Suppressing Grain Growth on Cu Foil Using Graphene

The effect of graphene coating on the growth of grains on bulk copper film was studied. When methane gas is catalytically decomposed on the surface of copper, and a carbon–copper solid solution is formed at high temperature, precipitated carbon on the copper surface forms graphene during rapid cooling through strong sp2 covalent bonding. The graphene layer can prevent the growth of grains by suppressing the diffusion of copper atoms on the surface, even after continuous heat treatment at high temperatures. The actual size of the copper grains was analyzed in terms of repetitive high-temperature heat treatment processes, and the grain growth process was simulated by using thermodynamic data, such as surface migration energy and the binding energy between copper and carbon. In general, transition metals can induce graphene growth on surfaces because they easily form carbon solid solutions at high temperatures. It is expected that the process of graphene growth will be able to suppress grain growth in transition metals used at high temperatures and could be applied to materials that are prone to thermal fatigue issues such as creep.


Introduction
Since the performance of a device (or system) using metal as a component is influenced by grain size [1,2], texture (preferred orientation) [3][4][5], and surface roughness [6][7][8], it is necessary to manage these characteristics well, in order to maintain constant device performance.However, because of the nature of metallic materials, these microstructures can be changed by heat from the environment or generated by the device itself (effects on the atomic movement based on Fick's Law) [9][10][11][12].Therefore, a technology that can maintain a constant microstructure for a long time, even at a high temperature, can be said to be very important from the viewpoint of the reliability of the device it has been applied to.
Traditionally, the method used to inhibit grain growth was to prevent the diffusion of atoms by pinning secondary precipitates in grain boundaries [13,14].However, since the types and sizes of the precipitates depend on the sort of metal or alloy, there has been a problem in that separate metal reaction studies have to be carried out for each variety of metal [15,16].
In this study, we propose a method to obtain high-temperature stability, which can be applied to various metals, using carbon precipitate (graphene) and we then test it experimentally by using copper (Cu).In particular, Cu thin films are widely used in the electronics industry because of their high thermal and electrical conductivity [17][18][19].This conductivity is particularly influenced by the size and Coatings 2018, 8, 334 2 of 9 arrangement of grains, so close attention to graining is required [20,21].It is known that the surface behaves similar to a liquid at a temperature close to its melting point of 1084 • C [22,23] at which Cu atoms actively move, and the grains grow in a crystalline orientation (described as the 111 orientation), in the most thermally stable direction [24,25].
When graphene is grown on Cu foil by chemical vapor deposition (CVD), graphene is weakly bound to the Cu foil (0.33 eV per atom, the magnitude of the binding energy between graphene and transition metal atoms varies, depending on how many electrons are filled in the outermost orbitals of the metal atoms) [26,27].Graphene synthesized onto the transition metal surface has been studied mainly in relation to chemical stability-such as corrosion and oxidation protection [28,29]-and has not been studied in terms of grain growth control, although the bulk graphene/Cu composite materials have been studied for mechanical strength [30][31][32].In this study, however, graphene was proposed for suppressing grain growth on metal in consideration of the excellent thermal conductivity and bonding energy associated with metal atoms.The repeated heat treatment of graphene-coated copper samples, synthesized at high temperatures, was investigated.

Sample Preparation
A commercial 25 µm copper foil (item no.13382, 99.8% purity, Alfa Aesar, Ward Hill, MA, USA) was first immersed in 0.1 M ammonium persulfate solution ((NH 4 ) 2 S 2 O 8 , item No. 248614, ACS reagent, 98.0%, Sigma-Aldrich, St. Louis, MO, USA) for 1 min for cleaning and to etch oxide from the surface.Next, graphene growth was induced by loading the pre-etched Cu foil onto a quartz tube reaction chamber.Each Sample (A-C) was subjected to a CVD process three times, with and without methane (CH 4 ) flow.
The CVD process used went as follows: • The pressure in the chamber was pumped down to 5 mTorr using a mechanical pump; • Hydrogen (H 2 ) gas was injected into the chamber at 40 sccm, the temperature of the Cu foil then raised to 1000 This CVD process was repeated up to two more times to examine the thermal effect on grain growth on Cu, as follows: Sample A had CH 4 injection three times, Sample B had CH 4 injection in the first process only, and Sample C did not have CH 4 injection at any time.

Results and Discussion
When the Cu foil reaches 1000 • C, the growth temperature of graphene, the surface of the Cu foil behaves like a liquid (Cu foil has a melting point of 1084 freely to adopt a (111) direction, which is a thermodynamically stable direction resulting in retention of the (111) preferred orientation after rapid cooling.
However, when graphene is formed on the surface of Cu by precipitated carbon, it exhibits a different crystal orientation.The Cu (100) crystal plane, which can be nucleated easily on the graphene surface as it is kinetically favored over the (111) orientation, is formed at the bottom of graphene [25,33].The (100) grains can suppress the bulk grain growth (111) orientation, as shown in Figure 1a.For an explanation of the sampling sequence, see also Figure 1b and Table 1.

