Grafted Low-Leakage Si/AlN p-n Diodes Enabled by Fluorinated AlN Interface

This paper demonstrates that fluorination-induced AlFx formation combined with SiNx passivation effectively suppresses defect-assisted leakage in grafted p-Si/n-AlN heterojunction diodes by removing RTA-induced oxides and stabilizing the interface, thereby enabling low-leakage ultrawide-bandgap power electronics.

The Gist

Imagine you are trying to build a super-efficient, high-speed highway for electricity. The material you want to use is Aluminum Nitride (AlN). Think of AlN as a "super-highway" because it can handle massive amounts of electrical pressure (voltage) and heat without breaking down, far better than the silicon roads we use in our phones and computers today.

However, there's a huge problem: AlN is incredibly sensitive and messy.

The Problem: The "Oxidation Rust"

To make this super-highway work, you need to attach metal wires to it. To get a good connection, you have to bake the metal onto the AlN at extremely high temperatures (like a pizza oven at 1,100°C).

But here's the catch: Aluminum is like a very reactive teenager. As soon as it gets hot, it grabs onto any oxygen in the air and instantly forms a thick, ugly layer of "rust" (oxide) on its surface.

• The Result: This rust layer is full of holes and cracks. When you try to send electricity through your new device, instead of flowing smoothly, it leaks out through these cracks like water leaking through a cracked dam. This is called leakage current, and it ruins the device's performance.

The Old Solutions (And Why They Failed)

Scientists tried two main things to fix this:

1. Washing it off: They tried using strong chemicals (like a harsh detergent) to scrub the rust away.
   • Why it failed: The rust formed during the baking was so hard and crystallized that the chemical couldn't get it all off. Plus, once you scrub it, the fresh AlN surface immediately starts grabbing oxygen again, re-forming the rust.
2. Leaving it alone: They tried to just accept the rust.
   • Why it failed: The rust was too thick and full of defects, causing massive electrical leaks.

The New Solution: The "Fluorine Shield"

This paper introduces a clever new strategy, which the authors call Fluorination. Think of it as a three-step "magic trick" to protect the AlN.

Step 1: The Gentle Scrub (Pseudo-ALE)

Instead of a harsh chemical wash, they use a very gentle, low-power plasma "airbrush."

• The Analogy: Imagine using a soft, targeted air stream to blow away only the top layer of dust (the bad rust) without scratching the paint underneath. This removes the damaged layer and leaves the AlN surface perfectly smooth and ready for the next step.

Step 2: The Fluorine Armor (XeF₂ Treatment)

This is the star of the show. They expose the clean AlN surface to a gas called Xenon Difluoride (XeF₂).

• The Analogy: Think of the AlN surface as a sticky velcro patch that desperately wants to grab onto oxygen (which causes the rust). The Fluorine gas acts like a super-strong, non-stick coating.
• The Fluorine atoms rush in and bond tightly to the Aluminum. These Aluminum-Fluorine (Al-F) bonds are like super-glue. They are much stronger than the Aluminum-Oxygen bonds.
• The Result: The Fluorine covers the surface, blocking oxygen from ever touching the Aluminum again. It's like putting a raincoat on the AlN so it can't get wet (oxidized).

Step 3: The Protective Cap (SiNₓ)

The Fluorine layer is great, but it's a bit fragile on its own. So, they add a tiny, invisible "cap" made of Silicon Nitride (SiNₓ) on top.

• The Analogy: This is like putting a clear, tough glass cover over a delicate painting. It seals the Fluorine layer in, protects it from the air, and gives it enough strength so that when they stick the next piece of the device (a silicon layer) onto it, the connection stays strong and doesn't break.

The Big Win

When the scientists built their diodes (the electrical valves) using this new "Fluorine Shield" method, the results were amazing:

• No Leaks: The electrical leakage dropped by millions of times compared to the old methods. It's like fixing a dam so that not a single drop of water escapes.
• Uniformity: Every single device they made worked exactly the same way, whereas before, some worked and some failed.
• Stability: The device can handle much higher voltages before it starts to leak, meaning it can be used in much more powerful electronics.

Why This Matters

This discovery is a game-changer for the future of electronics. By solving the "rust" problem on AlN, we can finally build:

• Smarter Power Grids: Devices that transmit electricity with almost zero waste.
• Faster Charging: Electronics that charge in seconds.
• Extreme Environment Tech: Devices that can run in the heat of a jet engine or the vacuum of space without failing.

In short, the authors figured out how to give a fragile, high-performance material a "super-suit" made of Fluorine, allowing it to finally live up to its potential as the next generation of electronic power.

Technical Summary

1. Problem Statement

Ultrawide-bandgap (UWBG) Aluminum Nitride (AlN) is a promising material for next-generation high-power and high-frequency electronics due to its wide bandgap (~6.1 eV) and high theoretical breakdown field (>12 MV cm⁻¹). However, its practical application is severely hindered by:

• Surface Chemistry Instability: AlN surfaces are highly reactive and prone to oxidation, even under mild processing conditions.
• Defective Interface Formation: The formation of low-resistance ohmic contacts on n-type AlN requires high-temperature Rapid Thermal Annealing (RTA, ~1100 °C). This process inevitably induces the growth of thick, partially crystalline, and electrically defective aluminum oxide (AlOₓ) layers.
• Leakage Mechanisms: These thermally induced oxides create high densities of interface traps, leading to severe reverse-bias leakage currents via Poole–Frenkel emission (PFE) at moderate fields and trap-assisted tunneling (TAT) at high fields.
• Limitations of Conventional Cleaning: Standard chemical cleaning (e.g., Buffered Oxide Etchant, BOE) fails to remove the crystallized AlOₓ domains formed during RTA, leaving behind a defective interface that degrades heterojunction performance.

