Scalable High-Mobility Graphene/hBN Heterostructures

Graphene-hexagonal boron nitride (hBN) scalable heterostructures are pivotal for the development of graphene-based high-tech applications. In this work, we demonstrate the realization of high-quality graphene-hBN heterostructures entirely obtained with scalable approaches. hBN continuous films were grown via ion beam-assisted physical vapor deposition directly on commercially available SiO2/Si and used as receiving substrates for graphene single-crystal matrixes grown by chemical vapor deposition on copper. The structural, chemical, and electronic properties of the heterostructure were investigated by atomic force microscopy, Raman spectroscopy, and electrical transport measurements. We demonstrate graphene carrier mobilities exceeding 10,000 cm2/Vs in ambient conditions, 30% higher than those directly measured on SiO2/Si. We prove the scalability of our approach by measuring more than 100 transfer length method devices over a centimeter scale, which present an average carrier mobility of 7500 ± 850 cm2/Vs. The reported high-quality all-scalable heterostructures are of relevance for the development of graphene-based high-performing electronic and optoelectronic applications.


■ INTRODUCTION
In recent years, hexagonal boron nitride (hBN) has attracted attention as a promising encapsulant for graphene 1,2 and other two-dimensional (2D) materials, 3 due to its remarkable structural, chemical, and electronic properties. Like graphene, hBN is a layered material with a hexagonal lattice, can be conveniently obtained via mechanical exfoliation from bulk crystals, and presents high chemical stability. Thanks to the small (∼1.8%) difference in lattice parameters between graphene and hBN 4 and its atomically flat surface, hBN can be integrated into graphene-based heterostructures with an effective minimization of extrinsic disorder. 5 Moreover, hBN presents a bandgap as large as 6 eV, 6,7 a dielectric constant of 3.4 8 comparable with that of silicon dioxide (SiO 2 ), and a very high breakdown voltage (i.e., 21 MV/cm 9 ), which make it a suitable dielectric for the realization of field-effect transistor (FET) devices. When used to encapsulate graphene and other 2D materials, hBN is effective in preserving the material quality and stability 1,10 and reducing the ambient induced contami-nation 11 with a beneficial effect on the electrical transport properties. 1,12 For most envisaged high-tech applications in the fields of photonics, optoelectronics, and spintronics, hBN has soon become the ideal encapsulant material, capable of yielding graphene-based devices with the required performances. 13,14 Therefore, the scalable synthesis of hBN has become a crucial field of research. hBN thin films have been obtained via chemical vapor deposition (CVD) and molecular beam epitaxy on several metallic substrates, such as copper, 15,16 platinum, 17 cobalt, 18 and nickel. 19 Indeed, the CVD synthesis of monolayer and few-layer hBN is by now an established technique and the material is presently commercially available. 20 However, the synthesis of hBN films with a thickness of tens of nanometers, suitable to be adopted for bottom and top graphene encapsulation, as well as serving as a gate dielectric in electronic and photonic devices, is still considered a challenge. In first place, there is an objective difficulty in obtaining hBN whose quality matches that of exfoliated flakes from bulk crystals. 1,21,22 Also, although progress has been reported for the CVD growth of thick hBN films both on metallic 23,24 and dielectric 25,26 substrates, there are significant challenges in identifying synthesis processes which comply with industrial requirements for CMOS integration such as metal contamination control (below 10 10 atoms/cm 2 ). 14,27 The use of insulating substrates for the synthesis of hBN offers advantages such as the absence of metal contamination, though temperatures as high as 1400°C 28 are often needed to obtain high quality hBN, which are not appealing from an industrial point of view. Other approaches have been explored to grow BN at low temperatures, such as microwave-assisted CVD 29 and plasma-enhanced atomic layer deposition, 30 but both these methods present safety limitations due to the use of toxic precursors such as n-ethylmethylamine. 30 The definition of a scalable and safe hBN growth approach yielding controlled thickness on insulating substrates would indeed be extremely attractive.
In this work, we report the realization of high-quality hBN/ graphene heterostructures by employing scalable techniques, which could be of potential interest for fab integration. 33,34 First, continuous films of nanocrystalline hBN with thicknesses of 10 nm are synthesized on SiO 2 /Si at 1000°C through a physical vapor deposition approach, namely, ion beam-assisted deposition (IBAD). Subsequently, arrays of monolayer graphene single-crystals are grown via CVD on copper and transferred with a semidry approach 33 on the target IBAD-hBN substrates.
Combined analyses of the spectroscopic, microscopic, and transport properties of the heterostructure indicate that IBAD-hBN is a promising substrate for graphene devices, as it provides a high-quality landscape for the graphene carriers. When measuring 109 devices, the room temperature carrier mobility in graphene on IBAD-hBN is found to average at ∼7500 cm 2 /Vs, and the residual carrier density at the charge neutrality point is ∼2 × 10 11 cm −2 . As-processed devices initially show displacement of the Dirac point and gate hysteresis (attributed to the presence of trapped charges at the hBN/SiO 2 interface), which are both significantly reduced by vacuum treatment of the heterostructure. ■ MATERIALS AND METHODS hBN Growth. Nanocrystalline hBN films were grown by IBAD on commercially available p-doped silicon substrates covered with 285 nm of thermally grown SiO 2 , using nitrogen gas and solid boron as sources. The films can be grown with thicknesses ranging from 1 to 100 nm; in this work, a thickness of 10 nm was used. The lateral size of the resulting hBN film is limited by the diameter of the ion gun, and in our setup, homogenous films up to 3″ wafers could be produced. Solid boron (Alfa-Aesar 12134) was evaporated using a 7 kV electron beam evaporator, while low energy nitrogen ions (average energy of 5 eV) were provided by a Kauffman ion gun fed with 5 standard cubic centimeters per minute (sccm) of high purity gas. The chamber base pressure was 10 −7 mbar, reaching 10 −4 mbar during the growth. The sample was maintained at 1000°C during the growth. With the adopted growth technique, it is possible to tune the properties of the material by changing the solid boron precursor from pure boron to boron-carbide (B 4 C), thus yielding BN films (i.e., BNC) with a limited content of carbon (<10%) and a different dielectric constant. 31,34,35 While both kinds of BN films (with and without carbon additive) were synthesized in this work, only pure hBN films were those ultimately adopted because of the higher crystallinity (see Supporting Information). Calibration of growth rates was done by contact profilometry, and the actual thickness was verified on test samples by UV−vis spectrometry and spectroscopic ellipsometry. The quality and orientation of the adopted hBN films, which were found to exhibit a basal plane parallel to the substrate, were determined by X-ray absorption near edge structure (XANES). 36 Graphene Growth and Transfer. Graphene single-crystal matrixes were grown by CVD in a deterministic pattern on electropolished copper foils (Alpha-Aesar 99.8%) in a commercially available cold-wall reactor (Aixtron 4″ BM Pro), as reported in previous work. 33 Specifically, the substrate was first annealed in non-reducing argon atmosphere for 10 min, and the growth was then performed at 1060°C with an argon flow of 900 sccm, 100 sccm of hydrogen, and 1 sccm of methane, with a base pressure of 25 mbar. The graphene crystal arrays were transferred on the target substrates (i.e., SiO 2 /Si with and without IBAD-hBN) through a deterministic semi-dry procedure. 33 The graphene on copper foil was covered with a double polymeric membrane of PMMA/PPC and baked at 90°C, 37 while a few millimeter-thick PDMS frame was applied on the edges of the sample, to ensure mechanical rigidity. The graphene was delaminated from the copper in a solution of 1 molar of NaOH 38,39 and transferred to the target substrate using a micromechanical stage to ensure the deterministic transfer. Once transferred, the polymer was removed by subsequent immersion in acetone and isopropanol. 2-step cleaning using remover AR 600-71 (Allresist) was performed to ensure the cleanliness of the graphene surface. 40 Atomic force microscopy and Raman Characterization. Atomic force microscopy (AFM) was used to investigate the topography of the samples; it was performed using an Anasys AFM+ tool in non-contact mode and a Bruker Dimension Icon microscope used in ScanAsyst mode. AFM micrographs were analyzed using the software Gwyddion 2.54.
