3.1. As-Deposited Pd/Ni/Au Contact on p-GaN
Following the previously described wafer-cleaning procedure, we employed the e-beam evaporation technique to deposit metal layers of Pd/Ni/Au with nominal thicknesses of 10/10/30 nm. The metallization in its as-deposited state showed a consistent and even surface morphology, as depicted in
Figure 1.
For the as-deposited metallization, distinct phases of pure Au, Ni and Pd were observed, with no detectable metal intermixing in XRD measurements (
Figure 2). No mixing of layers was observed.
Cross-sectional TEM analyses were conducted to describe the interfacial structure of the studied samples. TEM images reveal a clear demarcation between the metal and GaN layers (
Figure 3a). The metal layers exhibit a preferred orientation of crystalline structure and the presence of twinned structures within the deposited metal layers is observed. The interlayer spacing determined from the Pd/Ni/Au/p-GaN structure for the (111) plane measures 2.24 Å, 2.04 Å, 2.35 Å and 2.593 Å, respectively. The HR-STEM image displays well-defined metal-layer interfaces and thicknesses that closely correspond to the structure design.
The deposited metals are highly textured and show crystallographic orientation with respect to the substrate, described as follows:
in the growth direction and
,
in lateral directions (
Figure 3b).
The schematic view of the crystal lattice for GaN’s
and the metals’
zone axis orientations is shown in
Figure 4.
The crystal lattices are aligned in accordance with the experimental TEM data.
Figure 4b,c corresponds to the two observed types of twins in the TEM. The dashed rectangles correspond to the elemental crystallographic lattice for each structure.
The TLM technique was employed to characterize the contact resistivity (
Figure 5).
The I–V characteristics of the as-deposited contacts showed a strong non-linearity, with a resistivity of 7 × 10−1 Ωcm2.
3.2. Pd/Ni/Au Contact on p-GaN Annealed in N2 Atmosphere
The SEM image (
Figure 6a) of samples annealed under a nitrogen flow illustrates that the top surface of the metallization exhibits a more granular appearance compared to the as-deposited sample, while maintaining an overall smooth texture. STEM data still reveal three relatively well-defined layers, although the palladium layer displays a dense network of voids (
Figure 6b).
The EDX data demonstrate a more intricate STEM image of our annealed three-layer structure (
Figure 7a–f). Annealing evidently prompts the diffusion of Ga atoms from the GaN semiconductor into the adjoining Pd layer, thereby giving rise to PdGa clusters. This infusion of Ga atoms into the Pd layer and generation of voids of varying sizes lead to an almost twofold expansion in the thickness of the latter layer.
The precise mechanism behind the formation of voids adjacent to the semiconductor–metal interface is still a subject of debate; however, we can propose two hypotheses: the de-wetting process and the nitrogen release. The first is linked to the wettability of the semiconductor’s surface (or any ceramic material). It has been demonstrated that thin metal layers deposited on diverse dielectric materials (e.g., gold on SiO
2) can undergo a de-wetting process, associated with the interplay between surface energy and bulk cohesion [
16]. De-wetting involves the transformation from a continuous metal layer into droplets on the surface because of annealing. The second plausible factor contributing to void formation could be associated with the release of nitrogen from the subsurface of the semiconductor. Nitrogen, forming N
2 molecules [
15], might introduce a pressure component into the mechanism, driving void formation. Notably, voids are distributed randomly at the interface. The existence of these voids reduces the contact area between the metal and the semiconductor by almost 60%. This phenomenon inevitably leads to an elevation in the contact resistance that corresponds to the extent of the actual contact.
The diffusion of Pd into the Au layer occurs alongside the creation of an intermetallic solid solution of Pd in Au (
Figure 7g). The interaction between Pd, Au and Ga could be advantageous from an electronic structure perspective under the given thermal conditions. As highlighted in [
17], pure palladium has 9.4 electrons per atom in the d-band. To fully fill the d-band of Pd (5s
04d
10), an additional 0.6 electrons are necessary. The density of available d-states surpasses that of s-states. Consequently, the s electrons from both gallium (4s
24p
1) and gold (5d
106s
1) migrate into the d-band of palladium to complete its occupancy. The feasible bonding configurations involve either metallic Ga and Pd or covalent Ga and metallic Pd. This formation of a solid solution of Pd in Au contributes to reduced stresses due to a diminished lattice mismatch with GaN.
