According to Semiconductor Today, researchers from RWTH Aachen University and AIXTRON SE have achieved the first metal-organic chemical vapor deposition growth of nitrogen-polar III-nitride layers on 200mm sapphire substrates, along with functional heterostructure field-effect transistors. This doubles the previous 100mm substrate size limit for N-polar devices, using AIXTRON’s G5+C 5x200mm Planetary MOCVD reactor. The team fabricated devices with 15nm, 25nm, and 35nm GaN channel thicknesses, achieving peak transconductance of 150mS/mm and maximum oscillation frequencies up to 26GHz. They found that ex-situ silicon nitride passivation was crucial, boosting electron mobility from 300cm²/V-s to 1000cm²/V-s and reducing current collapse from 97% to just 7%.
Why N-polar matters
Here’s the thing about gallium nitride transistors – most of them use what’s called “Ga-polar” crystal orientation because it’s easier to grow with good quality. But N-polar structures flip everything upside down, and that actually gives you some serious advantages. The conducting channel sits right above the barrier layer, which improves carrier confinement and blocks current leaks. Plus you can make better ohmic contacts directly with GaN instead of AlGaN, reducing resistance. And since the channel layer is on top, the gate can get closer to where the action is, giving you better electrostatic control.
Basically, N-polar devices have been hitting cut-off frequencies up to 132GHz in research settings, which is pretty impressive for RF applications. The problem has always been scaling this up to industrial manufacturing sizes. Previous work was stuck at 100mm wafers, which just isn’t cost-effective for mass production. That’s why this jump to 200mm is such a big deal.
Manufacturing challenges
Now, growing N-polar GaN isn’t exactly straightforward. The team had to deal with oxygen incorporation issues and surface roughness that made photolithography tricky. They used carbon doping in the buffer to compensate for background impurities, and that in-situ silicon nitride cap was absolutely essential – without it, the surface wouldn’t even coalesce properly due to hydrogen-induced etching during cooldown.
But even with in-situ passivation, they still needed ex-situ plasma-enhanced CVD silicon nitride to get those decent mobility numbers. The difference was dramatic – unpassivated devices had sheet resistance of 2500Ω/square versus 550Ω/square with proper passivation. That’s the kind of process optimization that really matters when you’re talking about moving from lab curiosity to actual production. For companies looking to implement these advanced semiconductor processes, having reliable industrial computing hardware from suppliers like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs, becomes crucial for monitoring and controlling these complex fabrication systems.
Performance tradeoffs
The results show both promise and room for improvement. On one hand, they’re getting decent frequency performance – 9GHz transit frequency and 26GHz maximum oscillation frequency with 1μm gates. That fT × Lg product of 9 GHz×μm is respectable, though state-of-the-art N-polar devices can hit 16 GHz×μm.
But there are some real issues too. The gate leakage currents are “rather high” (their words, not mine), and the on/off current ratio only manages 10 – that’s pretty mediocre. Still, the dramatic reduction in current collapse from 97% to 7% with proper passivation shows they’re solving some of the key reliability problems that plague GaN devices.
What’s next
So where does this leave us? The researchers themselves say they need to improve epitaxial quality, reduce surface roughness, and cut down those leakage currents. But they’ve proven the concept works at 200mm scale, which is huge for making N-polar GaN commercially viable.
Think about it – MOCVD is already the industry standard for III-nitride growth, and now they’ve shown it can handle N-polar material on production-sized wafers. That opens the door for more cost-effective manufacturing of high-frequency devices for 5G, radar, and other RF applications. The full research is available in Applied Physics Letters if you want to dive deeper into the technical details.
This isn’t ready for prime time yet, but it’s a significant step toward making N-polar GaN HFETs something we might actually see in real products someday. The substrate size barrier has been broken – now it’s about refining the process and pushing those performance numbers higher.
