Breakdown-induced directional cracking in kilovolt-class β-Ga₂O₃ (001) vertical trench Schottky barrier diodes Open Access

Defect and reliability analysis of gallium oxide (Ga₂O₃) high voltage devices is becoming increasingly important due to their rapid advancement toward commercialization.^1–4 The development of melt grown Ga₂O₃ substrates using edge-defined film fed (EFG) growth,^5,6 Czochralski (CZ),^7,8 or vertical Bridgman (VB)^9,10 methods has enabled the availability of 4-in. wafers commercially, with 6-in. wafers expected to enter the market soon. However, the process of melt growth introduces crystallographic defects in the material, which might hamper the device performance and reliability. Such defect-induced reliability issues have been a significant concern in materials like silicon carbide (SiC), where defects, such as basal plane dislocations (BPDs), have adversely impacted the device reliability for many years.^11–13 Similar concerns have been raised for Ga₂O₃ where the defects can hamper the device performance.^14–18 Among these defects are plate-like nanopipes (PNPs),^19,20 which propagate toward the surface along the (100) plane in Ga₂O₃ EFG substrates with (001) orientation. In this work, we report a coincidence between the orientation of the PNP defects and the breakdown-induced cracking along the same orientation in kilovolt-class Ga₂O₃ vertical trench Schottky barrier diodes (TSBDs). By mapping the defect locations in the original material and the destructive breakdown sites in the devices using laser confocal microscopy and scanning electron microscopy, respectively, we explore whether these PNPs cause breakdown and subsequently are the fundamental cause of material cracking in the devices upon failure.

The β-Ga₂O₃ (001) samples used in this work were processed on 1 cm × 1 cm pieces, originating from a 4-in. commercial wafer from Novel Crystal Technology. The sample consisted of a 650 μm thick EFG Ga₂O₃ substrate with a ∼12 μm thick hydride vapor phase epitaxy (HVPE) grown unintentionally doped (UID) Ga₂O₃ drift layer on top with an effective doping concentration of 1.1 × 10¹⁶ cm⁻³. The trench Schottky barrier diode fabrication started with 1 μm wide trench/gap patterning and 1.5 μm deep trench etching, which was performed using inductively coupled plasma—reactive ion etching with a Ni hard mask. The Ni metal was etched using piranha; defects in the device regions were then imaged using a Zeiss LSM 900 laser confocal microscope through the wafer backside. The dielectrics have been the main limiting factor for achieving high breakdown voltage in TSBDs. So, an optimized 160 nm high-density aluminum oxide (Al₂O₃) layer was deposited using an Oxford Instruments FlexAL plasma enhanced atomic layer deposition (PEALD) tool;²¹ the oxide was etched on top of the trenches. Ni/Ti/Cr anode contacts were sputter deposited on the top side of the wafer, and Ti/Au cathode Ohmic contacts on the back side of the wafer. The fabricated devices did not have any edge termination, and the off-state breakdown voltage (V_br) was measured using an ipTEST Mostrak-2 10 kV testing kit operating in the current source mode. Post-breakdown, the devices were again imaged using a Zeiss Sigma scanning electron microscope (SEM) to locate cracks and delamination, and potential correlation to the pre-existing defects. Nanoindentation was performed on the sample using a Bruker Hysitron TI Premier Nanoindenter to study the material fracture properties.

Figure 1(a) shows a schematic of a PNP defect propagating through the Ga₂O₃ substrate, typically terminating at the interface of the EFG substrate and the Ga₂O₃ epi-layer on top. Figure 1(b) shows an image of the PNP defect acquired from a 3D scan of the Ga₂O₃ crystal using a laser confocal microscope, obtained by stitching the frames captured using a calibrated stage movement along the z-direction in 1 μm steps. These defects were imaged using a 640 nm red laser, which was illuminating the sample through its backside, and the resulting emission was captured in the range of 620–700 nm. The defect extended along the (100) plane and terminated ∼12 μm below the surface, which corresponds to the surface/epi-layer interface, confirmed by FT-IR measurements after material growth, with the interface not being distinguishable in the confocal microscope image due to the homogenous nature of the material stack. The length of these defects was observed to range from a few tens of micrometers to several hundred micrometers along the [010] direction. Although the PNP defects were captured in these images, other defects including dislocations extending from PNPs¹⁹ could not be imaged with the laser microscope.

Figures 2(a) and 2(b) show a cross-sectional schematic of the trench Schottky barrier diode (TSBD) and a top-view SEM image of the as-fabricated device, respectively. Figure 2(c) shows the forward characteristics of the device, demonstrating an on-current of 220 A/cm² at a forward voltage of 5 V, with a differential specific on-resistance (R_on,sp) of 15 mΩ cm². The breakdown of the device was induced by forcing a current pulse of 0.5 μA for a duration of 200 ms, and the reverse voltage and current between the anode and the cathode terminals were monitored with respect to time, as shown in Fig. 2(d). Due to the applied high reverse current, the voltage quickly ramps up to 3.8 kV in ∼160 ms, at which time the device breaks down destructively, leading to crack generation in the device along the (100) plane as shown in the cross-sectional SEM image in Fig. 2(e). The orientation of the breakdown-induced crack closely matches with the inclination angle of the PNP defect in Fig. 1(b), raising the question of whether these cracks originating after breakdown are directly linked to the PNP defects preexisting in the original material. The on-resistance and breakdown voltage values are benchmarked against previously reported values in the literature, as shown in Fig. 2(f). Utilizing high density PEALD Al₂O₃, the devices in this work achieved a superior V_br compared to edge terminated TSBDs in Refs. 22 and 23, highlighting the importance of dielectric optimization in high-voltage devices.

