Wrinkle-mediated CVD synthesis of wafer scale Graphene/h-BN heterostructures

The combination of two-dimensional materials (2D) into heterostructures enables their integration in tunable ultrathin devices. For applications in electronics and optoelectronics, direct growth of wafer-scale and vertically stacked graphene/hexagonal boron nitride (h-BN) heterostructures is vital. The fundamental problem, however, is the catalytically inert nature of h-BN substrates, which typically provide a low rate of carbon precursor breakdown and consequently a poor rate of graphene synthesis. Furthermore, out-of-plane deformations such as wrinkles are commonly seen in 2D materials grown by chemical vapor deposition (CVD). Herein, a wrinkle-facilitated route is developed for the fast growth of graphene/h-BN vertical heterostructures on Cu foils. The key advantage of this synthetic pathway is the exploitation of the increased reactivity from inevitable line defects arising from the CVD process, which can act as active sites for graphene nucleation. The resulted heterostructures are found to exhibit superlubric properties with increased bending stiffness, as well as directional electronic properties, as revealed from atomic force microscopy measurements. This work offers a brand-new route for the fast growth of Gr/h-BN heterostructures with practical scalability, thus propelling applications in electronics and nanomechanical systems.


Introduction
The discovery of over 2000 species of 2D materials so far provides a resource platform with limitless possibilities for a wide range of applications [1]. We may engineer any combination of properties we desire by constructing heterostructures with these nanomaterials [2,3]. To achieve this purpose, however, the assembly of 2D heterostructures must be done in a controlled manner. The most typical method for vertically constructed heterostructures requires repeated mechanical exfoliations as well as layer-by-layer transfer. Unfortunately, this approach has a very low yield and is extremely prone to interfacial contamination. As a result, chemical vapor deposition (CVD) is gradually gaining traction for the synthesis of large wafer-scale heterostructures, which are required in many technical areas [4].
The nucleation and subsequent growth of graphene in CVD is dependent on carbon supersaturation, which comes through either isothermal growth or precipitation on cooling [5]. The reaction is self-limiting for low carbon solubility catalysts, such as copper, and terminates when a monolayer covers the surface [6]. Nonetheless, inherent catalyst properties like as surface roughness and grain boundaries can cause defects in the asgrown material, lowering final product quality. The same restrictions apply to the formation of h-BN, which has a structure extremely similar to graphene [7]. Many techniques, such as hydrogen annealing and electropolishing, or the use of either monocrystalline foils or liquid metal catalysts as substrates, have been developed to address these problems [8][9][10][11][12][13][14]. Nonetheless, because to its monoatomic dimensions, graphene is mostly affected by the substrate, in which case the undesirable metallic catalyst is often etched away. Furthermore, because of their low bending stiffness, 2D materials grown on metals completely conform to the substrate morphology and deform quickly, giving rise to wrinkles [15]. As a result, insulating substrates are in high demand for applications, however wafer scale direct growth has proven difficult due to their catalytically inert nature [16]. Vertical graphene and h-BN heterostructures have greater carrier mobilities and improved device performance than typical graphene devices on SiO 2 /Si substrates due to structural similarities [17,18]. As a result, studies on these heterostructures have focused on band gap engineering, and promising results have already been widely presented [19,20]. Thus, it is of great importance to be able to facilitate their growth and create devices of wafer dimensions. Numerous attempts have been made for the successful synthesis of graphene/h-BN heterostructures [21] but most of which are not suitable for industrial applications. Some of these methods involve: pre-exfoliated h-BN flakes on SiO 2 and subsequent growth of graphene via CVD or molecular beam epitaxy [22,23], as well as, successive growth and deposition of 2D materials and metal catalysts [24]. These attempts, while achievable in a controlled setting, suffered from either extremely low yield or laborious and complicated techniques, despite the fact that heterostructure domain sizes are often quite tiny. Direct CVD synthesis, on the other hand, has brought us closer to the desired outcome [25][26][27][28][29].
In the current work, we report a two-step CVD procedure for the development of wafer-scale graphene/h-BN heterostructures. By taking advantage of the unavoidable wrinkle population emergent after the CVD growth and carefully engineering the process, wafer scale is readily achieved. The combined wrinkling from both CVD steps gives rise to out-of-plane corrugations making a final product of increased bending stiffness. Additionally, we present an extensive atomic force microscopy (AFM) characterization in order to probe topographic features as well as electrical and tribological properties of the synthesized heterostructure. This work disentangles the thought pathways for successful wafer scale CVD synthesis of 2D heterostructures and overlays a rationale for their industrial integration.