The thin layer of Ni-and-Au mixture situated between GaN and the metallization points to an ongoing interfacial reaction. GaN and Pd are divided by Au–Ni clusters. Unlike [
18], no Ni
xGa
y or Ni
xN
y phases were observed. It is well established that Ga-face p-GaN exhibits robust chemical resistance [
19].
The inert behavior displayed by Ga-face material finds its explanation in the state of surface bonding. Specifically, the outer layer consists of Ga atoms, forming bonds with three N atoms at the surface and one N atom in the adjacent layer. The chemical reactivity of the GaN substrate is influenced by the lack of electron lone pairs. Notably, our experimental results illustrate instances of GaN decomposition and the subsequent diffusion of gallium atoms into the underlying Pd layer. This phenomenon could potentially be clarified by considering semiconductor activation due to the presence of dangling or defective Ga bonds within regions abundant in defects.
We propose that the annealing process at a relatively low temperature (530
°C) renders GaN less stable. The presence of Pd species along with Ni at the interface acts as a potent thermodynamic driving force, leading to the rupture of Ga–N bonds, decomposition into liquid Ga and subsequent nitrogen release [
20].
The interfacial reaction between Pd and p-GaN induces covalent atomic interactions [
21]. Based on fast Fourier transform (FFT) analysis of HR-STEM images, PdGa, Ni and AuPd species exhibit growth in the same direction toward GaN. According to [
22], gallides of palladium are notably more stable compared to other group VIII metal gallides. Therefore, in our case, Pd is thermodynamically favored to interact with p-GaN.
Palladium, possessing a higher surface energy, agglomerates to minimize the total interfacial energy of the metal–semiconductor system. The thermal expansion coefficient of Pd is twice that of GaN (11.7 × 10−6 K−1 vs. 6 × 10−6 K−1). The annealing process could generate compressive stress exceeding the ultimate tensile strength of Pd, leading to layer de-wetting. The lateral non-uniformity of metallization upon annealing could negatively impact the electrical characteristics.
Furthermore, due to the constraints of the metal deposition technique employed, achieving an epitaxial metallization structure remains elusive. The evaporated metal is subjected to compressive stress [
23], contributing to the emergence of nanovoids on the GaN surface, resembling ceramic characteristics. This phenomenon is observed even without subjecting the samples to thermal treatment [
24]. Notably, upon annealing, the newly formed metallization composition did not exhibit the capacity to disband the nitrogen atoms or prevent void formation. Determining the potential mechanism behind the formation of these voids during contact treatment should consider the coalescence of nanovoids as a significant factor.
In a general sense, annealing through RTA within a nitrogen-rich environment serves to hinder the out-diffusion of Ni towards the upper surface of the metallization layer, subsequently preventing its oxidation by enabling its dissolution into the Au layer. Oxidation of Ni is a trait typical of Ni/Au contacts annealed in an O2 atmosphere. However, partial diffusion of Ni and Au to the PdGaN interface persists. Given the high diffusivity of metals, both Ni and Au species wet the GaN surface and segregate from the PdGa layer, as illustrated by the EDX maps.
The above EDX analysis was confirmed by XRD data (
Figure 8). XRD findings suggest that Pd is entirely dissolved and PdGa and PdAu phases are detected. Lowering (but not disappearance) of Au (002) and Ni (111) reflection intensity indicates that these layers have almost completely been consumed through interfacial and interlayer reactions.
Thermal treatment within the nitrogen atmosphere does not significantly enhance the contact quality. Current–voltage measurements reveal a non-linear characteristic (see
Figure 4). The specific contact resistivity decreases to 3.7 × 10
−1 Ωcm² compared to the as-deposited layer. However, due to the contact area being reduced by approximately two times, there is a substantial fourfold enhancement relative to the as-deposited contact. Nevertheless, the high contact resistance and its non-linear nature render this metallization unsuitable for practical applications. The obtained value is four times higher compared to the contact annealed at 500
°C in N
2 for 1 min as reported in [
25]. This discrepancy suggests that extended annealing periods may lead to the degradation of the electrical properties of Pd/Ni/Au metallization. On the contrary, it
is important to consider that such a short annealing time may not be sufficient for the stabilization of the contact structure.
3.3. Pd/Ni/Au Contact on p-GaN Annealed in Oxidizing Atmosphere
Annealing in an oxygen-containing environment was historically a pivotal step in producing semi-transparent, low-resistance Ni/Au contacts for early InGaN-based LEDs [
12]. Recognizing the critical impact of oxygen during the annealing process for Ni-containing metallization, we chose to follow this route for our palladium–nickel-based contacts. Drawing from our prior experience with oxidized Ni/Au contacts, we anticipate significant reactivity of species in this scenario as well. Our Pd/Ni/Au metallization underwent annealing in a mixture of N
2 and O
2 or in ambient air at 530 °C.
