Synthesis of large-area multilayer hexagonal boron nitride for high material performance — TEI-tagged view

Soo Min KimpersName, Allen HsupersName, Min Ho ParkpersName, Sang Hoon ChaepersName, Seok Joon YunpersName, Joo Song LeepersName, Dae-Hyun ChopersName, Wenjing FangpersName, Changgu LeepersName, Tomás PalaciospersName, Mildred DresselhauspersName, Ki Kang KimpersName, Young Hee LeepersName, Jing KongpersName (2015date). “Synthesis of large-area multilayer hexagonal boron nitride for high material performance.” Nature Communicationstitle·j.

Object of study — multilayer h-BNterm·dom films grown by CVDterm·mth on Fe foilterm·key and devices built on them

Research question — whether large-area multilayer h-BNterm·dom, so far absent, can be grown by CVDterm·mth

Contribution — a new synthesis technique for multilayer 2Dterm·dom materials and a high-quality h-BNterm·dom substrate for 2Dterm·dom nanoelectronics

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

Synthesis of large-area multilayer hexagonal boron nitrideterm·dom for high material performance Abstract Although hexagonal boron nitrideterm·dom (h-BNterm·dom) is a good candidate for gate-insulating materials by minimizing interaction from substrate, further applications to electronic devices with available two-dimensional semiconductors continue to be limited by flake size. While monolayer h-BNterm·dom has been synthesized on Pt and Cu foil using chemical vapour depositionterm·mth (CVDterm·mth), multilayer h-BNterm·dom is still absent. Here we use Fe foilterm·key and synthesize large-area multilayer h-BNterm·dom film by CVDterm·mth with a borazineterm·key precursor. These films reveal strong cathodoluminescenceterm·mth and high mechanical strength (Young's modulusterm·dom: 1.16±0.1 TPa), reminiscent of formation of high-quality h-BNterm·dom. The CVD-grown grapheneterm·key on multilayer h-BNterm·dom film yields a high carrier mobilityterm·dom of ∼24,000 cm 2 V −1 s −1 at room temperature, higher than that (∼13,000 2 V −1 s −1 ) with exfoliated h-BNterm·dom. By placing additional h-BNterm·dom on a SiO 2 /Si substrate for a MoS 2term·key (WSe 2term·key ) field-effect transistorterm·dom, the doping effect from gate oxide is minimized and furthermore the mobility is improved by four (150) times.

In addition to metallic grapheneterm·key and semiconducting transition metal dichalcogenidesterm·dom, hexagonal boron nitrideterm·dom (h-BNterm·dom) is an insulating two-dimensional (2Dterm·dom) material, in which all these materials serve as primary components for soft electronics with high flexibility and transmittance 1 2 . Since all the atoms in 2Dterm·dom layer are exposed to the surface, the related physical and chemical properties are strongly influenced by adjacent materials and sometimes surface corrugation 3 4 . Therefore, a special care is required to deal with atomically thin-layered materials. h-BNterm·dom has several unique physical and chemical properties of its own, which has potential applications as a dry-lubricant 5 , passivation layer 6 and deep ultraviolet emitter 7 8 9 . The h-BNterm·dom layer has been recently demonstrated to be an ideal substrate for 2Dterm·dom materials due to its atomic flatness, large optical bandgap, superb mechanical strength, absence of dangling bonds and low dielectric screening 10 11 12 13 14 15 16 .

So far, tremendous efforts have been provided to large-area synthesis of 2Dterm·dom materials. For example, a meter-scale polycrystalline and centimetre-scale monocrystalline monolayer grapheneterm·key have been synthesized on copper foil 17 18 . A centimeter-scale monolayer transition metal dichalcogenideterm·dom (TMdCterm·dom) has been synthesized on an insulator or Au substrate 19 20 21 22 . Several methods including ultrahigh vacuum chemical vapour depositionterm·mth (CVDterm·mth) on various single-crystalline transition metals including Ni (111) 23 24 , Cu (111) 25 26 , Pt (111) 27 and Ag (111) 28 have been reported to achieve large-area monolayer h-BNterm·dom. More recently, more scalable synthesis techniques on polycrystalline Cu and Pt foils were used to obtain monolayer h-BNterm·dom by CVDterm·mth 11 29 30 31 . Monolayer h-BNterm·dom due to its ultrathin nature is very useful for a variety of applications, such as growth templates 32 , tunnelling barriers 33 34 and atomic membranes 35 . Yet, multilayer h-BNterm·dom is in general highly desired 3 36 37 38 for numerous real device applications, such as dielectric layers, atomically flat and dangling bond-free substrates or mechanical membranes.

