Impact of AlN Layer Thickness on Interface and Electrical Performance of ALD‑AlN on c‑Plane GaN

Abstract

We examined how the thickness of atomic‑layer‑deposited (ALD) AlN influences the interfacial and electrical characteristics of Pt/AlN/n‑GaN MIS diodes. A 7.4‑nm AlN layer produced the highest interface and oxide‑trap densities, whereas a 0.7‑nm film exhibited dominant Al–O bonding and no clear AlN peak in XPS spectra. The presence of oxygen throughout the AlN layer contributed to interface‑state accumulation, especially for the thicker films. Thermionic‑emission (TE) analysis described the forward‑bias behavior of the 7.4‑nm device, while the 0.7‑nm sample required a different transport model. Reverse‑bias leakage for both thicknesses was best described by Fowler–Nordheim (FN) tunneling rather than Poole–Frenkel (PF) emission, underscoring the role of barrier inhomogeneity and defect‑assisted tunneling in these structures.

Background

III‑nitride semiconductors, with their wide bandgaps, high electron saturation velocities, and robust breakdown fields, are pivotal for both optoelectronic (LEDs, laser diodes, UV detectors) and high‑power electronic devices (HEMTs, power switches) [1–4]. Achieving low‑density, high‑quality metal/semiconductor interfaces is essential to minimize electron traps that pin the Fermi level and hinder barrier modulation [5, 6]. AlN, boasting a 6.2‑eV bandgap, high thermal conductivity, and a lattice mismatch of only 1.1 % with GaN, is a compelling passivation candidate. ALD of AlN at ~300 °C offers CMOS compatibility and enables the growth of uniform, high‑k dielectric layers [12–15]. However, ALD‑grown AlN often contains substantial oxygen impurities, which can degrade electrical performance [15, 16]. While high‑k oxides such as Al₂O₃ and HfO₂ have been employed as passivation layers, they can introduce deep interface states via Ga–O bond formation [19]. In contrast, thin AlN layers have been shown to modulate barrier heights in GaN contacts [20–23], yet systematic studies on thickness‑dependent interfacial properties remain sparse. This work addresses that gap by depositing AlN films of 0.7, 1.5, and 7.4 nm on n‑GaN and evaluating their interfacial chemistry and electrical behavior.

Methods

Materials and Device Fabrication

Commercial c‑plane bulk GaN wafers (300 µm, 5 × 10¹⁴ cm⁻³, 1.5 × 10⁷ cm⁻² threading dislocation density) were diced into 5 mm × 5 mm chips. After a 1:1 HCl:H₂O clean, samples were loaded into a CN‑1 thermal ALD reactor (trimethylaluminum and NH₃ precursors) at 350 °C. AlN thickness was tuned by varying ALD cycles, yielding 0.7, 1.5, and 7.4 nm films, as confirmed by FS‑1 ellipsometry. MIS diodes were formed by evaporating 50 nm Pt (500 µm diameter) as the Schottky top electrode and 100 nm Al as the backside contact. Reference Pt/n‑GaN Schottky diodes were also fabricated.

Characterization

Temperature‑dependent I–V (−5 V to +5 V, 80–400 K) was measured with an HP 4155B analyzer on a hot‑chuck stage. Capacitance–voltage (C–V) data (0.1–1 MHz) were acquired using an HP 4284A LCR meter. Cross‑sectional STEM images were obtained to verify layer thicknesses. Energy‑dispersive X‑ray spectroscopy (EDS) provided elemental depth profiles, while X‑ray photoelectron spectroscopy (XPS) with a monochromatic Al Kα source examined interfacial chemistry.

Results and Discussion

Structural Verification

STEM cross‑sections confirmed AlN thicknesses within ±0.2 nm of the ellipsometry values. The 0.7‑nm film appeared discontinuous in localized regions, whereas the 7.4‑nm layer was uniform and continuous.

