Self-Assembly of Graphene Nanoblisters Sealed to a Bare Metal

Feb 1, 2016 - Zamborlini, Imam, Patera, Menteş, Stojić, Africh, Sala, Binggeli, Comelli, and Locatelli. 2015 15 (9), pp 6162–6169. Abstract: We pr...
0 downloads 0 Views 7MB Size
Letter pubs.acs.org/NanoLett

Self-Assembly of Graphene Nanoblisters Sealed to a Bare Metal Surface Rosanna Larciprete,*,† Stefano Colonna,‡ Fabio Ronci,‡ Roberto Flammini,‡ Paolo Lacovig,§ Nicoleta Apostol,∥ Antonio Politano,⊥ Peter Feulner,# Dietrich Menzel,#,∇ and Silvano Lizzit§ †

CNR-ISC Istituto dei Sistemi Complessi, Via Fosso del Cavaliere 100, 00133 Roma, Italy CNR-ISM Istituto di Struttura della Materia, Via Fosso del Cavaliere 100, 00133 Roma, Italy § Elettra-Sincrotrone Trieste S.C.p.A., AREA Science Park, S.S. 14 km 163.5, 34149 Trieste, Italy ∥ National Institute of Materials Physics, Atomistilor 105b, 077125 Magurele-Ilfov, Romania ⊥ Department of Physics, University of Calabria, via ponte Bucci 31/C, 87036 Rende (CS), Italy # Physikdepartment E20, Technische Universität München, 85748 Garching, Germany ∇ Department of Chemical Physics, Fritz-Haber Institut, 14195 Berlin, Germany ‡

S Supporting Information *

ABSTRACT: The possibility to intercalate noble gas atoms below epitaxial graphene monolayers coupled with the instability at high temperature of graphene on the surface of certain metals has been exploited to produce Ar-filled graphene nanosized blisters evenly distributed on the bare Ni(111) surface. We have followed in real time the self-assembling of the nanoblisters during the thermal annealing of the Gr/ Ni(111) interface loaded with Ar and characterized their morphology and structure at the atomic scale. The nanoblisters contain Ar aggregates compressed at high pressure arranged below the graphene monolayer skin that is decoupled from the Ni substrate and sealed only at the periphery through stable C−Ni bonds. Their in-plane truncated triangular shapes are driven by the crystallographic directions of the Ni surface. The nonuniform strain revealed along the blister profile is explained by the inhomogeneous expansion of the flexible graphene lattice that adjusts to envelop the Ar atom stacks. KEYWORDS: Graphene, nickel, Ar intercalation, nanoblister, XPS, STM

I

copolymer self-assembling13 to manufacture supported graphene layers. Although these fabrication strategies have the potential to be eligible for several applications, due to the variety of chemicals and physicals agents employed in the numerous processing steps, issues might arise when chemical state and defect type in the quantum dots need to be known at the atomic level. A possible way to face this need is to turn to bottom-up strategies carried out in vacuum. Up to now, isolated quantum dots or nanosized domes have been achieved by using C60 fragmentation on transition metals,14 or incomplete growth of epitaxial graphene.1,15,16 Here, we profit from the extraordinary flexibility and mechanical strength of graphene coupled with its thermal instability on the surface of certain metals such as nickel to promote the dissolution of graphene in the substrate while forcing a small portion of it to keep afloat, owing to the

n addition to the impressive characteristics featured by extended, flat two-dimensional (2D) graphene, unusual intriguing behaviors appear when its electronic,1 tribological,2,3 and mechanical4 properties are modulated at the nanoscale. The fabrication and the manipulation of graphene nanostructures, possibly with size and edge control, allow the quantum confinement, the local charge density and the strain field to be varied to achieve bandgap opening,5 enormous pseudomagnetic fields,1 and tunable luminescence.6,7 Owing to these properties, the appeal of low-dimensional graphene pervades several technological fields. Graphene quantum dots are used as key components in physical and chemical sensors,8 are expected to yield new classes of bioimaging devices, and are used to open novel perspectives for energy related applications9 and in the medical field.10 In the last years, several methodologies have been envisaged to nanostructure graphene at the atomic level. Besides chemical routes able to readily produce nanosize graphene flakes starting from C-based materials,11 other fabrication approaches use lithographic patterning, 5 ion beam milling, 12 or block © XXXX American Chemical Society