For the synthesis of multilayer h-BNterm·dom, liquid–metal methodterm·mth 39 , solid-phase epitaxy growth 10 40 , direct growth in CVDterm·mth 13 and co-segregation method 41 have been reported. The liquid–metal methodterm·mth so far produces the highest crystallinity but the bulk crystals are produced, resulting in isolated small-size flakes of h-BNterm·dom after exfoliation 3 . Solid-phase epitaxial growthterm·mth involves the low-temperature deposition of precursors followed by a post-annealing process, which provides large-area h-BNterm·dom film with a controllable thickness; however, expensive single-crystalline substrates such as Ni (111) or Ru (0001) are required 10 40 . Other methods such as the direct growth with atmospheric pressure CVDterm·mth on copper foil 13 or the diffusion and co-segregation of boron and nitrogen atoms from a Fe–Ni alloy 41 leaves poor crystallinity of h-BNterm·dom, making it difficult to serve as a substrate for 2Dterm·dom materials. Therefore, improving the crystallinity of large-area h-BNterm·dom films on a cheap substrate is still required for electronic device applications.

Here we report large-area h-BNterm·dom films grown by CVDterm·mth using Fe foilterm·key. The thickness of the h-BNterm·dom (5–15 nm) is controlled by the cooling rateterm·dom, that is, the segregated boron and nitrogen atoms that are precipitated in Fe substrate at high temperature. X-ray diffraction measurements exclusively revealed the (0002), (0004) and (0006) peaks, indicating that the layers are well aligned perpendicular to the c axis. A Young's modulusterm·dom of 1.16 TPa for a 15-nm-thick h-BNterm·dom is measured by nanoindentationterm·mth using atomic force microscope (AFM) tip. All these values are in good agreement with theoretical predictions, indicating the high crystalline quality of the grown multilayer h-BNterm·dom films. Field-effect transistorsterm·dom (FETsterm·dom) with CVD-grown monolayer grapheneterm·key, monolayer MoS 2term·key and monolayer WSe 2term·key are fabricated on the grown multilayer h-BNterm·dom substrates, achieving carrier mobilities as high as ∼24,000, 40 and∼9 cm 2 V −1 s −1 at room temperature, respectively. The reported grapheneterm·key mobility is the highest value among those of the previous reports with CVD-grown grapheneterm·key samples on CVD-grown h-BNterm·dom substrates.

Results Synthesis and materials characterization of multilayer h-BNterm·dom To synthesize multilayer h-BNterm·dom on Fe foilterm·key, a single-zone CVDterm·mth furnace is equipped with a bubbling system for liquid borazineterm·key, as shown in Fig. 1a 29 . Typically, the growth was performed at a temperature of 1,100 °C for 30 min, followed by various cooling ratesterm·dom (see the Methods section for details). Figure 1b,c shows optical photographs of two 3 × 3-cm 2 area h-BNterm·dom samples of as-synthesized on Fe and after transfer onto SiO 2 /Si, respectively. The size of h-BNterm·dom is only limited by the CVDterm·mth chamber size. The scanning electron microscopyterm·mth (SEMterm·mth) images in Fig. 1d,e are taken of the as-grown h-BNterm·dom on Fe foilterm·key; regions of uniform thickness have typical lateral dimensions of tens of micrometre and without any deposited particles being observed. Several wrinklesterm·dom can be also seen due to the coefficient of thermal expansion mismatch between the film and the underlying substrate, similar to grapheneterm·key grown on metallic substrates 42 . Once the h-BNterm·dom film is delaminated from Fe substrate through a bubble transfer process 29 , the Fe substrate can be reused for future growths. Even after 10 synthesis and transfers of h-BNterm·dom on the reused Fe substrate, the morphology and quality of the growth results were confirmed to be almost identical ( Supplementary Fig. 1 ).

While reports from the literature have shown both surface-mediated growthterm·dom, as well as bulk precipitationterm·dom of h-BNterm·dom experimentally, the synthesis of h-BNterm·dom on Fe foilsterm·key is strongly dependent on the cooling rateterm·dom of the sample. This strongly suggests that the precipitationterm·dom plays an important (or even dominant) role in the growth. The h-BNterm·dom was grown for 30 min with a borazineterm·key precursor at 1,100 °C and then fast (30 °C min −1 ) and slow (5 °C min −1 ) cooled until 700 °C without supplying additional borazineterm·key ( Supplementary Fig. 2 ). Supplementary Figure 2b–f shows the SEMterm·mth images of h-BNterm·dom on Fe foilterm·key for fast ( Supplementary Fig. 2b,c ) and slow ( Supplementary Fig. 2e,f ) cooling, respectively. In the case of fast cooling ( Supplementary Fig. 2b,c ), a relatively thin h-BNterm·dom film (an average thickness of 11.9 nm with a s.d. of 3.3 nm) with triangular islands covered the Fe surface ( Supplementary Figs 3 and 4 ). The wrinklesterm·dom of the background h-BNterm·dom film are all merged together, indicating that the h-BNterm·dom film is continuous; on the other hand, slow cooling results in a thicker h-BNterm·dom film (an average thickness of 17.8 nm with a s.d. of 6.1 nm over the whole area ( Supplementary Fig. 2e,f ). Moreover, the grain size of h-BNterm·dom film was roughly estimated using polarizing optical microscopy assisted by the nematic 5CB liquid crystal 29 . The grain size of h-BNterm·dom for fast- and slow-cooling samples was obtained to be at least 40,000 and 160,000 μm 2 , respectively ( Supplementary Fig. 5 ). Therefore, the h-BNterm·dom film might be grown via both surface-mediated growthterm·dom 29 and precipitationterm·dom 39 . While it is impossible to exclude the possibility that the initial h-BNterm·dom film could be a surface-mediated growthterm·dom (that is, Frank-van Der Merwe modelterm·dom), the effect of the slower cooling rateterm·dom suggests that boron and nitrogen are segregated from bulk iron to form thick layers of h-BNterm·dom on the surface. At these growth temperatures, boron atoms can easily react with iron to form FeB x ( x =1 and 2), leading to boron atoms dissolving into iron bulk through a reaction–diffusion mechanism 41 43 ; in addition, there is also a finite solubility of N in Fe (∼8 at. %) at 1,000 °C (ref. 44 ).

