Barrier-channel intermixing and 2-dimensional electron gas degradation in Al-rich Al(Ga)N/AlGaN high electron mobility transistor heterostructures

This paper addresses the degradation of 2-dimensional electron gas in high-aluminum AlGaN/AlGaN heterostructures caused by high-temperature growth-induced interface intermixing, demonstrating that optimized growth schemes combined with X-ray diffraction analysis can restore sharp interfaces and achieve high-quality 2DEGs with sheet resistivities around 2,500 $\Omega/\Box$.

The Gist

Imagine you are trying to build a super-fast highway for tiny particles called electrons. In the world of high-tech electronics, this highway is called a "2D electron gas" (2DEG). To make this highway work, scientists stack different layers of special materials on top of each other, like a very precise sandwich.

The goal of this paper is to fix a problem where the "filling" of this sandwich gets messy, ruining the highway.

The Problem: The "Melting" Sandwich

The researchers were building a specific type of electronic device using materials rich in Aluminum (Al). To grow these materials properly, they usually need to cook them at extremely high temperatures (around 1,160°C).

Think of the layers in the device as two distinct flavors of ice cream: a hard, cold layer (the barrier) and a softer layer (the channel).

• The Goal: You want a razor-sharp line between the two flavors so the electrons know exactly where to go.
• The Issue: When they cooked the top layer at the usual high temperature, the heat was so intense that the two flavors started to melt into each other. Instead of a sharp line, they got a long, messy gradient where the flavors blended together.

In the paper, they call this "interface smearing" or "intermixing." It's like trying to pour hot chocolate over a scoop of vanilla ice cream and expecting them to stay perfectly separate; the heat makes them swirl together. This blending destroys the "polarization contrast" (the force that pushes the electrons into a fast lane), causing the highway to collapse. The electrons get stuck, and the device stops working.

The Investigation: Finding the Culprit

The team used a special X-ray camera (XRD) to look at their sandwiches.

• The Clue: When the layers were messy, the X-ray images showed a bright, fuzzy streak connecting the two layers. It was like seeing a long smear of paint between two distinct colors.
• The Test: They tried waiting a long time between laying down the bottom layer and the top layer, hoping the "gas" causing the mix would settle. It didn't help.
• The Realization: They realized the heat itself was the problem. The high temperature was causing the atoms to dance around and swap places, blurring the line.

The Solution: Cooking at a Lower Temperature

To fix the melting, they tried a simple trick: turn down the heat.

Instead of cooking the top layer at 1,160°C, they cooked it at a much cooler 850°C.

• The Result: When they looked at the X-ray images again, the fuzzy streak disappeared. The line between the layers became sharp and clean, like a perfectly cut slice of cake.
• Proof: They also used a super-microscope (SIMS) to look at the atoms. They found that at the high temperature, the "blended" zone was about 35 nanometers thick (roughly the width of a virus). At the lower temperature, this messy zone shrank to just 5 nanometers.

Did it Break Anything Else?

Usually, when you cook something at a lower temperature, you worry about it being "undercooked" or picking up dirt (impurities like carbon or oxygen). The researchers checked this carefully.

• The Good News: The lower temperature did not cause more dirt to get into the material. The levels of carbon and oxygen stayed the same. The "undercooked" fear was unfounded.

The Outcome: A Working Highway

Finally, they tested if the electrons could actually run fast on these new, sharp sandwiches.

• High-Temperature Samples: The electrons were stuck. The device had no conductivity (it was like a road with a giant hole in it).
• Low-Temperature Samples: The electrons flowed freely! They measured the resistance and found it was very low, meaning the electrons were zooming along efficiently. They achieved some of the best results ever reported for this specific type of material.

The Takeaway

The paper concludes that if you want to build these high-performance electronic devices using Aluminum-rich materials, you must grow the top layer at a lower temperature. If you use the standard high heat, the layers will melt together, and the device will fail. By cooling things down, they kept the layers sharp, restored the electron highway, and created a working, high-speed electronic component.

They also showed that you don't always need a super-microscope to see this problem; a standard X-ray scan can spot the "fuzzy streaks" that tell you the layers are mixed up.

Technical Summary

Technical Summary: Barrier–Channel Intermixing and 2DEG Degradation in Al-rich Al(Ga)N/AlGaN HEMTs

Problem Statement
High-aluminium content AlGaN/AlGaN heterostructures are critical for developing high-voltage and high-power-density High Electron Mobility Transistors (HEMTs) that utilize ultra-wide bandgap semiconductors. However, a significant barrier to their advancement is the degradation of the two-dimensional electron gas (2DEG) caused by interface smearing between the barrier and channel layers. While Metal-Organic Vapour Phase Epitaxy (MOVPE) requires high temperatures to achieve high crystal quality, these same conditions promote alloy intermixing. This intermixing smoothens the polarization contrast at the interface, severely degrading or completely destroying the 2DEG conductivity. Current literature lacks a comprehensive understanding of this specific intermixing mechanism in Al-rich AlGaN/AlGaN systems and effective strategies to mitigate it without compromising crystal quality.