Raman spectroscopy was used to characterize the crystalline quality of both graphene and hBN. Raman data were acquired using a commercial Renishaw InVia spectrometer, with a laser wavelength of 532 nm. The Raman setup is linked to a microscope with mechanically controlled stage, thus allowing to perform spatially resolved micro-Raman characterization with spot size in the order of 1 μm 2 , defined by the 100× magnification lens. Raman characterization of hBN was performed using a laser power density of ∼10 mW/μm 2 . An acquisition time of 600 s was needed to detect representative hBN peaks. 10 Graphene was measured with a laser power density of 1.7 mW/μm 2 . The statistics reported in the paper were obtained from spectra acquired on areas of 15 × 15 μm 2 with a step of 1 μm.
Device Fabrication. Optical lithography and metal thermal evaporation (50 nm of gold on top of 5 nm of chromium) were used to pattern an array of markers on top of the hBN substrate, before the transfer of graphene. Hall bar and transfer length method (TLM) devices were fabricated using standard e-beam lithography technique (EBL), with a Zeiss UltraPlus scanning electron microscope and Raith Multibeam lithography module. The graphene channels were defined with the first lithographic step. Reactive ion etching with Ar/O 2 atmosphere for 45 s was used to remove graphene from the patterned areas. Subsequently, the metal contacts were defined via a second EBL step and thermal metal deposition of 50 nm of gold on top of 5 nm of chromium.
Electrical Characterization. Electrical characterization was performed at room temperature and in air, using five micrometric positioners (MPI-corporation MP40) with 7 μm tungsten tips to provide signal and check the read-out. A Keithley 2450 sourcemeter, in high tension configuration, was used as a DC source for the gate potential, with constant reading of the current to check for eventual leakages from the back gate. DC measurements, both in two and fourterminal configuration, were performed through a second Keithley 2450 sourcemeter. To assess the real graphene electrical performance, avoiding any contribution to the resistivity arising from the contacts, 4-probe measurement, both in Hall bar and TLM devices, was performed. AC measurements were carried out with a Signal Recovery 7260DSP in low frequency (10−100 Hz) configuration and differential voltage read-out. The constant current was achieved using a large pre-resistor (4.7 MΩ) in series with the measured device. The orthogonal magnetic field in the Hall measurements (up to 1000 Oe) was provided through a commercial resistive electromagnet operated at room temperature.

■ RESULTS AND DISCUSSION
The morphology of the hBN film synthesized via IBAD was examined via AFM and compared to that of SiO 2 /Si used as growth substrate. As shown in Figure 1a, the hBN film shows uniform nanoscale flatness over an area of 10 × 10 μm 2 . The morphology is qualitatively comparable to that of the bare silica substrate (Figure 1b). Indeed, we retrieve an average root mean square roughness of 450 pm for the SiO 2 /Si substrate used as target for the growth and of 935 pm for the hBN (see Figure S1). In the inset of Figure 1a,b, we report a representative AFM line profiles for commercial SiO 2 /Si and IBAD-hBN. This result indicates that the IBAD growth process maintains the surface morphology in a range potentially suitable for high-quality graphene-based devices, even for thick hBN films. A very low roughness is in fact instrumental for a material to be used as a graphene substrate, since local strain variations are regarded as a major source of carrier scattering in graphene. 41 In Figure 1c, we have shown the angle-dependent study of the XANES for our hBN: around 192 eV, we have the energy transition from 1s to π* for boron; 32 the peak intensity in this transition follows a cosine square dependence with the  incident angle, 42 suggesting that the hBN crystals have a preferential orientation parallel to the substrate plane.