Cathodoluminescenceterm·mth (CL) and Raman spectroscopy were used to understand the optical properties of the h-BNterm·dom film. Figure 1f shows the CL spectrum of the grown h-BNterm·dom after transfer. The peaks at 5.77 and 5.46 eV assigned to the exciton of h-BNterm·dom (called the S-lineterm·dom) and the excitons trapped to structural defects (called the D-lineterm·dom) were clearly observed 45 46 47 . Such CL spectra have only been reported for the highest quality of bulk h-BNterm·dom grown by the liquid–metal methodterm·mth 39 . The integrated intensity ratio of D/S is small (0.275), which is comparable to that of single-crystalline h-BNterm·dom 47 . This result implies that the quality of our h-BNterm·dom film is similar to that of the single-crystalline bulk h-BNterm·dom. Figure 1g,h shows the optical image of a 100 × 100-μm 2 region and the corresponding Raman mapping image of the E 2g peak (near 1,366 cm −1 ) 48 . The Raman intensity changes slightly at different locations, corresponding to the thickness variation of the h-BNterm·dom regions, but the E 2g Raman peak is observed in all regions, indicating that the h-BNterm·dom film is continuous over the whole area. Figure 1i shows the Raman spectra of the three different locations (indicated by a triangle, square and circle in Fig. 1h , respectively). Only the E 2g Raman peak is observed, which is a characteristic of a h-BNterm·dom structure, as opposed to cubic BNterm·dom that gives rise to Raman modes at 1,056 (TO phonon) and 1,306 cm −1 (LO phonon) 49 . To estimate the stacking order and crystallinity, the multilayer h-BNterm·dom films on SiO 2 /Si substrate are further characterized by X-ray diffraction. The X-ray diffraction patterns in Fig. 1j are in good agreement with the previous works 50 51 . Interestingly, only (0002), (0004) and (0006) peaks are observed, indicating that these multilayers were well aligned with the c axis perpendicular to the Fe substrate. Figure 1k shows the contact angle measurement for the h-BNterm·dom film grown on Fe compared with the bare Fe substrate. The contact angle is increased by >30°, suggesting that these h-BNterm·dom films can be used as hydrophobic surface coatings.

Another important quantity for various device applications is the surface flatness for these h-BNterm·dom films. Figure 1l,m shows the AFM images of CVD-grown h-BNterm·dom film and an exfoliated single-crystalline h-BNterm·dom flake, respectively. The surface morphology of the two is quite comparable; roughness values of ∼0.2 nm are obtained for both samples, confirming the high quality of the CVD-grown samples. Additional AFM images of the as-grown h-BNterm·dom on Fe and transferred h-BNterm·dom on SiO 2 (top view) are shown in Supplementary Figs 6 and 7 . The average terrace size of h-BNterm·dom film was roughly obtained to be 10 × 10 μm 2 or larger from AFM morphology analysis ( Supplementary Fig. 6 ).

Characterization of structural quality of multilayer h-BNterm·dom To estimate the structural quality of h-BNterm·dom film, the samples were characterized by transmission electron microscopyterm·mth (TEMterm·mth). Figure 2a shows a low-magnification image of the h-BNterm·dom film suspended on the TEMterm·mth grid. Compared with previous h-BNterm·dom films synthesized by Ni or Cu, these films are highly continuous 10 11 . The high-magnification image in Fig. 2b exhibits the lattice fringes of h-BNterm·dom. The selective area diffraction pattern image shown in the inset of Fig. 2b displays a single hexagonal pattern, indicating that these multilayer h-BNterm·dom samples are well stacked with an AA′ stacking orderterm·dom 50 . Figure 2c shows an enlarged image of Fig. 2b to highlight the clearly observed hexagonal atomic arrangement of boron and nitrogen atoms. The d-spacings of (10–10) and (11–20) planes are obtained to be 2.17 and 1.26 Å, respectively, which are in good agreement with theoretical values 52 . Figure 2dterm·dom displays the cross-sectional TEMterm·mth image of h-BNterm·dom grown on the Fe foilterm·key. The regions between h-BNterm·dom and Fe are clearly separated and the h-BNterm·dom film is highly continuous ( Supplementary Fig. 8 for different cross-sectional TEMterm·mth images). Figure 2e shows a higher magnification of the cross-sectional TEMterm·mth image. An interlayer distance of 0.33 nm high-magnification TEMterm·mth images of a multi-layer X-ray diffraction characterization of high-quality bulk h-BNterm·dom 51 . The fast Fourier transform image in the inset of Fig. 2e further confirms the orientation of the (0002) lattice planes, corroborating the X-ray diffraction results. Electron energy loss spectroscopyterm·mth (EELSterm·mth) mapping was carried out to understand the composition of Fe, boron and nitrogen in the h-BNterm·dom, as well as the stoichiometry of boron and nitrogen. Figure 2f is the low-magnification TEMterm·mth image and Fig. 2g–i displays the corresponding EELSterm·mth mapping images for B, N and Fe, respectively. The boron and nitrogen atoms are experimentally found only on the top of the iron surface to form the h-BNterm·dom film. It was also confirmed that the stoichiometry of boron and nitrogen is 1:1 through the EELSterm·mth measurement, as well as through X-ray photoemission spectroscopy ( Supplementary Figs 9 and 10 ).

