Skip to content

Scalable Fabrication of 12-inch Polycrystalline Diamond Membranes

FieldValue
JournalNational Science Review
Published2026-10-14
AuthorsXiaoyu Yang, Haonan Li, Zhongqiang Wang, Yang He, 薛生杰
DOI10.1093/nsr/nwag645

A 12-inch freestanding polycrystalline diamond membrane was released in approximately 10 s, producing the largest-area diamond film reported by the authors. Its area was 36 times that of their earlier incomplete 2-inch membranes. The sheet retained circular 12-inch wafer geometry and reached approximately 1700 W m⁻¹ K⁻¹ through-plane thermal conductivity near 30 μm thickness. These results combine wafer-scale MPCVD growth, laser edge-cutting, automatic peeling, and reuse of the silicon growth substrate.

The process starts with a standard 12-inch silicon wafer prepared as the growth substrate. Diamond seeding is followed by an optimized MPCVD growth procedure. After deposition, a laser cuts a narrow annular groove near the wafer perimeter. The groove is approximately 20 μm wide and exposes the diamond-silicon interface while preserving the central membrane region. Edge loss is controlled to less than 1 mm.

The groove also defines the delamination boundary. It limits edge material loss and maintains the circular geometry required by present semiconductor handling systems. Using a standard 12-inch silicon wafer keeps the growth stage within an established wafer format. A photograph compares the released membrane with a 2-inch reference reported in 2024. The authors identify their 12-inch film as the largest-area diamond film reported to date.

A dedicated automatic peeling system replaces manual scribing and hand-assisted exfoliation. It synchronizes roller rotation with wafer translation, which spreads the peeling force across the membrane. The authors say this avoids local stress concentration and reduces operator-induced variability. The complete peeling cycle takes approximately 10 s. The membrane separates intact while its silicon substrate remains available for reuse.

The edge-truncated silicon wafer can return to MPCVD for later growth cycles. Because silicon dominates the consumable cost, reuse reduces amortized material expense. The route also removes extensive grinding and polishing from the release sequence. These choices address the paper’s stated manufacturing bottlenecks of membrane size, yield, and cost.

Substrate reuse makes the release part of a closed-loop growth cycle. The paper presents this sequence as a manufacturing-compatible route because the growth wafer remains available after the diamond film is removed. The automatic equipment controls peeling force over the full membrane instead of concentrating it at manually scribed locations.

Raman spectroscopy samples 16 selected regions arranged in a 4 × 4 array across the membrane. The average diamond first-order optical phonon shift is approximately 1333 cm⁻¹. The average full width at half maximum is approximately 5.9 cm⁻¹. Individual shifts concentrate between 1332 and 1334 cm⁻¹, while full widths concentrate between 5.0 and 6.5 cm⁻¹. The authors interpret these distributions as relatively uniform crystal quality and stress across the full area.

Raman mapping covered the same 16 regions to display crystal quality and stress distribution across the sheet. The narrow spread of both parameters supports the authors’ claim that automatic release preserved structural uniformity at wafer scale.

A weaker, diffuse sp² signal appears between 1450 and 1550 cm⁻¹. It remains weaker than the dominant sp³ signal, which the authors use to assign a low impurity content to the membrane. X-ray diffraction supports the spatial uniformity of grain size and microstrain. The diffraction peak assigned to the diamond 111 plane has negligible spatial dispersion, with a reported position of 43.82° ± 0.01°. Atomic force microscopy of the buried interface shows a smooth, featureless surface without apparent grain boundaries or topographic variation. Average Ra is 0.853 nm over 1 μm² and 1.353 nm over 100 μm².

Together, the Raman, X-ray diffraction, and AFM results indicate limited spatial variation in strain and defects after automatic release. The AFM result applies to the buried interface, rather than the exposed membrane face.

The authors fabricated membranes from 0.25 μm to 50.00 μm thick to study crystal quality and through-plane thermal conductivity. Thermal conductivity rises with thickness because grain growth reduces grain-boundary phonon scattering and improves crystalline order. The increase then gradually saturates. Membranes approaching 30 μm reach a near-saturated conductivity of approximately 1700 W m⁻¹ K⁻¹. Further thickening brings limited performance gain for the added growth time and cost. The authors therefore identify approximately 30 μm as a practical thickness window.

The thickness series spans films far below and above the selected operating region. Its lower endpoint is 0.25 μm, and its upper endpoint is 50.00 μm. The reported saturation therefore rests on a broad thickness study.

For material selection, this result makes thickness a direct tradeoff between thermal performance and manufacturing burden. It also shows why adding more diamond does not produce proportional gains above the saturation region. The supplied text does not specify the thermal-conductivity test method, test temperature, or measurement uncertainty.

The 12-inch format corresponds to a 36-fold area increase over the earlier 2-inch incomplete membranes. It also replaces the manual release used in the authors’ previous small-area approach. The current flow adds automatic force control and retains the silicon substrate after exfoliation. These process changes address the size, yield, and cost limitations identified for diamond thermal materials. The 36-fold comparison describes geometric area, not a measured production yield.

The paper motivates this work with the authors’ assessment that copper, aluminum nitride, and diamond-copper composites cannot always suppress near-junction hotspots. That comparison appears as background rather than a head-to-head test. The extracted results establish membrane size, release time, structural uniformity, and through-plane thermal conductivity. They do not report packaged-device thermal resistance.

Potential uses include chip-backside heat spreaders, through-diamond-via interposers, and thermally conductive insulating interlayers for multilayer electronic systems. The authors also identify optical windows, acoustic diaphragms, and radiation-resistant coatings as application areas. These uses are suggested by the paper and are not demonstrated in the reported fabrication experiment.

Handling remains a practical limitation. The released membrane is thin, flexible, and fragile during transfer and integration. The authors use protective carrier films to support the freestanding sheet. They state that full-line automation, bonding interfaces, and device-level integration still require further optimization. The carrier film supports handling but does not resolve the stated bonding and integration requirements.

The supplied text gives no numerical manufacturing yield, substrate reuse cycle count, or cost per wafer. It also omits the MPCVD gas chemistry, pressure, and deposition temperature needed to reproduce the growth step from the information provided. The material data establish a platform for device engineering, but device performance and manufacturing economics remain unresolved.

The paper’s thermal-platform concept requires freestanding MPCVD-grown polycrystalline diamond as a thermally conductive, electrically insulating heat-spreading layer. 6CCVD’s diamond materials catalog includes MPCVD-grown polycrystalline diamond plates and wafers up to 125 mm, with thickness from 0.1 to 500 μm, so it provides a smaller-format material source for prototypes. For laser edge-cutting and release development, the engineering support can be assessed against the paper’s 12-inch process needs, but the catalog’s stated size limit does not cover that target.

View Original Abstract

Scalable fabrication protocol produces 12-inch polycrystalline diamond membranes via microwave plasma chemical vapor deposition for advanced thermal management.