Tailored CVD graphene coating as a transparent and flexible gas barrier

The chemical vapor deposition (CVD) method to obtain tailored graphene as a transparent and flexible gas barrier has been developed. By separating nucleation step from growth, we could reduce early graphene nucleation density and thus induce better stitching between domain boundaries in the second growth step. Furthermore, two step growth in conjunction with electrochemical polishing of Cu foils achieved large graphene domains and improved graphene quality with minimized defects. The performance of resulting graphene as a gas barrier was superior to the graphene obtained by one-step growth on polished or unpolished Cu foils. The CVD graphene reported here could open up the possibility for exploring graphene-based gas barrier due to the minimized density of defect area.

For practical applications as protective barrier, preparation of large area, defect-free graphene and a suitable method for its integration on polymeric or other substrates are important requirements. Among many preparation methods, chemical vapor deposition (CVD) is the most effective method to reliably produce high-quality, wafer-scale monolayer graphene films. By virtue of this, CVD-grown graphene has been investigated as a protective barrier to prevent transport of gas molecular between the environment and the arbitrary surface such as polymers or metal layer [9][10][11][12][13][14][15][16][17] . This is a promising step towards exploring alternative materials, because conventionally used organic polymers for barrier application has limited potential due to their relatively high gas permeability as compared to high-criteria demands for modern packaging application and current display devices such as organic light emitting diodes and liquid crystal displays. However, despite the promising properties, CVD graphene layer as a protective gas barrier has been found to result reduced oxidation and corrosion resistance in protecting the metal surfaces 12,15,l8,19 . Also, a recent study showed that the corrosion rate of Cu is accelerated by graphene overlayer compared to bare Cu under exposure for long period 20 . This is because line defects and grain boundaries formed at the interfaces between two domains with different crystallographic orientations of polycrystalline Cu foil allow the transfer of gaseous molecules toward arbitrary surface covered by such graphene layer 15 . Furthermore, CVD graphene contains many defects, viz, point defects, wrinkles, folds, tears and cracks, due to the local variations in the properties of the Cu substrate such as its thickness, surface roughness, grain size and orientation, surface impurities, and transfer process. As a consequence, unfavorable gas and aerosol molecules can be transferred through these defects, causing local corrosive change in the surface and degradation of the protection ability of the graphene 18,19 . Thus, for barrier applications, it is very important to ensure complete surface coverage by minimizing any grain boundaries or structural defects. Numerous approaches have been developed to reduce the nucleation density and point defects as well as to increase the graphene domain size [21][22][23][24][25][26][27][28][29] . Alternatively, it has been demonstrated that the corrosion rate of metallic surfaces can be reduced by controlling the number of graphene layers and multilayer graphitic films formed by gentle chemical reduction of graphene oxide laminates with hydroiodic and ascorbic acids were found to furnish high-quality barrier films that obstruct all gases and liquids 12,16 . Hsieh et al. 15 introduced a method to selectively passivate the defects in graphene by using atomic layer deposition (ALD) of aluminum oxide, which offered enhanced corrosion protection to graphene. However, the above approaches have some disadvantages such that an increment in the number of graphene layers diminishes the transmittance and an additional transfer process reduces the production yields. Also, the ALD is very expensive and hence its broad application is still limited because of a lack of scale-up.
Here, we use a combined approach of electro-chemical polishing (ECP) and two-step growth methods to improve the gas barrier property of graphene without an increase in the layer numbers or additional fabrication process. ECP is known as a highly effective step to reduce the surface roughness of Cu foil and remove the impurity particles 30 . Both effects could decrease the point defects and wrinkles within the graphene layer and increase the graphene domain size. However, single-crystal terraces and step edges unintentionally exist on Cu foils during graphene growth in the CH 4 /H 2 mixture which act as nucleation seeds and can influence the gas barrier properties of graphene 26,31 . In particular, the two-step growth approach reported here is facile and simple. From the understating that the nucleation density is governed by carbon supply rate, a nucleation step was decoupled from the growth to achieve high quality graphene. By keeping the carbon supply rate low during the initial stage of graphene growth, a low nucleation density is achieved and a successive fast lateral growth is dictated by a high carbon supply rate to achieve continuous graphene. The simple two-step growth approach led to better stitching between domain boundaries.

