Graphene growth mechanisms and kinetics

Generally, graphene growth on metals by CVD can be divided into the following steps (see Figure 2 ): (1) Carbon precursors are transported to the metal surface; (2) the precursor is adsorbed and catalytically decomposed into active C species; (3) the active C species diffuse on the surface or into the bulk; (4) C aggregates on the surface for nucleation; (5) C attaches graphene nuclei for growth; (6) catalytic decomposition of C precursors terminates once the substrate is fully covered with graphene; and (7) more graphene layers form by C precipitation during the cooling process due to supersaturation of C in the metal.

Figure 2.

Schematic of graphene growth kinetics. (1) Gas phase transportation, (2) surface adsorption and decomposition of gas molecules, (3) diffusion on the surface and in the bulk of decomposed active species, (4) graphene nucleation, (5) graphene growth, (6) termination of catalytic decomposition of gas molecules, and (7) graphite precipitation.

The main issues with graphene growth are the control of the number of layers (and stacking order) and crystallinity. When C solubility in a metal is low (such as in Cu), there is almost no C precipitation (step 7) and graphene growth is surface mediated, preferentially growing monolayer graphene. Monolayer graphene growth also benefits from a self-limited growth mechanism (i.e., once the metal surface is fully covered by graphene, it loses its catalytic activity for C precursor decomposition and thus limits more graphene layer growth). ^{Reference Li, Cai, Colombo and Ruoff13} On the other hand, if C solubility in a metal is high (such as in Ni), graphene growth follows a C dissolution-segregation-precipitation (DSP) process (C first dissolves into the metal. When the whole metal, or the region near the surface, is saturated, a monolayer graphene forms on the metal surface by surface segregation. Then, more graphite layers precipitate out due to the decrease of C solubility with temperature decrease), ^{Reference Eizenberg and Blakely14} which can be used to grow multilayer graphene. ^{Reference Li, Cai, Colombo and Ruoff13} However, both monolayer and multilayer graphene can be grown by either surface mediation or DSP processes with appropriate kinetic control.

Control of layer number and stacking order

Few-layer graphene with a specific stacking order (e.g., AB-stacking) has unique electronic and optoelectronic properties as compared to monolayer graphene. ^{Reference McCann and Koshino15} