Breakdown Mechanisms and Mobility Optimization in Ultrathin-Channel AlGaN/GaN HEMTs Based on AlN Template +

AlN templates are available for AlGaN/GaN HEMT epitaxial growth. For detailed technical parameters, please contact us at [email protected].

In the field of high-voltage power electronics, GaN-based HEMTs have achieved commercial application at the 650V class, but scaling up to 1200V and beyond remains a formidable challenge. The voltage-blocking capability of conventional GaN-on-Si structures is constrained by the trade-off among buffer layer thickness, wafer bow, and crystal quality. Although SiC offers a wider bandgap, its cost impedes market penetration. Against this backdrop, the ultrawide-bandgap material AlN (bandgap 6.2eV) has become an ideal candidate for next-generation power devices owing to its theoretical breakdown field as high as 8.4MV/cm and thermal conductivity of 2.9W/(cm·K).

However, simply using AlN as the buffer layer does not directly solve all problems. The ultrathin-channel strategy has emerged: compressing the GaN channel thickness to below 10 nm brings the two-dimensional electron gas (2DEG) closer to the AlN buffer, thereby exploiting AlN’s high breakdown strength to suppress leakage paths in the substrate and buffer. Abid et al. (2019) successfully fabricated HEMT structures on AlN/sapphire templates featuring an 8-nm thin GaN channel and an Al₀.₉Ga₀.₁N barrier layer. The device exhibited a lateral breakdown voltage exceeding 10kV (the measurement system limit) at a contact spacing of 96μm, and a breakdown field as high as 5MV/cm at a short spacing (2μm) – a value far surpassing the theoretical limit of GaN.

Fig. 1 Cross-sectional schematic of the ultrathin-channel Al₀.₉Ga₀.₁N/GaN HEMT (left) and high-resolution TEM image of the MBE-grown active layers (right)

Yet this structure presents a significant contradiction: ultrahigh breakdown voltage coexists with low mobility. Hall measurements showed a 2DEG mobility of only 340cm²/(V·s), far below the typical 1500–2000cm²/(V·s) level in conventional GaN HEMTs. This performance bottleneck restricts the device’s application in high-frequency domains and poses a key obstacle to the practical adoption of ultrathin-channel technology. Bassaler et al. (2022) conducted systematic materials-physics research on this issue, using deep-level transient spectroscopy (DLTS) and transmission electron microscopy (TEM) to uncover the physical origins of the low mobility.

1. Breakthrough in Breakdown Performance: Voltage-Blocking Advantage of the AlN Buffer

Abid et al. (2019) compared two structures: the aforementioned 8-nm thin-channel structure and a 240-nm thick-channel structure (barrier layer with 29% Al content). Lateral breakdown tests were performed in Fluorinert solution to avoid air arcing, revealing dramatic differences.

At a short spacing of 2μm, the thin-channel structure exhibited a breakdown voltage slightly above 1000V, corresponding to a breakdown field exceeding 5MV/cm, whereas the thick-channel structure reached only 270V (<2MV/cm). This difference directly demonstrates the necessity of the ultrathin channel – when the GaN channel thickness is compressed to sub-10nm, the depletion region can extend into the AlN buffer, making the device’s voltage-blocking capability dominated by AlN’s wide bandgap. To the authors’ knowledge, this experiment was the first demonstration of high voltage using a thin GaN channel on an AlN-based HEMT.

At a long spacing of 96μm, the thin-channel structure achieved a breakdown voltage of 10kV (system limit) with a leakage current of only 300nA/mm, indicating that the heterostructure did not suffer from parasitic conduction and that full depletion from the channel down to the sapphire substrate was realized.

Fig. 2 Comparison of lateral breakdown voltage between thin- and thick-channel structures at a contact spacing of 96μm at room temperature

Notably, the breakdown field decreased with increasing contact distance for both structures. Abid et al. attributed this to regrowth interface states activated under extremely high electric fields. This suggests that even with high-quality AlN templates, interfacial defects introduced during MBE regrowth still manifest their influence under long-distance, high-voltage conditions.

2. Physical Origins of Low Mobility: Interfaces, Dislocations, and Traps

2.1 Structural Defects: from Dislocations to Interfacial Steps

TEM analysis by Bassaler et al. (2022) on the same structures revealed key clues. When viewed along the [11-20] plane, the AlN buffer contained numerous threading dislocations along the growth direction, and the GaN channel interface exhibited pronounced step-like roughness on this plane. In contrast, along the [10-10] plane, the channel morphology was more regular with significantly fewer steps. This anisotropic interfacial morphology directly affects 2DEG transport behavior.

