Anisotropic Etching of Hexagonal Boron Nitride ... - ACS Publications

Nov 14, 2017 - Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States ... Department of Chemistry and Biochemistry, New Mexico State...
0 downloads 0 Views 2MB Size
Letter pubs.acs.org/NanoLett

Anisotropic Etching of Hexagonal Boron Nitride and Graphene: Question of Edge Terminations Yijing Y. Stehle,*,† Xiahan Sang,† Raymond R. Unocic,† Dmitry Voylov,‡ Roderick K. Jackson,⊥ Sergei Smirnov,*,§ and Ivan Vlassiouk*,† †

Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States University of Tennessee, Knoxville, Tennessee 37996, United States ⊥ National Renewable Energy Laboratory, Golden, Colorado 80401, United States § Department of Chemistry and Biochemistry, New Mexico State University, Las Cruces, New Mexico 88011, United States ‡

S Supporting Information *

ABSTRACT: Chemical vapor deposition (CVD) has been established as the most effective way to grow large area two-dimensional materials. Direct study of the etching process can reveal subtleties of this competing with the growth reaction and thus provide the necessary details of the overall growth mechanism. Here we investigate hydrogen-induced etching of hBN and graphene and compare the results with the classical kinetic Wulff construction model. Formation of the anisotropically etched holes in the center of hBN and graphene single crystals was observed along with the changes in the crystals’ circumference. We show that the edges of triangular holes in hBN crystals formed at regular etching conditions are parallel to B-terminated zigzags, opposite to the N-terminated zigzag edges of hBN triangular crystals. The morphology of the etched hBN holes is affected by a disbalance of the B/N ratio upon etching and can be shifted toward the anticipated from the Wulff model N-terminated zigzag by etching in a nitrogen buffer gas instead of a typical argon. For graphene, etched hexagonal holes are terminated by zigzag, while the crystal circumference is gradually changing from a pure zigzag to a slanted angle resulting in dodecagons. KEYWORDS: 2D Materials, APCVD, edge termination, hydrogen etching, hBN, boron nitride, graphene

B

eing a two-dimensional (2D) insulator with alternating sp2 hybridized B and N atoms in a honeycomb structure, hBN not only has a similar structure to graphene (only 1.7% of lattice mismatch) but also shares many physical properties, such as high optical transparency, large thermal conductivity, high mechanical strength, and superior chemical inertness.1−9 CVD growth has been established as the most effective way to grow large area 2D materials and a better understanding of the growth process is imperative for increasing the material’s quality necessary for the large scale industrial applications. There are many similarities in the optimal conditions for CVD growth of hBN and graphene such as the choice of catalytic substrates and the need for hydrogen in the reactor’s atmosphere besides the precursor. It was shown that hydrogen plays a dual role in the graphene growth process, it is required as a cocatalyst for dehydrogenation of hydrocarbon but also etches the growing crystal edges.10 A competition between the two defines not only the overall rate but the shapes of the individual domains. The detailed role of hydrogen in growth of hBN remains elusive but it was reported that the nucleation density, crystallinity, domain size, morphology, edge structure, and the number of layers is affected by it.11−14 Focusing on just the etching process can provide the necessary details and help to delineate the overall growth © 2017 American Chemical Society

process which is the focus of this work. Previously was reported that etching of 2D materials can result in various etching patterns and a competition between various factors such as the growth atmosphere, synthesis temperature, and substrate’s chemical composition can define the dominant one.10,11,13,15−24 The etching is often anisotropic as has been demonstrated and extensively studied for graphene etching by hydrogen.10,25−27 Recently it was shown that etching of hBN on Cu by hydrogen could be also induced but required significantly higher temperatures than for graphene, greater than 1020 °C.11,13 Controlled etching can provide a scalable route for fabrication of highly desired nanostructures providing a convenient way for 2D material property design.17,19,26 For example, etching process dictates the edges structure which in turn have a profound effect on various properties of graphene and hBN nanostructures.28−31 Thus, control of termination through etching is a potential way of engineering edge structure and may provide a route for scalable 2D device fabrication. Here we present our study of the hydrogen-induced etching of hBN and graphene grown via atmospheric pressure chemical Received: July 5, 2017 Revised: November 13, 2017 Published: November 14, 2017 7306