Received: November 27, 2015 Revised: January 21, 2016

A

DOI: 10.1021/acs.nanolett.5b04849 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 1. Ar intercalation below graphene/Ni(111). (a,b) Selected C 1s and Ar 2p spectra measured on the Gr/Ni(111) surface exposed to increasing doses of 100 eV Ar+ ions. The best-fit curves and spectral components are also shown. In (b), the top Ar 2p curve was measured on the Ar/Ni(111) surface at 30 K and shows at 241.65 and 242.34 eV the contributions due to the first (FL) and to second physisorbed layer (SL), respectively. (c,d) Evolution of the C 1s and Ar 2p component intensities with the Ar+ dose. The intensities of the Ar 2p and C1s signals are expressed in terms of the Ar coverage below graphene (see Methods). (e,f) The 200 × 100 nm2 STM images of the graphene/Ni(111) surface taken (e) before (−0.01 V, 5 nA; inset 2 × 2 nm2) and (f) after (+0.02 V, 5 nA) the exposure to 0.6 MLE of Ar+ ions at RT. (g) STM image of the Ar+ dosed sample (50 × 50 nm2, +0.02 V, 5 nA); the inset shows the line profiles of three structures taken along the indicated lines. (h) High-resolution image (bottom) (12 × 6 nm2, + 0.02 V, 100 nA) and corresponding differentiated image along the x-direction (top) of a graphene bubble.

progressively assemble into 3D aggregates in the temperature range where graphene is stable on Ni. Above this point, the C atoms lying on the substrate suddenly penetrate into the bulk of the crystal, whereas the edges of the elevated graphene regions join with the Ni surface sealing the squeezed Ar atoms inside. The self-assembled GNBs remain stable when cooling down the system. We found that graphene wrapping the Ar aggregates forms GNBs that occupy nearly 10−20% of the sample surface and are surrounded by the bare Ni(111) surface. The starting sample is a graphene monolayer grown on Ni(111) by chemical vapor deposition using ethylene. For this surface, the XPS C 1s spectrum shown at the top of Figure 1a reveals a dominating (C0, 284.80 eV) and a weaker (C1, 284.40 eV) component assigned to strongly interacting and weakly interacting (in correspondence of rotated domains) graphene on Ni(111),25 respectively, and/or to the coexistence of top-fcc and bridge-top domains in graphene/Ni(111),26 plus a minimal additional contribution of Ni carbide (C2, 283.27 eV). The ratio between C1 and C0 can vary for different graphene growths

presence of Ar atoms properly assembled underneath. By exploiting the possibility to intercalate rare gas atoms below metal-supported 2D materials recently demonstrated for hBN17 and graphene,18−24 we have trapped remarkable quantities of Ar, equivalent to 15% of the substrate surface atoms, under epitaxial graphene monolayers on Ni(111) by means of irradiation with low energy Ar+ ions. At high temperature, the Ar atoms below graphene assemble into solid clusters.21−24 Therefore, while graphene in direct contact with the Ni surface decomposes, the floating regions encase the Ar aggregates forming graphene nanosized blisters (GNBs), evenly distributed on the bare substrate surface and stable in a wide temperature range. We used X-ray photoelectron spectroscopy (XPS) to monitor the Ar+ intercalation and follow in real time the interface evolution during thermal annealing and scanning tunneling microscopy (STM) to study the morphology and the structure of the GNBs at the atomic scale. The noble gas atoms stay locked between graphene and the metal surface and B

DOI: 10.1021/acs.nanolett.5b04849 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

