Why Non-Polar GaN? The 5G and Power Electronics Driver

With the rollout of 5G mobile communications, demand for III-nitride-based power electronics and radio-frequency (RF) devices has surged. Gallium nitride (GaN) offers a unique combination of high breakdown voltage, high saturated electron velocity, and exceptional mechanical hardness — making it the material of choice for high-power, high-frequency, and high-temperature operation.
To date, III-nitride electronic device development has centered almost entirely on AlGaN/GaN heterostructure field-effect transistors (HFETs) grown on c-plane GaN. The strong spontaneous and piezoelectric polarization at the c-plane AlGaN/GaN interface produces a two-dimensional electron gas (2DEG) with a sheet carrier density as high as 10¹³/cm² — all without intentional modulation doping. This is the core advantage of c-plane GaN technology.
Yet polarization is a double-edged sword:

1. Normally-off vs. normally-on dilemma: Strong polarization inherently yields depletion-mode (normally-on) devices, whereas power applications prefer enhancement-mode (normally-off) operation for safety reasons.

2. Limited 2DEG tunability: The 2DEG sheet density at the AlGaN/GaN interface is highly sensitive to polarization direction, making it difficult to engineer the channel density with the controllability afforded by conventional modulation-doped heterostructures.

3. Passivation reliability concerns: Any variation in strain — for example, during SiN surface passivation deposition — directly degrades HFET performance and can compromise long-term reliability.

4. Optoelectronic efficiency bottleneck: In LED quantum wells, the strong polarization field induces the quantum-confined Stark effect (QCSE), spatially separating electron and hole wavefunctions and severely suppressing emission efficiency in the green and amber spectral windows.

The Solution: Non-Polar Growth

Drawing on decades of experience with AlGaAs/GaAs high-electron-mobility transistors (HEMTs), a simple yet compelling approach is to grow modulation-doped AlGaN/GaN heterostructures along a non-polar crystallographic direction. This eliminates the polarization field entirely, allowing the 2DEG sheet density to be tuned simply by adjusting the AlGaN barrier doping — enabling enhancement-mode power devices and restoring high quantum efficiency in long-wavelength LEDs.

Why r-Plane Sapphire?

Although freestanding GaN substrates offer the ideal lattice match, they are prohibitively expensive for mass production. From an industrial cost perspective, r-plane sapphire is an attractive alternative:

1. Far lower cost than freestanding GaN substrates;

1. Compatible with existing MOVPE production lines;

1. Scalable to large-area manufacturing;

1. Crystallographically matched to (11–20) non-polar GaN.

Developing a method to grow high-quality non-polar GaN on industrially compatible substrates such as sapphire or silicon is therefore a critical task for the nitride community.

The Two Hard Problems: Defects and Strain Anisotropy

Directly growing non-polar (11–20) GaN on planar r-plane sapphire, however, confronts two major bottlenecks.

1. Extremely High Native Defect Density

Without any patterning or intervening process, as-grown non-polar GaN on r-plane sapphire exhibits poor crystalline quality:

1. Dislocation density on the order of 10¹¹/cm²;

1. Stacking fault density above 10⁶/cm;

1. X-ray rocking curve full-width at half-maximum (FWHM) reaching 1180 arcsec along [0001] and as high as 1880 arcsec along [1–100].

Dislocations and stacking faults act as leakage pathways and non-radiative recombination centers, degrading LED efficiency, increasing power-device leakage current, and eroding long-term reliability. A dislocation density of 10¹¹/cm² is far above the typical device-grade threshold of ~10⁸/cm².

2. Intrinsic In-Plane Strain Anisotropy

The lattice mismatch between r-plane sapphire and a-plane GaN differs dramatically between the two orthogonal in-plane directions:

1. Along [1–100]: lattice mismatch of approximately 16%;

1. Along [0001]: lattice mismatch of only about 1%.

This enormous mismatch asymmetry produces highly unequal biaxial strain. In the as-grown sample, the compressive strain along [1–100] (–0.291%) is nearly three times that along [0001] (–0.101%).
The consequences cascade through device design and characterization:

1. Crystallographically dependent device performance: Electrical properties vary with in-plane orientation, complicating device layout and design rules.

2. Loss of hexagonal symmetry: Lacking six-fold symmetry, non-polar GaN on a highly mismatched substrate undergoes orthorhombic distortion, rendering the standard c-plane GaN lattice parameter formulas inapplicable.

3. Difficult and error-prone characterization: Accurate determination of lattice constants requires measuring multiple diffraction planes, accounting for the basal-plane angle offset δ, and applying least-squares error minimization.

 

Takeaway: r-plane sapphire is a promising low-cost platform for non-polar GaN, but direct planar growth simultaneously suffers from high defect density and strongly anisotropic in-plane strain. The industry's solution is to pattern the GaN template on the sapphire substrate and exploit epitaxial lateral overgrowth (ELO) to improve both crystal quality and strain distribution in one process.