Figure 9 depicts the rugged surface morphology and microstructure of the metallization post-annealing. Remarkably, the morphology of the annealed Pd/Ni/Au contact closely resembles that of a standard oxidized Ni/Au contact.
Like the preceding case, we can observe the decomposition of GaN. TEM and EDS data provide evidence of a certain degree of separation between Pd and GaN by Au and Ni (
Figure 10). Towards the bottom of the contact, spheroidal islands exist with a thickness of around 15–20 nm, displaying irregular sizes. These islands primarily consist of Pd atoms and diffused Ga. The GaPd alloy is subsequently covered by Ni–Au mixture and/or AuPd species.
EDS map data illustrate a uniform distribution of Pd in the upper Au layer. EDX data indicate that the diffusion of Pd atoms into the Au layer is more noticeable, and Pd distribution in this layer appears more even compared to the sample annealed in N2. The thickness of the AuPd layer varies between 30–40 nm. The differing shades within the PdGa or AuPd layers correspond to the variations in composition within these layers.
Despite the decomposition of GaN, no GaxOy and NxOy phases are formed after annealing in an oxygen-containing atmosphere.
Two novel interfaces were formed: AuPd/NiO and GaN/Au/Ni/PdGa. The annealed metallization primarily consists of PdGa, with a smaller region featuring gold and nickel at the lower end, Au and AuPd in the middle and NiO at the top (
Figure 10g).
It is important to note that the contact structure annealed in an atmosphere containing O2 undergoes species mixing, but void generation was not observed. The absence of voids in the metallization annealed in an oxidized atmosphere prevents uneven current injection and the development of hotspots within the contact region.
EDX data were supported by XRD analysis (
Figure 11).
XRD findings confirm that after annealing in a N2 + O2 atmosphere, phases of NiO, PdGa and PdAu appear. The intensity of the Ni (111) reflection disappears, indicating either that too little of the metal is present to be detected by XRD or that the pure Ni is completely dissolved. Low intensities of Au (002) and Pd (111) reflections indicate that these layers have been almost completely consumed by interfacial and interlayer reactions.
Due to its catalytic activity, Ni can prompt the desorption of hydrogen from the Mg–H complex within the p-GaN substrate, a process that might not have occurred during post-growth Mg activation [
20]. This desorption could subsequently decrease the electrical passivation of the acceptors, leading to a reduction in the Schottky barrier width and, consequently, a decline in contact resistivity. Another supporting factor for this role of Ni is the absence of Ni
xGa
y intermetallic compounds at the interface, which generally exhibit a low work function and have the potential to elevate the Schottky barrier height.
Finally, we also conducted contact metallization annealing using an open-air resistive furnace. The distinctions between RTA in simulated air and furnace annealing in a laboratory atmosphere lie in the subtle variations in the temperature–time profile and the introduction of water vapor into the laboratory atmosphere. The outcomes of TLM characterization suggest a slightly lower specific contact resistance for the metallization annealed in the open furnace (
Table 2). While there is no definitive understanding of this, one can speculate about the potential role of water vapor in facilitating hydrogen removal.
3.4. Three Methods of Contact Preparation—Summary
Microscopic studies have indicated a significant correlation between the contact structure’s morphology and the atmosphere’s composition during annealing. As per the results, the nature of the gas atmosphere during heat treatment notably impacts Ga out-diffusion, intermixing of metals and void generation. Among the most notable effects are:
Extensive gallium diffusion into the palladium layer and oxidation of nickel when the metallization undergoes annealing in an O2-containing atmosphere;
Void formation in a nitrogen-containing atmosphere;
Existence of Au–Ni mixture that acts as an intermediate layer between the GaN and metal structure.
Table 2 presents the recorded contact resistances for various annealing methods.
The specific contact resistivity numbers presented for non-ohmic contacts (as-deposited and annealed in N
2) were calculated as averages. This is due to the challenges in precise measurement using the classical TLM method [
26].
