Growth of Epitaxial Oxide Thin Films on Graphene

The transfer process of graphene onto the surface of oxide substrates is well known. However, for many devices, we require high quality oxide thin films on the surface of graphene. This step is not understood. It is not clear why the oxide should adopt the epitaxy of the underlying oxide layer when it is deposited on graphene where there is no lattice match. To date there has been no explanation or suggestion of mechanisms which clarify this step. Here we show a mechanism, supported by first principles simulation and structural characterisation results, for the growth of oxide thin films on graphene. We describe the growth of epitaxial SrTiO3 (STO) thin films on a graphene and show that local defects in the graphene layer (e.g. grain boundaries) act as bridge-pillar spots that enable the epitaxial growth of STO thin films on the surface of the graphene layer. This study, and in particular the suggestion of a mechanism for epitaxial growth of oxides on graphene, offers new directions to exploit the development of oxide/graphene multilayer structures and devices.

: I-V curve measured on graphene after STO growth Fig. S1 shows I-V curve that was measured on Au/graphene of two corners across sample after STO growth. The estimated graphene resistivity was below ~100 µ·cm, which is not very far away from the values measured for conventional graphene layers.
Nevertheless, this is a clear indication that the graphene layer is still conductive after the STO thin film deposition using PLD.

SI. 3. Chemical evaluation of the Gr/STO interface on STO/Gr/STO sample
The chemical evaluation of the Gr/STO interface was performed on an aberrationcorrected (at the image plane) FEI Titan 80-300 scanning/transmission electron microscope (S/TEM) in STEM mode using a Gatan Tridiem electron energy-loss spectrometer (EELS) with an energy resolution of 0.7eV.
(a) (b) Figure S4: Energy electron-loss spectra acquired at the interface and on the STO substrate in (a) the core-loss region detailing the Carbon K, the Titanium L3,2 and the Oxygen K edges; and (b) in the low-loss region monitoring the plasmon loses; (c) quantitative analysis of the spectrum image; (d) taken along the thin film structure reveals high carbon content at the interface associate with an overall reduction both in Ti and O relative composition.
To evaluate the chemical character of the carbon-based interface, electron energy-loss spectra were acquired extending from the substrate to the interface. Fig. S4(a) depicts the core-loss excitations of the C K, Ti L3,2, and O K edges, which represent the local density of states of the material. The first C peak corresponds to the * transition and the shift to higher energy of the spectrum at the interface (285.4eV) compared to the substrate one (284.8eV) confirms the graphene-nature of the interface. [ 1 ] The * transition of the interface is highly reduced and this could be attributed to partly oxidized structure. Indeed, inspection of the plasmon loss region at the interface, Fig.   S4(b), reveals two peaks at 23.5 eV and 29.5 eV, which correspond to graphene oxide and graphene respectively. Furthermore, quantitative analysis of the spectrum image shown in Fig. S4(c) reveals that the relative amount of both Ti and O is highly reduced at the interface followed by an increase in C content. This can be visually and intuitively observed by the dark colour at interface in Fig. S4(d). It is noted that the cross-sectional TEM sample preparation process always results in carbonaceous amorphization of the sides of the lamellae. However, the results suggest that the crystalline nature of the graphene interfacial layer is preserved.

SI. 4. Density functional theory (DFT) modelling
All electron, local atom centred Gaussian basis set, calculations were performed using the CRYSTAL14 software [ 2 , 3 ] . The basis sets for STO were adapted for use in condensed systems and of triple valence quality (ie: three independent radial functions for each valence electron) with polarisation functions. These basis sets have been described and used in previous studies of SrO [4] and TiO2 [5,6] . For the graphene sheet a modified 631G* basis set was used. [ 7 ] Electronic exchange and correlation were described within density functional theory in the B3LYP approximation which has been used extensively in previous studies of titanates [5,6] and graphene derived nanostructures [7] . Long range London dispersion interactions were included through the empirical correction scheme proposed by Grimme [8] . For the C and O centres the atomic radius (Angstrom) and C6 coefficient (Jnm 6 mol -1 ) were set to (1.452, 1.75) and (1.342, 0.70) respectively and the overall scaling (s6) set to 1.05 which are the default values in the CRYSTAL14 code [2,3] . shrinking factor 2, which results in 3 irreducible k-points in the first Brillouin zone, was adopted as it is sufficient to define the total energy to within 10 -4 eV per cell. Structural optimization of internal coordinates only was performed by relaxing all atoms of the slab model, using the Broyden-Fletcher-Goldfarb-Shanno scheme, as implemented in CRYSTAL14 [2] . The thresholds for the maximum and the rms forces (the maximum and the rms atomic displacements) were set to 0.00045 and 0.00030 (0.00180 and