Min-Sik Kim‡
a,
Ki-Ju Kim‡a,
Minsu Kima,
Sangbong Leea,
Kyu Hyun Leec,
Hyeongkeun Kimc,
Hyun-Mi Kim*bc and
Ki-Bum Kim*ab
aDepartment of Materials Science and Engineering, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul, 08826, South Korea
bResearch Institute of Advanced Materials, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul, 08826, South Korea. E-mail: hyunmi@snu.ac.kr; kibum@snu.ac.kr
cKorea Electronics Technology Institute, 25 Saenari-ro, Bundang-gu, Seongnam-si, Gyeonggi-do 13509, South Korea
First published on 28th September 2020
The oxidation kinetics of Cu through graphene were evaluated from the surface coverage of Cu oxide (Fox) by varying the oxidation time (tox = 10–360 min) and temperature (Tox = 180–240 °C) under an air environment. Fox, as a function of time, well followed the Johnson–Mehl–Avrami–Kolmogorov equation; thus, the activation energy of Cu oxidation was estimated as 1.5 eV. Transmission electron microscopy studies revealed that Cu2O formed on the top of the graphene at grain boundaries (G-GBs), indicating that Cu2O growth was governed by the out-diffusion of Cu through G-GBs. Further, the effect of Cu oxidation on graphene quality was investigated by measuring the electrical properties of graphene after transferring. The variation of the sheet resistance (Rs) as a function of tox at all Tox was converted into one curve as a function of Fox. Rs of 250 Ω sq−1 was constant, similar to that of as-grown graphene up to Fox = 15%, and then increased with Fox. The Hall measurement revealed that the carrier concentration remained constant in the entire range of Fox, and Rs was solely related to the decrease in the Hall mobility. The variation in Hall mobility was examined according to the graphene percolation probability model, simulating electrical conduction on G-GBs during Cu2O evolution. This model well explains the constant Hall mobility within Fox = 15% and drastic Fox degradation of 15–50% by the concept that the electrical conduction of graphene is disconnected by Cu2O formation along with the G-GBs. Therefore, we systematically developed the oxidation kinetics of Cu through graphene and simultaneously examined the changes in the electrical properties of graphene.
Raman spectroscopy is among the most efficient characterization methods of graphene. The D peak (which appears at ∼1350 cm−1 on the Raman spectrum) represents the breakage of the sp2 bond by structural defects in the graphene lattice.14 Raman spectroscopy provides various quantitative information of graphene, such as the number of graphene layers (from the intensity ratio of the G and D peaks) and doping amount (from the shift of the G or 2D peak).15–17 However, Raman spectroscopy cannot provide information over the entire sample area because the laser beam size is limited to a few μm.2 Furthermore, the graphene must be transferred to an SiO2 substrate for accurate analysis. Transmission electron microscopy (TEM) is clearly advantageous because its atomic resolution enables direct observation of the graphene lattice.18 However, sample preparation is quite difficult, and the observed area is extremely small. The measurement of the electrical properties of graphene requires the transfer of graphene on the insulating substrate for fabricating a device. Because various defects (e.g., wrinkles, cracks, holes, and organic residues) are introduced during the transfer process,19 the measured property cannot be correlated with the quality of the as-grown graphene. Therefore, a method for the analysis of the as-grown graphene on Cu is required for proper tuning of the graphene-growth condition.
In this respect, the oxidation of Cu underlying graphene has been proposed by several researchers as a useful method to simply observe the graphene grain.20–26 Because the hexagonal arrangement of carbon atoms in graphene is sufficiently compact to inhibit the penetration of oxygen (and consequently the oxidation of Cu),27 Cu is selectively oxidized by oxidizing species, such as OH− or O−, which penetrate the G-GBs.22–24 These radicals are typically dissociated from H2O at elevated temperatures (approximately 200 °C) or under UV exposure.23,24 Energy calculations using density functional theory and TEM observations have shown that oxygen radicals penetrate the G-GBs without bond breaking. Oxygens penetrating the boundary at room temperature was reported to form Cu oxide at the graphene–Cu interface.25 Moreover, the lateral diffusion of oxygen at the graphene–Cu interface is hindered by van der Waals forces between the Cu and graphene.26 Although the results of oxidation kinetics depend on the states of the graphene (e.g., grain size, shape, and stitching between grains) and the Cu foil (e.g., crystallographic orientation),28 the oxidation of Cu through the grain boundary of the graphene layer has not been systematically investigated. Furthermore, the extent to which Cu oxidation damages the electrical property of graphene is unknown.