spectra manifests the response of the electrons of the surrounding media in screening the core hole generated by photoemission. For Ar atoms in direct contact with metals, substrate electrons are readily available for hole screening, whereas for the Ar atoms at the interior of a bubble the relaxation is less efficient and the photoelectron is emitted with lower kinetic energy resulting in a higher BE. Therefore, the BE shift of the Ar 2p components provides information on the shape and the size of the Ar aggregates.36,37 Figure 1b shows that at low coverage, which is in correspondence to isolated Ar atoms trapped below graphene, the BE of Ar0 is shifted by −0.4 eV with respect to the energy position of the Ar 2p spectrum that we measured for Ar atoms adsorbed on the bare Ni(111) surface at 30 K (top of Figure 1b). This clearly indicates that the extra-atomic screening of the core hole is more efficient for the Ar atoms squeezed between graphene and the Ni substrate than for atoms physisorbed on the metal. The progressive down-shift of Ar0 up to −0.22 eV with ion dose can be related to incremental changes of the local environment that for inner atomic orbitals mainly determine final state effects (variation of the extra-atomic relaxation by matrix electrons).37 At Ar+ doses higher than 0.2 MLE, the appearance of the component Ar1 shifted by +0.3 eV with respect to Ar0 (see Figure.S4) is a fingerprint of the formation of two-layer Ar clusters below graphene. As a matter of fact, the shift to higher BE indicates that the Ar atoms in contact with graphene but not with Ni are less efficiently screened than single layer atoms and thus photoemit at higher BE.21 The evolution of the total Ar 2p and C 1s intensities with Ar+ dose shown by the black curves in Figure 1c,d reveals that the C loss is negligible in the first part and then becomes linear with irradiation. Upon irradiation with the highest Ar+ dose, ∼25% of graphene has been removed. In contrast, the quantity of intercalated Ar initially grows almost linearly with the Ar+ dose. However, above 0.4 MLE ion irradiation becomes more effective in releasing some of the intercalated Ar rather than in trapping new atoms, likely due to the formation of larger holes in the graphene lattice. From the ratio between the Ar 2p and C 1s intensities, we have estimated that the maximal Ar coverage below graphene is 0.17 monolayer (ML) (1 ML = 1.86 × 1015 atoms/cm2), which gives an average number of intercalated atoms per impinging ion of 0.42, and is in good agreement with the value found for graphite32 and slightly higher than that reported for Gr/Ir(111).31 Figure 1f−h reports the STM images taken on the Gr/ Ni(111) exposed at RT to 0.6 MLE of Ar+ ions and, according to the curve of Figure 1d, intercalated with ∼0.15 ML of Ar. The large area image (Figure 1f) clearly shows that after the Ar+ implantation the surface roughness and the degree of disorder increase due to the presence of a large number of protrusions and depressed features (Figure 1g). The bright protrusions absent in the pristine Gr/Ni(111) (Figure 1e) arise from graphene lifted by the Ar atoms penetrated below. As the van der Waals diameter of Ar (3.76 Å)38 is larger than the graphene-Ni interlayer spacing (2.11−2.17 Å),39,40 the Ar interstitials deform elastically the graphene lattice and appear as bulges. The profiles taken on the high-resolution image in Figure 1g show that the smallest protrusions have widths of ∼2 nm and heights even smaller than 0.2 nm, whereas the biggest protrusions are ∼6 nm wide and ∼0.3 nm high. The smallest protrusions are likely due to the immobilization of isolated Ar atoms,41 which remain compressed under graphene causing only a limited out-of-plane bending.22,42 However, where

carried out in the same experimental conditions (see Figure.S1), possibly due to the uncontrollable presence of low quantities of defects in the Ni(111) surface that favor the nucleation of the different graphene structures.26 The complete graphene layer covering the substrate terraces is shown in the large area STM image of Figure 1e, whereas the atomic scale image reported in the inset shows the triangular structure typical of the top-face-centered cubic (fcc) or top-hexagonal close-packed (hcp) configurations of Gr/Ni(111).27,28 The volume between graphene and the Ni(111) substrate is loaded with Ar atoms by exposing the Gr/Ni(111) surface to a flux of 100 eV Ar+ ions (see Figure.S2). We found that an Ar+ dose of 9 × 10−3 MLE (1 monolayer equivalent (1 MLE) = 1.86 × 1015 atoms/cm2 corresponds to the Ni(111) surface atomic density) is sufficient for the appearance of some intensity in the Ar 2p region in correspondence of the Ar0 doublet (3/2 component at 241.26 eV). It is worth noting that while Ar atoms might accumulate below the surface of metal crystals irradiated with ∼1 keV Ar+ ions,29,30 low energy (≤100 eV) ion-surface collisions only displace at most a few surface atoms, producing vacancies and interstitial defects in graphene at room temperature (RT).31,32 We have fully excluded the penetration of Ar into the Ni substrate by verifying that the XPS spectrum measured on the bare Ni(111) surface exposed to a dose of 0.77 MLE of Ar+ (that is the maximal dose used in this XPS study) is flat in the Ar 2p region (see Figure.S3). Analogously, for Gr/Ir(111) exposed to 100 eV Ar+ ions STM measurements have excluded the implantation of Ar in the substrate.31 Therefore, because the zero RT sticking coefficient of Ar both on Ni and on graphene rules out that any noble gas atoms can bind to the sample surface, the presence of Ar in the Gr/Ni(111) sample has to be attributed to intercalation below graphene. While knocking out C atoms, the Ar+ ions remain trapped below graphene due to the fast rebonding of the C dangling bonds to the Ni substrate. The high-resolution C 1s and Ar 2p spectra shown in Figure 1a,b with the corresponding best fit curves and spectral components allow us to elucidate the modification of the Gr/ Ni(111) surface and to determine the quantity and the state of the trapped Ar. The C 1s spectra (Figure 1a) and component intensity curves (Figure 1c) show that the intercalated Ar decouples the graphene layer from the Ni surface. The C1 component grows progressively at the expense of C0. After an Ar+ dose of ∼0.2 MLE, C0 has decreased to less than 60% of its initial value, whereas C1 has more than doubled due to the formation of sizable regions where the graphene−Ni interaction is switched off. In parallel, the growth of the carbide component C2 points out that an increasing number of C atoms, likely those surrounding the vacancies, gets stabilized by making C− Ni bonds with the substrate. The new component C3 (283.93 eV), which arises around 0.04 MLE of Ar+, is attributed to the formation of single vacancies and larger defects in the graphene lattice.33 The progressive shift of the whole C 1s spectrum by ∼−0.5 eV (see Figure.S4) can be related to a variation of the overall graphene−substrate interaction due to the presence of the Ar intralayer. For Ar+ doses ≥0.6 MLE, C1 is located at 284.0 eV, which is close to the BE position found when graphene is decoupled from the Ni(111) substrate by intercalated Au atoms34 or additional graphene layers.35 Concerning the state of the trapped Ar, much insight can be gained from the evolution of the Ar 2p spectra. Rare gas atoms are not expected to make chemical bonds with the hosting materials. Nevertheless, the energy position of their core level C