2. Methodology

The authors developed a novel interface engineering strategy combining pseudo-atomic layer etching (pseudo-ALE), fluorination, and atomic layer deposition (ALD) to fabricate p-Si/n-AlN heterojunction diodes via a semiconductor grafting process.

• Substrate Preparation: An ~800 nm n-type AlN layer was grown on sapphire via MOCVD.
• Contact Formation: Ti/Al/Ti metal pads were deposited and annealed at 1100 °C (RTA) to form ohmic contacts, inducing surface oxidation.
• Defective Oxide Removal (Pseudo-ALE): A low-power ICP-RIE process using BCl₃/Cl₂ plasma was applied to selectively etch the defective, crystallized oxide without damaging the underlying stoichiometric AlN. This was followed by a BOE dip to remove any residual amorphous oxide.
• Fluorination: The cleaned surface was treated with XeF₂ gas at room temperature. This step replaced surface Al–O bonds with strong Al–F bonds, forming an ultrathin AlFₓ passivation layer.
• Stabilization (ALD): An ultrathin (~0.5 nm) SiNₓ capping layer was deposited via ALD to seal the fluorinated surface, prevent re-oxidation, and provide mechanical stability for the subsequent transfer.
• Device Fabrication: A p-type Si nanomembrane (NM) was released from an SOI wafer and transferred onto the treated AlN surface using a PDMS stamp. The stack was annealed at 350 °C to form chemical bonds.
• Comparison Groups: Three device structures were fabricated and compared:
  • Sample A: RTA-oxidized AlN (Thermal oxide interface).
  • Sample B: Pseudo-ALE + BOE cleaned AlN (Oxide removed, but unpassivated).
  • Sample C: XeF₂-fluorinated AlN with AlFₓ/SiNₓ interface (Proposed method).

3. Key Contributions

• Development of a Low-Damage Oxide Removal Strategy: Demonstrated that pseudo-ALE can effectively remove thermally induced, partially crystallized AlOₓ layers that are resistant to standard wet chemical etching, restoring a near-stoichiometric AlN surface.
• Fluorine-Based Interface Passivation: Established that XeF₂ treatment creates a chemically stable AlFₓ layer. The strong Al–F bond energy (~660–680 kJ/mol) effectively passivates Al dangling bonds and suppresses oxygen adsorption/re-oxidation, outperforming traditional Al–O bonds.
• Robust Bilayer Passivation Scheme: Introduced an AlFₓ/SiNₓ bilayer structure where the SiNₓ layer mechanically stabilizes the fluorinated surface and electronically passivates interface states, enabling successful grafting of Si nanomembranes.
• Mechanism Elucidation: Correlated surface chemistry directly to leakage transport mechanisms, showing that interface engineering can shift the dominant leakage mechanism from defect-assisted PFE to bulk-limited TAT.

4. Results

• Surface Characterization:
  • XPS analysis confirmed that pseudo-ALE reduced the Al–O ratio from ~23.6% (after RTA) back to ~5% (stoichiometric).
  • XeF₂ treatment successfully introduced Al–F bonds, with the AlFₓ fraction saturating at ~8.5% after 30 cycles.
  • AFM showed the pseudo-ALE process maintained a smooth surface morphology (~0.3 nm roughness).
• Electrical Performance:
  • Leakage Reduction: Sample C (AlFₓ/SiNₓ) exhibited a reverse leakage current density of ~10⁻⁸ A cm⁻² at -20 V, which is 4 to 6 orders of magnitude lower than Samples A and B.
  • Rectification Ratio: Sample C achieved an average rectification ratio of ~3 × 10⁷ at ±20 V, compared to ~475 (Sample A) and ~10⁴ (Sample B).
  • Uniformity: Sample C showed excellent device-to-device uniformity, whereas Samples A and B exhibited significant variability due to non-uniform defect distributions.
  • Forward Conduction: Forward bias characteristics (RON) were comparable across all samples, indicating that the interface engineering primarily affected reverse leakage without degrading forward transport.
• Leakage Mechanism Analysis:
  • Sample A & B: Leakage was dominated by Poole–Frenkel emission (PFE) at moderate biases (-5 to -18 V) due to high interface trap densities.
  • Sample C: The onset of PFE was suppressed. Leakage remained low until higher biases, where the mechanism shifted to Trap-Assisted Tunneling (TAT). The transition to TAT occurred at significantly higher voltages, suggesting the remaining leakage is limited by the bulk crystal quality (dislocation density) of the AlN rather than interface defects.
• Microstructural Analysis:
  • TEM and EDS confirmed the formation of a ~5 nm chemically intermixed interface (SiOₓ/SiON/AlFₓ) while maintaining single-crystalline integrity of both Si and AlN.
  • XPS confirmed the presence of Al–F bonds and the suppression of Al–O formation at the interface.

5. Significance

This work provides a critical pathway toward realizing high-performance, low-leakage ultrawide-bandgap heterojunction devices. By solving the long-standing challenge of AlN surface oxidation during high-temperature processing, the authors demonstrate that:

1. Fluorination is a superior passivation strategy for AlN compared to conventional oxide-based methods, offering higher bond stability and larger conduction band offsets.
2. Interface Engineering is the key to unlocking the theoretical potential of AlN, as device performance is currently limited by surface chemistry rather than bulk material properties.
3. The proposed AlFₓ/SiNₓ passivation scheme is robust, reproducible, and compatible with semiconductor grafting processes, paving the way for the integration of AlN with other semiconductors (like Si) for advanced power electronics and optoelectronics.