The ohmic contact characteristics can be linked to both the arrangement of the metallization structure and the interactions occurring at the metal–p-GaN interface. Another plausible factor contributing to the enhancement of contact resistivity is the potential increase in contact area and adhesion. Thermal treatment has the capacity to induce interface roughening through interfacial reactions, thereby promoting better adhesion. The diminished surface uniformity might arise from variations in material thermodynamic properties, possibly compounded by Pd disrupting some interfacial contamination. This disruption could facilitate enhanced current flow across the interface. The elevated contact resistance compared to that reported in [
27] suggests that, in the context of annealing with RTA in N
2 + O
2, a shorter treatment time might be more effective in reducing the specific contact resistance.
The reduced contact resistance observed in samples annealed in a furnace can be rationalized by a tighter contact structure. Annealing in an air environment encourages the formation of a more cohesive film, owing to the substantial melting impact of initially deposited sponge-like metals. This effectively prevents the out-diffusion of N2 and void formation. Additionally, this process governs the suboptimal wettability observed on the GaN substrate.
The outcomes from RTA annealing in artificial air and oven annealing in air are encouraging and have been employed as metallization techniques in laser diodes.
3.5. Influence of the Pd Thickness on the Contact Characteristics
To achieve a more uniform metal system, we enhanced the thickness of the Pd layer to mitigate the potential migration of Ni and Au to the GaN interface. The samples were subjected to annealing using both an oven in air and RTA in a N
2 + O
2 composition to study the impact of the thermal treatment method. Contacts based on thick Pd layers exhibit distinct morphologies after annealing in an oxidizing environment (
Figure 12). However, the specifics are contingent on the type of annealing apparatus employed, such as RTA versus oven.
In
Figure 12a, we observe the metallization with a thick (90 nm) Pd layer after annealing in artificial air. Evidently, we note the emergence of inhomogeneities in the metal layers akin to those described in the preceding sections.
Figure 12b demonstrates that the external appearance of the contact in the SEM images is significantly smoother for specimens annealed in the oven.
Obtained STEM images revealing thick Pd-based contacts annealed in an oxygen-containing atmosphere surprisingly exhibit the presence of voids at the metal–GaN interface (
Figure 13a,g).
According to EDX data, it is observable that the heating of contacts with a thicker Pd layer leads to a more extensive diffusion of Ni through the upper Au layer to the uppermost section of the metallization, where it subsequently undergoes oxidation (
Figure 13b–f).
The thickness of the Pd layer is increased to 130 nm, with approximately 50% of the layer near the interface containing Ga atoms. The varying shades of the Pd layer indicate the diverse composition of the solution, a fact corroborated by the fluctuations in the measured interplanar distances (refer to
Table 3). For the PdGa alloy, the latter increases from 2.18 to 2.20 Å with decreasing Ga concentration in the lower contact layer. According to EDX data, the lower Pd layer encompasses around 12% gallium. Roughly 18% of palladium diffuses into the upper Au layer.
When examining the intricacies of the metal-layer arrangement in both scenarios (annealing in oven and RTA), that is, with pure and uniform Pd layers, the Pd layer is consistently situated in the middle of the structure. However, heating in a N
2 + O
2 flow triggers the development of two distinct types of PdGa solutions (
Table 3).
In comparison to the PdGa mixture, the formation of the AuPd solution resulted in less alteration in the thickness of the initial metal layer. The upper Au layer is expected to impede the out-diffusion of Ni species to the surface; however, EDX data reveal that the 30 nm Au layer is insufficient in curbing the out-diffusion of Ni.
Towards the upper portion of the metal structure, a blend of Au, Ni and Pd is present, topped by NiO. The sequence of as-deposited Ni and Au is reversed after undergoing thermal treatment. In a broader sense, the migration of Ga and Au into Pd represents the predominant kinetic mechanism for altering the metal–substrate interfacial energy. Notable variations in the composition of the layers were observed for different sample treatments, evident through the lattice parameters (refer to
Table 3). These variations are more pronounced in the XRD data, where lattice parameters can be determined with higher precision; this is discussed later on.
It is important to highlight that annealing using RTA leads to more pronounced reconstruction and interaction between the metal and GaN species. The annealed samples exhibit more a disordered and deteriorated microstructure characterized by a higher number of voids compared to the sample subjected to furnace annealing.
XRD analysis was conducted to confirm EDX data (
Figure 14). XRD findings suggest that pure Au, Pd and Ni layers are consumed through interfacial and interlayer reactions, which are contingent on the structure and gas atmosphere. It is possible that Pd is entirely dissolved in samples with thin metal layers. Intensity of the Ni (111) reflection disappears which either means that too little of the metal exists to be detected by XRD or that the pure Ni is entirely dissolved. The lower intensity of Au (002) indicates that a small portion of pure Au is still present. Still existing after annealing, the intensity of Pd (111) reflection depends on the as-deposited thickness of Pd and is higher in the case of a thicker layer which indicates the presence of pure Pd in the contact design.