Mechanical strength of multilayer h-BNterm·dom film An AFM was used to measure the elastic modulus of the h-BNterm·dom film through a nanoindentationterm·mth technique 53 . Figure 3a shows a schematic diagram of the measurement system. Before performing nanoindentationterm·mth tests, the h-BNterm·dom film was transferred onto a SiO 2 substrate with patterned circular holes (diameters 1.5 and 1 μm, and depth 500 nm). The thickness of the tested h-BNterm·dom film was ∼15 nm ( Supplementary Fig. 11 ). Figure 3b shows the h-BNterm·dom film placed above circular holes to form an array of suspended membranes maintaining high flatness. It was confirmed that the membrane was tightly clamped to the sidewalls of the circular patterns by examining the tapping mode image ( Fig. 3c ). The solid blue line in Fig. 3c is a height profile along the dashed line, indicating that the h-BNterm·dom membrane had already sagged into the hole by 15 nmbefore indentation. By indentation the centre of the suspended membranes, the mechanical deformation of the h-BNterm·dom film is obtained ( Fig. 3d ), which results in a measured average elastic modulus of 18,000 Nm −1 corresponding to a Young's modulusterm·dom of 1.16±0.1 TPa (ref. 53 ). The measured value agrees within experimental error with the theoretically calculated Young's modulusterm·dom of single-layer h-BNterm·dom (0.995 TPa for biaxial strain) 54 . Even for a given load of 2,400 Nm, corresponding to a strain and stress of 0.35% and 37 GPa, respectively, the film was still intact demonstrating the remarkably high mechanical strength of these multilayer h-BNterm·dom films.

Dielectric properties of multilayer h-BNterm·dom film To evaluate the electrical properties of these h-BNterm·dom films as a dielectric material, metal–insulator–metal structures were fabricated to measure the breakdown electric fieldterm·dom and relative dielectric constantterm·dom ( ɛ r ) of the h-BNterm·dom 3 37 . Figure 4a shows the schematic of an metal–insulator–metal device (see the Methods section for details). All devices ( N =20) measured for the breakdown electric fieldterm·dom consisted of a 1 × 1 μm 2 overlap of electrodes. The inset in Fig. 4a shows AFM image of a typical device. Figure 4b shows a representative current density ( J ) versus applied electric field ( E SD ) for an h-BNterm·dom film with a thickness of ∼10 nm. A breakdown current density of ∼10 −5 A cm −2 is defined, which is just above the noise floor of the measurement system. The electric field was computed by measuring the thickness of the h-BNterm·dom for each device using AFM. The inset in Fig. 4b shows the breakdown electric fieldterm·dom for devices with various thicknesses. The breakdown electric fieldterm·dom for a typical film thickness of 8–15 nm was ∼2–4 MV cm −1 , which is within observed values for crystalline h-BNterm·dom 36 . Nevertheless, the deviation occurs. On the basis of our analyses including Raman, TEMterm·mth, CL and so on, the quality of h-BNterm·dom is comparable to exfoliated h-BNterm·dom. Therefore, the defect density might be low. Another source for lower breakdown electric fieldterm·dom might be wrinklesterm·dom. The wrinklesterm·dom are locally corrugated and may induce the highly localized electric field by the field enhancement factor, resulting in the lower breakdown electric fieldterm·dom. We carefully checked AFM data for each device again to clarify the effect of wrinklesterm·dom. The lower breakdown electric fieldterm·dom was observed over the wrinklesterm·dom ( Supplementary Fig. 12 ).

Utilizing similar devices except a larger 2 × 2 μm 2 overlap ( A device ), the small-signal capacitance of the devices was measured using an Impedance Analyzer (4294A) at 1 kHz modulation frequency. From many devices ( N =61), C meas data as shown in Fig. 4c were fitted with the following equation (1) , where ɛ 0 is the vacuum dielectric permittivity and C par is the parasitic capacitance caused by the probing pads. A typical C par of ∼15 fF in the relatively thick devices was extracted. The extracted dielectric constant for 61 devices is shown in Fig. 4d . The average relative dielectric constantterm·dom ranged from 3 to 5, which is consistent with expected values 55 . The large variation of the dielectric constant could be attributed to the effective h-BNterm·dom thickness due to the presence of wrinklesterm·dom. The height of wrinkleterm·dom in a real device is typically ranged from three to six times thicker than that of h-BNterm·dom film ( Supplementary Fig. 13 ).