Here, Cu oxidation through graphene was investigated as a function of temperature (180–240 °C) and time (10–360 min). The oxidation kinetics of Cu through the G-GBs was determined based on the Cu oxide coverage that was quantitatively estimated from the optical microscope (OM) images. To investigate the effect of Cu oxidation on graphene quality, all graphene samples were transferred on a polyethylene terephthalate (PET) film and then the electrical properties (i.e., sheet resistance, Hall mobility, and charge carrier density) of these samples were measured as a function of oxidation time and temperature by Hall measurement. We correlated the extent of Cu oxidation and electrical properties of graphene; these results enable the prediction of electrical properties through graphene visualization by Cu oxidation.
Fig. 1 Optical microscope images of the as-grown graphene after the oxidation at (a)–(d) 180 °C, (e)–(h) 200 °C, (i)–(l) 220 °C, and (m)–(p) 240 °C. |
The dark area in the sample oxidized at 200 °C for 120 min was analyzed using cross-sectional TEM (Fig. 2). The cross-sectional TEM samples were prepared perpendicular to the dark lines. Bright-field scanning TEM (STEM) revealed a radially grown morphology on the flat Cu surface (Fig. 2(a)). From the indexing of the selected area diffraction pattern (Fig. 2(c)) and composition analysis of energy dispersive spectroscopy (EDS, Fig. 2(e)), the species was determined as Cu2O (space group: Pnm, a = 4.288 Å). Judging from the low contrast in the high-angle annular dark field (HAADF) STEM images (Fig. 2(b)), the Cu2O had a lower mass density than the Cu substrate. Therefore, the interface between Cu2O and Cu was very clearly defined. The high resolution TEM (HRTEM) images in this area clarified the graphene layer at the Cu–Cu2O interface with an opening at the center (marked by arrows in Fig. 2(d)). The graphene layer was also found under other Cu2O particles (see ESI S-III†). The semi-circular shape of the Cu2O and graphene opening at the center of the interface indicate that Cu2O is formed from a point source of Cu. Furthermore, graphene existence between Cu2O and Cu evidenced that the Cu2O grew by out-diffusion of Cu.30 Indeed, Cu oxidation predominantly occur by the diffusion of Cu cations through Cu oxide.31,32 However, these results contradict a few previous reports, which claim that Cu2O nucleates and grows by oxygen in-diffusion through graphene at the graphene–Cu interface.23,24 According to these previous reports, oxygen atoms dissociate from H2O at the graphene vacancies and GBs then penetrate the G-GBs and form Cu2O at the graphene–Cu interface.24 The randomly oriented and shaped graphene grains are stitched together, forming abundant pentagon, heptagon, and other non-hexagonal carbon rings at the G-GBs.33 G-GBs with imperfect stitching and voids would crack during the CVD process. We believe that Cu2O nucleation was promoted at the non-hexagonal carbon rings or in the imperfect-stitching region of G-GBs but was suppressed at the relatively stable grain boundary. For instance, at 180 °C after 240 min of oxidation, the line width of the grown Cu2O reached 10 μm (Fig. 1(c)), and these locations are considered as the non-hexagonal carbon rings or imperfect-stitching; however, at certain G-GBs where relatively stable G-GBs, still no Cu2O was formed. We surmise that the initial nucleation process of Cu2O was generated by oxygen in-diffusion through the G-GBs as previously reported but that Cu2O grew by the dominant out-diffusion of Cu cations at the G-GBs. Fig. 2(f) shows the schematic sequence of Cu oxidation through graphene based on the results of the current study. During the oxidation, Cu oxide was initially formed at G-GB and further growth was proceeded by out-diffusion of Cu through the Cu oxide. Otherwise, if Cu oxide growth was proceeded by in-diffusion of oxygen to G-GBs, not by out-diffusion of Cu cation, Cu oxide would be encapsulated by the graphene layer. Consequently, the graphene layer undergoes tensile stress generated by 1.7 times of volume expansion during Cu oxide formation, and the estimated strain of graphene is about 80% based on Fig. 3(a). Considering fracture occurs on PMMA and PDMS supported graphene only about 8% of strain during the tensile test,34,35 it seems impossible to survive graphene without fracture during Cu oxide formation. However, no significant Rs change is observed even if almost G-GBs have been decorated by Cu2O (Fig. 1(c), (g) and (i)). Moreover, few samples of graphene seem to be isolated by grown Cu2O, but the measured Rs shows that graphene samples are still electrically connected (Fig. 1(d), (h), (k), and (m)). Therefore, Cu oxide growth proceeds without further graphene damage caused by the volume expansion during Cu2O formation because of oxidation proceeds by out-diffusion of Cu cation through formed Cu2O.
