Advances in Homoepitaxial Mosaic Single-Crystal Diamond - Interface Stress Regulation
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | Crystals |
| Published | 2026-07-10 |
| Authors | Rong Rong, Jie Bai |
| DOI | 10.3390/cryst16070448 |
Two-inch mosaic single-crystal diamond has reached a reported dislocation density of about 1.0 × 10⁵ cm⁻² and residual stress of approximately 0.04 GPa. The review reports these values for a 50 × 50 mm² wafer, compared with about 10⁷ to 10⁹ cm⁻² in Ir-mediated heteroepitaxial diamond. Aligned step flow and controlled geometry produced the low-stress junction, but the demonstrated mosaic format remains at 2-inch scale.
Mosaic growth breaks the seed size limit
Section titled “Mosaic growth breaks the seed size limit”Natural diamond crystals are generally too small and contain impurity concentrations unsuitable for electronic devices. High-pressure high-temperature synthetic diamonds are usually limited to several millimeters laterally.
Single-seed microwave plasma chemical vapor deposition has reached 12.6 × 13.3 mm². High-pressure epitaxial lateral overgrowth at 165 Torr produced a 59% area increase, while nitrogen-assisted growth produced about a 1.6-fold area gain. Both routes remain at millimeter to half-inch scale, and repeated growth can increase dislocation density.
Mosaic growth joins multiple small seeds before homoepitaxial overgrowth. The process enlarges the usable area without requiring one seed to grow across the full wafer. Its main barrier is crystallographic healing where adjacent growth fronts meet.
Seed preparation sets interface quality
Section titled “Seed preparation sets interface quality”The review favors high-quality CVD homoepitaxial seeds because epitaxial layers on CVD IIa-type substrates have about half the dislocation density of layers on HPHT Ib-type substrates. Lift-off can clone plates from one CVD crystal while preserving orientation. These cloned plates reached dislocation densities as low as 400 cm⁻². Seed defects cannot be removed by later overgrowth because they propagate into the deposited layer.
Laser-cut side faces introduce heat-affected layers, microcracks, and amorphous carbon. These features become nucleation sites and dislocation sources during regrowth. Plasma-assisted polishing removed about 100 µm of waviness from a 20 mm square mosaic substrate at 13.3 µm/h. Atomic force microscopy measured 0.361 nm Sq roughness, while Raman peak positions and widths remained nearly unchanged. The review cites 0.2 to 0.3 nm RMS as the post-growth target for microelectronic fabrication.
Geometry controls coalescence
Section titled “Geometry controls coalescence”Rectangular holes with {100} side surfaces reduced dislocation density from 6 × 10⁶ cm⁻² to 3 × 10⁵ cm⁻². The 20-fold reduction also lowered tensile stress by about 0.14 GPa. This result came from an epitaxial lateral overgrowth study, rather than a full mosaic wafer.
Consistent step-flow direction produced smooth fusion, while mismatched directions caused overgrowth and stress concentration. A 50 × 50 mm² mosaic wafer reached about 1.0 × 10⁵ cm⁻² dislocations and 0.04 GPa residual stress. An earlier 40 × 60 mm² wafer used 24 cloned plates and cracked when seed off-directions were parallel or perpendicular to the edges. Its center-to-edge Raman shift was about 0.5 cm⁻¹, corresponding to an internal stress difference of about 0.17 GPa.
A 60° interface angle produced the smoothest junction in another study, with 0.42 GPa residual stress. Orientation deviations below 1° enabled nearly seamless coalescence after polishing, with Raman full width at half maximum of about 2.83 cm⁻¹ and thermal conductivity above 2100 W/mK. Seed thickness differences of 50 µm or less maintained a smooth interface, while a 100 µm difference produced distinct defects.
Growth parameters regulate morphology and stress
Section titled “Growth parameters regulate morphology and stress”Open substrate holders permit higher growth rates but produce pronounced edge polycrystallization. Closed holders give a more uniform temperature field and suppress nonepitaxial crystallization. In one comparison, the closed holder retained a smooth surface through 24 growth cycles. The open holder developed central cracks after five cycles.
Methane concentration sets the step-flow morphology. Raising CH₄ from 6.0% to 8.0% changed a smooth surface into one covered by polycrystalline and nonepitaxial crystallites. The lowest reported RMS roughness was 0.573 nm at 6.4% CH₄. A reviewed high-quality process used about 3% CH₄/H₂, 1.5 sccm oxygen, and a growth rate near 15 µm/h. Nitrogen addition at 300 Torr produced rates as high as 165 µm/h while preserving good crystal quality. Higher chamber pressure contracts the plasma volume, which limits the area available for simultaneous multi-seed growth.
Three sources of interface stress
Section titled “Three sources of interface stress”Thermal mismatch develops as the wafer cools from roughly 900 to 1000 °C to room temperature. Small orientation differences make adjacent seeds contract by different amounts. Intrinsic growth stress comes from compression when lateral growth fronts meet and from defects formed under nonequilibrium conditions. Structural mismatch arises when amorphous carbon or graphite remains near the seam.
Raman mapping after 24 and 48 hours of growth found that interface full width at half maximum broadened from about 3.0 cm⁻¹ in the bulk to 8.0 cm⁻¹. The diamond peak shifted from 1332.37 to 1331.45 cm⁻¹. The local stress concentration zone extended about 200 to 400 µm from the junction.