Results
The initial stage of growth is critical to obtain large size domains and high quality graphene layer. In this study, three different graphene layers are grown as follows: The first graphene was grown on as-received Cu foil without any treatment by one step growth (1step-w/o ECP), the second graphene was grown on electrochemically polished Cu foil by one step growth (1step-ECP) and the third graphene was grown on ECP Cu foil by two step growth (2step-ECP). Figure 1(a-c) show the scanning electron microscopy (SEM) images of the three different graphene samples at early stage of growth after 60 sec, respectively. All graphene domains on either substrates show irregular shape, similar to previously reported graphene domains grown by CVD [21][22][23] . According to previous studies on the CVD growth of graphene, the CH 4 to H 2 flow ratio, total pressure, temperature, and Cu foil treatments are crucial parameters that influence the graphene domain shape [21][22][23][24][25][26] . Interestingly, the two-step growth on ECP Cu foil resulted in large graphene domains with an average size of approximately 110 μm 2 , twice or four times larger than the graphene domains obtained by one-step growth on ECP and unpolished Cu foils, respectively, as illustrated in Fig. 1(d). Furthermore, it is observed that the increase in domain size results in a decrease in the nucleation density, which is estimated to be 0.7 × 10 6 /cm 2 in the case of two-step growth on ECP Cu. It is well known that steps, wrinkles, defects, and grain boundaries in Cu foils have much lower nucleation barriers than flat surface. These could act as nucleation sites, triggering the formation of graphene nuclei and eventually give rise to small single-crystal graphene domains. On the other hand, ECP process reduces the surface roughness as well as defects and impurities that subsist on Cu foils, leading to the suppression of nucleation sites and hence reduced domain density of graphene. In addition, during the early stage of two-step growth, the low CH 4 concentration also mitigates the graphene nucleation and provides space for the growth of large graphene domains. As a result, two-step growth of graphene on ECP Cu foil gives large-size single crystalline graphene domains with a reduced domain density. Figure 2(a-c) show the SEM images of the three different graphene films on Cu foils. One can notice that all the three graphene films are continuous, uniform and clean without noticeable particles, but grain boundaries, Cu surface steps, and wrinkles are observed, which could influence the gas barrier property of graphene. This will be addressed further later with water vapor transmission rate (WVTR) results.
To investigate the quality and number of graphene layers, Raman measurements were performed on representative graphene films after being transferred on to SiO 2 /Si substrate. Figure 2(d) compares the Raman spectra of the three different graphene films. Two predominant peaks i.e., the G-band and the 2D-band associated with graphene are observed from all the samples at 1589 cm −1 and 2690 cm −1 , respectively. However, the G-band position is found to be blue shifted compared to the value 1582 cm −1 reported in the literature 32 . This is due to unintentional charge transfer induced p-type doping of graphene caused by the redox reaction between oxygen from air and adsorbed water at the surface of SiO 2 /Si substrate having hydrophilic property 33 . Raman spectra obtained from all samples exhibit a feature of monolayer graphene, i.e. a single Lorentzian peak with a full width at half maximum (FWHM) of 30 cm −1 and the intensity ratio of 2D to G greater than 2. Also, the so-called defects or disorder-induced D-band peak is seen around 1350 cm −1 at about half of the frequency of the 2D band with negligible intensity. Figure 3(a,b) show the opto-electrical properties of the three graphene films under study. In general, monolayer graphene absorbs nearly 2.3% of the incident light at the wavelengths over 500 nm 34 . As shown in Fig. 3(a), the transmittance of all graphene samples transferred on to glass substrate is about 97.4% at the wavelength of 550 nm, which deviates slightly from the expected value of 97.7%. However, this is a general trend since monolayer graphene grown on Cu foils by CVD contains some portion of bilayer, depending on growth conditions. Even though the transmittance of 2step-ECP graphene films shows no difference compared with that of 1step-w/o ECP or the 1step-ECP graphene films, sheet resistance change reflects graphene quality rather than thickness. In Fig. 3(b), the sheet resistance of 2step-ECP graphene films is estimated to be 388 ± 20 Ω/□ , a value lower than the values of 726 ± 30 Ω/□ and 627 ± 20 Ω/□ measured for the 1step-w/o ECP and the 1step-ECP graphene films, respectively. The observed low sheet resistance can be understood as follows: Line defects and grain