The relationship between the PNP defects and the cracking observed after off-state breakdown, both associated with the (100) planes, was explored by mapping defects in the original materials and the cracks after device failure. First, prior to depositing the metal contacts in the device fabrication process, the defects in Ga₂O₃ in locations where TSBDs were to be fabricated were imaged using a laser confocal microscope. Following TSBD fabrication, testing, and electrical breakdown, the device regions were imaged using an SEM to find the locations in the structure at which the cracks originate. The images obtained using the confocal microscope and the SEM were then compared to find whether there is a correlation between location of PNPs and material cracking induced during breakdown.

Figures 3(a) and 3(d) show confocal microscope images of the fins patterned in the HVPE Ga₂O₃ epi-layer for two different TSBDs. With the laser light focused on the surface, the PNP defects were not optically visible in these images, inferring that they do not propagate all the way to the surface. However, on shifting the focus toward the substrate using a calibrated z-movement in steps of 1 μm, the PNP defects come into focus at different depths for different devices. Figures 3(b) and 3(e) show the PNPs imaged at 30 and 92 μm below the surface for the devices in Fig. 3(a) and 3(d), respectively. Since this is a 2D image, these defects appear like lines propagating along the [010] direction. These defects are expected to result in dislocations that propagate, from the substrate/epitaxial layer interface, vertically along the (100) plane toward the surface.²⁰ The corresponding SEM images of these devices post breakdown are shown in Figs. 3(c) and 3(f). In both devices, the failure cracks were observed in the [010] direction, which would propagate down along the (100) plane as shown in Fig. 2(d). Moreover, these devices also exhibited delamination after cracking. While the device in Figs. 3(b) and 3(c) shows cracking approximately at the defect site, this was not the case for the majority of devices; for example, in the device shown in Figs. 3(e) and 3(f), the defect is located 13 μm away from the top of the fin region highlighted using the dashed circle, while the failure cracks appear below the center of the patterned fins. Out of eight devices with optically visible defects below the anode region, only two devices showed failure cracks near or at the defect site, indicating this is likely only a coincidence. Therefore, we infer that the material cracking post-breakdown is not necessarily a damage related to the preexistence of the PNP defects in the substrate.

To understand the reasons for material cracking along the [010] direction during device failure, nanoindentation was performed on a Ga₂O₃ sample without devices.³⁷ A diamond Berkovich tip with a high-load transducer was used to indent the surface with a fixed load of 800 mN to check the direction along which the material cracks. Figure 4(a) shows an SEM image of the nanoindented region; cracks were observed at multiple locations along the [010] direction. A FIB cross section was prepared along one of the cracks, shown using the dashed region in Fig. 4(a). SEM in Fig. 4(b) shows these cracks also propagate at angle of ∼73° with respect to the surface. This matches with the direction of crack propagation in the TSBDs after breakdown, corresponding to the (100) cleavage planes in Ga₂O₃. The (100) planes are typically inclined at an angle of 76.3° with the (001) plane. This small difference between the crack orientation and (100) plane may be accounted for by considering the surface tilt/miscut of ∼4° in the samples measured.

During device breakdown, arcing at the anode contact and the resulting high instantaneous discharge current leads to a sudden rise in the device temperature causing local material expansion and mechanical stress. The anisotropic coefficient of thermal expansion of Ga₂O₃³⁸ combined with arcing-induced damage is likely to result in cracking along the (100) cleavage planes extending along the [010] direction. Thus, the breakdown-induced cracking in TSBDs appears to be a thermo-mechanical failure of the material rather than a defect mediated process in most cases. TSBDs with field plates were also tested (not shown here), where similar failure patterns were observed, further confirming that field crowding near the edge does not affect the overall failure patterns. While the mechanical failure along the (100) planes is an inherent property of (001) Ga₂O₃ substrates, investigating alternative wafer orientations, such as (010) or (100), might offer potential solutions to mitigate failure-induced cracking.

In summary, Ga₂O₃ (001) vertical trench Schottky barrier diodes were observed to crack along the [010] direction following electrical off-state breakdown, coinciding with the direction where PNP defects are known to be present in Ga₂O₃. Mapping defects and breakdown-induced cracks in the device regions using confocal and scanning electron microscopy revealed no clear correlation between the PNP locations and the breakdown sites. The breakdown-induced directional cracking appears to be a consequence of thermo-mechanical failure due to the weak (100) planes, which was confirmed using nanoindentation.

This work was supported by the University of Bristol Cleanroom Facility. We acknowledge in part financial support from the Engineering and Physical Science Research Council under the Innovation and Knowledge Centre (IKC) REWIRE under Grant No. EP/Z531091/1. M. Kuball acknowledges financial support by the Royal Academy of Engineering through the Chair in Emerging Technologies Scheme. The authors would like to thank Carl Zeiss Ltd., Oxford Instruments, and ipTEST Ltd., for their assistance with defect mapping, oxide deposition, and breakdown measurements, respectively.