Low Resistance Non-Alloyed Ohmic Contacts to High Al Composition n-type AlGaN

Abstract

Ohmic contacts to high (>70\%) Al content n-type Al$_x$Ga$_{1-x}$N ultra-wide bandgap semiconductor layers in nitride electronic and photonic devices are typically fabricated by a lift-off process and high temperature ($>700^\circ$C) thermal alloying. These conditions often result in significant structural deformations of the fabricated structures and impose a harsh thermal budget on all other aspects of the device. Here, we report the fabrication of \textit{non-alloyed} \textit{as-deposited} ohmic contacts to 71\% n+AlGaN ($E_\text{g}\sim5.4$~eV) with a free carrier concentration of roughly $7\times 10^{19}$~cm$^{-3}$ and a resistivity of 4 - 5.5 m$Ω$cm (among the lowest reported for Al$_{0.71}$Ga$_{0.29}$N) with linear $I-V$ characteristics and a contact resistivity of $ρ_\text{c}=(4.4\pm1.0)\times10^{-4}$~$Ω$cm$^2$ (measured at zero voltage). Contacts with this quality are formed by two separate fabrication schemes: (i) metal-first patterning, and (ii) lift-off with an oxygen asher descum prior to metal deposition. Given the low threading dislocation density in the single-crystal AlN substrate used for epitaxy, the smooth morphology of the contacted epitaxial surface, and the non-alloyed nature of the contacts, this contact resistivity is attributed purely to thermionic field emission through the metal-semiconductor junction. Contact resistivity extraction at low current injection enables us to model these results using a thermionic field-emission model of contact resistivity, yielding a barrier height for Ti/Al$_{0.71}$Ga$_{0.29}$N of $(0.81\pm0.02)$ eV.

Figures (5)

• Figure 1: (a) Schematic cross-section of the n+AlGaN sample grown by molecular beam epitaxy on a 2-inch diameter metal-polar single-crystal AlN substrateMueller_2009 (with threading dislocation density $<10^4~\text{cm}^{-2}$). (b) Mobility (top) and carrier concentration (bottom) from temperature-dependent Van der Pauw Hall measurement. (c) Reciprocal space map across the asymmetric $(\bar{1}05)$ diffractions. (d) Wafer-scale contactless sheet resistance map. Data were collected within a 0.7-inch radius of the wafer center; the edge region of the map is intentionally left blank because no data points were measured in this area. (e) $5\times5$ µ m$^2$ AFM scan of AlGaN surface with 0.60 nm RMS roughness. (f) Measured (blue) and simulated (red) $\omega-2\theta$ X-ray diffraction scans across the (002) diffraction.
• Figure 2: Device schematic and fabrication process flow for lift-off (top; samples A-D) and metal-first (bottom; sample E) ohmic contact formation. The contact metal ($\Omega$) and oxygen asher descum treatment for each sample are specified in the upper right.
• Figure 3: (a) Current-voltage ($I-V$) characteristics of circular transfer length method (C-TLM) test structures from (a) samples A-E with a gap distance of $\sim$5 µ m. (upper-left inset) Summary of the surface treatment and contact metal for samples A-E. (lower-right inset) Circuit configuration for the current-sourced $I-V$ measurements. (b) Samples E, B, and D in the linear $I-V$ regime with gap distances from 2-35 µ m. (c) TLM plot of samples E, B, and D based on linear I-V shown in (b). (d) Box-and-whisker plot of the contact resistance $R_\text{c}$ extracted from 0 to 5 mA from all test structures on samples E, B, and D. Current-density-dependent $R_\text{c}$ analysisPiotrzkowski_2011Hu_2015Dill_2025Huang_2025Bhattacharya_2025 of the non-linear $I-V$ curves measured on samples A and C could not be performed as this analysis is only applicable to linear TLM test structures (but not C-TLM since a C-TLM contact pair does not share the same current density). The corresponding specific contact resistivity values, averaged across all test structures on each sample, are listed in Table \ref{['table']} (in units of $\Omega$cm$^2$). (e) (left) Microscope image of fabricated C-TLM pads probed in a four-point configuration and diagram indicating the dimensions of the inside and outside radii ($r_\text{i}$ and $r_\text{o}$, respectively) of the C-TLM test structures.
• Figure 4: (a) Specific contact resistivity $\rho_\text{c}$ as a function of doping density $N_\text{d}$, simulated for x=71% Al$_x$Ga$_{1-x}$N using the thermionic field emission model (see inset) defined in Eq. \ref{['Eq: rhoc_TFE']}. The average contact resistivity [see Fig. \ref{['Fig: 3_CTLM']}(c,d)] and Hall density (see Table \ref{['table']}) measured for sample B (Ti lift-off with O$_2$ descum) are plotted, suggesting a barrier height for Ti/n+Al$_{0.71}$Ga$_{0.29}$N of $\phi_\text{B}=(0.81\pm0.02$) eV, assuming an effective mass of 0.34 $m_0$, a dielectric constant of 8.6 $\varepsilon_0$, and a bandgap of 5.4 eV for Al$_{0.71}$Ga$_{0.29}$N (see Table \ref{['Table: AlGaN Material Parameters']}). (b) The inferred electron affinity $\chi_\text{e} = \phi_\text{M}-\phi_B/S$ (see Eq. \ref{['Eq: S Schottky Mott']}) from thermionic field emission modeling of sample B for $S$ values ranging from 1.0 to 0.25 (red), benchmarked against measurements of Al$_x$Ga$_{1-x}$N electron affinity from literature (black), obtained from Refs. Wu_1999($\varhexagon$), Levinshtein_2001 ($\pentago$), Grabowski_2001($\vartriangle$), Kozawa_2000 ($\medwhitestar$), and Lin_2012 ($\circ$).
• Figure 5: Benchmark plot of contact resistivity $\rho_\text{c} = (\frac{\partial J}{\partial V})^{-1}_{V=0}$, extracted at zero voltage, versus Al content from 50% to 100%. Data are taken from Refs. France_2007Srivastava_2009Tokuda_2010Yafune_2011Yafune_2014Park_2015Mori_2016Bajaj_2016Nagata_2017Douglas_2017Haidet_2017Armstrong_2018Bharadwaj_2019Sulmoni_2020Cho_2020Xue_2020Zollner_2021Xue_2021Zhang_2021Maeda_2022Cho_2023Kumabe_2024Zhou_2024Guo_2024Ebata_2024Liu_2025Guo_2025Bhattacharya_2025Guo_2025bShin_2025. TLM analyses with linear $I-V$ characteristics are plotted in full opacity. When possible, for publications that show TLM analysis with non-linear $I-V$ characteristics but report a $\rho_\text{c}$ value extracted at a non-zero current injection, the provided TLM data are scrubbed and re-fit at $V=0$. Where insufficient data are provided in the manuscript to perform such re-fitting, the reported contact resistivity is plotted, but with reduced opacity. $\circletfill/\blacksquare$ symbols mark contacts to unetched n-AlGaN surfaces, while ✚/✖ symbols mark those on etched surfaces for bulk/template substrates. Alloyed/non-alloyed contacts are indicated with full/top-filled markers.