Results and Discussion
When the Cu foil reaches 1000 °C, the growth temperature of graphene, the surface of the Cu foil behaves like a liquid (Cu foil has a melting point of 1084 °C).At this temperature, the Cu atoms move freely to adopt a (111) direction, which is a thermodynamically stable direction resulting in retention of the (111) preferred orientation after rapid cooling.
However, when graphene is formed on the surface of Cu by precipitated carbon, it exhibits a different crystal orientation.The Cu (100) crystal plane, which can be nucleated easily on the graphene surface as it is kinetically favored over the (111) orientation, is formed at the bottom of graphene [25,33].The (100) grains can suppress the bulk grain growth (111) orientation, as shown in Figure 1a.For an explanation of the sampling sequence, see also Figure 1b and Table 1.  1.

Injected gas 1 st Cycle 2 nd Cycle 3 rd Cycle
X means that no gas was injected.
Each Cu foil sample exhibited growth of grains.However, the grain growth tendency on Samples A and B, which were covered with graphene, was relatively less than that on Sample C (see Figure 2a showing the grains on Cu, taken by LM).Sample C exhibited a noticeably larger grain size when the CVD cycle was repeated.There was no significant change in grain size between each sample in the first CVD process.It is known that the CVD process for growing graphene made the Cu grain larger than annealing at the same temperature [34], but, in our experiment, there was an annealing time of 30 min before graphene growth.Since the grain growth effect of the 30 min annealing was greater than the effect of 10 min of graphene growth, each sample did not show a large difference in grain size after the first CVD process.
X means that no gas was injected.
Each Cu foil sample exhibited growth of grains.However, the grain growth tendency on Samples A and B, which were covered with graphene, was relatively less than that on Sample C (see Figure 2a showing the grains on Cu, taken by LM).Sample C exhibited a noticeably larger grain size when the CVD cycle was repeated.There was no significant change in grain size between each sample in the first CVD process.It is known that the CVD process for growing graphene made the Cu grain larger than annealing at the same temperature [34], but, in our experiment, there was an annealing time of 30 min before graphene growth.Since the grain growth effect of the 30 min annealing was greater than the effect of 10 min of graphene growth, each sample did not show a large difference in grain size after the first CVD process.
These results indicated that the graphene layer on the Cu surface suppressed grain growth on Cu.As mentioned in Figure 1a, the increase in Cu (111) orientation has a direct effect on the grain growth, so it was necessary to confirm the ratio of the (111) peak, using XRD analysis.As shown in Figure 2b and Table 2, repeating the CVD cycle did not gradually change the XRD peak.This is because Cu has a very large grain size distribution, and the penetration depth of the X-ray is deeper than the depth of the surface which is affected by graphene.The surface affected by graphene is atomically shallow, because graphene can only affect Cu atoms that are slightly bonded to it.These results indicated that the graphene layer on the Cu surface suppressed grain growth on Cu.As mentioned in Figure 1a, the increase in Cu (111) orientation has a direct effect on the grain growth, so it was necessary to confirm the ratio of the (111) peak, using XRD analysis.As shown in Figure 2b and Table 2, repeating the CVD cycle did not gradually change the XRD peak.This is because Cu has a very large grain size distribution, and the penetration depth of the X-ray is deeper than the depth of the surface which is affected by graphene.The surface affected by graphene is atomically shallow, because graphene can only affect Cu atoms that are slightly bonded to it.
We calculated the average grain size on each Cu foil.As the images in Figure 3a,b had lots of scratches, the computer program we used could not find the grain exactly.Therefore, we used the circle method from ASML E112 [35] (Incorrect ref order, 36 detected before 35.References should be cited in numerical order.) to distinguish the grain and scratches correctly.Equation (1) was used to calculate the grain size, by drawing a circle with a diameter of 200 µ m on an LM image magnified ×100, as seen in Figure 3a,b.
In Equation ( 1), Fk was the measured circle area on the LM image, z was the number of grains completely within the test circle, and n was the number of grains intercepted by the test circle.By assuming that the grains were circular, the average diameter of the grains was calculated using Equation (2).
Figure 3c shows the average grain area calculated with Equation (2).Because of the large grain size distribution of the Cu foil [36], 20 images per sample were taken and calculated to minimize the errors when selecting specific parts.Noting that the graphene was synthesized after annealing the Cu foil in the first CVD process, it was natural that there was no significant difference in grain size after the first cycle.From the second cycle, the grain size increase on the pure Cu foil with the CVD cycle was greater than that of the Cu foil with graphene.Furthermore, when the CVD cycle was We calculated the average grain size on each Cu foil.As the images in Figure 3a,b had lots of scratches, the computer program we used could not find the grain exactly.Therefore, we used the circle method from ASML E112 [35] (Incorrect ref order, 36 detected before 35.References should be cited in numerical order.) to distinguish the grain and scratches correctly.Equation (1) was used to calculate the grain size, by drawing a circle with a diameter of 200 µm on an LM image magnified ×100, as seen in Figure 3a,b.
In Equation (1), F k was the measured circle area on the LM image, z was the number of grains completely within the test circle, and n was the number of grains intercepted by the test circle.By assuming that the grains were circular, the average diameter of the grains was calculated using Equation (2).
