Crystal Orientation Dependent Oxidation Modes at the Buried Graphene–Cu Interface

We combine spatially resolved scanning photoelectron spectroscopy with confocal Raman and optical microscopy to reveal how the oxidation of the buried graphene–Cu interface relates to the Cu crystallographic orientation. We analyze over 100 different graphene covered Cu (high and low index) orientations exposed to air for 2 years. Four general oxidation modes are observed that can be mapped as regions onto the polar plot of Cu surface orientations. These modes are (1) complete, (2) irregular, (3) inhibited, and (4) enhanced wrinkle interface oxidation. We present a comprehensive characterization of these modes, consider the underlying mechanisms, compare air and water mediated oxidation, and discuss this in the context of the diverse prior literature in this area. This understanding incorporates effects from across the wide parameter space of 2D material interface engineering, relevant to key challenges in their emerging applications, ranging from scalable transfer to electronic contacts, encapsulation, and corrosion protection.


■ INTRODUCTION
Two dimensional layered materials have been extensively studied over the last 15 years. More recently, the importance of interface effects between 2D layers and 3D materials has gained increasing prominence. Graphene−substrate interactions play a governing role in the physical, electrical, and chemical characteristics of a graphene layer. Prominent examples include the graphene transfer from a catalyst, 1−5 electronic contacts, 6−8 encapsulation of graphene device channels, 9−13 diffusion barriers for Cu interconnects, 14,15 and corrosion protection effects that 2D materials can provide for 3D materials. 16,17 Depending on the device integration pathway and the nature of the application, precise tuning of the interaction strength between 2D and 3D materials is required. As such, intercalation at the graphene−metal interface has been widely studied as a means of tuning interaction strength. 18−22 Selective oxidation of the metal surface beneath graphene has been introduced as a route to weaken interaction and facilitate graphene transfer with minimized damage and contamination. 4,23−25 Alternatively, for applications where the 2D layer functions as a corrosion inhibitor, 17,26,27 a strong graphene−metal interaction is required to suppress oxidation. 16 Thus, engineering the graphene−substrate interface is of crucial importance for many 2D material applications.
Here we study the influence of crystallographic orientation on the interaction and chemistry at the 2D−3D materials interface, specifically focusing on ambient oxidation at the buried graphene−Cu interface. Our model system is chemical vapor deposited (CVD) graphene on poly-and single-crystalline Cu substrates. There have already been a large number of reports on various aspects of the graphene/Cu interaction and the related Cu oxidation/corrosion behavior. We therefore first summarize the salient points from these studies that we build on here, focusing particularly on room temperature conditions: for short time scales (minutes) graphene suppresses species reaching the Cu surface and thus protects Cu from corrosion in air and in corrosive liquids. 28−31 For longer time scales (hours to days) the presence of graphene can enhance corrosion due to galvanic corrosion. 23,29,32,33 For metals that interact strongly with graphene such that hybridization of the band structure occurs, the intercalation of species at the buried interface is suppressed. 16 Whereas for comparatively weakly binding, such as graphene on Cu, intercalation of species can more readily proceed. 34,35 Isotope labeling studies have shown that under certain conditions the oxygen that forms the Cu oxide at the interface is from adsorbed water and not from the molecular oxygen in air, 23 with similar results found on Ru(0001) at cryogenic temperatures, where dissociated water intercalates but not O 2 or H 2 . 36 The graphene/Cu interaction strength depends on the Cu crystal orientation as well as the epitaxial relationship, which leads to anisotropy in the Cu oxidation rate at the graphene/Cu interface for different Cu orientations. 25,37−40 Graphene on Cu (111) and Cu(311) can form an epitaxial system where the interaction of graphene and Cu is comparatively strong, 40−43 and the suppression of Cu surface oxidation seen for these systems has been attributed to this epitaxy. 37,40,44 Interestingly, however, other Cu facets covered with noncommensurate graphene, such as Cu(124), have also been found to be relatively inert to oxidation. 25 These previous studies on select graphene covered Cu facets motivate us here to systematically study ambient oxidation behavior for over 100 different (high and low index) Cu orientations. We employ complementary characterization by electron backscatter diffraction (EBSD), high spatial resolution (submicrometer spot size) X-ray scanning photoemission microscopy/microspectroscopy (SPEM, see Methods), Raman spectroscopy, and atomic force microscopy (AFM). Across the whole polar plot of surface orientations, we identify four general oxidation modes and rationalize the behavior of each with respect to the underlying mechanisms, and then further discuss their relevance and generality.