Synthesis and transfer of the heterostructure
h-BN synthesis was based on a previously reported recipe [30]. Graphene/h-BN heterostructure was produced by synthesis of graphene on h-BN by CVD method as follows: h-BN on copper substrate was inserted in an AIXTRON ® BM Pro CVD chamber. After the closure of the chamber, it was immediately pumped down to 0.1 mbar and then argon gas was introduced (250 sccm) under 25 mbar. The foil was heated to 980°C and was kept there for 5 min Afterwards the temperature was decreased to 905°C, while methane (CH 4 ) was introduced (10 sccm) as carbon feedstock to initiate the graphene growth on h-BN surface. After 10 min the chamber was cooled down to 650°C, CH 4 flow was terminated and finally the chamber was cooled down to room temperature under Ar atmosphere. The graphene/h-BN sample was transferred onto a SiO 2 substrate by a dry approach as developed previously [31]. The heterostructure is initially firmly stamped on a silicon/PET substrate, followed by the chemical etching of copper using ammonium persulfate. After the copper removal, the sample is cleansed in distilled water three to five times and is then left to dry under nitrogen flow for several hours. Finally, the heterostructure on the polymer is stamped and rolled onto the SiO 2 substrate, leaving the heterostructure on the SiO 2 .

Raman spectroscopy
Raman spectroscopy was employed to assess the quality of the heterostructure. Raman spectra were collected with an InVia Renishaw spectrometer with 1200 grooves mm −1 grating using an objective ×100 lens (0.85NA) and a 514 nm excitation laser line. Laser power was kept at 0.5 mW to avoid sample heating/damage and acquisition time was optimized for recording noise free spectra. All Raman peaks were fit using Lorentzian functions.

AFM characterization
All AFM measurements were performed on a Bruker Dimension Icon under ambient conditions. For electric measurements the transferred heterostructure onto SiO 2 was mounted on an aluminum plate and was electrically grounded using silver paste and copper tape.
PeakForce-quantitative nano-mechanical (PF-QnM): silicon nitride ScanAsyst-Air probes (R=2 nm, k = 0.4 N m −1 , f = 70 kHz) were used for topographic images. The forces exerted were as low as possible in order to get a better visualization of the finer sample features. Relative nanomechanical data were also acquired in conjunction and correlated with the topography.
Lateral force microscopy (LFM): silicon nitride SNL-10-C (R=2 nm, k = 0.24 N m −1 , f = 56 kHz) probes were used for the friction measurements. The spring constant was calibrated by a thermal tune method while the deflection sensitivity was calibrated on a sapphire standard sample (Bruker).
PeakForce-Kelvin probe force microscopy (PF-KPFM): doped silicon PFQNE-AL probes (R=5 nm, k = 1.5 N m −1 , f = 250 kHz) on a silicon nitride cantilever were used for surface potential information. A dual-pass mode was employed in order to obtain both topographic/nanomechanical data and contact potential difference, V CPD . In the first pass, the topography line profile is recorded in PeakForce tapping mode and subsequently the same line profile is retraced at a set lift height above the surface for the surface potential.
PeakForce-tunneling AFM (PF-TUNA): Pt-Ir coated SCM-PIT-V2 probe (R=25 nm, k = 3 N m −1 , f = 75 kHz) was employed for the conductivity mapping. The designated current sensitivity was 1 nA V −1 , the sample bias given was 1 V and the force was selected so as the electrostatic contact would not saturate the tunneling current.