Fig. 3 (a) HAADF-STEM images of the sample cross-section along the [11-20] and [10-10] planes; (b) Magnified views of the channel region, showing distinct steps at the AlN/GaN and Al₀.₉Ga₀.₁N/GaN interfaces along the [11-20] plane (marked by arrows), while the [10-10] plane is relatively smooth. The inset shows the few steps still observable along the [10-10] plane.

XRD rocking curves showed that the full widths at half maximum of the AlN (002) and (101) reflections were below 350arcsec and 500arcsec, respectively, corresponding to a dislocation density below 10⁸cm⁻². The GaN layer, being only 8nm thick, gave too weak a diffraction signal to accurately determine its peak width, but the researchers inferred that some dislocations extend into the thin channel. These dislocations are directly related to the 13% lattice mismatch at the AlN/Al₂O₃ interface.

More importantly, a quantitative correlation exists between interfacial steps and mobility anisotropy. Bassaler et al. measured the mobility of Hall bars along different crystallographic directions; the results are summarized in Table 1.

Table 1 Sheet resistance and electron mobility of the thin-channel structure along different crystallographic directions (data from Bassaler et al., 2022, Table II)

Crystallographic direction	Sheet resistance (Ω/□)	Mobility (cm²/(V·s))
[-12-10]	2419 ± 114	136 ± 14
[01-10]	2280 ± 97	144 ± 14
[11-20]	1795 ± 70	183 ± 17
[10-10]	1683 ± 73	195 ± 19

Mobility is highest along the [10-10] direction (195cm²/(V·s)) and lowest along the [-12-10] direction (136cm²/(V·s)). Comparing with the step distribution observed by TEM in Figure 3, the [11-20] plane is precisely the direction with the densest steps, while the perpendicular [10-10] plane has the fewest steps. This correspondence conclusively proves that surface scattering caused by interfacial roughness is a major cause of mobility degradation.

2.2 Deep-Level Traps: Energy Distribution and Scattering Mechanisms of Electroactive Defects

In addition to structural defects, electroactive traps that capture and release carriers are equally critical. Using DLTS, Bassaler et al. detected five electron traps (E1–E5) in the GaN channel, with activation energies ranging from 0.14 to 0.64 eV. By varying the reverse bias (Vᵣ = –5.4V, –6.0V, –6.6V), the researchers distinguished traps activated under different depletion conditions: at Vᵣ = –5.4V (2DEG not fully depleted), only shallow traps E1 and E2 were detected; at Vᵣ = –6.0V (channel begins to deplete), deep traps E3–E5 also appeared.

Fig. 4 (a) Normalized DLTS spectra under different reverse biases (–5.4V, –6.0V, –6.6V); (b) Capacitance as a function of temperature

Analysis of trap origins:

• E1 (0.14 eV): Associated with nitrogen vacancies (Vɴ). Similar traps were observed by Polenta et al. in irradiated GaN.
• E2 (0.24 eV): Attributed to Vɴ–VGa complexes.
• E4 (0.46 eV): Possibly originating from nitrogen vacancies or dislocations themselves.
• E5 (0.64 eV): A typical gallium vacancy (VGa)-related center.

It is worth emphasizing that because the conduction bands of the AlGaN barrier and AlN buffer lie far above that of GaN, the Fermi level remains more than 1eV away from the conduction bands of these wide-bandgap materials. Consequently, all detected traps are located within the GaN channel or at its interfaces. Bassaler et al. proposed that interfacial steps disrupt lattice periodicity, increasing the probability of vacancy formation. These electroactive traps further scatter 2DEG electrons and, together with interfacial roughness, degrade mobility.

3. From MBE to MOCVD: Optimization Pathways for Materials Growth

The key to solving the mobility limitation lies in optimizing growth processes to reduce interfacial defect densities and trap concentrations. Sena et al. (2025) systematically grew AlₓGa₁₋ₓN/AlN double heterojunctions on AlN/sapphire templates by MOCVD, with Al composition x ranging from 0 to 0.70. The resulting AlN buffer layer exhibited excellent quality: an XRD (002) symmetric rocking curve FWHM of only 15arcsec, better than most bulk AlN single-crystal substrates, corresponding to a screw dislocation density as low as 4.8×10⁵cm⁻²; atomic force microscopy gave a surface roughness (RMS) of 0.22nm over a 5×5μm² area, achieving atomic-level flatness.

