Mechanical Exfoliation of Large Area (100) β-Ga2O3 Onto Arbitrary Substrates for High Power Devices

Monday, 2 October 2017: 15:20
Chesapeake A (Gaylord National Resort and Convention Center)
M. J. Tadjer, M. A. Mastro (U.S. Naval Research Laboratory), A. Nath (George Mason University), L. E. Luna (U.S. Naval Research Laboratory), T. J. Anderson, K. D. Hobart (Naval Research Laboratory), and A. Kuramata (Tamura Corporation and Novel Crystal Technology)
The single-crystal monoclinic (β) phase of Ga2O3 is an advantageous material for high-power, high-temperature electronic device applications due to its ultra-wide energy gap (~4.9 eV) and high breakdown field (8 MV/cm), yielding a nearly ten-fold higher Baliga figure of merit than that of 4H-SiC (BFOMGa2O3 = 3444, BFOM4H-SiC = 300) [1]. Large diameter, high quality, cost effective native substrates grown from the melt are expected to be the main technology driver for the commercial adoption of this material. Even though substrates with full-width half-maximum values as low as 15 arcsec are routinely obtained by the edge-defined film fed (EFF) and Czochralski growth methods, it is expected that Ga2O3-based power devices will require low dislocation density epitaxial drift layers. While fast, high quality epitaxial growth of β-Ga2O3 is already available in small production volumes, high quality electronic devices can also be demonstrated by exploiting the relatively easy mechanical exfoliation of nano-flakes of β-Ga2O3 as a result of the large a lattice parameter of β-Ga2O3 (1.22 nm) [2, 3]. Indeed, a number of transistor reports have been recently published using mechanically exfoliated β-Ga2O3 nano-flakes as the device channel to demonstrate high current, normally off transistors [4-8]. Because Ga2O3 is not a two-dimensional (2D) material, where a number of monolayers are being held together by Van der Waals forces, mechanically exfoliated Ga2O3 typically yields much thicker (~300 nm) films, which could then be thinned by repeated mechanical exfoliation or dry etching [9]. In this work, we demonstrate that β-Ga2O3 exfoliation is scalable towards mm-sized flakes, and their transfer onto arbitrary substrates is readily feasible [10].

Mechanical exfoliation was performed using (100) and (-201) oriented β-Ga2O3 wafers, grown by the edge-defined film fed (EFF) growth method [11]. A film of medium-tack dicing saw tape was pressed and removed from freshly-cleaved facets of the wafers. The (-201) 2-inch wafer was cleaved in half in order to expose the (100) facet; thus, it was only possible to produce relatively narrow (~700 µm wide) flakes using this crystal orientation. However, the flakes obtained using this approach were more than 40 mm long and only the wafer size limited this dimension. They were also relatively thick (>500 µm) and thus could not be easily transferred onto another substrate. Using the (100) oriented wafer, on the other hand, allowed for ~5 mm wide, ~20 mm long flakes to be exfoliated as freestanding samples. Their yield decreased significantly upon transfer onto a substrate such as SiO2/Si. Nevertheless, about 4 mm long, 100 µm wide flakes were successfully transferred onto a diamond substrate. It was determined that the transfer area yield strongly depended on the force during transfer. While further controlled experiments are required to quantify this relationship, flakes of β-Ga2O3 were successfully transferred onto arbitrary substrates such as GaN, graphene/SiC, and single-crystal diamond, with various degrees of yield. The resulting transferred flakes were characterized by Raman spectroscopy, scanning electron microscopy, and other complementary techniques. Such a process could potentially open further avenues for novel device applications, for example thermally-managed Ga2O3 transistors on diamond.

Research at NRL was supported by the Office of Naval Research (ONR). A.K. acknowledges partial support by ONR Global (Dr. Ming-Jen Pan) under a NICOP contract. L.E.L. gratefully acknowledges postdoctoral fellowship support from the National Research Council.

[1] M. Higashiwaki, et al., Appl. Phys. Lett. 100, 013504 (2012).

[2] K.S. Novoselov, et al., Science 306, 666 (2004).

[3] J.B. Varley, et al., Appl. Phys. Lett. 97, 142106 (2010).

[4] W.S. Hwang, et al., Appl. Phys. Lett. 104, 203111 (2014).

[5] R. Mitdank, et al., Phys. Stat. Solidi A 211, No. 3, 543-549 (2014).

[6] S. Ahn, et al., Appl. Phys. Lett. 109, 062102 (2016).

[7] M.J. Tadjer, et al., ECS Jour. Solid State Sci. Technol., vol. 5, no. 9, P468-470, 2016.

[8] H. Zhou, et al., IEEE Electr. Dev. Lett., vol. 38, no. 1, pp. 103-106, 2017.

[9] Y. Kwon, et al., Appl. Phys. Lett. 110, 131901 (2017).

[10] J.D. Caldwell, et al., ACS Nano, 2010, 4 (2), pp 1108–1114.

[11] H. Aida, et al., Jpn. J. Appl. Phys. 47, 8506 (2008).