Skip to content

Numerical analyses of a Brewster-angle window featuring the largest diamond disk to date and an overview of integrity assessment against fracture failure

FieldValue
JournalKITopen
Published2026-10-26
AuthorsGaetano Aiello, A. Meier, T. Scherer, S. Schreck, D. Strauss
DOI10.5445/ir/1000198445

The authors report an optical-grade CVD diamond disk with a 180 mm major axis and a 95 mm minor axis as the largest diamond disk to date for this window application. They numerically verified a 1.1 mm thickness model for a 2 MW, 204 GHz gyrotron load under normal operation and separate 1 bar and 2 bar overpressure cases. The mapped loss tangent changed the absorbed power and stress distribution, while the fracture assessment remains incomplete.

The paper applies a CVD diamond Brewster-angle window to high-power, continuous-wave, electron cyclotron frequency-steering gyrotrons. The selected Brewster angle is 67.2°. The manufactured disk is about 1230 µm thick at its center and 1100 to 1150 µm thick at its edge, although 2 mm was the manufacturing target. The numerical model used 1.1 mm as a lower-bound thickness and represented the entire window without symmetry.

Normal operation combined thermal loads, gravity, and 10 bar coolant pressure. A Gaussian beam entered from the electron cyclotron gyrotron side, while an HE11 mode beam entered from the torus side. The overpressure cases added 1 bar and 2 bar. Modeling the full window allowed the analysis to capture the asymmetric stress and vertical deformation seen in the results.

Loss tangent was measured with Fabry-Perot resonators at 102.4 GHz over a Ø150 mm disk area. One disk side was polished, and the growth side was included in the procedure. The analysis also considered thickness variation. Mapped tanδ values were assigned to the finite-element mesh so that each element received its own absorbed power.

The earlier reference analysis instead used one uniform tanδ of 3.5 × 10⁻⁵ with a 2 mm disk and a 15 mm beam radius. At 2 MW and 204 GHz, that case predicted 1546 W of absorbed power. The slides present this as a separate baseline, not as the final 1.1 mm window analysis.

The loss tangent map was compared with uniform values of 2 × 10⁻⁵ and 3 × 10⁻⁵ under both beam profiles.

Beam modeUniform tanδMapped absorbed powerUniform absorbed powerUniform case versus mapped result
HE112 × 10⁻⁵670 W486 WAbout 13 °C and 14 MPa lower
HE113 × 10⁻⁵670 W729 WAbout 10 °C and 3 MPa greater
Gaussian2 × 10⁻⁵589 W486 WAbout 10 °C and 8 MPa lower
Gaussian3 × 10⁻⁵589 W729 WAbout 22 °C and 8 MPa greater

All comparisons used 20 °C water with a heat transfer coefficient of 5000 W m⁻² °C⁻¹, taken from an earlier CFD analysis. The structural cases included thermal loads only.

At 2 × 10⁻⁵, the uniform case gave absorbed power 184 W below the mapped HE11 result and 103 W below the mapped Gaussian result. At 3 × 10⁻⁵, it gave 59 W and 140 W more than the respective mapped results. The direction of the temperature and stress difference also changed with the uniform value and beam mode. The authors therefore treat the spatial map as a more realistic representation than one uniform loss tangent.

Final HE11 verification used the mapped tanδ plus an added 1 × 10⁻⁵ allowance against possible loss tangent degradation. The absorbed power was 913 W. Normal operation with 20 °C water produced a reported maximum stress of about 85 MPa. Stress and vertical deformation were antisymmetric. The reported stress rose to about 110 MPa with 1 bar overpressure and about 135 MPa with 2 bar overpressure. Submodeling confirmed the 2 bar result.

However, another slide states that tensile-side stress exceeded 200 MPa at 2 bar overpressure. The slides do not explain how that value relates to the about 135 MPa figure. The two values therefore cannot be treated as the same stress measure without further clarification.

The Gaussian verification also used the map plus the 1 × 10⁻⁵ allowance. Normal operation gave 832 W absorbed power, about 80 W below the HE11 case. Its maximum temperature was about 20 °C higher, while its maximum stress was about 75 MPa. Submodeling confirmed that stress value. The authors state that both beam configurations were verified under the stated assumptions.