Fig. 3 (a) Cu oxide coverage as a function of oxidation time and (b) oxidation reaction constant as a function of inverse oxidation temperature. |
The surface coverage of Cu2O (Fox) was assessed based on the OM images in Fig. 1. Here, the intensity of the OM image was normalized and the fraction of dark regions was estimated from the intensity histogram (for more details refer to ESI S-VI†). Fig. 3(a) shows the Fox as a function of tox and Tox. It is worth noting that the change in the Fox as a function of tox agrees well with the typical Johnson–Mehl–Avrami–Kolmogorov (JMAK) equation [eqn (1)] for predicting phase transformation kinetics based on nucleation and growth.36
F = 1 − exp(−ktn), | (1) |
Furthermore, the k value related to the transformation rate was also extracted based on the JMAK equation in Fig. 3(a) and is plotted as a function of 1/Tox in Fig. 3(b). The slope of the curve fitted by the Arrhenius equation was 1.54 eV, which reflects the activation energy of Cu2O growth through graphene, representing the lateral growth rate of Cu2O. Therefore, the growth of Cu2O through graphene is controlled by Cu out-diffusion through Cu2O because this activation energy is similar to that of Cu diffusion in Cu oxide estimated in the classical Cu oxidation experiment.37 Consequently, it is possible to predict the kinetics of Cu oxidation through graphene at a certain targeting temperature through the JMAK equation using a k value obtained by extrapolation.
Because the electrical properties depend on both Tox and tox, it is difficult to quantitatively determine the effect of Cu oxidation on the electrical properties of graphene. Therefore, by expressing Tox and tox in terms of Fox, the electrical properties of graphene were plotted as a function of Fox (Fig. 5), in which the dashed lines represent the values of the as-grown graphene. As mentioned, ns was independent of Tox and tox and remained constant until Fox = 85% (Fig. 5(b)). The Rs and μ had different curves for each Tox, but these curves were merged into one curve by plotting it as a function of Fox. It clearly indicates that electrical properties of graphene are strongly related to Fox. From Rs = 1/nsqμ, the change in Rs as a function of Fox is fully described by the change of μ, inversely corresponding to the change of μ, because of the constant ns. The change in μ as a Fox is divided into three regions. In the first region (under Fox = 15%), the value of μ stayed at the as-grown value, and then the value of μ decreases linearly with the increase in Fox in the second region (Fox = 15–50%). Finally, the value of μ was only a few tens of cm2 V−1 s−1 over Fox = 50%. Herein, Cu2O is formed on the graphene surface and hence is undamaged by volume expansion (1.7 times) from Cu to Cu2O. However, as a defect site of the G-GB, the initial Cu2O nucleation site is the pathway of Cu cation out-diffusion (see the HRTEM image in Fig. 2(d)). The initial nucleation site is considered as a degraded G-GB point even before the Cu2O growth; thus, it may negligibly affect the electrical properties of graphene up to Fox = 15% of Cu2O growth. However, the continuous out-diffusion of Cu cations through the nucleation sites can expand and propagate the Cu cation diffusion path along with the G-GBs. When Fox exceeds 15%, the propagation of the Cu cation diffusion path becomes severe and μ of graphene is reduced. At Fox values above 50%, most of the graphene grains seem to be electrically disconnected. As electrical measurement inevitably involves charge carrier transfer, the mechanically weak points of the damaged grain boundaries can be further damaged during the transfer process, thereby exaggerating the decreasing tendency of the μ (see ESI S-VI†). From these results, we tried to provide optimal conditions for G-GB visualization through the contour map of Rs as a function of Tox and tox as shown in Fig. 5(d). The optimal condition requires a Fox enough to measure the graphene grain size at the same time without changing the electrical properties of graphene. Therefore, as mentioned above, the maximum value of Fox can be defined as 15%. On the other hand, G-GB is sufficiently revealed for measuring the graphene grain size when the amount of oxidation is at least about Fig. 1(b) and (f), and the Fox at this time is 10%. The boundary for optimal Fox values is shown as dashed lines in Fig. 5(d). The difference in oxidation time indicated by the distance between dashed lines at 200 °C is 50 min, so the process margin is very wide and oxidation time is relatively short which is compared to 180 °C. Therefore, oxidation at 200 °C for 100 min, can be suggested as an optimal condition for G-GB visualization.