■ RESULTS
As a model substrate we use 1 mm thick, chemical mechanical polished (CMP), polycrystalline Cu (purity is 99.99%) onto which without further pretreatment graphene islands were synthesized by CVD (see Methods). In addition, we use single crystal Cu (111), (011), and (001) substrates. Note the Cu(111) sample recrystallized into higher index surfaces during the CVD conditions used here (see the Supporting Information, SI). The CVD process results in a graphene nucleation density of approximately 30 mm −2 , and using a growth time of 4 h leads to largely isolated graphene domains of roughly 250 μm in lateral size. 45 The graphene is predominantly monolayer, as has been shown previously. 45 After the CVD process the lateral Cu grain dimensions of our model substrate are on the order of 1 mm, as shown by the electron backscatter diffraction (EBSD) map in Figure 1(a).
Highlighted by the inverse pole figure coloration, we find a range of different Cu surface orientations. The as-grown sample was stored in a class 10 000 cleanroom for more than 2 years, with a controlled humidity level of 50% and a temperature of 21°C. Figure 1(b) shows a corresponding optical microscope image. The observed contrast patterns highlight a very diverse, Cu facet dependent oxidation behavior underneath and in between the graphene islands.
It is well-known that a graphene coating can change the Cu oxidation behavior, and this effect is widely used to optically characterize graphene coverage and defects. 37,46−49 The optical contrast and color can thereby be linked to the Cu oxide thickness, ranging from metallic white (unoxidized) to yellow, orange, and red with increasing thickness of the Cu oxide. 50−53 We verify the correlation between optical contrast and degree of Cu oxidation via a combination of X-ray photoelectron spectroscopy (XPS) microscopy, Raman spectroscopy, and AFM. For XPS analysis, the sample was annealed at 100°C in vacuum to remove surface contamination. In order to structure these large data sets, we employ image processing and color thresholding (see SI) on our optical microscopy data to extrapolate a quantitative parameter θ ox that represents the extent of Cu oxidation based on areal coverage under graphene for a given Cu facet. We thereby only consider graphene covered areas, and for each crystallographic Cu orientation the θ o x -v a l u e w a s a v e r a g e d o v e r a l l p i x e l s , w i t h θ = ox oxidized area under graphene graphene covered area . The relative degree of oxidation of bare Cu areas, without graphene coverage, is characterized separately (see SI Figure S7). Figure 2 shows θ ox plotted onto the corresponding inverse pole figure (IPF), and it reveals four distinct modes of graphene/Cu interface interaction. Cu orientations for which we observe the dominating influence of Cu/graphene corrugations/wrinkles on the oxidation are all marked with rhombuses. The data from {011} and {001} single crystal Cu substrates are shown by pentagons (see also SI Figure S6). We find that our Cu(111) single crystal recrystallizes into different higher index orientations under the conditions used, therefore we were not able to verify the oxidation behavior for this reference orientation. Instead, we include in Figure 2 literature values marked with a star for the {111} orientation, which is reported not to cause interface oxidation even after extended air exposure. 40,44,54 On Cu facets close to {111} orientations, Cu oxidation occurs via a strong contribution from wrinkles. On Cu facets close to {011} orientations such wrinkle oxidation is not observed, and instead an irregular oxidation  pattern appears under the graphene. Complete oxidation (i.e., θ ox approaching 1) is found on Cu facets close to {001} orientations. For Cu facets close to {113} orientations there is a region where we find values of θ ox close to 0, indicating that graphene completely inhibits Cu oxidation even after 2 years of air exposure. In the following we categorize these regions as "complete oxidation", "irregular oxidation", "inhibited oxidation", and "enhanced wrinkle oxidation", as mapped out and labeled in Figure 2. Data points representative of the oxidation behaviors of the four identified key interface characteristics are highlighted by black arrows in Figure 2. In the following we focus on detailed characterization of these representative points in order to develop an understanding of these distinct oxidation behaviors at the buried graphene/Cu interface. Figure 3 shows higher resolution optical microscopy images (a−c) combined with spatially resolved, synchrotron source SPEM maps (d−h). The optical microscopy data in Figure  3(a−c) highlight that local phenomena influence the oxidation behavior, which we will describe and characterize in detail in the following. Figure 3(a) shows a graphene island that crosses a Cu grain