DOI: 10.1021/acs.nanolett.7b02841 Nano Lett. 2017, 17, 7306−7314

Letter

Nano Letters

Figure 1. (a) Schematic diagram illustrating the overall process of hBN and graphene growth and etching. Optical images of the oxidized hBN samples on a CuNi substrate: (b) complete monolayer and (c) incomplete coverage with single crystals. Area covered with hBN shows gray/white color while the oxidized area, that is, without protective hBN overlayer, has an orange color. The images show unetched (left) and etched (middle) samples, as well as the corresponding histograms of the orientations of edges (right). Clear 60° separation between the peaks in the histogram suggest that epitaxial growth on CuNi substrates is similar to that on pure Cu substrates.12 (d) Optical images of graphene: complete monolayer (left), etched monolayer (middle), and etched single crystal (right).

confirm that its crystals terminated by ZZ edges during growth, slowly change to slanted 19° terminations during etching by hydrogen while the holes etched in the center of each crystal is always terminated by ZZ edge. Figure 1a shows the process schematics for hBN and graphene growth by APCVD and etching. Growth was performed similar to the previous works and discussed in detail elsewhere.10,12 CuNi alloy with 10% of Ni was chosen as a catalytic substrate due to its higher melting temperature (experiments were performed at 1065 °C unless specified otherwise) and lower nucleation density while having a negligible solubility of carbon similar to pure copper.6 Similar results were obtained on pure copper substrates at slightly different conditions: CuNi substrates required higher precursor concentrations to initiate material growth suggesting higher etching rates on CuNi substrates compared to pure Cu. After deposition, the foils were oxidized in air on a hot plate which provided a convenient way to visualize growth/etching outcomes; areas protected by hBN or graphene were not

vapor deposition (APCVD) on CuNi and pure Cu catalytic foils. The etching process is induced by stopping the flow of the precursor after synthesis while keeping the other parameters unchanged. Higher temperature compared to previous studies, 1065 °C, allowed greater etching rates and varying the growth conditions provided the controllable means for altering shapes of the structures to etch from single layer islands to continuous monolayer and multilayered structures. We illustrate that hydrogen activated by the catalytic substrate, plays a key role in the observed etching process resulting in triangular shapes of the hBN islands with their edges parallel to nitrogen-terminated zigzag (N-ZZ) but the holes etched in continuous monolayer or inside the islands have edges parallel to the boron zigzag (BZZ). Substitution of buffer argon by molecular nitrogen shifts equilibrium between the nitrogen species and results in holes which have N-ZZ termination. Significantly lower etching rates of hBN on other substrates, such as SiO2, strongly suggests that adsorbed atomic hydrogen is responsible for the observed etching. We compare the results with graphene, for which we 7307

DOI: 10.1021/acs.nanolett.7b02841 Nano Lett. 2017, 17, 7306−7314

Letter

Nano Letters

Figure 2. (a) SEM image of typical triangular hBN single crystals on a CuNi substrate. SEM allows for unambiguous discrimination between the adlayers, etched holes, and single layer crystals. (b) SEM image of a triangular hBN crystal transferred on a TEM grid. White dashed lines highlight the triangular hBN crystal; adlayers are also visible in the middle (c) STEM image of the same crystal; nitrogen atoms show a greater brightness. The triangles in both images illustrate the correlation of the microcrystal edges with the atomic structure suggesting the N-ZZ termination of the crystal edge. (d) Typical AFM image of an etched triangular hBN single crystal transferred onto SiO2/Si.

visualization. We found that, like with graphene,32 SEM is very efficient not only for counting the number of hBN layers but also for identifying the etched holes. For example, SEM image in Figure 2a shows various options: single layer hBN triangles with triangular hole and multilayered hBN with no holes in the middle. The etched hole in the center of hBN single layer islands show similar brightness as the substrate outside the island, whereas the multilayered islands have their layers with alternating contrast (darker or lighter appearance) depending on the layer number and the imaging conditions, and the second layer always has a different appearance compared to the uncovered substrate. Atomic force microscopy (AFM) topography provides an additional more direct method for distinguishing between the etched holes and adlayers corroborating the assignment from SEM. Figure 2d shows an AFM image of a typical etched hBN single crystal transferred onto SiO2/Si substrate with the representative topographic profile unambiguously confirming the same height in the central hole as on the substrate outside the domain. Observed high roughness of the hBN crystal is likely due to PMMA leftovers from the transferring procedure. Raman spectroscopy is another method typically used for distinguishing the holes from the adlayers in graphene but it is less effective for hBN due to its very weak Raman signal. Mutual orientation of graphene adlayers varies and, most of the time, the adlayers appear rotationally misalligned.10 Different from simple CVD approach is required for synthesis of Bernal-stacked graphene bilayers.33 hBN multilayers are different in always appearing with aligned layers having the sides of the triangular adlayers parallel to the first (bottom) layer. The resulting aligned hBN layers together with epitaxial growth suggest stronger interactions between the hBN layers and the CuNi substrate than in the case of graphene. Theoretical calculations suggest that hBN triangular crystals grown on Cu substrate should be terminated by the most stable nitrogen zigzag edges (N-ZZ).34 It is difficult to resolve