DOI: 10.1021/acs.nanolett.5b04849 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 2. Formation of Ar-filled graphene nanoblisters. (a) Sequence of C 1s and Ar 2p spectra measured simultaneously at increasing temperature while annealing at a rate of 0.4 K/s, the Gr/Ni(111) sample containing 0.15 ML of intercalated Ar. The dark gray curves highlight the dramatic changes occurring between 900 and 935 K, whereas the red and blue curves mark the stable line shapes measured at 940 K. The top and the bottom panels show the high-resolution spectra measured before annealing and on the sample heated to 940 K and then cooled to RT, respectively. (b) Evolution of the C 1s and Ar 2p component intensities versus time during thermal annealing. The top panel shows the sample temperature. (c) Large area STM image of the Gr/Ni(111) surface dosed with 0.6 MLE of Ar+ ions and annealed for 10 min at 1000 K (100 × 200 nm2, −0.02 V, 5 nA). (d) In-plane and (f) side-view 3D rendering of STM images. (e) GNB profiles taken along the lines indicated in (d). (g) Distributions of the GNB height for the samples dosed with 0.3 and 0.9 MLE of Ar+. The histograms are obtained by performing for each dose a statistical analysis on several STM images for a total sampled surface of 1.7 × 105 nm2.

possible several atoms clusterize to minimize the graphene buckling and form bigger protrusions with height comparable with the Ar atomic diameter. Two-layer Ar aggregates wrapped by higher graphene protrusions are rarely observed, which is in agreement with the low intensity of the second layer component Ar1 in the Ar 2p spectrum. The depressed features, which appear black in the image, can be attributed to defects induced in the graphene layer by the ion bombardment, which allow the intercalation of Ar ions beneath graphene.31 The evidence that the rough surface consists of a continuous graphene layer is provided by the high-resolution STM image reported in Figure 1h together with the differentiated image, which shows the atomic structure of graphene both on the top of the protrusion and in the surrounding region. The formation of GNBs was achieved by annealing the Gr/ Ar/Ni(111) sample. In order to follow the interface restructuring driven by the high temperature, we have used real-time XPS to monitor simultaneously the C 1s and Ar 2p spectra (Figure 2a) while heating a Gr/Ni(111) sample containing 0.15 ML of Ar. Starting from the spectra measured at RT (top), the whole series taken during annealing are shown in the central panels. The system evolution is precisely probed

