1. Mg-pretreated Ohmic Contact

An epitaxial structure used for Mg-pretreated ohmic contact process consisted of a 70-nm p-type GaN layer, a 15-nm Al_0.21Ga_0.79N barrier, a 0.8-nm AlN spacer, a 200-nm unintentionally-doped GaN layer, and a 5200-nm buffer grown on a Si (111) substrate.

Figs. 1(a)-(d) schematically illustrate the fabrication process of the p-type GaN ohmic contact. First, mesa isolation was carried out to provide device separation, and the TLM patterns were defined by inductively coupled plasma-reactive ion etching (ICP-RIE) using BCl₃/Cl₂ gas mixture to a depth of 450 nm, as shown in Fig. 1(a) ^[9]. Subsequently, a 50-nm Mg film was deposited by electron-beam evaporation [Fig. 1(b)], followed by furnace annealing at 800 ◦C for 1 hour in N₂ to promote Mg diffusion [Fig. 1(c)]. After annealing, wet etching was carried out using boiled aqua regia (HCl:HNO₃ = 3 : 1), followed by diluted HF (DHF, 1:10). This process removed the top metallic Mg layer and Mg-O related surface layers, while the underlying Mg-related contact layer that was non-reactive to the acid etchants remained. Finally, Pd/Ni/Au (30/20/150 nm) and Ni/Au (20/150 nm) ohmic metal stacks were deposited for comparison [Fig. 1(d)], followed by rapid thermal annealing (RTA) at various temperatures in an O₂ ambient.

Fig. 1. Fabrication process of Mg-pretreated p-type GaN ohmic contact formation.

Fig. 2. Current-voltage (I-V) characteristics of (a) Ni/Au and (b) Pd/Ni/Au ohmic contacts at various RTA temperatures.

As shown in Fig. 2, the lowest contact resistance for both metal stacks was achieved at 550 ◦C. In previous studies without Mg pretreatment, the optimal annealing temperatures were 450 ◦C for Ni/Au and 500 ◦C for Pd/Ni/Au ^{[10, 11]}. In contrast, with the Mg pretreated process, the optimum temperature was 550 ◦C regardless of the metal stack, indicating that Mg incorporation is the dominant factor governing ohmic contact formation. Although the Mg incorporation process facilitates ohmic contact formation, the I-V characteristics were not perfectly linear and still exhibited non-ideal rectifying behavior, as will be discussed in the following section.

To further investigate the role of Mg pretreatment, secondary ion mass spectrometry (SIMS) analysis was performed. For this experiment, an epitaxial structure grown on sapphire was used, consisting of a 10-nm p⁺-GaN layer, a 270-nm p-GaN layer, and a 3200-nm n-GaN layer. The fabrication steps were identical to those in Figs. 1(a)-(c). As shown in Fig. 3, the SIMS results confirmed that the Mg-diffused region exhibited a significantly higher Mg concentration near the surface compared with the surface without Mg incorporation. These findings clearly demonstrate that Mg diffusion occurs during annealing and plays a decisive role in the formation of the ohmic contact.

Fig. 3. SIMS analysis of surface magnesium concentration with and without Mg incorporation process.

2. Characterization of Non-ideal Ohmic Contact

While the Mg-incorporation process promoted ohmic contact formation, the electrical characteristics were not ideally ohmic. In such non-ideal contacts, when a voltage is applied across the two terminals, the energy-band profiles for hole injection are not identical. Consequently, as shown in Fig. 4, the contact resistances at the two ends cannot be regarded as equal; instead, they must be decomposed into forward- and reverse-biased components, R_C,FOR and R_C,REV, respectively, such that 2R_C = R_C,FOR + R_C,REV.

Fig. 4. Energy band diagram between two non-ideal ohmic contacts.

The measured I-V characteristics were fitted using a Schottky contact model implemented in TCAD. The TCAD simulations were performed using the commercial device simulator Silvaco ATLAS, employing a conventional drift-diffusion transport model with Fermi-Dirac statistics. The physical models included doping- and field-dependent carrier mobility, Shockley-Read-Hall (SRH) recombination, and Auger recombination in GaN.The metal/p-GaN interface was modeled as a Schottky contact governed by thermionic emission over the potential barrier, including image-force barrier lowering. To account for the high Mg acceptor concentration near the p-GaN surface and the resulting strong electric field at the interface, field-assisted tunneling mechanisms, namely thermionic-field emission and phonon-assisted tunneling, were incorporated using a unified Schottky tunneling model. Surface recombination at the metal/p-GaN interface was also considered.

The acceptor concentration and hole mobility in p-GaN were fixed to the values extracted from Hall-effect measurements, while the Schottky barrier height at the metal/p-GaN contact was treated as a fitting parameter to match the measured I-V characteristics. A Schottky barrier height of 0.4 eV provided the best agreement with the experimental data, as shown in Fig. 5(a).

Fig. 5. (a) Comparison of measured TLM I-V characteristics with TCAD fitting results, and (b) decomposition of the total contact resistance (2RC) into forward (RC,FOR) and reverse (RC,REV) components.

