Introducing Overlapping Grain Boundaries in Chemical Vapor Deposited Hexagonal Boron Nitride Monolayer Films

We demonstrate the growth of overlapping grain boundaries in continuous, polycrystalline hexagonal boron nitride (h-BN) monolayer films via scalable catalytic chemical vapor deposition. Unlike the commonly reported atomically stitched grain boundaries, these overlapping grain boundaries do not consist of defect lines within the monolayer films but are composed of self-sealing bilayer regions of limited width. We characterize this overlapping h-BN grain boundary structure in detail by complementary (scanning) transmission electron microscopy techniques and propose a catalytic growth mechanism linked to the subsurface/bulk of the process catalyst and its boron and nitrogen solubilities. Our data suggest that the overlapping grain boundaries are comparatively resilient against deleterious pinhole formation associated with grain boundary defect lines and thus may reduce detrimental breakdown effects when polycrystalline h-BN monolayer films are used as ultrathin dielectrics, barrier layers, or separation membranes.

Commercial Fe foils (0.1 mm, Alfa Aesar, 99.99% purity) were employed as catalyst material. CVD was undertaken in a customized Aixtron BM3 cold-wall reactor (base pressure 1×10 -6 mbar). After loading of as received foils into the reactor and pumping to base pressure, 3 (reduced pressure) or 4 (high pressure) mbar of NH 3 (or, for reference, 4 mbar H 2 ) were introduced and samples were heated in the pretreatment gas at 100 ºC/min to 750 ºC and then at 50 ºC/min to ~900 ºC. Estimated uncertainty of quoted temperatures is ±25 °C. After reaching ~900 ºC the NH 3 (or H 2 ) was removed. To initiate h-BN growth, borazine (B 3 N 3 H 6 ) vapor was then dosed into the reactor through a leak valve (from a liquid borazine reservoir) to a total pressure of 6×10 -4 mbar. Borazine exposure times for formation of a closed film from NH 3 pretreatment was 480 s, with shorter growth times (45 s) to arrest h-BN growth in the nucleation stage. To end the h-BN growth the borazine leak valve was closed to return the chamber to base pressure and the heater was turned off. Samples then naturally cooled (initial cooling rate ~300 °C/min) in base pressure vacuum.

Transfer of h-BN films onto transmission electron microscopy (TEM) grids
For (scanning) transmission electron microscopy characterization ((S)TEM) measurements the h-BN films were transferred via the electrochemical bubbling method. 2 Films on the catalyst were spin coated with a scaffolding layer of polymethylmethacrylate (PMMA). PMMA-coated samples are then placed on the cathodic side in an electrolysis setup with NaOH solution (1M). During electrolysis H 2 bubbles emerge at the h-BN/catalyst interface, thus releasing the PMMA-coated h-BN from the Fe catalyst foil. The released PMMA/h-BN stack is then washed in deionized water and scooped onto holey carbon TEM grids with regular hole arrays (Quantifoil, ~1 µm hole diameters). The PMMA is then removed by immersing the h-BN covered TEM grids in acetone followed by isopropanol, although we note that this PMMA removal step is incomplete and leaves residues on the films. 3,4

(Scanning) transmission electron microscopy characterization ((S)TEM)
For bright-field (BF) and dark-field (DF) TEM 5 at a large field of view (tens of nm to several µm) and selected area electron diffraction (SAED) a Philips CM200 TEM at 80 kV electron acceleration voltage was employed. The obtained SAED patterns were rotation-corrected with respect to the BF-/DF-TEM images via calibration against atomically resolved STEM images of the same sample locations. False color coded composites of DF-TEM images were prepared by translational and rotational correction of image stacks via registration at common features (e.g. small holes in the film, residual particulate contamination) in BF-and DF-TEM data.
For atomically resolved STEM a Nion UltraSTEM 100 was employed at an electron acceleration voltage of 60 kV, which reduces knock-on damage to the h-BN monolayers. 6,7 In this system the sample rests in ultra-high-vacuum conditions (UHV) of ~10 -9 mbar, which reduces electron-beam-initiated chemical reactions with residual gas species during STEM imaging. High angle annular dark field (HAADF, 80 to 200 mrad) and medium angle dark field (MAADF, 40 to 80 mrad) signal were simultaneously acquired during STEM. Typical beam currents of ~30 pA result for a spot size of ~1 Å 2 in typical electron dose rates directly under the beam of ∼5×10 8 e − Å -2 s -1 , which in turn equate to average dose rates of ~2×10 5 e − Å -2 s -1 for continuous scanning of a 5 nm × 5 nm area as in Fig. 3. Before loading into the STEM, samples were heated in a vacuum of ~10 -5 mbar at ~140 °C for 8 h in order to remove hydrocarbon adsorbates from storage in ambient air.