by the C 1s and Ar 2p spectral component intensities plotted in Figure 2b as a function of the annealing time and derived by analyzing the XPS series. Between 300 and 880 K, the intensity of the Ar 2p signal decreases. This is due partly to the deintercalation promoted by the high temperature and partly to the aggregation of Ar atoms into multilayer clusters43 with a consequent decrease of the XPS signal from the deeper layer atoms. In fact, the progressive continuous high-BE shift of the Ar 2p spectrum together with the conversion of Ar0 into Ar1 indicates the ripening of the Ar aggregates into 3D clusters.21−24,43 Because part of the Ar spacer is lost, graphene tends to recover the pristine interaction with Ni(111) revealed by the increase of C0. At around 940 K, the C 1s intensity starts to rapidly decay owing to the abrupt dissolution of graphene into the substrate. By keeping the temperature fixed at 940 K, in 5 min a metastable state is reached after the loss of ∼90% of the C 1s intensity. The curves in Figure 2b show that C0, the component due to the graphene strongly interacting with the Ni substrate, vanishes almost completely, whereas C1, arising from the regions lifted by the intercalated Ar atoms, partly survives. This behavior describes the formation of Ar filled GNBs. In correspondence to the D

DOI: 10.1021/acs.nanolett.5b04849 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters sudden decomposition of graphene, the Ar coverage drops moderately with the remaining Ar atoms impeding the residual graphene from dissolving into the substrate. The abrupt squeezing of Ar atoms into a smaller volume is unequivocally signaled by the appearance of the new component Ar2 (242.43 eV) in the Ar 2p spectrum (bottom right panel of Figure 2a). Ar2 is due to the inner atoms in the cluster, which being distant both from graphene and Ni substrate, behave like second layer (SL) atoms in the physisorbed multilayer (top of Figure 1b). At this stage, the C 1s spectrum (bottom left panel of Figure 2a) consists of the narrow C1 peak due to decoupled graphene accompanied by a negligible residual of the C0 component and a minute C2 peak due to the interfacial bonds with Ni, which seal the blister edges preventing Ar leaks. Further graphene decomposition is triggered by increasing the temperature to 1050 K. In a few minutes, this leads to the complete disappearance of C and Ar, whose XPS signals decay exponentially.21Alternatively, cooling the sample from 940 K to RT freezes the Ar-filled GNBs on the Ni surface, proving the high quality of the graphene skin and the absence of large defects that are the only places where Ar deintercalation could take place. It is worth noting that we did not observe any C segregation during sample cooling, which followed an exponential law with time constant τ = 1.4 min. However, for lower cooling rates the segregation of carbide phases and small graphene islands was observed to initiate around 680 K and mostly saturate below 550 K (see Figure.S6). The total quantity of segregated C is equivalent at most to less than 40% of the graphene forming the GNBs, thus much lower than the amount that dissolves into the substrate during the annealing of the Arintercalated Gr/Ni(111). It is interesting to evaluate the fraction of the initial amount of intercalated Ar that remains trapped in the GNBs. The Ar2ptot curve in the central panel of Figure 2b, which traces the total Ar 2p signal, decreases to 60% of its initial value after annealing at 940 K. However, if the Ar 2p components are properly weighted by taking into account that the ripening of the Ar atoms from nearly 2D to 3D clusters damps the contribution of the deeper Ar atoms to the XPS spectrum, it turns out that only ∼15% of the implanted Ar is lost during the thermal annealing. We found that neither the efficiency of Ar intercalation below graphene nor the yield of GNBs upon thermal annealing changes significantly for different initial ratios between the C1 and C0 components of the C 1s core level (see Figure S7 and connected discussion). It is worth noting that a recent study has shown that similar GNBs can be obtained by annealing graphene monolayers grown on a Ni(111) crystals implanted with 1 keV Ar+ ions, which at high temperature diffuse to the surface and remain trapped below graphene.21 In that case, however, the initial Ar coverage below graphene was nearly one tenth of the value obtained with the method described here. A further proof that the residual graphene is decoupled from the Ni substrate is provided by the near edge X-ray absorption (NEXAFS) spectra at the C K-edge measured on the Gr/Ni(111) and GNBs/Ni(111) samples at normal (θ = 90°) and grazing (θ = 20°) incidence and shown in Figure 3. The two geometries correspond to having the photon electric field E in turn parallel or nearly perpendicular to the substrate plane and for a flat graphene layer enhance the intensity of the C 1s electron transitions to the empty σ* or π* orbitals, respectively. The double peak structure appearing in the 1s → π* region for Gr/Ni(111), which results from the transition of the C 1s core electrons to two unoccupied C−Ni hybridized

Figure 3. NEXAFS monitoring of graphene decoupling. X-ray absorption spectra at the C K-edge measured for Gr/Ni(111) and GNBs/Ni(111) at grazing and normal incidence. The inset shows the corresponding C 1s core level spectra measured for the two samples.

