Search

The study of interfaces for both epitaxial and wafer-bonded systems draws from materials science, electrical engineering, and mechanical engineering and involves advanced materials characterization techniques. Low temperature wafer bonding has been leveraged to produce a wide array of materials combinations, most notably silicon-on-insulator structures. However, bonded interfaces can impact the electrical or thermal transport across such interfaces. In this presentation, we provide a few examples in semiconductor-based systems to address the ability to study and modify different, technologically important, interface combinations as a function of processing, such as annealing. The materials combinations primarily involve wide bandgap materials combinations including GaN|Si to b-Ga2O3 | SiC as well as bonding to single crystal diamond. In addition, we describe the role of the relative lattice orientation across semiconductor-semiconductor interfaces on transport properties and polarization engineering to producing high performance heterointerfaces. Our main goal is to be able to study and engineer the interfaces to optimize properties and ultimately, device performance.

In this work, the effects of the substrate defects on the reverse leakage behavior of vertical GaN Schottky and p-i-n diodes are investigated. A direct connection between the reverse leakage behavior of GaN based vertical devices and the defect characteristics of the substrate was achieved. The difference in the leakage current can be as high as 6 orders of magnitude for vertical GaN p-i-n diodes at -200V, using HVPE substrate with inhomogeneous defect distributions (with dot-cores). For comparison, using HVPE substrate with uniform defect distribution (no cores), the p-i-n diodes show much more uniform leakage behavior at -200V, varying within only an order of magnitude. The wafers with inhomogeneous defect distribution possess periodically patterned core-centers with higher defect density (up to 108 cm-2) than regions in between cores, which have very low defect concentrations (as low as 104 cm-2). The full width at 0.01 maximum of the triple axis rocking curves (GaN (0004)) shows a large variation ranging from 60 to 930 arcsec. In comparison, wafers without the periodic cores have a uniform defect density (~106 cm-2) and show a small variation in the rocking curve FW0.01M, ranging only from 70 to 280 arcsec. For Schottky diodes fabricated on substrate with cores, positioning the devices away from the core-centers results in a reduction of the reverse bias leakage by 2-4 orders of magnitude at -10 V. Similar trend are also observed in the p-i-n diodes, with device further away from the core-centers showing lower leakage. The devices in the low defect concentration region (> 200 um away from core centers) on the wafer with cores outperform the devices from the wafer without cores, showing up to 2 orders of magnitude lower leakage current at -200 V. The results from this study show that the substrate defect density and distribution play an important role on the device leakage current and avoiding highly defective regions on the substrate will improve the performance of vertical GaN devices.

Vertical GaN power devices have emerged to become promising candidates for next-generation high power applications due to superior material properties such as high breakdown voltage, low on-resistance, and high mobility compared to devices based on Si and SiC. GaN-based p-n junction switching devices enable higher voltage power with significantly higher efficiencies with added advantages of reduced size and weight systems. A technological limitation of GaN, however, has been the inability to achieve high p-type doping in a planar, vertical device. Here, we will focus on recent developments to achieve high p-type efficiency though ion implantation, novel high temperature annealing schemes, and the importance of defects and morphology in native substrates and epitaxial layers GaN epitaxial layers for subsequent p-type ion implantation can be grown on sapphire substrates but vertical devices are optimal when using GaN homoepitaxial structures. Unlike most other semiconductor substrates, GaN substrates are not grown from the melt. Hydride vapor phase epitaxy or ammonothermal growth are predominantly used to produce substrates and these substrates will experience higher temperatures during subsequent epitaxy and ion implant activation annealing than occurs during substrate formation. In both types of substrates, the threading dislocation distribution can vary by orders of magnitude – these variations correlate well with leakage currents in vertical devices. The activation of the implanted p-type dopants requires high temperature annealing. For most semiconductors, an annealing temperature of ~ 2/3 the melting point leads to a high activation fraction (> 95%) of the implanted species. The implantation process introduces large concentrations of native defects as well as the p-type implant (typically Mg), so it is necessary to remove the defects and for the dopants to locate on substitutional sites (Ga sites for Mg). We demonstrate how the formation of inverted domain defects at temperatures below 1400 °C correlate with Mg-compound formation, whereas annealing temperatures 1400 °C and above do not lead to the formation of such compounds but rather high Mg activation efficiencies. These recent developments, partly through the ARPA-E PNDIODES program as well as international efforts, have brought understanding of the key processing steps and substrate requirements to achieve high activation efficiency p-type doping for planar, vertical device structures in a scalable framework.

In this study, thermal and structural characteristics of direct wafer bonded (0001) GaN on (001) Si were examined as a continuation of a previous effort1. EVG® ComBond® equipment was used for bonding under high vacuum (~10-8 mtorr) at room temperature by bombarding the GaN and Si surfaces with an Ar ion beam to remove unwanted native surface oxides.1,2 The samples are placed face-to-face and pressure is applied to initiate the bond. The [1010] GaN edge was aligned parallel to the Si [110] edge. An X-ray diffraction reciprocal space map of the (004) Si and (0004) GaN revealed that there is a ~0.2° tilt between the GaN and Si layers. Bombardment of the bonded surface causes an amorphous region at the interface that has been seen in many other bonded systems.3-5 High resolution transmission electron microscopy revealed a ~2 nm amorphous region at the bonded interface, which resides mostly in the Si as confirmed by energy dispersive x-ray spectroscopy. This is consistent with SRIM6 simulations that predict a shallower damage profile in GaN as compared to Si. The thermal boundary resistance of the GaN|Si interface was measured to be 3.9 m2K/GW,1 an interfacial thermal resistance lower than any of the current state-of-the-art GaN-based heterojunctions, including GaN|Si,3 GaN|SiC,4 and GaN|diamond.5 Subsequent annealing is expected to alter the thermal properties as the bonded interface is allowed to recrystallize10. The intermediate steps of the recrystallization at the bonded interface were examined using TEM to give insight on the interface recovery and reconfiguration mechanism of this ion bombarded GaN|Si interface. Dragoi, et al., ECS Trans., 86(5), 23 (2018) Flötgen, et al., ECS Trans., 64(5), 103 (2014) E. Liao, et al., ECS Trans., 86(5), 55 (2018) Xu, et al., Ceramics International, 45, 6552 (2019) Mu, et al., Appl. Surf. Sci., 416, 1007 (2017) F. Ziegler, et al., The Stopping and Range of Ions in Solids vol 1 (Oxford, Pergamon), (1985) Yates, et al., ASME InterPACK (2015) Cho, et al., Phys. Rev. B, 89, 115301 (2014) Sun, et al., APL, 106(11), 111906 (2015) Mu, F. et al., (2019). ACS applied materials & interfaces, 11(36), 33428-33434. Figure 1

Wide bandgap materials are suitable for high power and high temperature applications. However, these materials, and metal layers that are part of device structures must effectively transport heat from the active device regions. Structural defects, chemical defects, and interface properties can degrade thermal transport, but relatively little attention has been paid to the precise nature of these defects and how they impact thermal transport. Electron microscopy and x-ray scattering based techniques are employed to provide insight into the defects present in these materials. When these techniques are integrated with thermal transport measurements and theoretical insights gained from molecular dynamics simulations, a much better understanding of thermal transport in wide bandgap materials can be realized. Examples of how defects impact transport for diamond boundary interfaces, AlN layers, and GaN heterostructures will be presented.