🔬 applied physics

Microwave Performance of all MOCVD-grown AlScN/GaN MIS-HEMTs on Semi-Insulating GaN Substrates

This paper reports the successful design, fabrication, and comprehensive microwave characterization of long-gate, all MOCVD-grown AlScN/GaN MIS-HEMTs on semi-insulating GaN substrates, demonstrating robust high-voltage operation, excellent interface quality, and superior RF performance including high output power, efficiency, and low noise figure.

Original authors: Can Cao, Vijay Gopal Thirupakuzi Vangipuram, Abdul Mukit, Motahareh Helli, Jinwoo Hwang, Hongping Zhao, Wu Lu

Published 2026-04-07

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Can Cao, Vijay Gopal Thirupakuzi Vangipuram, Abdul Mukit, Motahareh Helli, Jinwoo Hwang, Hongping Zhao, Wu Lu

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are building a super-fast highway for tiny electronic cars (electrons) to travel on. This paper is about building a new, better type of highway using a special material called AlScN/GaN, and proving that it works perfectly even when built with standard, mass-production tools.

Here is the breakdown of their achievement using simple analogies:

1. The Goal: A Better Highway for Electronics

In the world of electronics, we need devices that can handle high speeds (like 5G and 6G) and high power (like radar or satellite transmitters). The current "gold standard" for these highways is a material called GaN (Gallium Nitride).

However, the researchers wanted to upgrade the "guardrails" of this highway. They used a special mix called AlScN (Aluminum Scandium Nitride). Think of AlScN as a super-charged guardrail. Unlike the old guardrails, this one creates a much stronger "magnetic push" that forces more electronic cars to travel in a single, tight lane. This means more traffic can flow faster with less friction.

2. The Challenge: Building it in a Factory

There was a problem. While scientists could build these super-highways in a high-tech, expensive laboratory (using a method called MBE), they couldn't easily build them in a regular factory using the standard, cheaper method called MOCVD (Metal-Organic Chemical Vapor Deposition).

It's like being able to bake a perfect cake in a Michelin-star kitchen, but failing to make it in a standard home oven. The researchers wanted to prove they could bake this "perfect cake" in the standard oven.

3. The Innovation: The "MIS" Gate (The Traffic Cop)

The biggest issue with the standard factory-built version was that the "traffic cop" (the gate that turns the device on and off) was too leaky. Electrons were sneaking through when they shouldn't, causing the device to waste energy and overheat.

To fix this, the team built a MIS-HEMT.

Analogy: Imagine the old gate was a wooden fence with holes in it. The new MIS (Metal-Insulator-Semiconductor) gate is like a solid steel door with a rubber seal. It completely stops the leaks.
The Result: The device turned on and off with incredible precision. It had a "subthreshold swing" of 63 mV/dec. In plain English, this means the switch is so sharp it's almost perfect, wasting almost no energy when switching states.

4. The Foundation: A Smooth Road

They built this highway on a special type of ground called a Semi-Insulating GaN substrate.

Analogy: Most roads are built on gravel or dirt, which causes bumps and potholes (defects). These researchers built their highway on perfectly smooth, solid concrete.
Why it matters: Because the ground was so smooth (low defects), the electronic cars didn't get stuck or slowed down. This allowed the device to handle high voltages (63 Volts) without breaking, which is crucial for power applications.

5. The Performance: The Test Drive

The team put their new device through a rigorous test drive, and the results were impressive:

Speed (RF Performance): Even though the gate was relatively long (1 micrometer, which is like a wide lane), the device could still switch on and off 25.8 billion times a second (25.8 GHz). If they made the gate shorter (like a race track), it could theoretically go even faster (up to 170 GHz).
Power (The Engine): When they pushed it hard, it could output 4.04 Watts of power for every millimeter of width. That's like a small engine producing the power of a large truck relative to its size.
Efficiency: It converted about 23% of the energy it used into useful work. While not perfect, it's a great start for a factory-built device.
Quietness (Noise): One of the most unique findings was how "quiet" the device was. It produced very little static noise (below 2.5 dB).
- Analogy: Imagine trying to hear a whisper in a crowded room. This device is so quiet that you can hear the whisper clearly, even when the room is loud. This is vital for sensitive communication equipment.

6. The Big Picture: Why This Matters

This paper is a milestone because it proves that AlScN/GaN isn't just a lab curiosity.

Before: We could only make these high-performance devices in expensive, slow labs.
Now: We can make them in standard factories (MOCVD) with high quality, low leakage, and great performance.

In summary: The researchers successfully built a "Ferrari engine" (high-performance AlScN/GaN) using a "Toyota factory" (standard MOCVD process). They proved that this new material system is robust, efficient, and ready to power the next generation of 6G communications, radar, and space technology.

1. Problem Statement

While Aluminum Scandium Nitride (AlScN) has emerged as a superior barrier material for Gallium Nitride (GaN) High-Electron-Mobility Transistors (HEMTs) due to its strong polarization and lattice-matching capabilities, most high-performance demonstrations have relied on Molecular Beam Epitaxy (MBE), which is not scalable for mass production.

Manufacturing Gap: Metal-Organic Chemical Vapor Deposition (MOCVD) is the industry standard for GaN production, but MOCVD-grown AlScN/GaN devices have historically suffered from high off-state leakage currents and lower on/off ratios compared to MBE-grown counterparts, primarily due to Schottky gate limitations.
Device Architecture: There is a need to demonstrate that MOCVD-grown AlScN/GaN heterostructures, when combined with a Metal-Insulator-Semiconductor (MIS) gate structure and high-quality semi-insulating (SI) GaN substrates, can achieve the low leakage, high breakdown voltage, and excellent microwave noise performance required for next-generation high-power and high-frequency applications (e.g., 6G, space, quantum devices).