states,44 in the case of GNB/Ni(111) transforms into the sharp π* feature typical of graphite and nearly free-standing graphene,45 demonstrating that for the large majority of the C atoms the interaction with the Ni substrate is switched off. The contribution of the vanishingly small fraction of C atoms bonded to the Ni surface at the periphery of the GNBs (compare the C 1s spectrum at the bottom of Figure 2a) cannot be discerned in the NEXAFS spectra. The limited dichroism of the π- and σ-resonances in comparison to flat graphene45 derives from the curved blister morphology that causes the orientation of the π* orbitals and hexagon planes with respect to E to vary continuously along the surface. The formation of GNBs evenly covering the Ni surface is demonstrated by the STM images shown in Figure 2c,d, taken on the Gr/Ni(111) surface dosed with 0.6 MLE of Ar+ ions and annealed for 10 min at 1000 K. The surface morphology, strongly different from that observed before annealing, shows GNBs arranged on the flat substrate and along the terrace edges. The GNB in-plane size distributions (Figure.S8) are peaked around 20 and 30 nm2 for Ar+ doses of 0.3 and 0.9 MLE, respectively, values that in the approximation of round shapes would correspond to equivalent radii of 2.5 and 3 nm. In samples dosed with 0.3 MLE of Ar+, the GNBs have a surface density of ∼4.2 × 10−3 nm−2 and cover slightly less than 10% of the Ni surface. Larger coverage values up to ∼20% are found for more heavily dosed samples. Typical line profiles shown in Figure 2e indicate heights spanning between 0.7 and 1.5 nm, which is in agreement with the formation of multilayer Ar aggregates as inferred from the XPS results. Figure 2f shows a side view of the GNBs where the heights appear grouped around discrete values, which are traced by the horizontal lines. A more detailed information on the height distribution of GNBs in differently dosed samples is provided by the E

DOI: 10.1021/acs.nanolett.5b04849 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 4. Atomic structure of the GNBs on Ni(111). (a) STM image of the sample surface (100 × 100 nm2, 0.02 V, 5 nA). (b) Close up of the GNB indicated by the yellow circle in (a) (10 × 10 nm2, 0.02 V, 5 nA); the crystallographic directions are referred to the underlying Ni(111) substrate. (c) Close up of the bare Ni(111) substrate (10 × 10 nm2, −0.005 V, 150 nA). (d) Enlarged view of the region marked by the red square in (b) after the background subtraction. (e) Line profile measured along the green line in (d). (f) Line profile of the GNB taken along the dashed red line in (b).

nm in the GNB height distribution shown for 0.3 MLE in Figure 2g indicates an effective height h ≃ 0.46 nm for the Ar atom stacks (graphene layer thickness 0.335 nm), corresponding to Ar bilayers with d111 = h/2 ≃ 0.23 nm (a ≃ 0.40 nm), which is in agreement51 with the pressures of the several hundreds of kbar calculated above. The Ar layers above may be less ordered due to the flexibility of the Gr membrane, but the layered structure is likely maintained for blisters containing only a few layers. The ticks in the 0.3 MLE histogram, which mark the heights expected for similarly compressed Ar multilayers, in a few cases correspond to structure emerging over the histogram envelop, providing the indication that also inside bigger GNBs the (111) stacking is maintained and that the Ar compression does not change substantially with respect to the Ar bilayers. A less evident indication for Ar layer ordering is provided by the histogram relative to the Ar dose of 0.9 MLE that shows a smoother height distribution where the bilayer Ar stacks have almost vanished. By considering a packing density of 0.74 for the Ar fcc lattice, it comes out that the inner volume of a typical GNB contains nearly 103 Ar atoms. In Figure 4a a 100 × 100 nm2 STM image showing several GNBs is reported, where it is possible to observe that the GNBs do not have a round shape in-plane but nearly truncated triangular shapes.52 By inspecting the high-resolution image of a bare substrate region (Figure 4c) it is noted that the surface is a perfectly ordered Ni(111) and that the GNBs edges are mainly oriented along the ⟨110⟩ high-symmetry directions. The presence of the ordered Ni(111) lattice in the space free from GNBs demonstrates that, where not supported by Ar