boundary (highlighted by an arrow) and upon crossing of the Cu grain boundary the color and homogeneity of the copper oxide changes. We observe this phenomenon across the entire sample and graphene islands that cross a Cu grain boundary show an oxidation pattern that is dominated by the underlying Cu surface orientation (for details see Figures   S1 and S2) and apparently rather independent of the graphene flake and its relative orientation to the underlying Cu. A brighter area in Figure 3(a), marked by the letter "A" which corresponds to a bilayer graphene region remained unoxidized [all other graphene covered regions in Figure 3 are monolayer graphene (see SI section 3)]. On the left side of the domain boundary in Figure 3(a) Cu shows an area of "irregular oxidation", whereas on the right side a region of "complete oxidation" is shown. Figure 3(b) shows bright contrast across all graphene covered regions corresponding to unoxidized Cu below the graphene layer except for narrow wrinkles with orange contrast. We label this as "inhibited oxidation". In strong contrast to the narrow wrinkles in Figure 3(b), Figure  3(c) shows parallel lines of dark red color where enhanced oxidation is found, and the remaining graphene covered region appears brighter in optical contrast and unoxidized. We label this region "enhanced wrinkle oxidation". In Figure 3(c) we observe that starting from a wrinkle the oxide preferentially grows anisotropically in the direction perpendicular to the wrinkle direction. Furthermore, the graphene island edges are also preferentially oxidized, and oxidation seems to proceed from the graphene island edge inward, in agreement previous reports. 55,56 Conventional XPS with a spot size of several hundred μm is not suitable for resolving local oxidation behavior as observed here. Therefore, to unequivocally determine the oxidation state Chemistry of Materials pubs.acs.org/cm Article of Cu under the graphene layer with a resolution small enough to resolve local effects such as wrinkles and Cu grain boundaries, we here use SPEM (see Methods) that has submicrometer spatial resolution. Figure 3(d−f) shows the C 1s intensity maps acquired at 285.3 eV binding energy corresponding to approximately the same locations as the optical images (Figure 3a−c). A clear contrast is visible in the C 1s maps between the graphene-covered and uncovered areas.
When measuring the C 1s signature on Cu and Cu 2 O there is also a change in background intensity, resulting in the contrast seen in Figure 3(d−f). The reason for this is, that the Cu signals have a higher intensity in the case of metallic Cu which results in a higher number of secondary electrons and thus a higher background level at the C 1s peak position. At the Cu grain boundary in Figure 3(d) the signal intensity is increased which is not related to oxidation, but to a topography artifact.
Since there is only a small binding energy difference in the Cu 2p 3/2 core level peak for metallic Cu and Cu 2 O, we focus here on the Cu LMM Auger-Meitner signature which is more sensitive to changes in the Cu oxidation state. 34,58,59 Spatial mapping of the Cu LMM fingerprint region allows one to obtain chemical contrast maps which show the ratio of metallic to oxidized Cu (more details see Methods). They are shown in Figure 3(g,h) and confirm that locations in white in the optical microscope image match to a Cu LMM chemical contrast intensity corresponding to unoxidized regions.
The letters "A−J" in Figure 3 mark points were XPS and Raman point spectra were taken to fully characterize these locations. Upon exposure to air Cu mainly forms Cu 2 O, CuO, and Cu(OH) 2 . Note that prior to the XPS measurement, the sample was annealed at 100°C in vacuum to remove surface contamination but which could also begin to reduce any CuO present to Cu 2 O, 60 given its instability in vacuum. 61 Figure 4 shows the Cu LMM spectra of locations "A−J" and the peak positions of the most intense peak corresponding to Cu 2 O, CuO, and Cu(OH) 2 , whose kinetic energies are shifted with respect to the metallic peak by −1.04, −1.62, and −2.36 eV, 57 respectively. The Cu LMM signature is very complex, and in Figure 4 we only mark the location of the most intense peak for the species indicated (for further details see Figure S10). In line with the optical microscopy images, points "A", "C", "G", and "J" show a Cu LMM spectra corresponding to unoxidized Cu, whereas the remaining spectra show Cu LMM spectra corresponding to oxidized Cu (peak fitting confirms this; see SI section 2). Cu LMM peak fitting (see SI Figure S10) reveals that the presence of a graphene overlayer alters the oxide composition. Graphene covered and oxidized regions "B", "D", and "I" consist of only Cu 2 O, whereas locations "E", "F", and "H" that correspond to graphene uncovered regions are composed of a mixture of Cu 2 O and CuO.