boundaries formed at the interfaces between two domains with different crystallographic orientations of polycrystalline Cu foil can scatter charge carriers and disturb their transport toward lateral direction. This suppresses the ballistic transport path length and leads to high sheet resistance for the graphene films with a higher domain density. By increasing the domain size, the influence of grain boundaries can be reduced because the probability for carriers encountering the grain boundaries decreases, which as a consequence could reduce the sheet resistance of graphene. Since the ECP-2step graphene had relatively large domain size with reduced domain density, its sheet resistance decreased remarkably. Previous studies also demonstrated similar enhancement in the electrical conductivity of graphene as a result of Cu pretreatment and growth conditions 30,35,36 and the lowest sheet resistance reported to date is around 210 Ω/□ for a CVD graphene on electroplolished Cu 30 , while the reported best value is still around 125 Ω/□ for a 30 inch graphene produced by CVD 35 . The relatively higher sheet resistance of our graphene could be likely due to difference in growth conditions, the transfer, and measurement conditions. Figure 4(a,b) show WVTR results and water permeability with time for three different graphene samples. All the three samples reached a steady-state of WVTR within 24 hours. Corresponding data for pure poly ethylene terephthalate (PET) is also shown for comparison. As can be seen in Fig. 4(a) that the formation of graphene on PET leads to a decrease WVTR compared to bare PET, indicating that graphene acts as a protecting layer for PET against water owing to its impermeable property. In particular, 2step-ECP graphene on PET exhibits the lowest WVTR value of 0.665 ± 0.046 g/m 2 -day, which is 57.9% less than that of the value of 1.58 ± 0.029 g/m 2 -day measured for the bare PET. Figure 4(b) depicts the WVTR fitting results through the Fick's second law of diffusion 37 . The water vapor permeability (P) values for the 1step-w/o ECP, 1step-ECP, and 2step-ECP graphene films on PET are found to be 6.608 × 10 −14 , 5.806 × 10 −14 and 4.224 × 10 −14 mols −1 m −1 atm −1 , respectively. It is evident that the lowest WVTR and P values are obtained for monolayer graphene formed by two-step growth on ECP Cu foil without any additional treatment or control of graphene number.
Though Raman analysis can well assess the defect distribution and quality of graphene through the G-to-D band intensity ratio, it is hard to measure the spatial distribution of defects that spread across over large area. To give a further insight into the structural quality of graphene on large scale, a film-induced frustrated etching  (FIFE) is carried out, which is a simple and facile method to observe the structural defects spread on large area of graphene 38 . This method allows us to easily analyze the differences in the sample quality. To perform this study, a commercial copper etchant, ammonium persulfate [(NH 4 ) 2 S 2 O 8 ], is dropped onto graphene synthesized on Cu foils. After waiting for about 5 sec, the dropped etchant is removed through rinsing process with deionized water and subsequently the surface is examined by atomic force microscope (AFM), as shown in Fig. 5(a-c). It is expected that etching of copper occurs in regions that are not covered by graphene, which include partially grown areas, lattice or line defects, and particles on the Cu surface. The etch pit densities for 1step-w/o ECP, 1step-ECP, and 2step-ECP graphene are found to be 4.4 × 10 −7 ± 3.3 × 10 −6 /cm 2 , 2.3 × 10 −7 ± 2.8 × 10 −6 /cm 2 , and 2.2 × 10 −6 ± 6.0 × 10 −5 /cm 2 , respectively. It is evident that the graphene grown by two-step method on ECP Cu has the lowest etch pit density, signifying that the two-step growth method reduces the defect density to a great extent compared to conventional one-step growth. Figure 5(d,e) show the correlation between ratio of graphene/ PET WVTR (W) to bare PET WVTR (W PET ) and etch pit density and etched to the total area ratio (Δ) obtained by a commercial image processor (ImageJ 1.42q, NIH) from corresponding AFM images. The Etch pit density shows a linear decrease from the first sample (1step-w/o ECP) to the third one (2step-ECP), which is in accordance with the SEM results shown in Fig. S5. Moreover, Δ of graphene obtained by one-step growth is found to decrease when the Cu foil is subjected to ECP. A further considerable decrease in Δ is obvious for graphene grown by two-step method on ECP Cu foil, as shown in Fig. 5(e). This result signifies that the enhanced barrier property of graphene stems largely from the two-step growth process rather than electrochemical polishing. It should be noted that Δ curve is similar to that of WVTR results, indicating that barrier performance of graphene actually corresponds to the etched area than etched defect density.