histograms of Figure 2g derived for the samples intercalated with the lowest (0.3 MLE) and highest (0.9 MLE) Ar+ doses investigated by STM. A broad band centered at 1.5 nm appears in both histograms, the one corresponding to 0.3 MLE showing also a narrow peak around 0.8 nm. The pressure inside a typical GNB can be roughly estimated by using the relation between the pressure difference Δp across a graphene circular bubble and its mechanical and geometrical properties46 Δp = K(ν) Et(h3/R4), where K(ν) is a coefficient that depends on the Poisson’s ratio ν, E is the Young’s modulus, t is the graphene thickness, and h and R are the height and the radius of the bubble, respectively. By using Et = 347 Nm−1,46 K(ν = 0.16) = 3.0946 and for h and R the most frequent values deduced from the distributions in Figure 2g and Figure.S8 (h = 1.5 nm, R = 3 nm), Δp ≃ 450 kbar.23 For inert gas implanted in metals slabs several publications have demonstrated that the atoms are confined in solid precipitates aligned epitaxially with the crystal lattice of the matrix up to elevated temperatures (fcc Kr and Xe in Al matrix, hcp Xe in Zn matrix).47,48 In view of the high pressures estimated above, we can safely assume that also inside the GNBs the Ar is solid (probably as an fcc crystal) considerably above RT and that it is likely that at least the first Ar layers on Ni align to the substrate in this range. If the Ar atoms in the GNBs piled up forming fcc crystals oriented as the Ni(111) substrate, the lattice parameter a = 0.530 nm49 (and the interlayer distance d111 = a/√3 = 0.3 nm) measured for the uncompressed Ar lattice should significantly decrease under high pressure.50,51 Considering that, as it will be shown below, the GNBs emerge above the bare Ni surface, the peak at ∼0.8 F

DOI: 10.1021/acs.nanolett.5b04849 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 5. GNB stability in air. (a) Survey and (b) Ar 2p XPS spectra taken on a GNBs/Ni(111) sample before (black curves) and after (red curves) the exposure to air for several hours. The spectra of the air-exposed sample were acquired after the reintroduction in UHV. In each panel, the spectra are vertically shifted for clarity. (c−f) STM images taken on a GNBs/Ni(111) sample (c) before and (d−f) after exposure to air. In this latter case, the sample was annealed for 2 h at 470 K before image acquisition. The inset in (f) shows a magnification of the framed region after the subtraction of a polynomial background to take into accounts the GNB curvature. (c,d) 200 × 200 nm2, 0.02 V, 5 nA; (e) 50 × 50 nm2, 0.02 V, 5 nA; (f) 10 × 10 nm2, 0.005 V, 15 nA.

atoms, graphene is destroyed and its C atoms diffuse into the metal and dissolve in the crystal bulk. The GNB shape is better imaged in the high-resolution STM image reported in Figure 4b showing the GNB marked by the yellow circle in Figure 4a. The GNB presents an arched shape (see Figure 4f) with the inplane edges determined by the ⟨110⟩ crystallographic directions of the Ni substrate. In the enlarged view of the region marked by the red square (Figure 4d), the honeycomb lattice of graphene is detected after the subtraction of a polynomial background that takes into account the GNB curvature. The image profile collected along the green line traced in Figure 4d and shown in Figure 4e accounts for an expanded graphene lattice parameter of about 0.28 nm, corresponding to a strain ϵ of ∼14%. Such a large ϵ value is detected only locally on the GNB surface, whereas a lower value of 4% is obtained from the relation ϵ = (L − L0)/L0,53 where L0 and L are the lengths of the cord (flat graphene) and of the arc (blister), respectively, as deduced from the line profile of Figure 4f. The presence of highly stressed areas in the GNB can be attributed to the inhomogeneous expansions of the flexible graphene lattice, which adjusts to envelop the Ar atom stack. They should be usable to modify the reactivity of the graphene in analogy to what has been shown for metal bubbles.30 In order to explore the versatility of the GNBs, we have tested their stability in atmosphere, which is a technologically relevant issue. GNB/Ni(111) samples were removed from the ultrahigh vacuum (UHV) chamber and kept in air for several hours. Afterward, the samples were reintroduced in UHV and examined by XPS and STM. Figure 5a compares the survey XPS spectra taken before and after the exposure to ambient atmosphere. In the latter spectrum, taken without any thermal annealing of the sample, the Ni spectral features appear