To explain the propagation of damage on G-GBs and electrical disconnection in the view point of Hall mobility variation as a function of Fox, the percolation probability was introduced in a simple schematic model of Cu oxidation at G-GBs. The percolation probability is used for conductivity prediction as a fraction of metallic component in composite materials.38 Because there is no mathematical formula for calculating the percolation probability, the percolation probability of graphene is simulated through a simple model, i.e., the graphene is disconnected as Cu oxide randomly nucleates and grows in G-GBs. The graphene grains in the 2D matrix were assumed to have square shapes to simplify the calculation (dashed line in Fig. 6(a)). Based on the morphological change illustrated in Fig. 1, we assumed the following steps. Step-I-nucleation, at the beginning, Cu oxide randomly nucleates at the G-GBs and there is no additional nucleation during the growth step (Fig. 6(b)). Step-II-oxygen impingement, the oxygen impingement is random (Fig. 6(c)). Step-III-Cu oxide growth, Cu oxide grows when the oxygen impinged near the already existing Cu oxide (Fig. 6(d)). After step-I-nucleation, step-II and step-III were repeated to grow Cu oxide. The random nucleation and continuous growth of Cu2O according to the iteration was provided in ESI S-VII† in the form of moving images. The connected graphene and Cu oxide coverage were estimated in each growth step using MATLAB software. The percolation probability, P(Fox), was 1 when the conducting path of graphene exists in both the x and y directions, 0.5 when the conducting path of graphene exists only in one direction (x or y), and 0 when the graphene is isolated by Cu oxidation. Fig. 6(e) and (f) show the graphene percolation probability and Cu oxide coverage, respectively, as a function of growth iteration. Each step function in Fig. 6(e) indicates 20 times repeated percolation probability with different initial positions of Cu oxide nucleus, and its average is shown as a gray solid curve. As shown in Fig. 6(f), the Cu oxide coverage increases with growth iteration, regardless of the initial position of the Cu oxide nucleus, whereas the graphene percolation probability is strongly affected by the initial position of the Cu oxide nucleus. Moreover, to investigate the effect of the amount of initial Cu oxide nucleus that might be related to G-GBs quality, the percolation probability was plotted as a function of the Cu oxide coverage [P(Fox)] by averaging 200 simulation units and varying the initial nucleus density (F_nc) from 20 to 200 ea/grain (as the solid lines in Fig. 6(g)). As oxidation progressed, there was a region where P(Fox) = 1 was maintained, after which P(Fox) decreased rapidly and eventually the conducting path of graphene disappeared [P(Fox) = 0]. P(Fox) = 1 reflects that the presence of the 2-directional conducting path in graphene and no disturbance to the charge transfer in the graphene. As P(Fox) < 1, the conducting path between graphene grains start to disappear, indicating the disturbance of the charge transfer in the graphene. The simulation results revealed that the Fox range at P(Fox) = 1 was determined by the initial nuclei density of Cu oxide at G-GB, which is related to the defect concentration in G-GB. This change in P(Fox) at F_nc = 40 ea per grain is very consistent with the change in the measured Hall mobility as shown in Fig. 6(h). Therefore, the relationship between the oxidized graphene morphology on Cu and its electrical properties was determined and the extent of morphological variation was assessed without affecting the graphene electrical properties. This simple modeling of Cu oxidation through graphene enables the calculation of the percolation probability of spanning graphene as a function of surface coverage of the Cu oxide, which is closely related to the carrier mobility within graphene.
Temperature (°C) | Oxidation time (min) | |||
---|---|---|---|---|
180 | 60 | 120 | 240 | 360 |
200 | 30 | 60 | 120 | 240 |
220 | 15 | 30 | 60 | 120 |
240 | 10 | 20 | 30 | 60 |
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0ra06301k |
‡ These authors contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2020 |