Chapter 2-SiC Coating Corrosion Resistance Data and Lifetime Optimization Practices in MOCVD Epitaxy

Author: Dr. Zhang, Senior Engineer, Semicera Semiconductor Materials
15 years of experience in CVD coating and hot-zone material R&D, having led coating selection and failure analysis projects for more than 30 MOCVD/PVT customer production lines, and participated in formulating the company’s internal SiC/TaC coating quality control specifications.

I. Engineering Background

In low-to-medium temperature epitaxial processes such as GaN MOCVD, UV LED, and gallium oxide epitaxy, the SiC coating on the surface of graphite susceptors and trays is exposed long-term to atmospheres of NH₃, HCl, and H₂, and must withstand heating/cooling thermal cycles in every furnace run. A common issue engineers encounter: after about 100+ uses, imported trays first show a rise in particle counts at the edge region, forcing premature scrapping — but few people immediately clarify whether this is chemical corrosion or thermal-stress cracking that triggers it first.

II. Mechanism Analysis: Why Does the Edge Fail First?

SiC coatings have good chemical inertness toward common MOCVD atmospheres such as HCl, NH₃, and H₂. This is rooted in the high covalent bonding character and large bond energy of the Si-C bond (about 451 kJ/mol); typical MOCVD process temperatures (900–1150°C) are far below the critical range (starting at 1550–1650°C) at which significant volatilization/decomposition occurs in a hydrogen atmosphere. Therefore, under most low-to-medium temperature epitaxial conditions, the chemical corrosion rate of the SiC coating body itself is actually quite slow. What truly determines lifetime is often not “the coating being consumed by gas,” but rather another, more hidden failure chain.

When SiC coatings are deposited by CVD, the reactive gases exhibit a concentration gradient along the flow direction inside the chamber; regions near the chamber walls consume precursor faster and have a thicker boundary layer, causing the coating thickness at the tray edge to be inherently lower than at the center — with measured deviations commonly reaching 20%–30%. Thinner regions have lower heat capacity, so local temperature fluctuations are noticeably amplified during the heating/cooling cycles of each furnace run — this corresponds directly to the stress formula given in Chapter 1, σ ≈ E·Δα·ΔT/(1−ν): the larger ΔT and the thinner the coating, the higher the thermal stress accumulated at the interface, and the more likely micro-cracks are to nucleate preferentially at the columnar grain boundaries at the edge.

Once a micro-crack appears, the problem shifts from “thermal stress” to a compound failure of “chemical corrosion + particle contamination”: the crack provides a direct pathway for atmospheres such as NH₃ and HCl to reach the graphite substrate, and the local corrosion rate becomes noticeably higher than at a dense coating surface; at the same time, debris generated by crack propagation is itself a direct source of rising particle counts. In other words, the root cause of edge failure is usually not “a harsher chemical environment at the edge,” but rather “a thinner edge coating with larger temperature fluctuations and earlier crack formation” — chemical corrosion is the result that gets amplified after the crack opens, not the initial trigger.

III. Data Support

Take as an example a tray optimization project Semicera completed for a UV LED customer in Q3 2024: the customer’s MOCVD process temperature was about 1100°C, and the original imported SiC-coated tray lifetime was stable at around 120 uses, with cracks always appearing at the edge first. We performed on-site sample analysis at the customer’s production line using cross-sectional SEM, and the results showed the coating thickness differed by nearly 30% between the center and edge, with the edge region noticeably thinner (this customer is under an NDA; production line and equipment model information has been anonymized).

Data source note: The following thickness, lifetime, and particle count data all come from joint testing records by the Semicera lab and the customer’s production line (eddy-current thickness gauge + cross-sectional SEM + inline particle counter), not issued by a third-party testing agency, and are for reference only; actual values will vary across different furnace types and atmosphere ratios — it is recommended to rely on your own production line measurements.

Under this process condition, the optimized SiC coating can reliably withstand more than 150 thermal shock cycles at ΔT = 1100°C without noticeable cracking. The core variable driving the lifetime improvement is not a change in material grade, but the narrowing of thickness uniformity and the increased thickness margin at the edge. Based on the customer’s 6 MOCVD units, the annual tray procurement cost savings total more than $100000.