oxidized, while unprotected bare metal substrate changed its color due to oxidation. For example, Figure 1b (left) demonstrates a CuNi surface fully covered by an hBN monolayer and thus gives a uniform white appearance under optical microscope even after oxidation. On the other hand, the identically prepared but etched sample shows multiple triangular red-colored holes (Figure1b, middle). Figure 1c shows individual triangular-shaped hBN crystal samples which were synthesized during shorter time, that is, when the individual hBN crystals did not coalesce yet into a complete hBN monolayer. Similarly prepared and etched afterward sample is shown in Figure 1c (center). Both etched samples with separated triangular hBN crystals (Figure 1c) and etched triangular holes in monolayer (Figure 1b) show uniform orientations of the edges within an individual CuNi domain. Histograms on the right in Figure 1b,c confirm the same orientation of triangles with 60° separation suggesting epitaxial growth of hBN on CuNi substrates for the growth conditions used here. Similar epitaxial growth was seen for hBN grown on pure Cu foils at lower temperatures.8,12 Close analysis of Figure 1c reveals that the majority of separated single hBN crystals (white color) after etching have triangular etched holes (red color) in their centers. These triangular holes are in the centers, that is, apparently had originated at the same sites that served as the nucleation points for the crystals’ growth. The same nucleation points of growth and etching are observed for graphene as well (Figure 1d). In both cases, they are likely the surface defects or uncontrollable contaminants. For graphene, the etched holes always have hexagonal shape. When hBN is grown at a higher rate due to a higher local concentration of borazane precursor, the crystals have multiple layers appearing as “wedding cake” structures also centered in the middle of hBN crystals. The top layers in the multilayers are not visible in optical microscopy after oxidation and scanning electron microscopy (SEM) is more suitable for their 7308

DOI: 10.1021/acs.nanolett.7b02841 Nano Lett. 2017, 17, 7306−7314

Letter

Nano Letters

Figure 3. Influence of the buffer gas composition on hBN etched holes’ shapes. Optical microscope (left) and SEM (middle) images along with the crystal and holes’ termination schematics (right). (a) Etched islands with argon used as a buffer gas. The outer crystal edge is terminated by the nitrogen terminated zigzag (N-ZZ), while the inner edge is always parallel to the boron terminated zigzag (B-ZZ). (b,c) Etched islands with nitrogen used as the buffer gas demonstrate more options for terminations of the etched hole from triangular with B-ZZ, to truncated triangular/hexagonal with both B-ZZ and N-ZZ (b), to fully inverse triangular with N-ZZ (c).

experimentally the atomic structure of the crystal edge but one can correlate the orientation of the 2D crystal from the atomically resolved STEM crystal structure with the edge direction from SEM,35 as shown in Figure 2b,c. The atomic structure unambiguously suggests that the edge of hBN triangle grown on CuNi catalyst is parallel to the N-ZZ termination edge. Thus, experimentally derived crystal terminations (N-ZZ or B-ZZ) by correlation of STEM images with observed crystal shapes can only point to the edges’ directions and not to exact terminations. When using argon as a buffer gas, the etched triangular holes in the hBN crystals always have their sides parallel to the outer sides of the triangular crystal (Figure 3a) which points to their different termination, as illustrated by cartoon on the right. Because the triangular hBN crystals are terminated by edge parallel to N-ZZ (Figure 2b,c) then the etched holes must be terminated by the edge parallel to boron zigzag (B-ZZ). The reason for such inner (B-ZZ) and outer (N-ZZ) terminations is puzzling. When using molecular nitrogen instead of argon as the buffer gas, truncated triangular and hexagonal holes were