Location "A" refers to a bilayer region and Figure 4 shows that "A" is unoxidized although the remaining Cu areas beneath monolayer regions of this graphene domain are oxidized. This highlights that bilayer graphene shows a very different interface oxidation behavior compared to monolayer graphene. Figure 5 shows Raman mapping data corresponding to the same sample regions as in Figure 3, performed to resolve the effect of Cu orientation and oxidation on the graphene doping concentration and strain. We present the Raman data as scatter plots of G peak wavenumber (ω G ) vs 2D peak wavenumber (ω 2D ) for the four key regions (Figure 2), to highlight the different local strain (ε) and doping (|n|) concentra-tions. 13,62−64 The color bar indicates the full width halfmaximum (fwhm) of the 2D peak, which is an indication of local strain homogeneity. 64 There is a clear trend in how the strain in the graphene relates to the Cu oxidation state irrespective of the Cu orientation. Oxidized Cu regions result in tensile strain (ε > 0) in the graphene layer ( Figure 5(a, b, and some regions in d)), whereas unoxidized regions are under compressive strain (ε < 0) (( Figure 5(c and parts of d)). Irrespective of the Cu grain, we observe a very widespread in graphene strain that ranges approximately from 0 ≲ ε ≲ 0.8% on oxidized areas and from −0.8 ≲ ε ≲ 0% on unoxidized areas. The graphene doping concentration also correlates with the underlying Cu oxidation state: in the case of "inhibited oxidation" we find the overall lowest doping concentration of approximately |n| ≅ ∼0−0.5 × 10 13 cm −2 . For "complete oxidation" |n| appears to converge to a value of approximately | n| ≅ 1 × 10 13 cm −2 , and for "irregular oxidation" |n| is inbetween the oxidized and unoxidized case and shows a larger spread from |n| ≅ ∼0−1 × 10 13 cm −2 . This is in line with the C 1s spectra provided in Figure S9, where a change in the level of charge transfer between the Cu and graphene results in a binding energy shifts of the C 1s core level. Here, a shift toward higher binding energies of approximately 0.15−0.25 eV suggesting n-type doping for locations "A", "B", "C", "D", and "I" is observed but no binding energy shift is observed for position "G" and only a small shift of approximately 0.05−0.1 eV is seen for location "J".
We further note that the Cu crystallographic dependent oxidation behavior described above is not limited to graphene islands, but the same Cu grain dependent oxidation behavior is observed for samples with full graphene coverage (see SI Figure S18). Furthermore, the results presented here are not  Figure S10). 57 Chemistry of Materials pubs.acs.org/cm Article limited to CMP polished Cu substrates but equally apply to cold rolled thin (25 μm) Cu foils that are substantially rougher (R a ≈ 300 nm). 45 SI Figure S19 shows that Cu grain dependent oxidation behavior following the 4 different oxidation categories identified in Figure 2 is also observed on these rough cold rolled Cu foils on which graphene was grown and which were exposed to air. Whereas our study primarily focuses on the oxidation at the graphene/Cu interface under ambient conditions, we also performed a wet oxidation treatment to investigate the Cu orientation dependence under such different, comparatively harsher conditions. 23,65 In order to use the same sample, the Cu interface oxidation was reduced by hydrogen annealing. 34 Annealing for 3 h at 250°C and 5 mbar H 2 reduces the oxide at the graphene Cu interface (see SI section 5). 34,35 After reduction, we immerse the Cu/graphene sample in deionized water at 50°C for 3 days, which causes wet-oxidation of the graphene/Cu interface. 4 In contrast to the above-described oxidation in ambient conditions, the microscope color contrast shows that wet oxidation at these conditions leads to uniform oxidation across all Cu crystallographic orientations (see SI section 5). Compared to oxidation in air, interface oxidation in water is significantly accelerated. 23,66,67 Furthermore, Raman spectroscopy analysis shows very low residual strain after water oxidation compared to air oxidation and in the case of water oxidation also graphene bilayer regions can be intercalated and oxidized.