Synthesis and topographic features of the heterostructure
As it is known, 2D materials grown on metallic substrates, such as graphene and h-BN, will present a plethora of structural defects and out-of-plane deformations in the form of wrinkles. This effect results from the difference on the thermal expansion coefficients of the synthesized material and its substrate [32,33], as well as the release of the substrate's thermal expansion and is still a research topic under investigation [34]. The first CVD step induced wrinkling and crumpling of h-BN from the initial cooling down of the copper substrate, which helped with the formulation of a surplus of expansion points for graphene. By introducing the h-BN/copper for graphene growth, the wrinkle density was initially downsized with a high temperature annealing in order to reduce the roughness of the substrate dramatically. Annealing close to the growth temperature of h-BN will flatten most of the wrinkles and make the surface smoother [35]. Subsequently, the temperature was slightly decreased in order to tune the wrinkle density of the substrate and respectively the nucleation density for graphene. It is well known that the thermal expansion coefficient of copper is 16.7×10 6 K −1 [36], while for h-BN as well as graphene, is negative [37]. This will induce a compressive strain to h-BN if it cooled down from 1000°C to room temperature. The critical compressive strain for wrinkle formation in graphene is less than 1% [38] and it is expected to be similar in h-BN. When h-BN/Cu is annealed for the growth of graphene, the system is under tensile strain again which will decrease the height of the wrinkles but will not cancel them out completely. The starting point for the 2D heterostructure synthesis was selected above 900°C, which is near the lower limit temperature for methane decomposition [39], and below 1000°C , which was the growth temperature of h-BN. The closer we get to 1000°C, the more defects will arise from h-BN such as cracks and tears at the points of wrinkles due to locally weakened mechanical properties [40]. That is the reason the temperature range selected was between 905°C and 980°C. For larger grains we need higher wrinkle heights and wavelengths, while smaller grains require smaller wrinkles. The carbon precursor was then introduced, which was also carefully engineered in order to avoid any undesirable multilayer formation. No other catalytic constituent, such as Cu vapors [41], aided the production of the heterostructure which makes our synthesis a clearly intrinsic procedure. The final product was then cooled initially with continuous precursor feed to avoid any cracks and discontinuities and subsequently to room temperature. This two-step CVD procedure is illustrated in figure 1. The robustness of the synthesized material allowed for its reliable transfer on SiO 2 /Si substrates for further characterization.
The synthesis of graphene on h-BN was constructed upon previously reported synthesis of graphene on copper [42,43]. The parameters were tuned from the standard growth in order to achieve uniform wafer scale heterostructure synthesis. Two crucial modifications were implemented regarding the CVD conditions: the complete absence of hydrogen and extended duration of methane feeding. It is well known that hydrogen can cause etching of h-BN or graphene at high temperatures [44][45][46], so no hydrogen was inserted into the chamber to avoid any alterations or removal of h-BN. The only available hydrogen was supplied from the decomposition of methane, which was enough to improve the grown graphene quality and produce a bilayer heterostructure as confirmed by AFM and Raman spectroscopy. Additionally, since the catalytic role of copper is compromised as it was covered by h-BN, an extended duration and higher concentration of carbon supply were required in order to enhance the kinetics of graphene growth. The carbon introduction was carefully selected because it is known that increased supply favors considerably the formation of multilayers [47]. The temperature of the heterostructure growth was selected between 905°C to 980°C. These boundaries make sure that the synthesized graphene will be less defective, and the wrinkle density will be sufficient for nucleation tuning. Also, above 900°C, methane decomposes more controllably allowing for better management of the process [48].
It is interesting to note here that, what makes the synthesis of graphene problematic on insulating substrates, such as SiO 2 or sapphire, is mainly the lack of control in creating innate nucleation points which is further exacerbated by their low catalytic characteristics. In our case, the role of nucleation sites on the ultrathin h-BN insulator is partially served by the presence of CVD-induced defects of the substrate. In other words, a disadvantage of the CVD method for the synthesis of monolayer materials becomes an advantage for the synthesis of epitaxial 2D heterostructures. Substrate hailing defects can have significant impact on heterostructure nucleation and growth. Dangling bonds at grain boundaries, vacancies, and curved sp 2 π-bonds at wrinkles, are overly reactive which can attract adsorbates [49][50][51][52][53]. Furthermore, thinner growth substrates are shown to be advantageous for the ultrafast growth of graphene [54].