Methodology
The study employed MOVPE growth using an AIXTRON CCS 3x2 showerhead reactor with trimethylgallium (TMGa), trimethylaluminium (TMAl), and ammonia (NH₃) as precursors. The research utilized a comparative approach involving:

• Substrates: Samples were grown on both standard AlN templates (2.0–2.5 µm thick on sapphire) and bufferless auxiliary substrates (sapphire with a thin 25 nm sputtered AlN nucleation layer) to isolate interface effects from buffer-related strain relaxation.
• Growth Variables: The channel layer (600 nm) was grown at a constant 1,160 °C. The barrier layer (50 nm nominal thickness) was grown at varying temperatures (1,160 °C, 1,000 °C, and 850 °C) and V/III ratios (1,000 and 50) to isolate the cause of intermixing.
• Characterization:
  • X-Ray Diffraction (XRD): Reciprocal space maps (RSMs) of asymmetric 105 reflections were used to assess composition, strain, and interface sharpness. The presence of high-intensity streaks connecting barrier and channel peaks was analyzed as an indicator of graded interfaces.
  • Secondary Ion Mass Spectroscopy (SIMS): High-resolution depth profiling quantified the thickness of the compositionally graded transition layer.
  • Transmission Electron Microscopy (TEM): Cross-sectional imaging provided direct visualization of the heterointerfaces.
  • Electrical Characterization: Contactless sheet resistivity measurements (Semilab LEI-88) were used to evaluate 2DEG conductivity, circumventing the difficulty of forming ohmic contacts on Al-rich materials.
  • Simulation: One-dimensional Schrödinger-Poisson simulations modeled the impact of composition gradients on carrier density and mobility.

Key Findings and Results

1. Temperature-Dependent Intermixing: XRD analysis revealed that growing the Al-rich barrier at high temperatures (1,160 °C) results in a high-intensity streak in RSMs connecting the barrier and channel peaks. This indicates a thick, graded interface (approximately 35 nm) rather than a sharp heterojunction. In contrast, growing the barrier at 850 °C eliminated these streaks, leaving only low-intensity truncation tails, indicative of a sharp interface.
2. Mechanism Verification: The intermixing was confirmed to be thermally driven rather than a result of gas-phase residence time or V/III ratio. A 5-minute growth pause between channel and barrier deposition did not improve interface sharpness at high temperatures, ruling out precursor recirculation as the primary cause.
3. Impact on 2DEG Conductivity: The presence of a graded interface correlated directly with 2DEG collapse. Samples with barriers grown at 1,160 °C showed no measurable conductivity (sheet resistivity above detection limits). Conversely, samples grown with the 850 °C barrier strategy exhibited clear 2DEG conductivity.
4. Performance Metrics: The low-temperature growth strategy yielded sheet resistivities ($R_{sh}$) of approximately 2,500 Ω/□ for AlN/Al₀.₇₅Ga₀.₂₅N structures and 5,500 Ω/□ for Al₀.₇₅Ga₀.₂₅N/Al₀.₅₅Ga₀.₄₅N structures. These values are consistent with the best-reported literature for comparable compositions.
5. Impurity Levels: SIMS analysis confirmed that lowering the barrier growth temperature to 850 °C did not significantly increase carbon or oxygen impurity incorporation, suggesting that the low-temperature approach does not introduce detrimental point defects.
6. Simulation Insights: Schrödinger-Poisson simulations indicated that unintentional composition grading reduces the effective barrier thickness, leading to a significant decrease in total carrier density ($n_{es}$). While grading might theoretically increase mobility by reducing alloy disorder scattering, the reduction in carrier density dominates, resulting in overall conductivity degradation.

Significance and Claims
The paper establishes that standard high-temperature MOVPE growth of Al-rich barriers leads to unintentional barrier-channel intermixing, which is detrimental to HEMT performance. The authors claim that a low-temperature barrier growth strategy (850 °C) is a critical and effective solution to prevent this intermixing and restore high-quality 2DEG conductivity.

Furthermore, the study demonstrates that X-Ray Diffraction (XRD) analysis, specifically the observation of high-intensity streaks in reciprocal space maps, serves as a reliable, non-destructive method for assessing interface sharpness. This allows for the detection of intermixed layers without requiring destructive techniques like SIMS or TEM. The work provides a validated growth protocol for realizing high-performance, Al-rich AlGaN/AlGaN HEMTs, addressing a key material limit in the development of ultra-wide bandgap power electronics.