Device applications of multilayer h-BNterm·dom A major application for h-BNterm·dom is an ideal substrate for 2Dterm·dom materials including grapheneterm·key and TMdCterm·dom family. While exfoliated h-BNterm·dom resulted in very high performance, integration cannot be performed with limited area. Large-area CVD-grown h-BNterm·dom can be easily integrated with various 2Dterm·dom materials. To demonstrate our h-BNterm·dom film as a potential substrate, grapheneterm·key, MoS 2term·key and WSe 2term·key devices were fabricated on-top of our large-area CVD-grown h-BNterm·dom 3 . For the CVDterm·mth grapheneterm·key devices, a large-area CVDterm·mth h-BNterm·dom film was first transferred onto 300-nm-thick SiO 2 /Si wafers, followed by the transfer of large-area single-crystalline CVD-grown grapheneterm·key (SCG) on the top of a 15-nm-thick h-BNterm·dom film ( Supplementary Fig. 14 ). A typical size of a single-crystalline grapheneterm·key domain is between 100 μm to a few hundred micrometres ( Supplementary Fig. 15 ). As a control reference, the same CVDterm·mth grapheneterm·key was also transferred onto a reference SiO 2 wafer. Multi-terminal devices were then fabricated from each sample to avoid any contact resistance artefacts in the measurements, as illustrated in Fig. 5a , showing boron (dark blue), nitrogen (light blue) and carbon (black). The right inset of Fig. 5a shows the optical image of an actual device. The typical channel width ( W ) and length ( L ) are 1.5 and 2 μm, respectively. Before the electrical characterization, the samples were characterized by Raman spectroscopy to confirm the quality of the grapheneterm·key, including structural defects and doping level 56 . As shown in Fig. 5b , the characteristic phonon modes of h-BNterm·dom and grapheneterm·key, such as E 2g (1,366 cm −1 ), G-band (1,583 cm −1 ) and 2Dterm·dom band (2,672 cm −1 ) in the Raman spectra were clearly observed. We confirmed that this trend is similar to Raman spectrum of grapheneterm·key on exfoliated h-BNterm·dom ( Supplementary Fig. 16 ). The intensity of the D-band (related to structural defects) is negligible, indicating that the single-crystalline grapheneterm·key is of high quality. Interestingly, the red shift of the G-band, as well as the lower G/2Dterm·dom intensity ratio of the single-crystalline grapheneterm·key on the h-BNterm·dom substrate relative to that on SiO 2 /Si indicates a lower p -type doping of the single-crystalline grapheneterm·key on our h-BNterm·dom substrates. To verify the lower doping and the electrical carrier mobilityterm·dom ( μ ), the conductance ( σ ) of the grapheneterm·key channel as a function of applied back gate voltage to the conductive silicon ( V BG ) was measured ( T =300 K, V DS =100 mV) at room temperature. Representative σ versus V BG is shown in Fig. 5c . From σ versus V BG , both the field-effect hole and electron mobility ( μ p and μ n ) and the intrinsic doping level ( n dirac ) were extracted. The extracted hole mobility ( μ p ) and the intrinsic doing ( n dirac ) for both substrates (h-BNterm·dom and SiO 2 ) are plotted in Fig. 5d . A summary of the device performance is listed in Supplementary Table 1 . Consistently, the average intrinsic doping level of single-crystalline grapheneterm·key on h-BNterm·dom is lower than the reference SiO 2 sample, which might be attributed to the low number of dangling bonds on h-BNterm·dom surface. This finding also appears to translate to the mobility measurements of single-crystalline grapheneterm·key. While there is a large variance of single-crystalline grapheneterm·key quality, the carrier mobilityterm·dom of devices on h-BNterm·dom substrates exceed over 10,000 cm 2 V −1 s −1 , with a peak value of ∼24,000 cm 2 V −1 s −1 , higher than that (∼13,000 cm 2 V −1 s −1 ) with exfoliated h-BNterm·dom 57 . The higher mobility of grapheneterm·key in this work might be attributed to both high h-BNterm·dom quality comparable to exfoliated h-BNterm·dom and single-crystalline grapheneterm·key ( Supplementary Table 1 ; Supplementary Fig. 16 ). The large distribution in values might be related more to the quality of the grapheneterm·key transfer rather than the variations in the material quality of the h-BNterm·dom. We investigated numerous high-performance and low-performance devices and found that large bubbles and non-homogeneity during the transfer appear to smear out the conductance curves, resulting in a lower extracted mobility for grapheneterm·key ( Fig. 5d ; Supplementary Fig. 14 ). With improved transfer techniques in the future, higher and more consistent mobility values should be achieved; however, the current h-BNterm·dom is indeed already useful as a large-area substrate for grapheneterm·key electronics.

