6-Inch Silicon-Based AlN Thin Film Customization Case: How MOCVD Solved the Heteroepitaxial Stress-Cracking Problem？

Summary: From 500nm Down to 100nm, Delivered On-Spec in One Pass Within 3-4 Weeks

Core takeaway: What the client originally wanted was a 500nm-thick film. What we ultimately delivered was a validated, risk-controlled 100nm solution — the point of this case isn’t just hitting a spec, it’s surfacing the risk before production started.

A company that develops precision semiconductor inspection equipment came to Semicera’s Wafer Business Division needing an AlN thin film grown on a 6-inch silicon wafer via MOCVD, for experimental validation of a core module. The client initially wanted a 500nm film thickness, chasing stronger performance. After our process team evaluated the request, we found that this thickness carried a high risk of cracking on a silicon substrate, so we recommended switching to 100nm and delivered on that basis.

From inquiry to delivery took 3-4 weeks total. Film thickness uniformity, internal stress, and surface quality all met spec, and the client passed validation on the first try with no rework needed.

Client Background: Why an Inspection Equipment Company Came to Us for Custom AlN Film

Core takeaway: The companies that come to us for non-standard thin-film customization are often not wafer fabs themselves, but — as in this case — inspection/metrology equipment developers. What they need isn’t volume supply; it’s a batch of structurally stable, traceable experimental sample wafers.

The client is a company focused on precision semiconductor inspection equipment, whose core business is providing high-precision inspection solutions for the wafer manufacturing process — in short, the equipment fabs use to “check up” on their own products’ yield, defects, and film performance often comes from companies like this.

It was the client’s R&D department that reached out to us, working on a new generation of inspection modules. Before such a module can go to market, it has to run through a full round of selection validation and performance testing on real functional thin-film samples — what’s being tested isn’t “is the film good,” but “can our inspection equipment accurately measure whether the film is good.” In other words, the quality of the sample itself directly determines how trustworthy the inspection data is; if a sample has a problem, every conclusion drawn from that round of R&D validation comes into question.

Customers like this tend to share a few common purchasing characteristics:

• Low volume: usually just a few to a few dozen pieces per order, far below a fab’s standard batch threshold;
• Consistency requirements are actually higher: because these are baseline samples used for precision validation, any single piece with a micro-crack or abnormal stress can distort an entire round of experimental data, pushing back the project timeline — and the cost of rework far exceeds the value of the material itself;
• Parameters are often non-standard: the thickness or substrate combination the client specifies is frequently derived by working backward from theoretical performance targets, and isn’t necessarily the most process-stable option.

According to client feedback, some suppliers they’d approached before either declined outright because the volume was too small, or simply “took the client’s parameters as given” — no process feasibility assessment, just accepting the order as long as it could be produced. This approach looks efficient in the short term, but it leaves the risk with the client: if there’s a hidden problem in the parameters themselves, it only surfaces later during equipment validation, by which point the cost of troubleshooting and lost time is much higher.

That’s precisely why many research institutes, university labs, and precision optics/inspection equipment companies, when looking for a wafer material supplier, aren’t really looking for a “standard catalog product” — they’re looking for a partner that genuinely understands the process, is willing to make a judgment call before accepting an order, and is willing to tell a client “this parameter carries risk” when necessary. That’s where this collaboration started.

The Challenge: Why a 500nm Target Was a Hidden Risk for Silicon-Based AlN Film

Core takeaway: 500nm is already close to the critical thickness for a silicon-based AlN system. Forcing it through might not crack immediately, but delayed failure could still show up later during the client’s equipment validation or storage — a risk the client wasn’t initially aware of.

The client’s original technical requirement was clear: grow an AlN thin film via MOCVD on a 6-inch silicon wafer, target thickness 500nm. The reasoning was practical too — a thicker film theoretically means stronger piezoelectric response and dielectric properties. Taken purely from the experimental goal, there was nothing wrong with the request.

But from a process-feasibility standpoint, there was an easily overlooked risk hiding in it:

• AlN and silicon have a significant lattice-constant mismatch and differing thermal expansion coefficients — this is a classic heteroepitaxial system;
• Internal film stress keeps accumulating as thickness increases; once it exceeds the critical thickness, the film warps, develops micro-cracks, and in severe cases the whole wafer cracks;
• If a batch develops a structural defect, the loss isn’t limited to that batch of material — if wafers with hidden defects make it into the client’s equipment validation stage, the downstream test data becomes unreliable, and the client may make R&D decisions based on faulty data, a cost far higher than the material itself.

Producing directly to the client’s original parameters would have looked like meeting the requirement, but it would actually have shifted the risk downstream into the client’s own experimental process.

The Solution: Going from 500nm to 100nm — How We Adjusted the MOCVD Process Parameters

Core takeaway: We didn’t simply produce to the original 500nm spec, and we didn’t simply reply “we can’t do it” either. Instead, we ran a risk assessment first, walked the client’s engineering team through the reasoning, and then settled on 100nm together.