2. Methodology

The researchers designed, fabricated, and characterized long-gate (1 µm) MIS-HEMTs to isolate material quality and interface electrostatics from short-channel effects.

Epitaxial Growth (MOCVD):
- Substrate: Ammonothermal, Mn-doped semi-insulating (SI) GaN substrates with low threading dislocation density (~2×10⁵ cm⁻²).
- Structure: 100 nm unintentionally doped (UID) GaN channel, followed by an AlN/AlScN barrier layer.
- Precursors: A customized high-temperature setup was used with (MCp)₂ScCl (heated to 120°C) as the Sc precursor, TMAl for Al, and TMGa for Ga.
- Gate Stack: An in-situ SiNₓ cap (~2 nm) was deposited immediately after AlScN growth, followed by a total barrier thickness of ~11 nm. This results in an Equivalent Oxide Thickness (EOT) of 4.6 nm.
- Composition: The Sc composition was estimated to be up to ~13%, with a graded profile to maintain strain.
Device Fabrication:
- Process: Direct-write optical lithography, ICP-RIE mesa isolation (175 nm depth), HF wet etching to remove the SiN cap, and Ti/Al/Ni/Au ohmic contacts (annealed at 865°C).
- Gate: Ni/Au Schottky gates were deposited, followed by 200 nm PECVD SiN passivation.
- Dimensions: Gate length ( $L_g$ ) = 1 µm, Gate-Source ( $L_{gs}$ ) = 0.5 µm, Gate-Drain ( $L_{gd}$ ) = 0.9 µm.
Characterization:
- Structural: STEM imaging, EDX mapping, and Reciprocal Space Mapping (RSM) to verify crystallinity and strain.
- Electrical (DC): Transfer characteristics, breakdown voltage, pulsed I-V (to measure current dispersion), and C-V/G-f measurements for interface trap density ( $D_{it}$ ).
- RF/Microwave: Small-signal S-parameters (1–45 GHz), Large-signal load-pull at 10 GHz, and Noise Figure measurements (2–18 GHz).

3. Key Contributions

First MOCVD-grown AlScN MIS-HEMTs on SI GaN: Demonstrated a fully MOCVD-grown device architecture that overcomes the leakage issues typical of Schottky-gate MOCVD AlScN devices.
Interface Quality: Achieved an exceptionally low interface trap density ( $D_{it} \approx 2.11 \times 10^{11} \text{ cm}^{-2}\text{eV}^{-1}$ ) and a near-ideal subthreshold swing, validating the quality of the MOCVD-grown AlScN/GaN interface.
Comprehensive Microwave Noise Data: Provided the first reported microwave noise performance for AlScN/GaN HEMTs, showing competitive noise figures suitable for low-noise amplifier (LNA) applications.
Scalability Validation: Proved that the material system supports high-voltage operation and low current dispersion, establishing a baseline for future scaling to shorter gate lengths (sub-100 nm) for mm-wave applications.

4. Key Results

Structural & Material Quality:

STEM and RSM confirmed the AlScN/AlN barrier is fully strained to the GaN channel with no strain relaxation, despite the ~11 nm thickness.
Surface roughness (Ra) was measured at 0.372 nm, indicating smooth morphology.
Hall measurements showed a sheet carrier density of $1.12 \times 10^{13} \text{ cm}^{-2}$ and mobility of 813 cm²/V·s.

DC Performance:

On/Off Ratio: $2 \times 10^5$ (an order of magnitude improvement over previous MOCVD AlScN reports).
Subthreshold Swing (SS): 63 mV/dec (near the Boltzmann limit), attributed to low $D_{it}$ and excellent gate control.
Breakdown Voltage: 63 V (3-terminal), corresponding to a critical electric field of 0.75 MV/cm.
Leakage: Off-state gate leakage was limited to ~ $10^{-3}$ mA/mm.
Current Dispersion: Extremely low at 7.8% at 10 V, indicating suppressed trapping effects due to the dual SiN passivation (in-situ and PECVD).

RF & Microwave Performance:

Small-Signal:
- Cut-off frequency ( $f_T$ ): 25.8 GHz.
- Maximum oscillation frequency ( $f_{max}$ ): 51.1 GHz.
- These values were achieved with a relatively long 1 µm gate, suggesting that scaling to 150 nm could theoretically yield $f_T \approx 170$ GHz.
Large-Signal (10 GHz Load-Pull):
- Output Power Density ( $P_{out}$ ): 4.04 W/mm.
- Power-Added Efficiency (PAE): 22.7% (Class-AB operation at $V_d=30$ V).
Noise Performance:
- Minimum Noise Figure ( $NF_{min}$ ): < 2.5 dB over a wide drain current range (100–700 mA/mm) at frequencies below 6 GHz.

5. Significance

This work represents a critical milestone in the commercialization of AlScN/GaN technology. By successfully transitioning from MBE to MOCVD and utilizing a MIS gate on semi-insulating substrates, the authors have demonstrated that AlScN/GaN can meet the rigorous requirements for both high-power and high-frequency applications.

High-Power: The high breakdown voltage (63 V) and low current dispersion make these devices ideal for power amplifiers in 5G/6G infrastructure and radar.
High-Frequency: The competitive $f_T/f_{max}$ and low noise figure confirm the material's potential for mm-wave and sub-THz communications.
Manufacturability: The use of MOCVD and standard lithography processes suggests a clear path toward mass production, addressing a major bottleneck in the adoption of AlScN technology.

The results confirm that the unique polarization benefits of AlScN can be fully leveraged in a manufacturable process, paving the way for next-generation electronics that require high efficiency, high power density, and low noise.