The fracture work centered on ITER electron cyclotron heating and current drive torus windows made from Ø70 mm, 1.11 mm disks. The authors describe fracture as the main failure mode and note that qualification had relied on tanδ measurements alone. They also note limited published fracture toughness data for microwave plasma assisted CVD diamond.

A specific extended focal imaging procedure was developed for 60 ITER torus disks. It used a single image and a z-stack scan to identify surface features and estimate their depth. The slides report a 322 µm equivalent diameter criterion and a maximum microcrack extension of about 300 µm. A backup table also lists a 447 µm feature in the brazing area.

KIT’s double torsion method used a custom in-house test setup. The specimen geometry used (L = 30) mm, (W = 15) mm, thickness of 1.1 mm, and crack length (a) of about 11 mm. The authors report consistency with the ITER disks and agreement with the limited literature on optical-grade diamond. The method can also provide slow crack growth behavior, but it has not been standardized.

For normal operation with 2 bar overpressure, the analysis used about 84 MPa maximum principal stress on the growth side. It assumed a crack of (a = 160) µm and a first-order geometry factor (\beta = 1). The equation

[ K_I = \beta \sigma_0(\pi a)^{1/2} ]

gave (K_I) of about 1.9 MPa m¹ᐟ². The double torsion result was (K_{Ic} = 8.0 \pm 0.1) MPa m¹ᐟ². The authors therefore report no crack propagation because (K \leq K_{Ic}). The assumed defect had an equivalent diameter of about 320 µm.

This result applies to the stated crack size, stress state, and toughness value. The paper does not establish a safe flaw limit for every crack shape or size, and it presents continued fracture assessment as future work.

The numerical verification covers optical-grade CVD diamond for the analyzed 180 mm by 95 mm window geometry under the stated assumptions. It does not remove the need for optical characterization. Loss tangent was mapped at 102.4 GHz, while the operating beam was at 204 GHz. The slides do not state how the measured map was converted for the operating frequency. Raman spectroscopy and polarimetry remain planned for the large disk.

A design calculation should preserve the measured loss tangent distribution and the actual thickness profile. It should also reproduce the stated coolant temperature, heat transfer coefficient, beam mode, gravity, coolant pressure, and overpressure. A single average tanδ cannot reproduce both the absorbed power and the spatial stress results shown in the paper.

Integrity assessment should include fracture toughness, measured flaw sizes, and slow crack growth rather than tanδ alone. The authors plan continued structural assessment and detailed microscopy of fractured specimens to determine crack paths. Their stated next steps are Raman spectroscopy, polarimetry, continued fracture analysis, and crack path investigation.

View Original Abstract

The diamond Brewster-angle window is a compact broadband window solution for high power, continuous wave (CW), frequency steering gyrotron operations. A large diamond disk with minimum diameter of 180 mm is required for a waveguide aperture of 63.5 mm, compatible with a mm-wave power of 2 MW-CW. In 2020, a first-of-its-kind disk was manufactured with 180 mm diameter and thickness of about 1230 µm at centre and in the range of 1100-1150 µm at the edge. The first part of this work shows thermal and structural analyses run to check the window performance with the large 180 mm disk at the minimum achieved thickness of 1.1 mm. The analyses were carried out using the experimentally measured loss tangent distribution over the disk as input and considering HE11 and Gaussian modes for the torus and gyrotron side of the electron cyclotron (EC) system, respectively. The window design was then verified for normal operation, considering also the off-normal event of overpressure on the torus side. The second part of the work deals with the path to assessing the integrity of the diamond disks on the base of fracture mechanics. Main failure mode of the disks is the failure to fracture. On one side, the fracture toughness of the diamond of interest shall be measured to check the resistance to crack propagation. On the other side, stresses acting on the disks and size and positions of the microcracks generated by the diamond manufacturing process shall be evaluated. The recent developments of this approach are reported with reference to the ITER EC torus diamond disks. These disks feature a diameter of 70 mm and thickness of 1.11 mm.