Low contact resistivity at the 10⁻⁴ Ω cm² level fabricated directly on n-type AlN

Aluminum nitride (AlN), known for its ultrawide bandgap, high breakdown strength, superior thermal conductivity, and exceptional stability, is favored for high-power electronic devices.¹ Thus, the pursuit of efficient AlN-based power electronics has garnered increasing attention. Presently, metal-organic chemical vapor deposition (MOCVD) is the predominant method for commercial AlN growth.² The MOCVD-grown n-type AlN's high dopant ionization energy limits its carrier concentration to below 10¹⁶ cm⁻³ at room temperature (RT).^3–5 Coupled with the low electron affinity, AlN suffers from high Ohmic contact barriers, which cause non-linear behavior and elevated resistivity. To achieve low contact resistance, the current leading strategy uses n-type graded AlGaN layers on n-type AlN films.^6–8 However, this approach is highly dependent on MOCVD equipment conditions and necessitates meticulous control of parameters like the Ga gradient and precursor flow rate. Additionally, it involves extra growth and etching steps. In contrast, metallization process optimization is more straightforward. Strategies include choosing appropriate primary contact layers (like vanadium and thin silicon nitride) to reduce the contact barrier^9–12 and inducing nitrogen (N) vacancies at the AlN/metal interface via the rapid thermal annealing process (RTP) to create extra donor states.^9,13–15 In n-type high Al content AlGaN, N vacancies serve as donors, creating a thin, highly doped layer under the metal after annealing and boosting interface net carrier concentration. This increases the probability of carrier tunneling through the space charge region at the metal/semiconductor (M/S) interface, which is crucial for lowing Ohmic contact resistance.¹⁶ However, the high dissociation energy of AlN from the surface renders the formation of N vacancies post-thermal annealing more challenging.^17,18 To date, despite significant efforts, the contact resistivity at RT on n-type AlN still exceeds 10⁻¹ Ω cm²,^11,19,20 not matching the graded n-AlGaN method's advantage (10⁻² Ω cm²).⁶ 

In this study, optimized metallization strategies that achieved a record low contact resistivity at 10⁻⁴ Ω cm² level on n-type AlN were demonstrated, circumventing the necessity for AlGaN gradient layers. Comprehensive investigations into the electrical and structural attributes of Ohmic contacts were conducted. These findings highlight the great promises of superior Ohmic contacts directly on n-type AlN hold for advancing high-performance AlN-based power electronics.

AlN layers were epitaxially grown on (0001) AlN/sapphire substrates via MOCVD. The epitaxial structure comprised a 100 nm undoped AlN regrowth layer and a 250 nm Si-doped AlN layer (Si: 6.4 × 10¹⁸ cm⁻³). Two Ohmic contact metal stacks were evaluated, TiAlTiAu (20/120/50/200 nm) and TiAlTi (20/120/80 nm), prepared via sputtering and subjected to RTP in a N₂ atmosphere, with temperatures ranging from 800 to 950 °C and durations from 30 to 120 s. After annealing, the contact characteristics were determined at RT using the circular transmission line method (CTLM). The conductivity and ionization energy of n-AlN at RT are evaluated to be 4.2 × 10⁻³ Ω⁻¹ cm⁻¹ and ∼320 meV, respectively. The schematic structure and fabrication flow are shown in Fig. 1(a). Two annealed samples with distinct metal stacks were prepared using a focused ion beam technique for cross-sectional transmission electron microscopy (TEM) analysis in a Titan ST microscope. High-resolution TEM (HRTEM) and high-angle annular dark-field scanning TEM (HAADF-STEM), along with energy-dispersive x-ray spectroscopy (EDXS) and electron energy loss spectroscopy (EELS), were employed to investigate the structure and compositional uniformity of the M/S interface post-annealing. To evaluate its impacts on diodes, the Schottky barrier diodes (SBDs) utilizing a NiAu (40/100 nm) stack for the Schottky metal were fabricated, with an electrode diameter of 150 μm and a gap of 3 μm. The electrical characterization was performed using a Keithley 4200-SCS analyzer. The surface morphology of annealed metal stack was obtained by atomic force microscopy (AFM) scans.

Surface morphology comparisons of both metal stacks under identical conditions (950 °C for 60 s), shown in Figs. 1(b) and 1(c), indicated a rougher surface for the TiAlTiAu stack due to gold's propensity for interdiffusion at high annealing temperatures, and low melting point of Au and Al.^13,15,21 I–V characteristics for both metal stacks under varying annealing conditions are depicted in Figs. 2(a) and 2(b). It revealed that all contacts exhibited moderate rectifying behavior. The TiAlTi stack demonstrated a closer to linear I–V characteristic with a significant increase in current. Specifically, the knee voltage (V_knee) for the TiAlTi stack was reduced to merely 1.5 V, substantially lower than the 5 V observed for the TiAlTiAu stack, suggesting a lower contact barrier.

