and 20 μm thin single crystals. These platelets are bonded side-by-side to 100 mm silica wafers and exhibit a record shear strength of 45.1 MPa for (100)-oriented diamond, surpassing all previously reported bonding attempts. Evidence indicates that the bonding is dominated by van der Waals interactions, likely arising from mismatched protonation mechanisms between Si-OH and C-OH surface terminations, rather than from covalent-bond-driven mechanisms. Despite this non-molecular nature, the heterostructures remain stable through liquid immersions and standard nanofabrication steps. Because the method depends primarily on surface cleanliness and roughness rather than specific chemistries, it is broadly transferable across wafer materials. This capability to parallel-bond ultrathin SCD films onto large-area substrates provides a scalable route to high-performance platforms spanning nanophotonic quantum technologies, high-power electronics, MEMS, and biotechnology.