Discussion
The solution of the Fick's second law of diffusion is used to fit the measured WVTR, by assuming the water vapor concentration is close to zero at the exit side of the film as given below 37 . where P 0 = Jd, P = J s d = SD, J and J s are the water vapor molar flux at time t and at steady state, d is the sample thickness, P is permeability, D is diffusivity, and S is solubility. If the whole area of PET film is fully covered by impermeable graphene layer, the permeability should be zero. However, the CVD graphene layer consists of defect-holes through which water molecules may diffuse. The water molecules in the PET film may diffuse out through the defect-holes. The surface defect-hole density, ψ , can be defined as the ratio between the defect-hole area and the total area. Inside of the defect-holes, the inhibition of water diffusion by graphene layer may be negligible, when the sizes of the holes are much larger than those of the water molecules and the thickness of hole is much smaller than the mean free path of water molecules. In this case, the permeability can be simply expressed as because graphene without defects can be assumed as an impermeable membrane in the simple model where P eff , P G and P° represent the permeability of graphene layer/PET film, bare graphene film, and pure PET film, respectively. However, P G is not zero but a tiny value due to undetected defects, and thus the equation can be approximated in the form.
where B is a correction constant arising from undetected defects by the FIFE. It is reasonable to assume that the Δ is proportional to ψ such as ψ = AΔ. Then the reduced permeability can be related to Δ as follows; In conclusion, a combination of electrochemical polishing and two-step growth is demonstrated to obtain high-quality monolayer graphene by low pressure CVD. This method allows achieving large size graphene domains by controlling the nucleation density of graphene on Cu foils, a crucial starting step for the realization of high-quality graphene. The characteristics of the graphene were compared to that of graphene produced by one-step growth on both unpolished and electrochemically polished Cu foils. Two-step growth on electrochemically polished Cu also led to improved permeability property. This is attributed to the decrease in structural defects and suppression of nucleation of graphene at the initial stage. A simple and accurate model has been proposed to examine the barrier property of CVD monolayer graphene. Our study also brings into picture that the effective permeability of CVD monolayer graphene is proportional to defect area rather than defect density.

Methods
Electro-chemical polishing (ECP). ECP  Graphene synthesis. Large-scale graphene layer studied in this work was synthesized on 35-μm-thick Cu foils (Nippon Mining) by low pressure chemical vapor deposition (LPCVD). The Cu foil was then set into a 2 inch quartz tube and heated by split-tube furnace. Meanwhile, quartz tube was pumped down to 0.072 Torr, and H 2 gas was flown through the reactor at 15 standard cubic centimeters per minute (sccm) during the temperature ramp-up up to 1030 °C. Then, the Cu foils were annealed for 50 min. In the one-step growth, CH 4 of 13 sccm and H 2 of 15 sccm were supplied at 1030 °C for 23 min under a growth pressure of 0.160 Torr. In the case of two-step growth, at first, large size graphene domains were formed on Cu foil in a mixed CH 4 /H 2 (5 sccm/100 sccm) ambient for 1 min at 1030 °C under a growth pressure of 0.365 Torr. This is the optimal condition for obtaining uniform and large domains of graphene, as shown in the Figure S3 of supporting information. The growth of continuous graphene was then achieved during the second growth step when the flow rate of CH 4 and growth time were increased to 13 sccm and 8 min, respectively, while maintaining the H 2 flow rate and temperature unchanged. Finally, the samples were cooled down rapidly to room temperature with a cooling rate of 30 °C min −1 at 15 sccm of H 2 flow by moving the furnace out of the substrate region.
Graphene transfer. Polymethyl methacrylate (PMMA) was spin-coated onto graphene surface at 4200 rpm for 50 s to protect and transfer the graphene layer onto other substrates of interest. Prior to graphene transfer, graphene unintentionally formed on the back side of the Cu foil was removed by O 2 plasma etching. Then, the Cu foil with PMMA-covered graphene was dipped in 0.1 M Ammonium Persulfate [(NH 4 ) 2 S 2 O 8 ] solution for 4 h to etch the Cu foil. Thereafter, PMMA/graphene layer was transferred onto various substrates, such as 300 nm SiO 2 /Si substrate, PET film, and glass, to investigate the opto-electrical and gas barrier properties. The PMMA was removed by using acetone and the sample was annealed in an Ar/H 2 gas mixture at 400 °C for 3 h to eradicate remained PMMA residues. Characterization. Field emission scanning electron microscopy (FESEM, NovaSEM 450) was used to evaluate the surface morphology of the graphene samples studied in this work. The surface topography of graphene/ Cu foils after FIFE test was probed by atomic force microscope (AFM, Park NX10) in tapping mode. The Raman spectra were recorded using 514 nm-line of an Ar ion laser as an excitation source. In order to observe the change in electrical conductivity as a function of graphene domain size, the sheet resistance was measured by Van der Pauw method using Hall measurement system (Lakeshore 7500 serises).