More specifically regarding growth mechanisms, the growth of h-NB has been extensively reported elsewhere [30]. Concerning graphene growth on h-BN, it has been proved that charge transfer from the Cu foil via the h-BN film is critical to the catalytic capabilities of the h-BN film surface, which are thus considerably different from those of bulk h-BN surfaces [55]. As a result, the surface points with the higher charge transfer become mainly the nuclei for the graphene growth. Furthermore, as discovered at this work and discussed at the next section, the wrinkles of h-BN layer act as areas of increased electron mobility and this is the reason why their existence is crucial for the graphene growth, as they play the role of the major nucleation spots ( figure S6).
Regarding the structural matching of the two different nanomaterials, 2D heterostructures grown by CVD follow and align (twist angle=∼0°) with the crystal lattice of the 2D material underneath [4]. This usually results in prestrained heterostructures since the top lattice tries to align with the bottom one by stretching its bonds. In our case, graphene and h-BN have a very small lattice mismatch of 1.7% [3] and thus a very small pre-stress. Raman mapping which is presented below shows compressive strain which arises mainly from the cooling down of the CVD process.
In figure 2(a) the Raman spectrum of the heterostructure is presented. The spectrum shows clearly that the synthesized graphene is a monolayer based on the line-shape of the 2D  peak and the intensity ratio of the 2D/G peaks [56]. Moreover, despite the overlap with the D peak of graphene, the G-peak of h-BN can be distinguished at 1364.8 cm −1 (inset of figure 2(a)) with intensity of ∼30 counts, in very good agreement with other studies [33,57]. We note that the corresponding intensity for a monolayer h-BN of high quality produced by mechanical exfoliation and measured under identical conditions is ∼35 to 50 counts as given in figure S2 and confirms the presence and thickness of the CVD h-BN. The synthesized graphene contains structural alterations as evident by the presence of the D peak at 1650 cm −1 and the appearance of distinct D΄ signal at 1622 cm −1 in all Raman spectra. We ascribe these peaks to the defective structure of the underlying h-BN, since for CVD graphene synthesized with similar conditions a much less defective graphene is produced [42]. Furthermore, large area Raman mapping confirmed the uniform cover of h-BN with monolayer graphene (as also seen by the AFM measurements), which is under compressive strain. The slope of the pos(2D)/pos(G) is ∼0.94 which indicates that the graphene is n-doped [58], as seen in figure 2(b).
PF-QnM mapping provided information on the topographic features of the graphene/h-BN heterostructure. Figure 3(a) shows a wide area scan of the synthesized material while figure 3(b) shows the wrinkle structure. AFM height measurements confirm the bilayer nature of the heterostructure at ∼1.8 nm, which is an accepted value considering the higher mean roughness of CVD grown 2D materials [26,59] (figure S1). Due to the two-step process of the growth, both graphene and underlying h-BN present wrinkling. h-BN wrinkles appear to have smaller wavelength than graphene wrinkles due the higher roughness of the copper substrate. When h-BN/copper went back into the CVD chamber, the rising temperature caused the flattening of most wrinkles and resulted in a much smaller roughness. Reduced roughness is ideal for 2D material growth while substrate inhomogeneities such as wrinkles can induce active regions for nucleation which consequently translates into faster material growth. Due to the poor carbon solubility and catalytic activity of h-BN, graphene grows as a monolayer due to the prudently selected carbon supply. The cooling down of the final material induces wrinkling with higher width and periodicity. This high aspect ratio of the wrinkles makes the heterostructure possess significant bending stiffness due to nanoscale corrugations [60]. 2D heterostructures are innately well-suited for devices due to their ability to modulate their foldability through bending stiffness tuning. Such changes and control might be applied to modify the mechanical properties of 2D devices while simultaneously tuning their electronic characteristics [61,62]. Most of these corrugations arrange naturally into superstructures which are scattered across the material and are more prominent closer to the edges of the sample (figure 4). Some cracks and tears of the heterostructure as well as residues were a result of the transfer process.
There are two distinct wrinkle morphologies visible in the AFM images, both hailing from h-BN: one with high aspect ratio and higher wavelength and one with low aspect ratio and short wavelength. The first is a result of graphene's nucleation and the latter graphene's conformation to the growth substrate (in this case h-BN). Since wrinkles are used as nucleation sites, the reactivity is increased as the wrinkle height is increased [63,64]. For that reason, the high aspect ratio wrinkles are points of increased out-of-plane deformation due to their role as active sites for nucleation and display higher concentration of stress, thus appearing more prominently. The smaller wrinkles come from the cooling down of the heterostructure. The interface between graphene and h-BN in-between the nucleation points at the time of growth is perfect. As the temperature is decreased, delamination of the heterostructure occurs due to the difference in the thermal expansion coefficients of cu, h-BN as well as the difference in the volume of cu and the 2D heterostructure. This effect is partially suppressed from the smoothening of the substrate from high temperature annealing, thus resulting in shorter wrinkle wavelengths. Since delamination happens at the Cu/ h-BN interface, these wrinkles originate from h-BN and graphene follows the underlying topography.