■ DISCUSSION
The results herein show that for the graphene−Cu interface, there are several effects that influence the interface oxidation behavior, and these effects are strongly linked to the Cu surface orientation. The characteristic oxidation patterns observed can be associated with regions in the IPF plot rather than discrete Cu orientations (Figure 2). On the basis of this we can exclude epitaxial effects between graphene and the underlying Cu as a major influence on the observed oxidation behavior. Furthermore, Figure 3(a) (see also Figure S1) shows that if a graphene island crosses a Cu grain, then the oxidation pattern drastically changes at the Cu grain boundary to resemble that of the other islands on the new Cu grain, which indicates that the underlying Cu is the dominant factor in oxidation rather than a particular graphene orientation relative to the Cu. The dark red color in the optical microscope images indicates that some areas under the graphene develop a thicker Cu-oxide compared to the bare Cu surface, hinting at a galvanic corrosion mechanism. 23,29,32,33 Galvanic corrosion is driven by a work function difference between graphene and Cu. The work function of Cu is highly dependent on the Cu surface. 68 73 The work function of (undoped) graphene is around 4.5 eV. 74−76 Hence, if a surface, for example the {111}, {100}, and {211} orientations, have a work function higher than or equal to graphene, then no galvanic corrosion would be expected in this simplified model. In contrast, if the work function of the Cu surface is lower than the work function of graphene, e.g., vicinals of the {110} orientation, then galvanic corrosion effects are expected. Indeed, the overall picture in Figure 2 resembles some of these trends, i.e., vicinals of the {111} orientation with the highest work function mostly show oxidation along the wrinkles but no oxidation in unwrinkled areas (see Figure 3c,f) and the inhibited oxidation region corresponds to an area with work function of approximately Figure 5. Scatter plot of G peak position (ω G ) and 2D peak position (ω 2D ) as well as the full width half-maximum (fwhm) of the 2D peak indicated by color. ε refers to strain in % and |n| to the carrier concentration in cm −2 . (a) Refers to the data acquired at the region "irregular oxidation", the data points labeled BLG correspond to the bilayer graphene region, (b) to the region "complete oxidation", (c) to the region "inhibited oxidation", and (d) to the region "enhanced wrinkle oxidation".
Chemistry of Materials pubs.acs.org/cm Article 4.5 eV. Vicinals of the {110} orientation have relatively lower work functions, and we indeed see increased oxidation including the complete oxidation region, consistent with galvanic corrosion. Further evidence for this argument is that the acquired Raman spectra on areas corresponding to the "inhibited oxidation" region of the IPF do not show doping effects (see Figure 5(c)), indicating negligible difference between the Cu and graphene work functions. This is also the case after water oxidation (see Figure S17(b)). We note that the graphene covered surfaces shown in AFM maps of Figure S13 have a surface roughness (R a ) ranging from approximately 5 to 20 nm. The Cu work function decreases with increasing surface roughness and therefore the actual surface work function will be a convolution of the surface roughness and facet specific work function. 77 Also note that the electron beam penetration depth used for EBSD here is on the order of 50−100 nm. 78 Therefore, the graphene layer may locally be in contact with a range of other surface facets due to the surface roughness and surface reconstructions that are not captured by the EBSD measurement. 79 Nevertheless, there is a strong correlation of the interface oxidation behavior to the EBSD measured surface orientation despite the graphene layer being in contact with a rough surface. This not only holds for the flat samples used here, but also for comparatively rough (R a ≈ 300 nm) 45 Cu foils (see Figure S18). Further evidence for the strong contribution of galvanic corrosion are comparative studies of insulating hexagonal boron nitride (h-BN) monolayers on Cu, which are shown to be better long-term corrosion barriers than graphene, as insulating h-BN layer cannot drive galvanic effects. 80−82 Assuming a defect free graphene layer is impermeable to gas molecules, 83,84 for oxidation to occur oxygen species must intercalate beneath the graphene layer to react with the underlying Cu surface. We do not detect a Raman D peak away from the edges of the graphene domains on any Cu orientation (see Figure S11), however defect densities (n D ) on the order of less than n D ≈ 5 μm −2 on Cu substrates are very challenging to resolve with Raman. 85 Therefore, for Cu oxidation to occur under graphene, oxygen species need to diffuse through the graphene layer via defects or intercalate via the edges. For oxidation to proceed, the diffusion pathway must remain active and not be self-passivating upon oxidation. Additional AFM images in the SI (see Figure S13) show that there is no obvious correlation between surface roughness and the oxidation behavior with the exception of wrinkles ( Figure S13(f,g)).