Figure 3c shows the average grain area calculated with Equation ( 2).Because of the large grain size distribution of the Cu foil [36], 20 images per sample were taken and calculated to minimize the errors when selecting specific parts.Noting that the graphene was synthesized after annealing the Cu foil in the first CVD process, it was natural that there was no significant difference in grain size after the first cycle.From the second cycle, the grain size increase on the pure Cu foil with the CVD cycle was greater than that of the Cu foil with graphene.Furthermore, when the CVD cycle was repeated, the increase in the average grain size in the graphene-covered Cu foil became extremely small, and the grain area became constant.repeated, the increase in the average grain size in the graphene-covered Cu foil became extremely small, and the grain area became constant.Figure 4a shows the Raman shift of Samples A1, A2, B1, B2, and C1.In graphene, three peaks corresponding to the positions of approximately 1350, 1600, and 2700 cm −1 appeared, which were called D peak, G peak, and 2D peak, respectively [37,38].In Samples A1, A2, B1, and B2, there was a small D peak which indicated high quality and small defects of the graphene.The number of layers and quality of the graphene could be analyzed with the other two peaks.In all samples except Sample C, the intensity ratio of the two peaks (I2D/IG) was between 2 and 3, which means that a single-layer graphene was synthesized.The intensity ratios of Samples A2 and B2 were not different from those of A1 and B1, meaning that repeated heating did not damage nor affect the number of graphene layers.Sample C, which was not injected with CH4 during the CVD process, did not have any evidence of graphene.The graphene grain images obtained by repeating the CVD cycle once (A1) and three times (A3) were observed using PLM, and can be seen in Figure 4b [39].The colors of A1 and A3 are different because of the different thicknesses of the liquid crystal.Sample C1 does not show any difference at various angles because the liquid crystal could not be aligned without graphene.Figure 4a shows the Raman shift of Samples A1, A2, B1, B2, and C1.In graphene, three peaks corresponding to the positions of approximately 1350, 1600, and 2700 cm −1 appeared, which were called D peak, G peak, and 2D peak, respectively [37,38].In Samples A1, A2, B1, and B2, there was a small D peak which indicated high quality and small defects of the graphene.The number of layers and quality of the graphene could be analyzed with the other two peaks.In all samples except Sample C, the intensity ratio of the two peaks (I 2D /I G ) was between 2 and 3, which means that a single-layer graphene was synthesized.The intensity ratios of Samples A2 and B2 were not different from those of A1 and B1, meaning that repeated heating did not damage nor affect the number of graphene layers.Sample C, which was not injected with CH 4 during the CVD process, did not have any evidence of graphene.The graphene grain images obtained by repeating the CVD cycle once (A1) and three times (A3) were observed using PLM, and can be seen in Figure 4b [39].The colors of A1 and A3 are different because of the different thicknesses of the liquid crystal.Sample C1 does not show any difference at various angles because the liquid crystal could not be aligned without graphene.
The grain size of the graphene did not change considerably.However, in Sample A3 in Figure 4b, we can see a rougher surface, which was caused by the movement of the Cu particles on the surface as a result of the repeated CVD cycles.The grain size of the graphene did not change considerably.However, in Sample A3 in Figure 4b, we can see a rougher surface, which was caused by the movement of the Cu particles on the surface as a result of the repeated CVD cycles.
We used the formula in Equation (3) below to calculate the theoretical grain size [40,41].
In Equation (3), D was the grain size at time t, D0 was the initial grain size, n was the grain growth exponent, and k was a temperature-dependent and grain boundary mobility relative constant, which was expressed with Arrhenius equation (Equation 4).
In Equation ( 4), k was the rate constant, T was the absolute temperature, k0 was the preexponential factor, Ea was the activation energy, and R was the universal gas constant.We first determined the value of k0 by substituting the grain size value from Samples C1-C3.The plot of k for 1000/T of pure Cu using this value is shown in Figure 5a.The plot of graphene-covered Cu was drawn by applying the C-Cu binding energy (3182 J/mol) [26,27] and the activation energy of pure Cu.With this plot, the k value for 1000 °C was calculated as 0.05 µ m 2 s −1 .
In Figure 5b, the theoretical grain diameter was calculated by substituting the k value.In the case of graphene-Cu, it behaved like pure Cu at the first CVD cycle (0 to 2400 s) because graphene was synthesized after 40 min of heating (30 min of annealing and 10 min of injecting CH4).Thus, the (D−D0) value was calculated with the activation energy of pure Cu for the first cycle and with the activation energy of graphene-Cu (which contains C-Cu bonding energy) as well as the activation energy of pure Cu for the second and third cycles.After the first cycle, the theoretical value and the experimental value were significantly different for graphene-Cu, while the values were almost in agreement in pure Cu.We used the formula in Equation ( 3) below to calculate the theoretical grain size [40,41].
In Equation (3), D was the grain size at time t, D 0 was the initial grain size, n was the grain growth exponent, and k was a temperature-dependent and grain boundary mobility relative constant, which was expressed with Arrhenius equation (Equation 4).
In Equation ( 4), k was the rate constant, T was the absolute temperature, k 0 was the pre-exponential factor, E a was the activation energy, and R was the universal gas constant.We first determined the value of k 0 by substituting the grain size value from Samples C1-C3.The plot of k for 1000/T of pure Cu using this value is shown in Figure 5a.The plot of graphene-covered Cu was drawn by applying the C-Cu binding energy (3182 J/mol) [26,27] and the activation energy of pure Cu.With this plot, the k value for 1000 • C was calculated as 0.05 µm 2 s −1 .
In Figure 5b, the theoretical grain diameter was calculated by substituting the k value.In the case of graphene-Cu, it behaved like pure Cu at the first CVD cycle (0 to 2400 s) because graphene was synthesized after 40 min of heating (30 min of annealing and 10 min of injecting CH 4 ).Thus, the (D−D 0 ) value was calculated with the activation energy of pure Cu for the first cycle and with the activation energy of graphene-Cu (which contains C-Cu bonding energy) as well as the activation energy of pure Cu for the second and third cycles.After the first cycle, the theoretical value and the experimental value were significantly different for graphene-Cu, while the values were almost in agreement in pure Cu.
The reasons for this can be seen in Figure 5c.First, since graphene has high thermal conductivity [42,43], it takes away the heat from the Cu foil.Therefore, the Cu foil cannot receive the full thermal energy equivalent of 1000 • C; therefore, the corresponding grain growth does not occur.The second reason is that the C atoms are dissolved in the Cu foil at high temperatures.The remaining C atoms in the Cu foil suppress the potential shift of Cu dislocations and consequently interfere with grain growth [44,45].