For (001) diamond under a biaxial stress assumption, σ = −0.61Δω GPa, with Δω in cm⁻¹. Compressive stress produces a blue shift, and tensile stress produces a red shift. Confocal Raman imaging maps local stress. X-ray topography resolves dislocations and strain fields, X-ray diffraction measures lattice bending through rocking curves, and transmission electron microscopy images interface defects. Chemical etching with scanning electron microscopy counting remains a standard method for quantifying dislocation density. No single method identifies every defect type.
Stress control targets device performance
Section titled “Stress control targets device performance”Process control reduces stress before it reaches the seam. A lateral-to-vertical growth rate ratio above 1 bends dislocations into lateral paths. Oxygen adds OH radicals that etch nondiamond phases, while pulsed growth alternates growth and etching. In-situ high-temperature annealing promotes atomic diffusion across the interface.
Geometry can redirect the defect path. A SiO₂ seam mask lets later growth bypass damaged material, while chamfered seed edges reduce edge stress concentration. Periodic epitaxial lateral overgrowth masks can still create strain modulation with amplitudes typically below 0.1 GPa, which may compromise device uniformity.
The device consequence is measurable. Molecular dynamics simulations found 30% to 50% lower thermal conductivity in severely stressed regions and an order-of-magnitude loss of local heat spreading near the interface. These are simulation results, not direct measurements of every mosaic wafer.
The review also treats strain as a possible control variable. Calculations predict that 3% biaxial tensile strain lowers the B-doped ionization energy from 0.33 eV to 0.12 eV. This is a material-level proposal, not a demonstrated mosaic-interface process.
Limits on scale and process transfer
Section titled “Limits on scale and process transfer”The review reports no new experimental data. Its synthesis identifies seed-cloning throughput, closed-loop monitoring, and scaling from 2-inch to 4-inch wafers as the main manufacturing barriers. In-situ Raman monitoring and machine learning are proposed for process control, but this workflow is not yet established.
Most geometric results depend on the reported growth conditions and reactor configuration. The review questions transferability across MPCVD systems. The coupled effects of interface angle and CH₄ concentration on step flow also need systematic study. Researchers still lack a quantitative relation between interfacial stress size, defect density, and thermal degradation.
Current evidence therefore supports 2-inch-class mosaic diamond when low dislocation density and low residual stress matter more than maximum area. Device qualification still needs to include the junction zones, where defects and thermal resistance concentrate.
Materials and support
Section titled “Materials and support”Power and quantum devices discussed in this work require low-defect single-crystal diamond plates or wafers with controlled orientation and interface stress. The diamond materials catalog includes MPCVD-grown single-crystal plates and wafers up to 125 mm, which provides a size match for the paper’s mosaic development. 6CCVD’s engineering support can review intended thickness, finish, and crystallographic requirements against a selected mosaic design.
View Original Abstract
Single-crystal diamond is regarded as one of the most promising semiconductor materials for next-generation high-power electronic devices, quantum technologies, and extreme environmental applications, owing to its ultra-wide bandgap, exceptionally high carrier mobility, ultra-high breakdown electric field, and excellent thermal conductivity. However, the lateral dimensions of both natural and synthetic single-crystal diamond are limited, which severely restricts their large-scale industrial application. Mosaic growth, in which multiple small single-crystal seeds are laterally arranged and fused at the interfaces through homoepitaxial growth, offers a promising approach to overcoming the size limitation of seed crystals and producing inch-scale single-crystal wafers. This review systematically covers the entire mosaic growth process, including seed crystal preparation, geometric design, growth parameter optimization, and innovative processing methods. Particular emphasis is placed on the mechanisms of interfacial stress generation, along with characterization techniques and stress control strategies. Finally, future perspectives on the fabrication of large-size, low-stress single-crystal diamond wafers are outlined.
Electronic Grade Intrinsic Diamond Epitaxial Wafer H Terminated Superior 100 Mosfet Ready Substrate
View Specs →
Advanced Thermal Simulation Service Diamond Heat Spreader Performance Analysis
View Specs →
High Concentration Quantum Grade Single Crystal Diamond
View Specs →Original source
Section titled “Original source”References
Section titled “References”- 2018 - Thermal Conductivity of High Purity Synthetic Single Crystal Diamonds [Crossref]
- 2008 - Diamond as an Electronic Material [Crossref]
- 2018 - Recent Advances in Diamond Power Semiconductor Devices [Crossref]
- 2016 - Single Crystal Diamond Wafers for High Power Electronics [Crossref]
- 2016 - Diamond Epitaxy: Basics and Applications [Crossref]
- 2020 - Chemical Vapour Deposition Diamond Single Crystals with Nitrogen-Vacancy Centres: A Review of Material Synthesis and Technology for Quantum Sensing Applications [Crossref]
- 2022 - Progress of Structural and Electronic Properties of Diamond: A Mini Review [Crossref]
- 2024 - Recent Progress in Homoepitaxial Single-Crystal Diamond Growth via MPCVD [Crossref]
- 2022 - Past, Present, and the Future of the Research and Commercialization of CVD Diamond in China [Crossref]
- 2022 - Chemical Vapor Deposition Single-Crystal Diamond: A Review [Crossref]