In Figure S1c, we report a representative Raman spectrum of the synthesized hBN film. We observe two main Raman modes: the one at ∼1370 cm −1 is attributed to the characteristic E 2g vibrational peak of hBN, while the Si third order transverse optical (3rd TTO) peak 43 is located at ∼1450 cm −1 . The full-width-at-half-maximum (FWHM) of the E 2g (hBN) peak, an indication of the material crystallinity, is 37 cm −1 , higher than that measured for single crystal exfoliated hBN (∼8 cm −1 ), 44 but comparable to that reported for CVDgrown hBN. 45 Figure 2a reports representative Raman spectra for graphene single-crystals transferred on SiO 2 (black) and on hBN (green). The characteristic graphene 2D and G Raman peaks are observed around ∼2675 and 1582 cm −1 , respectively, while the D-peak (∼1350 cm −1 ) is absent, indicating that defects are negligible. 46 The 2D-peak can be fitted with a single Lorentzian, as expected for monolayer graphene, 47 with comparable FWHM values averaging at 25 and 23 cm −1 on SiO 2 and hBN, respectively, suggesting a low amount of strain fluctuations (Figure 2b). The FWHM of the G-peak is found to average at ∼12 and 10.5 cm −1 for SiO 2 and hBN, respectively, as shown in Figure 2c. The A(2D)/A(G) values, reported in Figure 2f, suggest a carrier concentration within the intrinsic limit for graphene on hBN and close to 100 meV for graphene on SiO 2. 48 Also, the increased I(2D)/I(G) value for graphene on hBN (Figure 2e) indicates a reduction of the doping level. In Figure 2d we report the correlation plot between the 2D and G peak position. 49,50 Although the data collected for graphene on SiO 2 /Si present a narrower dispersion than those on hBN, compatible with the higher roughness of the hBN measured from the AFM, both are indicative of slight compressive strain. Figure 3a shows a sketch of the typical graphene FETs (g-FETs) fabricated to investigate the transport properties of graphene when transferred on top of the IBAD-hBN. Figure 3b shows a representative transfer curve for a graphene/hBN device: employing the constant-mobility model (R for a 285 nm thick SiO 2 plus 10 nm thick hBN, used in the constant-mobility model. The increase in the carrier mobility correlates with a reduction of the residual carrier density close to charge neutrality, as reported in Figure 3e: n* were retrieved to be 2 × 10 11 and 3.8 × 10 11 cm −2 for graphene on hBN and SiO 2 , respectively. Although the morphology of the IBAD-hBN is slightly rougher than SiO 2 , these results indicate that a higher-quality potential landscape for the graphene carriers is provided by the nanocrystalline substrate. Furthermore, the Dirac point for this device is retrieved at 5 V, which corresponds to a charge density of 3.8 × 10 11 cm −2 , in line with the Raman estimation. Overall, both Raman spectrosco- py 41 and electrical measurements indicate a reduced doping and residual carrier density for graphene on hBN (see Figure  2b, Figures 2e and 3e), which explain the improved transport properties measured in the graphene/hBN heterostack. It should be mentioned that the device above was measured after keeping the structure in static vacuum (∼10 mbar) for prolonged time (>4 months) after fabrication. When measured immediately after fabrication, the devices presented a pronounced hysteresis of the transfer, as shown in red curve in Figure 3f, while after storage in vacuum the hysteresis was strongly reduced, as shown by the black curve in Figure 3f. Concurrently, we also observed a shift in the Dirac point to lower doping values (i.e., V Gate < 10 V). Also, no significant variation in the carrier mobility was observed after vacuum storing, as reported in Figure S4. All the electrical measurements reported have been performed in ambient (not vacuum) conditions, and it should be mentioned that there was no reappearance of hysteretic behavior in the sample after vacuum storage. The gate hysteresis can be attributed to charge traps that partially screen the back-gate potential. These traps can be present either at the SiO 2 /hBN or at the hBN/graphene interfaces. We report that these traps are characterized by slow charging and discharging times, as we observe different transfer curve behaviors for different gate sweeping rates ( Figure S3).
We were able to induce reduction of the hysteresis also by annealing the sample at 130°C overnight in high vacuum (10 −9 mbar). However, annealing at higher temperature for shorter times affected the transport properties of the device with a consequent mobility reduction of ∼50% (see Figure  S5). Moreover, the carrier concentration in the graphene shows a linear dependence on the gate voltage (see Figure 3c), with no sign of direct charge transfer.
To further assess the location of the charge traps, we performed transport measurements in a top-gated exfoliated hBN/graphene/IBAD-hBN/SiO 2 heterostructure. When using the top-gate (that is, applying the gate potential through the exfoliated hBN flake), we did not observe significant hysteresis, nor gate sweeping speed dependence. The electrical behavior of the fully encapsulated device was found to be qualitatively compatible with that measured for similar devices fabricated on SiO 2 /Si substrates (see Figure S7b and S7d): the fully encapsulated graphene presented a significantly higher carrier mobility of ∼15,000 cm 2 /Vs (at 5 × 10 11 cm −2 ) and lower residual carrier density of n* = 8.5 × 10 10 cm −2 , as the top hBN protects from environmental contaminations. Instead, measuring the same device in back-gated configuration led to the observation of a large hysteresis (see Figure S7c) comparable to that reported by the red curves in Figure 3f. This confirms that the charge traps are present at the remote hBN/SiO 2 interface, likely forming during the IBAD growth process. In Figure 3d, we show a representative schematic of the expected effect that charge impurities present at the interface hBN/SiO 2 may have on the transfer curve: when a gate potential is applied, these impurities can act as traps for the electrons, leading to a non-linear change of the electric field on the graphene with the applied gate potential. Vacuum storage appears effective in removing such traps, hence ultimately eliminating hysteretic effects.