Forward‑Bias Transport

Semilog I–V curves (Fig. 2a) show that the 0.7‑nm device exhibits higher leakage than the reference, while the 1.5‑ and 7.4‑nm devices suppress current. TE analysis yields barrier heights of 0.77 eV (reference), 0.61 eV (0.7 nm), 0.83 eV (1.5 nm), and 1.00 eV (7.4 nm). Ideality factors mirror this trend: 1.63 (ref), 4.19 (0.7 nm), 1.83 (1.5 nm), 1.57 (7.4 nm). The 0.7‑nm thickness thus represents a turning point where the barrier drops and recombination increases.

Capacitance–Voltage Behavior

Low‑frequency C–V (≤10 kHz) for 0 and 0.7 nm devices shows inversion, indicating defect‑induced deep depletion—absent in thicker AlN layers. EDS depth profiles reveal Pt diffusion into GaN for the reference, but this is suppressed when AlN ≥1.5 nm, confirming AlN’s barrier role. High‑frequency C–V hysteresis (Fig. 5 inset) reveals trap densities of 4.2 × 10⁹ cm⁻² eV⁻¹ (reference), 9.3 × 10⁹ cm⁻² eV⁻¹ (1.5 nm), and 3.6 × 10¹¹ cm⁻² eV⁻¹ (7.4 nm). The 7.4‑nm film therefore harbors the largest interfacial and oxide traps.

Interface Chemistry (XPS)

For the 0.7‑nm sample, the Al 2p spectrum shows peaks at 74.1 and 75.6 eV (AlOₓ and Al–OH), indicating incomplete AlN formation and predominant Al–O bonding. The 7.4‑nm film displays a clear AlN peak at 73.6 eV and a minor Al₂O₃ component (~531.8 eV in O 1s). Ga 2p₃/₂ spectra show a 1117.4 eV component (Ga–Al bonding) for the thick film, whereas the thin film shows 1118.0 eV (GaN) and 1119.2 eV (Ga₂O₃) peaks. Oxygen concentration is markedly lower near the AlN/GaN interface in the 7.4‑nm sample, but persists throughout the film, suggesting oxygen‑related defect states that elevate D_it.

Reverse‑Bias Leakage

Temperature‑dependent J–V data (Fig. 7) reveal that the 7.4‑nm device’s reverse leakage grows more steeply with temperature than its forward current. Modified Richardson plots confirm that barrier inhomogeneity, modeled by TE, fits the 7.4‑nm data (σ₀ = 0.204 V) but not the 0.7‑nm case (σ₀ = 0.147 V). FN tunneling dominates both devices: barrier heights of 1.67 eV (0.7 nm) and 0.78 eV (7.4 nm) at 300 K. The reduced barrier for the thicker film aligns with its higher trap density, facilitating trap‑assisted tunneling. Poole–Frenkel fitting yields unphysically large permittivities, ruling out PF emission as the primary leakage mechanism.

Implications for Device Design

Our findings suggest that AlN layers ≥1 nm effectively suppress Pt diffusion and reduce deep‑level trapping, yet too thick a film (>5 nm) introduces significant interface and bulk traps that lower the barrier and increase FN leakage. A 1.5‑nm AlN provides a balance: modest barrier elevation, low trap density, and acceptable leakage. These insights inform the optimization of passivation layers in GaN‑based HEMTs and power devices, especially when integrating with CMOS processes.

Conclusions

We demonstrated that AlN thickness critically governs interface quality and leakage in ALD‑AlN/n‑GaN MIS diodes. A 7.4‑nm film presents the highest interface/oxide‑trap densities, whereas a 0.7‑nm film suffers from Al–O bonding and lacks a robust AlN layer. Reverse‑bias leakage is dominated by Fowler–Nordheim tunneling for both thicknesses, with the thicker film exhibiting a lower effective barrier due to oxygen‑related defects. Optimizing AlN thickness—ideally around 1–1.5 nm—balances barrier enhancement against trap formation, paving the way for high‑performance GaN devices.

Abbreviations

ALD

Atomic Layer Deposition

AlN

Aluminum Nitride

C–V

Capacitance–Voltage

FN

Fowler–Nordheim

J–V

Current Density–Voltage

PF

Poole–Frenkel

TE

Thermionic Emission

XPS

X‑ray Photoelectron Spectroscopy