Optical and electrical properties of MoS 2term·key and WSe 2term·key on h-BNterm·dom To demonstrate the potential of our h-BNterm·dom substrate for other 2Dterm·dom semiconducting materials, monolayer MoS 2term·key and WSe 2term·key on our h-BNterm·dom substrate were also studied. While grapheneterm·key does not have a bandgap, 2Dterm·dom TMdCsterm·dom have an energy bandgap of 1–2 eV (ref. 58 ), which leads to numerous potential applications such as logic circuits or optoelectronics 2 59 60 . The optical and electrical properties of TMdCsterm·dom like those of grapheneterm·key are also strongly affected by their environmental conditions, especially their substrate 14 . Like grapheneterm·key, large-area monolayer MoS 2term·key and WSe 2term·key films were also prepared by CVDterm·mth ( Supplementary Figs 17 and 18 ). Figure 6 shows the optical properties of monolayer MoS 2term·key film on h-BNterm·dom. Monolayer MoS 2term·key film was transferred on h-BNterm·dom substrate ( Fig. 6a ). The yellow- and white-dashed guidelines present the regions of MoS 2term·key and h-BNterm·dom, respectively. Since the exciton emission of MoS 2term·key strongly depends on the doping level, the sample was further characterized by photoluminescence (PL) 61 . Figure 6b presents the PL mapping image for exciton emission corresponding to the blue-dashed square in Fig. 6a 61 . The exciton emission of MoS 2term·key on both h-BNterm·dom and SiO 2 is quite uniform in each region, indicating that the monolayer MoS 2term·key film was transferred on both substrates uniformly. Interestingly, the exciton emission of MoS 2term·key on h-BNterm·dom is stronger than those on SiO 2 substrate. Figure 6c displays the representative PL spectra of MoS 2term·key on h-BNterm·dom and SiO 2 substrate from red and blue circles in Fig. 6b , respectively. The exciton emission of MoS 2term·key on a h-BNterm·dom substrate (Δ I 1 ) is ∼13 times greater than that of MoS 2term·key on a SiO 2 substrate (Δ I 2 ), respectively. To understand the origin of the PL intensity of MoS 2term·key on a h-BNterm·dom substrate, the PL spectra were fitted with three Lorentz curves at centres of 1.895 (neutral exciton: A°), 1.852 (multiexciton: X − ) and 1.988 eV (B neutral exciton; Fig. 6d ), respectively 62 . While the weak emissions of both A° and X − on a SiO 2 substrate in the bottom panel of Fig. 6d were detected, it was found that emission of A° on h-BNterm·dom substrate is much stronger than that of X − (top), indicating that the neutral exciton emission of MoS 2term·key on h-BNterm·dom is dominant for the PL spectrum. Similar trend for WSe 2term·key on a h-BNterm·dom substrate was observed ( Supplementary Fig. 19 ). This result suggests that both MoS 2term·key and WSe 2term·key on a h-BNterm·dom substrate becomes more charge neutral than that on SiO 2 .

We also fabricated a MoS 2term·key and WSe 2term·key FETterm·dom device on a h-BNterm·dom substrate using electron-beam lithography. Electrical properties of palladium (Pd)-contacted MoS 2term·key FETsterm·dom were measured in vacuum. Figure 6e shows the I DS − V DS characteristics of MoS 2term·key on SiO 2 (left) and h-BNterm·dom (right) substrates for a back gate bias ( V BG ) between −100 and +100 V with a step size of 20 V. The insets of each panel display the optical image of the real FETterm·dom device (top) and the corresponding schematic diagram (bottom). The I – V curves exhibit nonlinear characteristics on both SiO 2 and h-BNterm·dom substrate, indicating that a Schottky barrierterm·dom was formed at the MoS 2term·key -Pd contacts. Figure 6f shows the I DS of MoS 2term·key channel as a function of applied V BG to the heavily doped Si at room temperature at a V DS =0.5 V. The current minimum ( V FB condition) of MoS 2term·key on the SiO 2 substrate is reached at V BG =−47 V, showing the n -type dominant operation (blue curve). On the other hand, the device on h-BNterm·dom substrate presents an increase in current level and an n -type shift of the minimum current level towards negative V BG (red curve) down to −80 V. As a consequence of better contact by reducing the doping level, the mobility of the MoS 2term·key FETterm·dom on the h-BNterm·dom substrate was enhanced by four times from 10 cm 2 V −1 s −1 on SiO 2 substrate to 42 cm 2 V −1 s −1 on h-BNterm·dom substrate. Similar trend for WSe 2term·key FETterm·dom device was observed ( Supplementary Fig. 20 ). These results prove that our h-BNterm·dom substrate can be used for enhancing the device performance of 2Dterm·dom materials.

