Scaling growth rates for perovskite oxide virtual substrates on silicon

The availability of native substrates is a cornerstone in the development of microelectronic technologies relying on epitaxial films. If native substrates are not available, virtual substrates - crystalline buffer layers epitaxially grown on a structurally dissimilar substrate - offer a solution. Realizing commercially viable virtual substrates requires the growth of high-quality films at high growth rates for large-scale production. We report the stoichiometric growth of SrTiO3 exceeding 600 nm hr−1. This tenfold increase in growth rate compared to SrTiO3 grown on silicon by conventional methods is enabled by a self-regulated growth window accessible in hybrid molecular beam epitaxy. Overcoming the materials integration challenge for complex oxides on silicon using virtual substrates opens a path to develop new electronic devices in the More than Moore era and silicon integrated quantum computation hardware.


Supplementary Fig. 2|
Intrinsic lattice parameter of homoepitaxial SrTiO3 thin films grown by molecular beam epitaxy using experimental data from shutter deposition and co-deposition with different Sr content from Ref. 19 and data set from Ref. 24. The out-of-plane lattice parameter taken from Refs. 19 and 24 were converted to intrinsic film lattice parameter using a Poisson ratio for SrTiO3 of 0.244 25 . The difference in the lattice parameter for Sr rich films were attributed to a different degree of SrO planar fault orientation. The linear interpolation was used to link lattice parameter reported for SrTiO3 in Refs. 1-23 to their degree of nonstoichiometry.

Supplementary Fig. 3|
Reflection high-energy electron diffraction (RHEED) images during mapping of the growth window for a Sr flux of 2.5x10 13 cm -2 s -1 , taken along the <100> and <110> azimuth of SrTiO3. The stoichiometric conditions were found for a TTIP gas inlet pressure between 57 and 69 mTorr, which is converted to a TTIP beam equivalent pressure pTTIP of 1.06×10 -6 Torr and 1.77×10 -6 Torr, respectively, using the calibration curve of beam equivalent pressure and gas inlet pressure shown in Supplementary Fig. 4.

Supplementary Fig. 4|
The calibration curve of titanium tetraisopropoxide (TTIP) beam equivalent pressure pTTIP and gas inlet pressure. The data set was fit using a second order polynomic expression. A good agreement was achieved with an R 2 value of 0.99.

Supplementary Fig. 5|
Growth rate extracted from GenX 26 fit of the X-ray data shown in Supplementary   Fig. 7 for SrTiO3 films grown on LSAT as a function of Sr flux measured by the quartz crystal monitor at sample position in units of 10 13 cm -2 s -1 of Sr atoms. The relation can be approximated by a linear function, the fit has an R 2 value of 0.94. For Sr flux values exceeding 2.00×10 14 cm -2 s -1 the effusion cell has been found to become rather unstable with sizeable flux drifts of about 4% per hour resulting in a much smaller growth rate than expected from Sr flux calibrations using the quartz crystal monitor prior to the growth.
The high Sr flux rates needed to demonstrate a growth rate of 600 nm hr -1 were generated from two Sr effusion cells operated in tandem. The significantly lower flux compared to the growth rate determined from film thickness measurements detailed in Supplementary Fig. 7 were attributed to a different tooling factor of the quartz crystal monitor.  Fig. 10| a, HAADF-STEM imaging of an un-annealed SrTiO3 film grown on 3" silicon at 240 nm hr -1 . Low defect, single crystalline film was found at all observed areas, in agreement with xray results. Defects were found more closely concentrated near the interfacial region, and disappeared in the bulk of the film. The surface roughening is due to ion beam damage induced during sample preparation process. b-g, HAADF and energy dispersive x-ray spectroscopy (EDSX) imaging taken at the interface of the sample. All scale bars are 4 nm. Supplementary Fig. 11| Secondary ion mass spectrometry (SIMS) of SrTiO3 on silicon is obtained using a PHI nano TOF. A primary 30 keV Bi3 + ion beam was rastered over a 100 x 100 µm area. The etch rate is approximately 3 nm cycle -1 . Profile covers approximately 85 nm, and begins after the first 75 nm of film has been etched away. The small increase in the carbon signal at the start of the profile is due to adsorption of background gas species in the analytical chamber. A small carbon signal slightly above the detection limit is present throughout the film. The carbon concentration is estimate this to be in the mid-10 17 atoms cm -3 based on previous studies into carbon incorporation of films growth by hMBE 27 . Carbon concentration increases at the interface due to lower cracking efficiency of the carbon-containing precursor at lower temperatures during deposition of the buffer layer.