Finally, to prove the scalability of our approach, we realized and electrically characterized more than 100 graphene devices, transferred on IBAD-hBN. As previously reported, 33 we have developed approaches to grow single-crystal graphene in a deterministic pattern and to precisely transfer such matrixes on a desired substrate; thus, we can scale-up the statistics on the device number by realizing a matrix of TLM devices. Each TLM device was realized on a different graphene single-crystal of the array, as shown in Figure 4a−c. Fabrication of each TLM channel (width 10 μm; length varying from 10 to 25 μm, in 5 μm steps) within the same graphene crystal allowed us to isolate the contact and channel contribution to the resistance and thus estimate the graphene mobility. 53 The devices are realized with the same orientation, which means appreciatively the same crystallographic orientation, as previously reported; 54 however, the electrical behavior of the devices should not be influenced by the crystallographic orientation in our regime of measure. Figure 4d shows the transfer curves for a representative TLM device. In Figure 4e, we report a plot of the carrier mobility as a function of the gate voltage position of the Dirac point. The mobility obtained from the constant mobility model in the array was found to be as high as 9000 cm 2 /Vs, with average μ at ∼7500 ± 850 cm 2 /Vs. The TLM sample was subjected to a shorter (few weeks) vacuum treatment with respect to the gFET, and for this reason, the average Dirac point is found to be significantly higher than expected, that is, 20.5 ± 4 V, while the hysteresis is still reduced. The mobility values reported indicate a substantial improvement with respect to large-scale characterization of the same CVD graphene crystals on SiO 2 (average μ ∼ 5000 cm 2 / Vs), 52 in line with the 30% improvement reported for the gFET devices. It should be mentioned that these graphene crystals were also proved to display ballistic transport (with low-temperature mobility only limited by the device physical edges) when top and bottom encapsulated with exfoliated hBN via van der Waals dry assembly. 55 The results reported here were obtained on a 1 × 1 cm 2 chip but could be straightforwardly extended to wafer-scale via multiple tile transfer of the graphene matrixes. 52

■ CONCLUSIONS
In summary, we have realized and characterized scalable vertical hBN/graphene heterostructures which allow for the realization of devices with promising electronic performances. Nanocrystalline hBN was grown via IBAD on SiO 2 /Si and presents a thickness of 10 nanometers, which makes it suitable to be used both as an encapsulant and as a gate dielectric. Microscopic characterization was performed to investigate the surface morphology of the scalable hBN, which was found to be comparable in terms of flatness to those of the SiO 2 growth substrate. Spectroscopic and transport measurements were carried out to compare the properties of graphene when transferred on commercial SiO 2 /Si and hBN. We observe a relevant improvement in terms of residual carrier density and carrier mobility, which indicates how the adopted hBN provides a high-quality landscape for the graphene carriers. Also, we demonstrate that the hysteretic behavior observed in heterostructure realized with as-received hBN can be significantly reduced by vacuum treatment of the material. To prove the scalability of our approach, we tested an array of graphene crystals, transferred on IBAD-hBN over centimeter scale, and obtained reproducible mobility values exceeding 7500 cm 2 /Vs. Future developments might concern a deeper investigation of the charge trapping mechanism at the hBN/ SiO 2 interface, as well as the optimization of the IBAD-hBN transfer to realize fully encapsulated and scalable hBN/ graphene/hBN heterostructures.  Complete contact information is available at: https://pubs.acs.org/10.1021/acsami.3c06120

Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
The research leading to these results has received funding from the European Union's Horizon 2020 research and innovation program under grant agreement no. 881603-Graphene Core3.