Complementary h-BN characterization
Scanning electron microscopy (SEM, Zeiss SigmaVP, 2 kV) was done on as grown h-BN samples on the Fe catalyst foils (i.e. before transfer to TEM grids). Additionally, h-BN films were checked after transfer onto SiO 2 (300 nm)/Si wafers using optical microscopy (Nikon eclipse ME600L) and Raman spectroscopy (Renishaw Raman InVia microscope, 532 nm laser excitation), both of which confirmed h-BN film quality equal to ref. 1.

Supporting Figures
Supporting Figure S1: Raw DF-TEM data to Fig. 2(c-e). (a) replots the false colored DF-TEM composite image of a grain boundary (GB) in the closed h-BN film from Fig. 2(c) and the inset replots the corresponding SAED pattern in Fig. 2(d). (b) shows the individual DF-TEM images which were used to produce the false colored composite in (a) and Fig. 2(c). The DF-TEM images were acquired via the correspondingly colored/labelled SAED reflections in the inset to (a) and Fig. 2(d). (c) replots the composite overlay of the DF-TEM data from Fig. 2(e) and was produced by intensity thresholding and a Boolean AND operation to the DF-TEM data in (b). S4 Supporting Figure S2: Unfiltered HAADF (a,b) and MAADF (c) STEM data corresponding to Fig. 2(f). (d) shows the Fourier transforms (FTs) to (c), with the spots indexed and color circled corresponding to Fig. 2(c,d), and clearly confirms that two monolayer h-BN grains merge in a turbostratic bilayer overlapping GB region of limited width. Figure S3: (a) Double Gaussian filtered 6 HAADF STEM image of a region below the overlapping GB in Fig. 2(f). (b) Intensity profile extracted from (a) along the yellow line indicated in (a). The intensity profile was extracted (after double Gaussian filtering to reduce contribution from probe tails 6 ) by subtracting the remaining averaged intensity at intensity minima between atoms followed by normalizing the intensity at suspected B sites to 1. This yields an intensity ratio between B and N sites consistent 6 with monolayer h-BN, thus confirming the monolayer nature of the two merging grains in Fig.  2(f). (c) Double Gaussian filtered 6 HAADF STEM image of a region above the overlapping GB in Fig. 2(f), where after extended imaging a hole to vacuum was sputtered into the h-BN layer. The direct step-free sputtering to vacuum further confirms the monolayer nature of the h-BN grain. 8 S6 Supporting Figure S4: FT of the (unfiltered) STEM-MAADF data corresponding to the bilayer region in the overlapping GB in Fig. 3(b). S7 Supporting Figure S5: SEM micrograph of h-BN growth on Fe catalyst foils after H 2 pretreatment (4 mbar) and 480 s borazine exposure. In contrast to the closed h-BN monolayer film from NH 3 pretreatment (Fig. 2(b)), N-free pretreatment with H 2 leads to growth of isolated multilayer h-BN pyramids. 1 S8 Supporting Figure S6: Raw TEM data to Fig. 4(a-b). (a) replots the false colored DF-TEM composite image of an overlapping GB from Fig. 4(a) and the inset replots the corresponding SAED pattern from Fig. 4(b). (b) shows the BF-TEM corresponding to (a). (c) shows the individual DF-TEM images which were used to produce the false colored composite in (a) and Fig. 4(a) via the correspondingly colored/labelled SAED reflections in the inset to (a) and Fig. 4(b). (d) shows a composite overlay of the DF-TEM data from (c) after intensity thresholding and processing via a Boolean AND function to highlight the overlapping region in the GB, corresponding well to the STEM data in Fig. 4(c).