attenuated and modified, which is in agreement with the presence of the intense O 1s core level, which together with the O KLL features attests the occurrence of surface oxidation. In contrast, the Ar 2p spectra taken before and after the exposure to air show similar line shapes and comparable intensities, indicating that the quantity of Ar and the GNBs did not change appreciably. In fact, the STM images taken on the air-exposed sample (Figure 5d) show that the GNB surface density is equivalent to that observed before the exposure to air (Figure 5c). Instead, even after a mild annealing to 470 K, the Ni surface of the air-exposed sample (see Figure 5e) appears much rougher than that of the pristine sample (Figure 4c) due to the occurrence of metal oxidation. The intense C 1s peak appearing in the XPS spectrum taken after the exposure to air contains the contributions of contaminants that overlap the pristine graphitic line shape due to GNBs (see Figure.S9). In this respect, it is worth noting that the limited attenuation of the Ar 2p peak intensity after air exposure (Figure 5b) reveals that only exiguous quantities of contaminants are bonded to the GNBs. The integrity of the graphene lattice in the air-exposed blisters is attested by the high-resolution image of Figure 5d, where the inset shows the framed region after background subtraction. We verified that the air-exposed GNBs can be safely heated in UHV up to 500 K as reported in Figure.S10. The method described here generates evenly distributed GNBs surrounded by bare metal surface, whose termination might be intentionally modified by chemical treatments. The known thermal instability of graphene on several transition metals such as Fe(110),54 Pd(110),55 Rh(111),56 Co(111),57 Re(0001),58 and probably on their alloys prefigures the extension of this method to a variety of other substrates. Different GNB shapes and lateral dimensions might be G

DOI: 10.1021/acs.nanolett.5b04849 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

STM measurements were performed at ISM-CNR in an UHV system composed of a preparation chamber vacuum connected to a analysis chamber housing an STM stage (Omicron LT-STM). The reported images were all collected at room temperature in constant current mode by using electrochemically etched W tips cleaned by in situ electron bombardment before the use. In order to investigate the stability of GNBs in ambient atmosphere GNB/Ni(111), samples were removed from UHV and kept in air for several hours. Afterward, the samples were reintroduced in UHV and examined by XPS and STM. The behavior of the air exposed GNBs upon thermal processing was tested by monitoring the Ar 2p and C 1s spectra during prolonged heating at 470 K and during annealing up to 570 K. The effect of air exposure on the GNB morphology was studied by comparing the STM images taken on a GNBs/Ni(111) sample before and after the exposure to air for several hours. In this case, the air exposed sample was annealed for 2 h at 470 K, treatment that, in agreement with the indication obtained by the XPS measurements, safely removed a large portion of surface contaminants allowing the acquisition of the STM images.

achieved by changing the size of the intercalated rare gas atoms. The dependence of the chemical reactivity of graphene on the local strain field59,60 suggests that chemical reactions inhibited on the flat graphene might take place at the blisters and discloses the possibility to produce an even distribution of functionalized GNBs. Therefore, the GNBs/Ni(111) system studied here, together with a whole class of similar systems that can be envisaged, represents a clean, stable, easy to handle benchmark for fundamental studies and application-oriented investigations in many fields including thermomechanics, nanotribology, optoelectronics, nanomagnetism, and chemistry of graphene. Finally, it is noteworthy to mention that the encapsulation of rare gas atoms in the GNBs provides at hand at room temperature a solid rare gas target ready to be probed with spectroscopic and structural tools. Methods. The XPS, NEXAFS, and STM measurements were performed in two different laboratories on samples prepared on the same Ni(111) crystal by following the same procedure. The Ni(111) crystal was cleaned by repeated sputtering cycles at 1 keV followed by annealing up to 1020 K. Graphene was grown by dosing ethylene at 5 × 10−7 mbar onto the Ni(111) surface kept at 890 K. The sample quality was checked by means of low-energy electron diffraction (LEED) and by verifying that the spectral features relative to C and O contaminants were absent in the XPS or in the Auger spectra. Ar atoms were implanted below graphene by exposing at RT the sample to a flux of 100 eV Ar+ impinging normally onto the surface at an Ar background pressure of 1 × 10−6 mbar. During irradiation, the sample current was 0.3 μA and different doses were obtained by varying the irradiation time. The Ar+ fluence and the Ar coverage below graphene are given in monolayer equivalents (MLE) and monolayer (ML), respectively, where 1 MLE = 1 ML = 1.86 × 1015 atoms/cm2, which is the Ni(111) atomic surface density. The photoemission and X-ray absorption experiments were performed at the SuperESCA beamline of the synchrotron radiation source Elettra (Trieste, Italy). High-resolution C 1s and Ar 2p core level spectra were measured at a photon energy of 400 eV with an overall energy resolution of