IV. Practical Recommendations

Actionable troubleshooting and improvement steps engineers can implement directly:

• ● During incoming inspection, use eddy-current thickness gauging or cross-sectional SEM to sample at least 3–5 points at the center, mid-ring, and edge; the thickness uniformity target should be controlled within ±5% of the nominal value — batches exceeding this tolerance should not be accepted;
• ● For structurally thinner areas such as the edge and near the gas inlet, selectively increase the coating thickness margin (e.g., from the standard 100 μm to 130–150 μm) rather than uniformly thickening the entire coating, to balance cost and thermal stress control;
• ● Establish a weekly or monthly particle count trend curve, focusing on whether the slope enters a sustained upward phase, rather than only responding once values exceed the limit;
• ● When a micro-crack first appears at the edge, simultaneously check the measured temperature fluctuation amplitude in that region (verifiable with thermocouples or infrared thermometry), rather than defaulting to attributing it to the coating batch or material issue;
• ● Include thickness uniformity testing alongside the incoming GDMS purity report as acceptance criteria; either one exceeding tolerance should trigger a re-inspection.

V. Applicable Boundaries and Limitations

It should be noted that the “edge thickening to 130–150 μm” approach given in this article is mainly aimed at low-to-medium temperature MOCVD processes in the 900–1150°C range (such as GaN MOCVD and UV LED), on the premise that the atmosphere is primarily H₂/NH₃ with no abnormal leakage. If the HCl concentration in the process atmosphere is significantly higher, or other strongly corrosive components are present, the relationship between coating thickness margin and corrosion rate needs to be re-validated, and this article’s data should not be directly applied.

In addition, local thickening is not without cost — coating thickness is not simply “the thicker the better.” Excessive thickness increases internal coating stress and deposition cost. The 130 μm figure in this article’s case is the result of repeated tuning for that specific customer’s operating conditions (furnace type, gas flow distribution, original thickness baseline); the optimal thickness for different production lines needs to be determined based on that line’s own measured data — directly copying a fixed number carries risk.

VI. FAQ

1. Why do SiC-coated trays typically develop cracks at the edge region first?

This is mainly related to the edge region’s coating being thinner and having larger local temperature fluctuations. Stress produced by thermal cycling tends to nucleate micro-cracks preferentially at the columnar grain boundaries at the edge, after which the cracks further intensify corrosion and particle contamination.

2. How much does SiC coating thickness uniformity affect MOCVD tray lifetime?

In the case documented here, optimizing thickness uniformity from about ±15% to within ±5% increased tray lifetime from about 120 uses to about 180 uses, and reduced particle count per furnace run by about 40%.

3. What is the recommended thickness for MOCVD SiC coatings?

For the 900–1150°C low-to-medium temperature MOCVD process in the documented case, the thickness margin can be increased in structurally thinner areas such as the edge — for example, from 100 μm to 130–150 μm. However, thicker is not always better; the optimal thickness needs to be validated based on furnace type, gas flow distribution, and actual operating conditions.

4. Does a continuous rise in particle count necessarily mean the SiC coating purity is a problem?

Not necessarily. In the documented case, GDMS testing showed total impurity content <5 ppm, consistent with a qualified batch. It was ultimately found that the main issue came from significant temperature fluctuations near the gas inlet and micro-cracks in the coating. Therefore, particle anomalies need to be investigated across the coating, hot zone, and gas flow distribution simultaneously.

VII. Downloadable Resources

For engineers’ convenience, Semicera has compiled the following materials, available by leaving your contact information:

• ● “SiC Coating Thickness Uniformity Incoming Inspection Checklist” (including thickness measurement point diagram and acceptance criteria)
• ● “SiC Coating Corrosion Reference Data Sheet for Common MOCVD Atmospheres (NH₃/HCl/H₂)”
• ● “Particle Count Trend Warning Threshold Setting Template” (ready-to-use Excel)

If you’re not sure whether your production line’s current coating thickness uniformity meets the standard, feel free to contact Semicera’s technical team for a free process evaluation.