observed (Figure 3b). Furthermore, some triangular etched holes were even fully inversed (up to 10% of the total number of holes), see Figure 3c. Thus, changing the buffer gas from argon to nitrogen caused transition for the hole termination from B-ZZ to N-ZZ. To the best of our knowledge, anisotropic etching process of hBN by hydrogen has not been previously discussed and thus requires special consideration. As a matter of fact, the etching process and its role in the overall growth is typically not included in the theoretical picture. We start by comparing with a better studied case of graphene. Graphene crystals grown at high H2/CH4 ratio always have hexagonal shapes terminated by zigzag (ZZ) edge.10,36 Similar to hBN, etching of graphene crystals results in hole formation in their centers (Figure 4). During initial stages of the etching process, the holes always have edges parallel to the crystal’s hexagonal circumference which points that both, circumference and the hole, have the same ZZ termination, confirming previous results.10,15 The etching rate is linear and for our conditions equals to ∼0.7 μm/min. The observed ∼4 7309

DOI: 10.1021/acs.nanolett.7b02841 Nano Lett. 2017, 17, 7306−7314

Letter

Nano Letters

Figure 4. Evolvement of graphene crystal shapes during etching. (a) SEM images of graphene crystals etched for different time (left) and hole etching kinetics (right). Holes always have hexagonal shape with ZZ termination (highlighted by blue color on the last image). Circumference is hexagonal for large crystals/small etching times with rounded edges parallel to ZZ. The shape gradually changes to dodecagon having edges ∼19° from ZZ with time. Change is faster for smaller initial crystals (highlighted by red color on the last image). (b) SEM image of graphene crystal etched for 5 min (left) and optical images of graphene crystals etched for 20 min (right). On the optical images, red/yellow areas are oxidized CuNi catalyst, while white areas are CuNi protected by graphene. During etching, the large crystals maintain their initial hexagonal shape, while smaller crystals transform faster to dodecagons.

previously also illustrated by Ma et al. on Pt substrate by varying the flow of methane at constant hydrogen and allowing to switch from growth to etching.15 It is important to note that initially larger hexagonal crystals require longer etching time to fully change their shape to dodecagons because there is more material that needs to be etched away as is clearly seen in Figure 4b for the two neighboring crystals of different sizes. The Wulff construction for hBN growth/etching is schematically illustrated in Figure 5d. During growth, the slowest N-ZZ edge defines the triangular shape of single crystals. During etching, enlarging of a triangular hole in the crystal center should be defined by the slowest etching edges, while the circumference should transform as dictated by the fastest etching edge. Obviously, the case of hBN is more complicated compared to graphene due to a binary composition which makes the assignment of the slow and fast edges dependent on the relative concentrations of free B and N and thus the shape of crystal circumference and its hole changes with time and location in the CVD tube. The surface concentrations of B and N species are not necessarily the same inside the hole and in the space between individual crystals. Because nitrogen product of etching more readily departs from the surface in the stable

min lag time corresponds to the time required to displace the precursor at the sample location in the CVD tube (Figure 4a). The shape of circumference during etching depends on the initial size of the crystal; smaller crystals change their shape to dodecagons (with ∼19° to the ZZ edge) faster than the large ones (Figure 4b). The kinetic Wulff construction model15 suggests that during growth the edge is terminated by the slowest growing crystal direction. For graphene, it is the ZZ edge which leads to the typical hexagonal shape of individual graphene domains with ZZ termination. Note that experimentally this is observed at sufficiently high concentrations of hydrogen, that is, when the overall growth rate is balanced with noticeable etching. The same Wulff construction model suggests that during etching the hole should be etched with the slowest etching edges as well but the circumference of individual domains should have the fastest etching edges instead (Figures 5a,b). In graphene, the fastest growing edges are the fastest etching edges as well and the etched islands thus have hexagonal holes with ZZ edges and the circumference gradually changes with time from the original ZZ terminated hexagon to a dodecagon with 12 identical fastest edges slanted at ∼19° from ZZ (Figures 4 and 5c). It was 7310

DOI: 10.1021/acs.nanolett.7b02841 Nano Lett. 2017, 17, 7306−7314

Letter

Nano Letters

Figure 5. Schematics of crystal circumference/hole shape evolution during growth/etching (kinetic Wulff construction model); fast edges are shown in red, whereas slow edges are blue. (a) During etching process, circumference of a crystal is dominated by the fastest etching edges (red), while the slowest etching edges (blue) will eventually disappear, opposite to the growth step. (b) Hole evolution during etching is dominated by the slowest etching edges instead. (c) Graphene crystal growth/etching schematics. During growth, a crystal is terminated by the slowest ZZ edges resulting in hexagonal shape. Etching exposes the fastest edges on the circumference (∼19° to ZZ) resulting in dodecagon, while the formed hole is terminated by the slowest ZZ edges.15 (d) Evolution for hBN. Growing crystal is terminated by the slowest edges (usually N-ZZ, at least for stoichiometric B/ N). Etching should result in exposure of the fastest edges on the circumference and the slowest edges should terminate the hole, that is, opposite orientation of triangles for the same stoichiometry.