Electrical characterization of the heterostructure
The completely smooth van der Waals interface that h-BN can generate with other 2D materials is one of the key advantages of its incorporation in electronic devices. This is a distinct benefit over traditional 3D insulators like SiO 2 or HfO 2 , which often have high concentrations of dangling bonds and charged impurities at their interfaces with 2D materials. These interfacial imperfections operate as scattering centers, substantially reducing electron mobility and creating a variety of instabilities. However, h-BN, like any other material, has atomic flaws, and films made using various processes exhibit a wide range of defect types and concentrations. Most critically, these defects significantly boost leakage currents via trap-assisted tunneling through thin h-BN layers. In order to visualize these nanoelectric phenomena as the size of electronic devices is scaled down to a few nanometers, the interest in applying more sophisticated electrical stresses at the nanoscale has rapidly grown recently [65,66].
Further confirmation of the successful growth of graphene on top of h-BN, was obtained by current distribution mappings with tunneling AFM. The topography depicts areas of both the heterostructure and the underlying h-BN while the peak current mapping shows areas of much higher conductivity on the heterostructure since graphene is on top of the h-BN (figure 5). Small currents are also observed in the h-BN sample which are attributed to leak currents through defects such as wrinkles and vacancies in the material [67,68].
KPFM has the advantage of non-contact electrical mapping with a spatial resolution in the range of nanometers. It has been widely used to characterize 2D systems and can easily distinguish between differences in nanoscale characteristics such as wrinkles [69]. Figure 6 presents a surface potential map of the heterostructure along with its respective topography. The surface potential variations apparent in the heterostructure, and more prominently on the wrinkles, are a result of two main reasons: concentrated strain at the wrinkle structure and increased doping levels originating from the substrate as well as the CVD synthesis [63,64,70]. Also, inhomogeneous charge distribution arising from the wrinkle formation induces local variation of surface potential, thus establishing potential barriers for electrons [71,72]. The potential variation is more evident in the case of h-BN  wrinkles, where, similarly to graphene, the morphology of the corrugation is vital [73]. The work function of h-BN is strongly dependent on the SiO 2 substrate. On wrinkle sites, we notice a decrease in surface potential, hence a decrease in work function, which is attributed to concentrated compressive strain [74], and is further corroborated by Raman. Furthermore, as shown from the CVD synthesis, wrinkles are highly reactive regions that can be doped easier [75]. The strain is induced externally by thermal treatment, which is uniform for the whole structure, but the heterostructure is strained less than the monolayer h-BN regions due to its higher stiffness. Substate doping is less effective on the wrinkles due to delamination occurring around the nucleation areas. In contrast, flatter areas are less affected by strain but since they are in contact with the substrate, they show increased doping levels, hence the increased work function of the heterostructure. The work function is homogeneous across all the flat areas, meaning that the two materials have a very good interaction and behave as a unified structure. Furthermore, the pathways of the underlying wrinkles and their variation in surface potential denotes that graphene on top, even with ample corrugations formulated by the two-step CVD process, remains mostly unaffected by the strain effects occurring in h-BN wrinkles, as if it is covering h-BN like a veil.
The electrical properties and quality of graphene devices are strongly linked to the substrate underneath. Defects arising from the monolayer h-BN translate into decreased graphene device efficiencies. Multilayer (ML) h-BN could provide a better platform for graphene electronics than monolayer since it has better insulating properties and higher rigidity. Direct growth of CVD graphene on ML h-BN suffers also from all the aforementioned detriments, which are bypassed by the proposed wrinkle-mediated growth. Another route of increasing device performance is by carrying out more effective transfer procedures and post-treatments [76]. Ways to combat the excess defective sites from the transfer of these materials, could involve strategies as: lower surface tension solvents for wet transfer, pre-strained target substrates for wrinkle neutralization [77], as well as ironing of the wrinkles with external parameters such as temperature [78]. 3.3. Tribological characterization of the heterostructure LFM measurements were used to probe the nanoscale frictional behavior of the heterostructure. The fundamental understanding of nano-tribology [79] is critical for the design Wrinkle surface potential is evidently reduced. (b) Magnification of a tear in the synthesized material. From the wrinkle path inside the slit, we can concur that the difference in wrinkle surface potential comes from the underlying h-BN wrinkles, while the graphene on top that follows its topography is less affected of substrate effects. and integration of these materials to advanced applications such as nano-electrical and mechanical systems (NEMs) [80,81]. The frictional characteristics of graphene and other 2D materials on SiO 2 /Si substrate are substantially dependent on the number of layers. Monolayer sheets tend to display more friction than multilayer sheets, which can be attributed to stronger elastic compliance. Monolayer materials locally buckle in the probe contact area which results in increased friction. This phenomenon is known as the 'puckering' effect [82]. As the thickness of 2D materials grows, so does the bending stiffness [83,84], which reduces puckering and lowers friction. A complementary aspect to the concept of true contact area governing the friction of 2D materials is related to evolving commensurability of the interface [85]. Another factor that plays a crucial role in the friction behavior of 2D materials is their interlayer shear strength. For the case of h-BN/graphene heterostructure it has been shown that this heterojunction possesses a robust superlubric behavior independent of the shearing direction [86]. Thus, the large area heterostructures synthesized herein can be used as lubricant coatings.
A region where h-BN, graphene/h-BN heterostructure and SiO 2 are present was chosen in order to discern between their relative friction. The scan lines were taken perpendicular to the wrinkles at 90°to avoid anisotropic friction distribution [87]. From correlating the topography and LFM images, we can easily distinguish between high friction areas like bare SiO 2 and wrinkles hailing from both h-BN and the heterostructure. As expected, the higher friction is observed for bare SiO 2 , which reduces for the area of SiO 2 covered with the h-BN. Further reduction of friction is observed for the heterostructure, which is evident not only in the flat areas but is also more prominent in the wrinkles of each system. As explained earlier, the heterostructure graphene/h-BN has shown to present superlubric behavior due to the low interlayer shear stress of this system [86], in agreement with our results. One significant finding is the pronounced reduction of friction over the wrinkled areas of graphene/h-BN compared to the wrinkled h-BN areas. The wrinkles of the h-BN increase significantly the friction which is comparable to this of the SiO 2 for highly wrinkled/fold areas. On the contrary, the results show that the wrinkles of the heterostructure suppress friction much more efficiently. Since it is quite challenging to transfer wrinkle free large area CVD graphene or h-BN for its use as a lubricant coating, this issue can be overcome by the use of graphene/h-BN CVD heterostructures such those synthesized in the present work. Furthermore, the wrinkled topographic features bestow an increased bending rigidity to the whole heterostructure, which aids in the reduction of the friction coefficient of the system. On the other hand, this demonstration of wrinkles increasing the frictional force can convert them into sites of reduced wear resistance (SI) [40].