Figure 2(c) exhibits the contact resistivities derived from the I–V characteristics' linear regions. Across all stacks, higher annealing temperatures led to lower contact resistivities and reduced rectifying tendencies, especially above 900 °C. This enhancement is due to the accelerated dissociation of surface AlN and the formation of N vacancies under increased temperatures. For the TiAlTiAu stack, annealing at 900 °C for 60 s resulted in the minimum contact resistivity ρ_c of 0.036 Ω cm². In contrast, increasing the annealing temperature of the TiAlTi stack to 950 °C remarkably lowered its rectifying behavior. Prolonging the annealing duration to 90 s for the TiAlTi stack further reduced its contact resistivity ρ_c and a minimum value of 5.82 × 10⁻⁴ Ω cm² with the contact resistance R_c of 82.7 Ω mm. Figure 2(d) shows the I–V characteristics of SBDs with both metal stacks post-optimal annealing, highlighting notable enhancements in conductance and threshold voltage. Specifically, the diode on-resistance dropped from 110.9 to 11.1 Ω cm⁻², and the desired voltage to reach 1 μA decreased from 3.9 to 2.5 V.

To clarify the contact formation mechanisms, representative HAADF-STEM images of the annealed metal contacts were shown in Figs. 3(a) and 3(b). Notable phase differences were identified between the two metal stacks by the EDXS analyses. Compared to the TiAlTi stack, noticeable metal clusters were observed in the TiAlTiAu stack, which is attributed to the rough surface. In addition, in the TiAlTiAu stack, gold's interdiffusion was observed, leading to an Al–Au alloy formation and prompting titanium's out-diffusion. No mutual diffusion between Al and Ti was observed in either stack. Moreover, the M/S interfaces were remarkably smooth, with no protrusions or metal diffusion into the grown-AlN layer. EDXS analyses of contact's interface [Figs. 3(c) and 3(d)] revealed the formation of a thick (25–35 nm) Al–N interlayer at the TiAlTiAu/AlN interface after annealing, covering the whole interface, whereas a much thinner (1–6 nm) Al–Ti–N layer was identified at the TiAlTi/AlN interface. Figure 3(e) displays the EELS spectra extracted from the interface. The EELS spectra and atomic percentage (in the supplementary material) of these interfacial layers corroborated their primary constituents, aligning with EDXS findings. Noticeably, Ti–N was detected on the surfaces of both metal stacks, indicating a significant presence of N, likely originating from the N₂ gas atmosphere. In the TiAlTiAu stack, the N in the thick Al–N layer may also partially derive from N₂, as the presence of metal clusters brings the metal/AlN interface closer to the external gas. Moreover, the imperfect vacuum (3–4 Torr only) of the RTP equipment makes Al more prone to oxidation, hence the observation of Al(O) in the EDXS results. Similar phenomena have also been reported in the literature on Al-rich AlGaN.⁹ 

In Figs. 3(f)–3(m), the HRTEM images and Fast Fourier transform (FFT) diffraction patterns showed partial nitride crystallinity in both interfaces, indicating N extraction from grown-AlN and formation of N vacancies. However, in the TiAlTiAu stack, a thicker Al–N interlayer hindered the further extraction of N by metal stack during annealing, which could explain the unimproved resistivity despite the temperature rise from 900 to 950 °C. Concurrently, this defective thicker interlayer inhibited carrier tunneling at the M/S interface, inversely rising the contact barrier and intensifying rectifying behavior. In contrast, the TiAlTi stack's thinner interlayer at 950 °C presented lower contact resistivity, further decreased by prolonged annealing duration, which boosted N vacancy creation and carrier tunneling. However, for the TiAlTi stack, a rise in contact resistivity following extended annealing duration to 120 s suggests possible interlayer thickening or oxidation of Al,⁹ warranting further study. Figure 4 benchmarks the reported contact resistivity of metal/AlGaN direct Ohmic contact vs the Al mole fraction.^{9,11–13,20,22–28} As observed, the contact resistance experiences a rapid increase beyond an Al mole fraction of ∼70% in AlGaN. Therefore, achieving a contact resistance in the 10⁻⁴ Ω cm² range through direct contact metallization on AlN is deemed promising for providing a comprehensive understanding of contact formation on Al-rich AlGaN and AlN films.

In summary, this work has demonstrated a low contact resistivity of 10⁻⁴ Ω cm² level for direct contact on n-AlN. The minimum contact resistivity of 5.82 × 10⁻⁴ Ω cm² was obtained with TiAlTi metal stack annealed under 950 °C for 90 s. The achievement of such low contact resistivity is attributed to the optimal metallization at the interface. The optimal metallization processes facilitated an increased net carrier density at the interface, enhancing carrier tunneling. These results enhance the comprehension of the interface characteristics and electrical properties of direct metal/AlN Ohmic contacts, offering valuable insights for the design and application of AlN-based power electronics and optoelectronic devices.