ference termination; in general, the edge tends to be terminated by the species which are in excess on the catalytic surface. On the basis of the analysis of Zhang et al.,34 at N-rich conditions the growth at the N-ZZ edge needs to overcome a substantial nucleation barrier and becomes the slowest among all edge directions, thereby shaping the h-BN islands into N-terminated triangles (see Figure 5d). Under B-rich condition, the hBN shapes can evolve into truncated triangles or hexagons with additional B-terminated edges. As was described before,12 in our conditions of growth with solid borazane precursor, the triangular shaped islands were typically observed in the front end of the deposition tube where the B/N ratio is close to one. Further into the tube the relative amount of nitrogen gradually gets depleted and under such growth conditions the individual crystals change their shapes into the truncated triangles with the shorter sides corresponding B-ZZ terminations. Thus, Brich conditions in the hole lead to the B-ZZ edges. We can

molecular form, its relative local concentration is lower in the hole or relative boron concentration is higher in the hole. In the outside area after long etching (large free area of metal catalyst), both B and N concentrations are similarly low because B can diffuse away from the edge. The interaction with the substrate also affects the outcome: on the copper (111) substrate, the N-ZZ edge is energetically favored at N-rich or equilibrium condition, whereas the B-ZZ is preferred at B-rich condition. However, at a very low relative concentration of B, extra N atoms can reconstruct the B-ZZ edge after attaching into a more stable Klein B (KB) edge. Similarly, B attachment at high B concentrations alters the N-ZZ edge into the Klein N (KN) edge.34,37,38 Similar to graphene,15 the kinetic Wulff consideration is likely defining the observed shapes for hBN on Cu and CuNi alloy. Interestingly, both thermodynamic and kinetic arguments give similar answers to the question of circum7311

DOI: 10.1021/acs.nanolett.7b02841 Nano Lett. 2017, 17, 7306−7314

Letter

Nano Letters

Figure 6. Schematic (top row), SEM images (middle row), and their zoomed-in versions with higher magnification (bottom row) of (a) hBN monolayer with multiple adlayers, (b) etched sample presented in (a) clearly shows the triangular holes in hBN monolayer parallel to edges of adlayers (6 min etching time); (c) similar to (b) but with longer etching process (8 min etching time) resulting in complete disappearance of hBN monolayer having only adlayers on bare CuNi substrate. Argon was used as a buffer gas. Reported etching times include 4 min “lag” time required for precursor displacement in the CVD tube (Figure 4a).

substrate is necessary for the high etching rate. Furthermore, etching was not observed when hBN was transferred onto a SiO2/Si substrate confirming that hydrogen dissociation on metal foil surface is responsible for the etching, similar to the case of graphene. The rate of etching is strongly temperature dependent. It takes ∼3 min (beyond the 4 min lag) to etch the whole monolayer hBN film away at 1065 °C but longer than 15 min (∼1.5 μm/min) is required at 1020 °C. No evidence of hBN etching was observed after 30 min at 900 °C. The hydrogen-induced etching process of hBN and graphene crystals was evaluated experimentally for 2.5% H2 in either Ar or N2 buffer gas at atmospheric pressure and compared to the predictions of the kinetic Wulff construction model. Etching process initiates on the defects at the crystal centers. We showed that hBN hole termination during etching strongly depends on the B/N ratio on the catalyst surface which can be altered by the choice of a buffer gas. The edges termination in the holes is changed from B- to N-terminated zigzags upon addition of molecular nitrogen to the CVD reactor (similar to what was previously observed for hBN crystal circumference termination during growth).12 Change in the hole edge termination is observed as evolution of the hole shape from pure triangular (B-ZZ) to hexagonal (both N-ZZ and B-ZZ) and further to the inverted triangular (N-ZZ). We show that at normal conditions the etching rates of the hole B-ZZ termination (∼10 μm/min, 1065 °C) is at least 1 order of magnitude faster than that for the outer N-ZZ edge (