XRD data analysis confirms that after annealing in the air, in both cases of metallization, with thin and thick palladium layers, new phases of NiO, PdGa and PdAu appear. The positions of reflections for these phases vary in XRD profiles, indicating changes in their compositions. The composition assessment results of the PdAu alloy, according to Vegard’s law, are presented in
Table 4.
As shown in
Table 4, annealing in nitrogen has the strongest effect on the formation of the PdAu phase. In the case of the heating method, treatment in the oven has a lesser impact on the reaction between Au and Pd species. This could be attributed to the gas flow during RTA, which promotes accelerated diffusion of metal species. Considering the variation in Pd concentration in both the PdGa and AuPd mixtures, it can be inferred that the post-reaction products likely exhibit characteristics of a mechanical solution rather than a chemical compound.
In contrast to findings in [
6,
12] and despite the high diffusion activity of Ni, it does not react with the out-diffused Ga atoms, and Ni–Ga–O phases were not observed. In [
9], authors illustrated that the smooth surface of the Pt/Ni/Au contact post-annealing can be attributed to the formation of nickel oxide. In our case, however, annealing in a N
2 + O
2 environment results in the roughening of the upper Ni layer, whose thickness varies from 5 to 30 nm.
The formation of NiO is energetically favored by its negative Gibbs free energy (ΔG = −83.7 kJ/mol) at 500 °C [
28]. This rapid oxidation process serves as a driving force for the further out-diffusion of Ni atoms until all Ni is converted into the observed continuous NiO surface layer. The presence of a small amount of pure Ni indicates incomplete conditions (time, available oxygen) for complete Ni oxidation. As a diffusion barrier, NiO can enhance the stability of the metallization structure, preventing degradation of the surface and electrical parameters of the contact after annealing [
29]. A NiO film covers the top of the metallization, ensuring close contact of Au with the Pd layer and the p-GaN surface [
30].
The NiO phase of the oxidized Pd/Ni/Au system remains separate from the p-GaN. Furthermore, the applied temperature is too low for oxide diffusion onto GaN. In our experiments, Au acts as a template for NiO growth due to their quite similar lattice constants (a = 0.408 nm for Au and a = 0.417 nm for NiO) [
31]. The epitaxial growth of a NiO structure could generate a single barrier determining carrier scattering through different barriers. Moreover, good species matching eliminates grain boundaries that act as traps and recombination centers for carriers. In fact, more Ni species are oxidized during oven heating, contributing to a lower contact resistivity compared to samples subjected to short-time annealing.
Fast Fourier transform analysis of the metal areas revealed that Pd, Au and NiO species grew in alignment with p-GaN, akin to samples with a thin Pd layer. The orientation relationships between Pd, NiO, Au and p-GaN are akin to those described in [
12].
For comparison, Pd/Ni/Au contacts with differing thicknesses of the palladium layer were uniformly annealed in a N
2 + O
2 atmosphere composition. Changing the Pd layer thickness has minimal impact on the electrical properties (refer to
Table 5). The specific contact resistance slightly diminishes in comparison to contacts with a thin palladium layer. In line with [
32], the higher contact resistance of the thin Pd system is attributed to the formation of a larger number of voids.
Experiments indicate that a thick Pd layer effectively serves as a barrier to prevent the diffusion of Au and Ni species to the metal–semiconductor interface. However, increasing the thickness of the Pd layer also leads to a more pronounced decomposition of GaN and the formation of voids.
The small but visible enhancement of the ohmic properties can be attributed to the heightened protection provided by the thicker Pd layer against interfacial reactions involving Ni, Au and the semiconductor. The findings obtained do not align with the suggestion made in [
33] that the formation of Pd–Ni–Ga- or Au–Ga-related compounds is necessary to reduce the contact resistance.
The presence of voids diminishes the contact area and results in non-uniform current distribution at the compromised interface. Nevertheless, I–V measurements affirm that even in the presence of voids, the Pd/Ni/Au contact system demonstrates favorable ohmic characteristics.
The fabricated LD (laser diode) structures featuring Pd/Ni/Au metallization were characterized by an operating voltage of approximately 4.1 V, observed at 50 mA or 3.3 kA/cm².