The areas under the graphene wrinkles are oxidized, which indicates that the Cu−graphene interaction is weakened in these areas and oxygen species can more easily penetrate the interface here. For the Cu grains classified as "enhanced wrinkle oxidation" (Figure S13(f,g)) we also observe a regular surface corrugation pattern on unoxidized areas, and these corrugations are perpendicular to wrinkles. The surface corrugations correlate to the anisotropic oxide growth rate that is observed in direction of the corrugation and may be explained by an accelerated oxide species diffusion rate along the corrugation compared to oxide diffusion across the corrugation. These surface corrugation have previously been described as Cu step bunches. 86 The existence of such Cu step bunches, wrinkle formation and variations in strain are commonly rationalized via a thermal expansion coefficient mismatch between graphene and Cu. The thermal expansion coefficient mismatch causes the graphene film to be under compressive stress and the Cu surface layer to be under tensile stress. 43,86−90 The Cu surface can be considered as highly mobile or "pre-molten" at the CVD conditions used, however such premelting behavior is also dependent on the Cu crystallographic orientation. 91 Upon cooling the tensile stress in the Cu surface layer is relaxed by the formation of step bunches. 86 The compressive strain in graphene is relaxed by out of plane wrinkle formation, which again is dependent on the Cu crystallographic interaction strength of the graphene− Cu interface. 43,86−90 This is consistent with our measurements in Figure 5(d) (see also Figures S11 and S12), where we observe reduced strain along a graphene wrinkle. Upon intercalation and interface oxidation there is a volume expansion due to Cu oxidation and thus the compressive strain in graphene is released and graphene is under tensile strain (see Figure 5). Previous reports describe how such Cu corrugations can relax in graphene wrinkles which occur perpendicular to the Cu surface corrugation. 86 Corrugations perpendicular to wrinkles are consistent with our observation in Figure S13. Previous reports also observed oxidation along wrinkles for orientations close to Cu {111} and {001} but not on surfaces closest to {011}, which is in line with our findings. 92 Indeed, the Cu thermal expansion coefficient is dependent on the Cu crystal orientation but these differences in thermal expansion coefficient are relatively small. 93 Figure S13 shows how different graphene covered Cu orientations reconstruct and as a result the surface roughness becomes markedly different. Differences in surface roughness can have notable effects on the hydrophobicity of a surface and thereby drastically change the oxidation rates. 94−98 It is wellknown that hydrophobic Cu surfaces exhibit increased corrosion protection when exposed to atmospheric conditions. 99 Furthermore, water diffusion at the graphene substrate interface was shown to be significantly increased by increasing the hydrophilicity of the substrate. 100 However, the fact that we observe similar surface roughness for the "inhibited oxidation" and "complete oxidation" case hints that the difference is not pronounced when Cu is stored in air. However, prolonged immersion in water at 50°C eventually leads to oxidation on all surface orientations (see SI section 5), in line with previous reports that have shown accelerated oxidation rates in water. 23,66,67 By isotope labeling Luo et al. 23 have shown that water, and not oxygen species from air, is the main contributor to the interface oxide. Our data are consistent with this. An acceleration of the oxidation rate can be obtained by supplying water vapor or by immersion in water.