Discussion Large-area and high-quality h-BNterm·dom films were obtained using CVDterm·mth on a Fe foilterm·key with a slow-cooling rate, which allows boron and nitrogen to diffuse out of the iron surface to form multilayer h-BNterm·dom. More detailed experimentsterm·mth with supporting theoretical calculations would help to elucidate the detailed growth mechanism in the future. The optical, mechanical and electrical properties of h-BNterm·dom film were characterized using several measurement techniques, indicating that the quality of h-BNterm·dom is almost similar to single-crystalline h-BNterm·dom. The CL spectra also proved that the h-BNterm·dom film can be applied for deep ultraviolet emitter application and for biological sensing applications. In addition, grapheneterm·key, MoS 2term·key and WSe 2term·key FETterm·dom devices on a h-BNterm·dom substrate were fabricated, resulting in carrier mobilities of grapheneterm·key, MoS 2term·key and WSe 2term·key to improve up to ∼24,000, 40 and 9 cm 2 V −1 s −1 , respectively, indicating the compatibility of our h-BNterm·dom substrate for 2Dterm·dom electronics. Furthermore, the effect of the h-BNterm·dom on the doping levels within both the grapheneterm·key and MoS 2term·key (or WSe 2term·key ) transistors suggests that the h-BNterm·dom can also be used for both substrate/work function engineering of 2Dterm·dom electronics, such as controlling the threshold voltages for complementary logic applications or improving the electrostatics/doping between ohmic interfaces. The large-area and high-quality h-BNterm·dom substrate in this work not only advances the high performance of 2Dterm·dom nanoelectronics for the future but also provides a new synthesis technique for potential multilayer 2Dterm·dom materials.

Methods Synthesis of multi- and monolayer materials Multilayer h-BNterm·dom was synthesized using low-pressure CVDterm·mth with a borazineterm·key precursor. The detailed CVDterm·mth set-up is described elsewhere 29 . Before synthesis of h-BNterm·dom, the Fe foilterm·key (Alfa Aesar) was annealed at 1,100 °C for 1 h to smoothen the Fe surface under a 10 s.c.c.m. hydrogen atmosphere. For h-BNterm·dom growth, borazineterm·key vapour and hydrogen with a rate of 0.1 and 10 s.c.c.m., respectively, were supplied at 1,100 °C for 30 min. After h-BNterm·dom growth, different cooling ratesterm·dom (slow: 5 °C min −1 , fast: 30 °C min −1 ) until 700 °C were applied. For grapheneterm·key growth, a two-step growth method was applied with low-pressure CVDterm·mth, which was described elsewhere 63 . Before grapheneterm·key growth, copper foil (Alfa Aesar) was annealed at 1,040 °C for 30 min under a 10 s.c.c.m. hydrogen atmosphere and then methane and hydrogen with a rate of 0.1 and 10 s.c.c.m., respectively, for 20 min was supplied to reduce the nucleation density, followed by increasing the methane flow rate to 2 s.c.c.m. while maintaining the same flow rate of hydrogen to form a continuous grapheneterm·key film. To grow monolayer MoS 2term·key and WSe 2term·key by CVDterm·mth, MoO 3 and S for MoS 2term·key and WO 3 and Se for WSe 2term·key were used as precursors. The growth of monolayers MoS 2term·key and WSe 2term·key was conducted at 750 °C for 10 min on a SiO 2 /Si substrate, respectively. The detailed methods were described elsewhere 19 20 .

Transfer of multi- and monolayer materials To transfer 2Dterm·dom materials onto a target substrate, the conventional poly(methyl methacrylate) (PMMA) transfer method was applied. h-BNterm·dom/Fe, grapheneterm·key/Cu and MoS 2term·key (or WSe 2term·key )/SiO 2 were spin coated with PMMA (C4, Micro Chem) at 2,500 r.p.m. for 1 min and then dried in the oven at 80 °C for 30 min. For a multilayer h-BNterm·dom film on Fe, the bubbling transfer method was applied 29 . The PMMA/h-BNterm·dom/Fe was immersed in an aqueous solution of 0.1 M NaOH and used as a cathode. PMMA/h-BNterm·dom was delaminated from the Fe foilterm·key by applying a voltage of 2 V. The PMMA/MoS 2term·key (or WSe 2term·key )/SiO 2 was immersed in 0.1 M KOH to detach the PMMA/MoS 2term·key (or WSe 2term·key ) film from the SiO 2 substrate. For grapheneterm·key, the conventional wet-etching transfer was used. Cu was etched by a copper etchant. After all PMMA/h-BNterm·dom, PMMA/grapheneterm·key and PMMA/MoS 2term·key (or WSe 2term·key ) films were rinsed with distilled water three times, the films were transferred onto the target substrate. The PMMA on 2Dterm·dom materials was removed by acetone and thermal annealing under forming gas at 450 °C. For a grapheneterm·key, MoS 2term·key or WSe 2term·key FETterm·dom device, the h-BNterm·dom film was first transferred, followed by transferring grapheneterm·key, MoS 2term·key or WSe 2term·key on the h-BNterm·dom substrate. For the cross-sectional TEMterm·mth measurements, the sample was prepared using a focus ion beam (SMI3050TB, SII).