Fig. 5 Quality characterization of a 1-μm-thick AlN buffer layer on a sapphire substrate: (a) Optical microscopy image; (b) AFM image (5×5μm²); (c) (002) symmetric rocking curve; (d) (103) asymmetric rocking curve.

Sena et al. (2025) further revealed the systematic influence of Al composition in the AlGaN channel on 2DEG transport properties. When x = 0 (pure GaN channel), the 2DEG density reached as high as 3.49×10¹³cm⁻²; as x increased, the 2DEG density slowly decreased, but for most compositions it remained above 10¹³cm⁻². Hall mobility decreased with increasing Al composition, consistent with theoretical expectations based on alloy scattering and increased effective mass. Notably, the mobility of the MOCVD-grown samples reported by Sena et al. was significantly higher than that of MBE samples with comparable compositions, attributed to the lower interfacial roughness and fewer residual impurities achieved by the MOCVD process.

XRD reciprocal space mapping further showed that at x = 0.70, the AlGaN channel was fully strained relative to the AlN buffer (relaxation degree R = 0%); as x decreased, the degree of relaxation gradually increased, reaching 41.1% at x = 0. This indicates that lattice-mismatch stress in low-Al-composition channels is partially relieved by dislocation slip, but may also introduce additional defects.

4. Conclusions and Prospects

Based on the above studies, the following core conclusions can be drawn:

(1) The ultrathin-channel strategy significantly enhances breakdown performance. AlGaN/GaN HEMTs with a sub-10 nm GaN channel on AlN/sapphire templates achieved a lateral breakdown voltage >10kV and a breakdown field of 5MV/cm, demonstrating that the AlN buffer can effectively overcome the voltage bottleneck of conventional GaN-on-Si.

(2) Low mobility arises from synergistic effects of multiple defects. Surface scattering due to interfacial steps, scattering centers provided by threading dislocations, and deep-level traps related to nitrogen/gallium vacancies (E1–E5 detected by DLTS, with E4 associated with dislocations and E5 with VGa) collectively inhibit 2DEG transport.

(3) MOCVD offers advantages in interface engineering. Compared to MBE, MOCVD-grown AlN/AlGaN heterojunctions exhibit lower interfacial roughness and higher mobility, providing a viable route for optimizing material quality.

Therefore, achieving performance breakthroughs in ultrathin-channel devices requires multi-dimensional synergistic optimization: (a) reducing the dislocation density in AlN templates (e.g., by high-temperature annealing or epitaxial lateral overgrowth); (b) optimizing MBE/MOCVD growth parameters to suppress interfacial steps; (c) employing in-situ passivation or doping compensation to neutralize vacancy traps; (d) balancing Al composition and mobility to find the optimal window between breakdown voltage and conduction loss. Moreover, Abid et al. noted that a non-optimized SiN passivation layer limited the transistor-level breakdown voltage to 600 V, indicating that device processing must also be improved concurrently.

Ultrathin-channel AlGaN/GaN HEMTs exemplify a technological path of “breaking device limits through materials innovation.” Although the current mobility issue restricts high-frequency applications, deepening the understanding of interface physics and optimizing growth processes will enable this technology to occupy a significant position in the >1200V high-voltage market.

Whether you need AlN templates for research or for industrial applications, please contact us email at [email protected] and [email protected].

 

References:

1. Abid, I., Kabouche, R., Bougerol, C., Pernot, J., Masante, C., Comyn, R., … & Medjdoub, F. (2019). High lateral breakdown voltage in thin channel AlGaN/GaN high electron mobility transistors on AlN/Sapphire Templates. Micromachines, 10(10), 690.
2. Bassaler, J., Comyn, R., Bougerol, C., Cordier, Y., Medjdoub, F., & Ferrandis, P. (2022). Transport properties of a thin GaN channel formed in an Al0. 9Ga0. 1N/GaN heterostructure grown on AlN/sapphire template. Journal of Applied Physics, 131(12).
3. Sena, H., Lawton Siukola Thurston, M., Meng, C., & Chowdhury, S. (2025). Systematic investigation of AlGaN channels on AlN/sapphire substrates using metal–organic chemical vapor deposition (MOCVD): Toward higher crystallinity and lower surface roughness. APL Materials, 13(5).