After receiving the request, Semicera’s Wafer Business Division technical team handled it in three steps:

Step 1: Process Feasibility Assessment

Based on a stress-accumulation model for AlN/Si heteroepitaxy, combined with the team’s prior MOCVD thin-film growth process data, we predicted the cracking risk at 500nm and confirmed that, under the existing silicon-substrate conditions, this thickness carried a clearly elevated risk with no way to guarantee stable yield.

Step 2: Transparent Communication with the Client’s Engineering Team

We shared the basis for our risk assessment, the possible failure modes, and alternative options with the client’s engineering team, so they could adjust their experimental design based on real process data rather than theoretical performance targets alone. Both sides held several online meetings during this period and jointly confirmed that 100nm was the optimal range for balancing film structural integrity, stress controllability, and the client’s experimental goals.

Step 3: Optimizing Process Parameters for the 100nm Target

We made targeted adjustments to key MOCVD process parameters — growth temperature, V/III ratio, buffer-layer conditions, and others — to ensure stable, uniform film growth on the 6-inch silicon wafer while maintaining surface quality and batch-to-batch consistency.

500nm Initial Request vs. 100nm Final Solution

Results: Delivered in 3-4 Weeks, Film Uniformity and Stress Control Met Spec in One Pass

Core takeaway: The real result of this collaboration wasn’t just three metrics meeting spec — it was that the upfront risk conversation helped the client avoid a hidden problem that could have distorted their experimental data.

•  Film thickness uniformity, internal stress, and surface quality all met the client’s validation standards — passed testing on the first try, no rework needed;

Delivery timeline: from inquiry to delivery took 3-4 weeks total — a fast turnaround for non-standard custom thin-film material.

• Building on this, the client went on to complete performance validation for the related module, giving them a reliable data foundation for their subsequent R&D plans.

This is also a principle Semicera’s Wafer Business Division consistently holds to when working with research institutes, universities, and precision equipment companies: professional judgment comes before “taking every order as specified.”

Frequently Asked Questions

Q1: Why not just deliver the 500nm the client wanted — doesn’t hitting the number prove the technology is strong?

Actually, it’s the opposite. A thicker film doesn’t by itself demonstrate technical strength — being able to grow it is only the first step; being able to grow a defect-free film stably and consistently across batches is what actually matters. In a heteroepitaxial system with as much mismatch as AlN/Si, 500nm already exceeds the stress range this silicon substrate can reliably withstand. Even if it doesn’t crack immediately, delayed failure can still show up later. Giving the client a validated, feasible thickness range demonstrates more professionalism than simply doing what was asked.

Q2: Besides 6-inch silicon wafers, can you make AlN film on other sizes or substrates?

Yes. Semicera supports MOCVD thin-film customization on 2/4/6/8-inch and other sizes of silicon, sapphire, and other substrates. The achievable thickness and process window vary depending on substrate type and crystal orientation. We recommend telling us your application needs first (substrate type, target thickness, downstream process) — we’ll run a process feasibility assessment before quoting.

Q3: What’s the typical minimum order quantity and lead time for this kind of non-standard custom film?

Semicera’s Wafer Business Division primarily serves research institutes, university labs, and precision optics/inspection equipment companies, so we’re well set up for small-batch, multi-spec R&D-type needs — we accept orders starting from a single wafer. Lead time depends on process complexity; the 3-4 weeks in this case is a reasonable reference point for a typical custom order.

Q4: Besides AlN, what other compound semiconductor thin-film materials does Semicera’s Wafer Business Division work with?

Beyond AlN, we also handle custom MOCVD growth for GaN, AlGaN, and other III-nitride thin films, as well as other functional thin films, and we support PVD-based customization as well. We can match substrate type and process route to the specific application — piezoelectric, dielectric, optical, buffer layer, and so on.

Q5: What if I’m not sure exactly what film thickness parameters I need?

That’s exactly why we recommend talking it through before production. Many clients base their requests on theoretical performance targets, but the actually achievable thickness is affected by the substrate, process route, and other factors. Semicera’s Wafer Business Division will first run a process feasibility assessment based on your experimental goals and material system, then recommend a range that balances performance and stability, before moving into formal production.

Conclusion and Next Steps

Core takeaway: The value of custom wafer materials isn’t just about whether something can be made — it’s about whether we can give validated, professional recommendations that balance process feasibility with the client’s actual needs.

The value of custom wafer materials isn’t just about whether something can be made — it’s about whether we can give validated, professional recommendations that balance process feasibility with the client’s actual needs.

If your team is also looking for a supplier of compound semiconductor thin films, heteroepitaxial materials, or other non-standard custom wafer materials, feel free to reach out to Semicera’s Wafer Business Division — we’ll start with a process feasibility assessment based on your specific application, then provide a solution that’s genuinely deliverable.