High thermal conductivity and thermal boundary conductance of homoepitaxially grown gallium nitride (GaN) thin films

Gallium nitride (GaN) has emerged as a quintessential wide band-gap semiconductor for an array of high-power and high-frequency electronic devices. The phonon thermal resistances that arise in GaN thin films can result in detrimental performances in these applications. In this work, we report on the thermal conductivity of submicrometer and micrometer thick homoepitaxial GaN films grown via two different techniques (metal-organic chemical vapor deposition and molecular beam epitaxy) and measured via two different techniques (time domain thermoreflectance and steady-state thermoreflectance). When unintentionally doped, these homoepitaxial GaN films possess higher thermal conductivities than other heteroepitaxially grown GaN films of equivalent thicknesses reported in the literature. When doped, the thermal conductivities of the GaN films decrease substantially due to phonon-dopant scattering, which reveals that the major source of phonon thermal resistance in homoepitaxially grown GaN films can arise from doping. Our temperature-dependent thermal conductivity measurements reveal that below 200 K, scattering with the defects and GaN/GaN interface limits the thermal transport of the unintentionally doped homoepitaxial GaN films. Further, we demonstrate the ability to achieve the highest reported thermal boundary conductance at metal/GaN interfaces through in situ deposition of aluminum in ultrahigh vacuum during molecular beam epitaxy growth of the GaN films. Our results inform the development of low thermal resistance GaN films and interfaces by furthering the understanding of phonon scattering processes that impact the thermal transport in homoepitaxially grown GaN.

Figure 1

Schematics of the (a) MOCVD-grown and (b) MBE-grown GaN thin film samples measured in this study.

Figure 2

Room-temperature thermal conductivity as a function of film thickness for MOCVD-grown (red symbols) and MBE-grown (blue symbols) GaN thin films of this study. For comparison, we also include the literature reported thermal conductivities of other homoepitaxially [6] and heteroepitaxially (GaN/sapphire [15] and GaN/$4\mathrm{H}$-SiC [36]) grown GaN films. The dashed line represents the FPLD predictions of defect-free and single-crystalline GaN. Solid and open symbols represent measurements taken in this work and literature values, respectively.

Figure 3

Temperature-dependent thermal conductivity of the $1.9\mu \mathrm{m}$ UID GaN, and $0.8\mu \mathrm{m}$ UID and Mg-doped GaN films. The dashed line represents the $A/T^{1.35}$ ($A=4.4\times {10}^5$) function used to fit the $1.9\mu \mathrm{m}$ GaN thermal conductivity from $\sim 200$ to 495 K. The red and blue symbols represent MOCVD-grown and MBE-grown samples, respectively.

Figure 4

(a) Room-temperature metal/GaN thermal boundary conductance as a function of elastic modulus for different metals. The solid symbol represents the MBE-grown GaN films measured in this study and open symbols represent different metal/GaN interfaces [52, 53, 54, 55, 56] compiled from Donovan et al. [56]. For cases where an adhesion layer is present at the interface (Au/Ti and Au/Cr), the elastic modulus [57, 58, 59, 60] corresponds to the layer in contact with the GaN. (b) Temperature-dependent Al/GaN thermal boundary conductance for the MBE-grown films. For comparison, we also include the Au/Ti/GaN and Al/GaN thermal boundary conductances from literature [56]. The dashed line represents the thermal boundary conductance predicted by the modal nonequilibrium Landauer method for an Al/GaN interface.