While the above has focused on graphene covered Cu, it is also known that the oxidation rate varies with surface orientation for bare Cu. 101−106 The oxidation rate of a bare Cu surface correlates with the surface step density during the initial phase of oxidation and therefore an orientation dependent oxide formation rate is expected. 101−104 However, the subsequent oxide growth process is complex and depends on various factors that cannot be solely explained by the surface step density. 105,106 Figure 1(b) shows an orientation dependent color contrast also for bare Cu regions. The relative orientation dependent oxide rate is further quantified in Figure  S7(b) and mapped in an IPF plot in similar fashion to Figure 2. Comparing Figures 2 and S7(b), a strong correlation between the oxidation behavior of graphene covered and bare Cu regions is observed. The orientations that are marked as inhibited oxidation under graphene in Figure 2 also show relatively low oxidation for bare Cu regions. Orientations with irregular and complete oxidation under graphene also show the Chemistry of Materials pubs.acs.org/cm Article highest degree of oxidation for bare Cu regions (see Figure  S7(b)). A notable difference in oxidation extent is the "enhanced wrinkle" orientations, where without the presence of graphene a relatively low degree of oxidation is observed. This highlights the argument above, that wrinkles enable oxygen species to diffuse and cause oxidation locally. Whereas graphene covered and bare Cu oxidation mostly show a similar trend when comparing patterns across the IPF, the oxide thickness and oxide composition are distinctly different. Our XPS data show that a graphene layer changes the Cu oxide composition and underneath graphene only Cu 2 O is observed, whereas bare Cu regions are composed of a mixture of CuO and Cu 2 O (see Figures S9 and S10). 55 Previous reports have shown that when Cu is exposed to air, CuO is only observed at the surface in direct contact with the atmosphere and a thicker Cu 2 O layer is found underneath the CuO layer. 107 Thus, for graphene covered oxide regions the graphene layer is protecting the Cu 2 O layer from contact with the atmosphere and therefore prevents the Cu 2 O layer from further oxidizing to CuO. We here focus on monolayer graphene films, but note that for areas covered with bilayer graphene, no Cu oxidation is observed at ambient conditions (see Figure 3a,d,g). Mismatched defects between the layers in bilayer graphene hinder diffusion of oxygen species in the vertical direction compared to monolayer graphene. 108 Furthermore, we do not observe wrinkles in the bilayer region (see Figure S13c), which when present provide additional pathways for oxygen species to access the Cu surface. This may explain why we do not observe oxidation under bilayer graphene (see Figure 3a,d,g), even on a Cu orientation where otherwise higher oxide coverages are observed. However, in more corrosive oxidation environments, e.g., by water immersion (50°C for 3 days) even bilayer regions become oxidized (see Figure S15(*)). This suggests that bilayer graphene does not offer a different corrosion protection mechanism beyond fewer penetration pathways for oxygen species, as eventually the oxidation front will intercalate from the graphene island edges inward irrespective of defect density.

■ CONCLUSIONS
We have synthesized graphene islands on flat polycrystalline Cu substrates with a range of different Cu orientations. After exposure to ambient conditions for 2 years, Cu grain orientation dependent oxidation was observed that can be classified as complete-, irregular-, inhibited-, and enhanced wrinkle-interface oxidation. The characteristic oxidation patterns observed are associated with regions in the IPF plot rather than discrete Cu orientations (see Figure 2). Our results clearly demonstrate that under ambient conditions, the Cu orientation is the dominant factor in determining the oxidation behavior under graphene, as our experimental design accounts for factors such as surface roughness, different growth conditions, and different graphene quality on each domain, allowing these to be excluded. We expect the observed classification to be relevant to different oxidizing environments, such as saturated water vapor now widely used for graphene transfer. Indeed, our generalized groupings and mapping onto the polar plot of Cu surface orientations consistently consolidates a large body of previous literature on select individual Cu orientations, all under slightly different conditions. Our example of immersion in water, however, shows that for very corrosive conditions ultimately all Cu orientations can be oxidized underneath the graphene (mono-or bilayer), i.e., the described Cu orientation dependent effects should be understood with respect to given exposure conditions. Recent reports have demonstrated the production of high and low index Cu foils with desired orientation, 109 which enables the findings of this paper to be translated beyond individual grains. We expect our study to facilitate future optimization of process technology tailored to diverse emerging application needs for graphene and other 2D materials that face analogous challenges, ranging from complete and uniform intercalation and interface oxidation as required for example for optimized 2D layer transfer to complete oxidation protection by an atomic monolayer for instance for plasmonics.

■ METHODS
Chemical mechanical polished (CMP), polycrystalline Cu metallic substrate (dimensions 1 × 10 × 10 mm 3 ) as well as single crystals with orientation (111), (011), and (001) were used as graphene growth substrate. According to distributor specifications (PiKem Limited) the Cu purity is 99.99% and the surface roughness is R a < 3 nm. CVD was performed in a commercial Aixtron Black Magic 4 in. system. The Cu substrate was loaded into the CVD system without further surface treatment. CVD synthesis was performed by heating in Ar atmosphere at 100°C/min to 1065°C. The sample was annealed for 30 min which leads to Cu grain growth for the polycrystalline substrates, followed by carbon precursor injection (gas flow rates Ar:H 2 :CH 4 (0.1% diluted in Ar) = 250 sccm:26 sccm:12 sccm) for a growth time of 4 h after which the heaters were switched off and the chamber was cooled to room temperature in an Ar atmosphere. 45 During all process steps, the chamber pressure was 50 mbar.