The contact resistances for the forward- and reverse-biased terminals were then extracted by modeling one side as an ideal ohmic contact. As shown in Fig. 5(b), the reverse-biased terminal overwhelmingly dominated the total contact resistance, while the forward-biased terminal contributed only marginally. This finding underscores that contact resistance cannot simply be approximated as half of the TLM-extracted 2R_C.

To further clarify the transport mechanism at the Mg-incorporated p-GaN contact, temperature-dependent I-V measurements of the TLM structure were performed over a temperature range from 150 to 475 K. Arrhenius plots of ln(I) versus 1/T at different applied voltages are shown in Fig. 6(a). For V ≥ 0.5 V, the ln(I)-1/T data in the temperature range of 200-475 K are well described by linear relationships, and the slope of each line was used to extract an effective activation energy, E_A. Fig. 6(b) summarizes the extracted activation energy as a function of the applied voltage. E_A is approximately 0.13 eV at V = 0.5 V, but it rapidly decreases to 0.11-0.115 eV as the voltage increases to around 2-3 V. This decrease reflects a reduction in the effective energy barrier due to barrier lowering and field-assisted tunneling at the reverse-biased contact. At higher voltages (V > 3 V), E_A increases only slightly and remains within a narrow range of approximately 0.11-0.12 eV, indicating that carrier transport is governed by a combination of thermionic emission and field-assisted processes. In the back-to-back Schottky configuration of the TLM structure, one contact is reverse biased while the other is forward biased; therefore, the extracted E_A should be regarded as an effective activation energy representing the superposition of thermionic emission at the forward-biased contact and field-assisted tunneling and hopping conduction at the reverse-biased contact. Overall, the bias dependence of E_A supports a transition from a thermionic-emission-dominated regime at low bias to a mixed transport regime in which thermionic-field emission becomes increasingly important as the applied bias increases.

Fig. 6. (a) Arrhenius plots of ln(I) versus 1/T at different bias voltages. (b) Extracted activation energy EA as a function of applied voltage.

When a non-ideal p-type GaN ohmic contact is integrated into a PN diode, the p-type GaN contact operates under reverse bias during forward-biased diode operation. Therefore, the effective contact resistance should be regarded as approximately 2R_C, not R_C, and this must be explicitly incorporated into the contact resistance model.

Another critical point is that contact resistance is not constant but strongly dependent on the applied bias, owing to the presence of the effective Schottky barrier at the non-ideal ohmic interface. The energy-band bending of the barrier varies with the electric field at the contact surface. As the applied voltage increases, the electric field enhances barrier lowering, which in turn reduces the effective SBH, accelerates thermionic emission, and increases current conduction. Thus, the contact resistance must be characterized as a function of current density rather than treated as a fixed value. Fig. 7 shows the contact resistance (2R_C) extracted from TLM measurements as a function of current density, clearly demonstrating this dependence.

Fig. 7. Contact resistance as a function of current density.

3. On-resistance Analysis

To assess the impact of contact resistance on the on-state conduction of PN diodes, vertical PN diodes were fabricated on an epitaxial structure consisting of a 425-nm p-GaN layer and a 15-µm n-GaN layer grown on a heavily-doped n-type GaN substrate where the bottom n-type cathode contact was formed by Ti/Al/Ni/Au (20/120/25/50 nm) and the top p-type anode contact was formed by Mg-pretreated Ni/Au contact [Fig. 8(a)].

The measured forward I-V characteristics of the fabricated diode are shown in a solid line in Fig. 8(b). The total on-resistance (R_on = V_A/I_A) comprises contributions from the p-GaN ohmic contact resistance, p-GaN drift resistance, pn junction resistance, n-GaN drift resistance, substrate resistance, and n-type GaN ohmic contact resistance. The contact resistance as a function of current density was de-embedded from the measured R_on. The contributions of R_D+ R_substrate, R_PN, and R_C to the total R_on are represented by the pink, blue, and green regions, respectively, in Fig. 8(c). In the low-current regime, R_C contributes negligibly to the overall R_on due to the large turn-on voltage of the PN junction. However, as the current density increases, the contribution of R_C increases and reaches approximately 50% of the total R_on. Based on this analysis, the I-V characteristics of a PN diode with an ideal ohmic contact were extracted and are plotted as the dashed line in Fig. 8(b).

Fig. 8. (a) Cross-sectional schematicof vertical PN diode used for on-resistance analysis, (b) forward I-V characteristic of the fabricated PN diode, and (c) de-embedded p-type GaN ohmic contact resistance extracted from Ron.

The non-ideal ohmic contact characterization proposed in this study provides important insights for accurate device modeling and design. In conventional TLM analysis, the two-terminal contact resistances are assumed to be equal, and the total resistance is simply taken as 2R_C. However, this assumption can lead to overestimation or underestimation depending on whether the contact operates under forward or reverse bias. In practical devices, when the diode is forward-biased, the p-type GaN ohmic contact is effectively reverse-biased; therefore, the contact resistance should be treated as the full 2R_C extracted from TLM measurements. Furthermore, when R_C is extracted as a function of current density, the relative contribution of the contact resistance to device performance can be quantitatively evaluated. This enables not only the derivation of ideal I-V characteristics but also the determination of the ultimate performance limit of the device.