Figure 1 .
Figure 1.(a) Changes in Cu grains at the surface during the chemical vapor deposition (CVD) process; (b) CVD process conditions.At the time indicated in blue, CH4 was injected as indicated in Table1.

Figure 1 .
Figure 1.(a) Changes in Cu grains at the surface during the chemical vapor deposition (CVD) process; (b) CVD process conditions.At the time indicated in blue, CH 4 was injected as indicated in Table1.

Figure 2 .
Figure 2. (a) Light microscopy (LM) images of Samples A-C, after each CVD cycle.The red shape on each LM image is the shape of the grain (b) X-ray diffraction (XRD) intensity ratios In/I1 (n is the number of CVD cycle) for the (111) peak, for Samples A-C.

Figure 2 .
Figure 2. (a) Light microscopy (LM) images of Samples A-C, after each CVD cycle.The red shape on each LM image is the shape of the grain (b) X-ray diffraction (XRD) intensity ratios I n /I 1 (n is the number of CVD cycle) for the (111) peak, for Samples A-C.

Figure 3 .
Figure 3. LM images of (a) Sample A, and (b) Sample C. A blue dot represents a grain in the red circle, and a green dot represents a grain that spans the red circle.(c) Calculated grain area after each CVD cycle.

Figure 3 .
Figure 3. LM images of (a) Sample A, and (b) Sample C. A blue dot represents a grain in the red circle, and a green dot represents a grain that spans the red circle.(c) Calculated grain area after each CVD cycle.

Figure 5 .
Figure 5. Pure Cu foil and graphene-covered Cu foil.(a) Plot of lnk against the annealing temperature (1000/T), (b) (D−D 0 ) against time, (c) schematic representation of the reasons of the gap between experimental and theoretical values for the graphene-covered Cu in (b).
• C).At this temperature, the Cu atoms move

Table 2 .
Intensity ratio and full width at half maximum (FWHM) related to Figure2b.

Table 2 .
Intensity ratio and full width at half maximum (FWHM) related to Figure2b.