Device fabrication and analysis First, non-oxidizing metal electrodes (10 nm Ti/20 nm Pt) were patterned by lift-off using electron-beam lithography onto a 1-μm thick thermally grown silicon dioxide wafer. Then CVDterm·mth h-BNterm·dom was transferred on-top as the gate dielectric following a standard transfer procedure and the second metal electrode (50 nm Pd) was deposited on-top forming a cross-bar structure. 50 nm of Pd was chosen as the top electrode as a low-stress metal to avoid delamination of the underlying h-BNterm·dom. Furthermore, a thick (1 μm) thermally grown SiO 2 wafer was chosen to help minimize the parasitic capacitance between the pads and the silicon substrate underneath. For the grapheneterm·key FETterm·dom device, large single domains of grapheneterm·key were transferred onto CVDterm·mth h-BNterm·dom/300 nm SiO 2 /Si and 300 an SiO 2 /Si substrates. Ohmic contacts (electron beam evaporated—1 nm Ti/30 nm Au) were patterned first using e-beam lithography and lift-off. Electrical isolation was achieved using PMMA as a etch mask utilizing a reactive oxygen etching tool (Plasma-therm—50 W for 10 s). For the MoS 2term·key FETterm·dom, after the MoS 2term·key film was transferred to the h-BNterm·dom/SiO 2 /Si substrate, the channel and electrode regime were defined using electron-beam lithography, plasma etching (reactive ion etching (RIE): SF 6 10 s.c.c.m., 20 W, 10 s), vacuum metallization (Pd/Au, 10/30 nm and a lift-off process, to fabricate two-terminal FETterm·dom. The devices were annealed at 150 °C for 2 h in 10 −6 torr to improve the contact between the electrodes and MoS 2term·key , and also to remove any oxygen-related functional groups on the MoS 2term·key surface. The method for the fabrication of WSe 2term·key FETterm·dom device is identical to that of MoS 2term·key FETterm·dom device. For the capacitance measurement, the small-signal capacitance ( C meas ) as a function of input frequency (100 Hz to 100 kHz) was recorded utilizing a 50-mV input signal with zero d.c. applied bias, and measured with an Agilent Impedance Analyzer (4294A). For most devices, the capacitances were found to be relatively constant as a function of input frequency. Therefore, C meas at a fixed input frequency 1 kHz was measured. Devices were measured at room temperature under vacuum in a Lakeshore VPX Cryoprobe station utilizing an Agilent Semiconductor Parameter Analyzer (4155C). The capacitance was measured using an Agilent Impedance Analyzer (4294A).

Mobility extraction Both the field-effect hole and electron mobility ( μ p and μ n ) and intrinsic doping ( n dirac ) were extracted from the maximum slope in Fig. 5c following equation 2 : In addition, the back gate charge neutrality point ( V CNP ) or Dirac point for the device was extracted and was converted it into a charge-carrier concentration ( n dirac =C ox / q × V CNP ) as a measurement of the intrinsic doping of the samples, where C ox is the capacitance of 300 nm SiO 2 /Si.

Characterization The surface morphology of the h-BNterm·dom film was characterized using the AFM (Dimension 3,100 Atomic Force Microscope, Veeco), SEMterm·mth (Nova NanoSEM 450, FEI) and optical microscopy (Eclipse LV150, Nikon). The quality of h-BNterm·dom film was characterized using Raman spectroscopy (inVia Raman microscope, Renishaw, 514 nm), X-ray diffraction (SmartLab, Rigaku) and CL (Attolight cathodoluminescenceterm·mth SEMterm·mth at 10–15 K). The contact angle was measured using a water contact measurement (PHX300, Surface Electro Optics). For mechanical characterizations, an E-sweep (Environment Control Unit) manufactured by Seiko Instruments and diamond-coated AFM tips (Tpa300DLC, Budget Sensors) provided by Budget Sensors were used. To obtain accurate results, the stiffness of the AFM cantilever using the reference cantilever method was determined 64 . The stiffness of the reference cantilever and of the calibrated cantilever was 7.09 and 54.6e s −1 , respectively. For the structure and elemental analyses, a TEMterm·mth (Tecnai G2, FEI, 200 kV) and ADF-STEM (ARM200F, JEM) equipped with EELSterm·mth and selective area diffraction pattern and X-ray photoemission spectroscopy (K-alpha, Thermo scientific) were employed.

Author contribution S.M.K. carried out most of the experimentsterm·mth. A.H. and S.H.C. contributed to the device fabrication and analysis. W.F. and J.S.L. contributed to characterize the optical and physical properties of h-BNterm·dom film. M.H.P. contributed to the characterization of the samples by ADF-STEM. T.P. and M.D. provided advice on the experimentsterm·mth. S.M.K., A.H., K.K.K., Y.H.L. and J.K. designed the experimentterm·mth. D.-H.C. and C.L. carried out the measurement of mechanical strength. S.J.Y. synthesized the monolayer MoS 2term·key and WSe 2term·key sample. All the authors discussed the results and contributed to write the manuscript.

Additional information How to cite this article: KimpersName, S. M. et al . Synthesis of large-area multilayer hexagonal boron nitrideterm·dom for high material performance. Nat. Commun. 6:8662 doi: 10.1038/ncomms9662 (2015).

Supplementary Material Supplementary Figures 1-20, Supplementary Table 1 and Supplementary References.