Conclusions
In conclusion, we presented a two-step growth of wafer scale graphene/h-BN heterostructures by exploring the presence of wrinkles on an insulating substrate as nucleation sites for graphene. The process is broken down to easily comprehensible concepts and by sensibly tweaking each parameter one can engineer wafer scale production. The synthesized heterostructure possesses a robust interface which allows reliable transfer onto other substrates as demonstrated in the main text. A variety of AFM based techniques for the characterization of the physical properties of the graphene/h-BN, including electrical and frictional properties, were employed. Electrical properties along wrinkles vary significantly and can play a vital role in the design of on-demand devices. It was also shown that the heterostructure can be used as an efficient lubricant coating despite the presence of high-density wrinkles for coating large areas. More importantly, the synthesis presented here could be readily translated to other 2D heterostructures and can offer a paradigm in wafer scale synthesis of epitaxial 2D materials. Such control will be critical in the development of new types of devices that take use of atomically thin films' low intrinsic bending stiffness to construct highly flexible devices, such as nanoscale wrinkleengineered apparatuses.

Acknowledgments
Authors acknowledge the support from 'Graphene Flagship Core Project 3', SGA: 881603 which is implemented under the EU-Horizon 2020 Research & Innovation Actions (RIA) and is financially supported by EC-financed parts of the Graphene Flagship. Authors also acknowledge and thank Mr Matteo Ceccanti for his artistic input in the illustrations of the article.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.

Competing interests
The authors declare no competing interests.

Author contribution
CG designed the research and supervised the project. GT synthesized the heterostructures and MK assisted in the transfer techniques of the samples. MD performed the AFM measurements and CA conducted Raman characterization. MD, GT, CA and CG analyzed the data and wrote the manuscript.