After synthesis the Cu substrate was stored in a class 10 000 cleanroom atmosphere for 2 years. The temperature of the cleanroom was regulated to 21°C, and the air humidity was regulated to 50%.
The model system employed here allows us to rule out a series of uncertainties. Unlike the commonly used cold rolled Cu foils for graphene synthesis, we used polished Cu substrate to rule out surface roughness induced effects on oxidation. Analyzing graphene islands which have previously been established to be predominantly single graphene domains, 45 we can rule out the effect of graphene domain boundary induced defects as pathways for intercalation of oxygen species, 46,110 or sites for enhanced dissociation of intercalating species. 36 Furthermore, a Cu sample which is only partially covered with graphene islands allows direct comparison between the oxidation rate of the uncovered and graphene-covered areas. Using a single polycrystalline substrate rather than a series of single crystalline substrates ensures that the growth conditions were equal on all Cu grains as gas flux and temperature dependence can be assumed to be negligible over the 1 cm 2 sample area in a 4 in. cold wall showerhead CVD reactor. 111 At the growth conditions used here CVD graphene defect densities are on a similar order for all Cu grains 112 and further we can observe graphene single domain islands that cross Cu grain boundaries to clearly show the effect of underlying Cu orientation.
Raman analysis was performed in a Renishaw inVia Raman microscope with a 532 nm laser with 100× objective and a 1800 l/ mm grating. Scanning electron microscopy (SEM) analysis was performed using a Carl Zeiss SIGMA VP at an acceleration voltage of 2 kV.
The EBSD was conducted on a FEI Nova NanoSEM scanning electron microscope with an acceleration voltage of 20 kV. The sample was tilted at 70°toward the EBSD detector, which was approximately 20 mm from the surface of the sample. The contrast and the background subtraction of the electron backscatter diffraction patterns were optimized to maximize signal strength prior to each mapping measurement. Multiple maps were taken and later stitched together to form the image maps used in this work.
Scanning X-ray Photoelectron Microscopy/microspectroscopy (SPEM) measurements were carried out at the Escamicroscopy Chemistry of Materials pubs.acs.org/cm Article beamline of the Elettra synchrotron facility (Trieste, Italy). 113 Samples were first annealed in vacuum to ∼100°C to remove residual surface contamination that otherwise obscures the signal from the Cu/graphene interface, with the temperature chosen to avoid significant changes to the Cu oxidation state 60 although the instability of CuO may lead to its partial reduction to Cu 2 O particularly under X-ray irradiation. 61 The X-ray beam was focused to a ∼180−200 nm spot by a Fresnel zone plate based optics. For both imaging and submicron spectroscopy, a SPECS-PHOIBOS 100 hemispherical electron energy analyzer with an inhouse customized 48 channel delay line detector was used. 114 A photon energy of 1074 eV was employed. The SPEM can mainly work in two modes: (i) microspectroscopy mode, i.e., the typical energy scanning mode employed in any standard XPS system, using a 180− 200 nm X-ray spot size, and (ii) imaging mode can map the photoelectrons emitted within a selected kinetic energy window by scanning the specimen with respect to the focused X-ray beam. In the imaging mode, each channel of the detector corresponds to slightly different energy, resulting in a 48 data point spectrum at each pixel. All binding energies are referenced to the Cu 2p peak energy which was set to be 932.6 eV. This approach is feasible on our sample because there is only a modest change between the metallic Cu and the Cu 2 O peak position and only small amounts of CuO are present. 59 To acquire the map of the C 1s peak in Figure 3(d−f) the energy window of the analyzer was centered at a binding energy of 285.3 eV and for the Cu LMM it was set to the kinetic energy of 917.0 eV. To obtain the Cu LMM chemical contrast maps shown in Figure 3(g,h) the integral ratio of the spectral regions containing mainly metallic (918.1−919.9 eV) and oxidized (915. 8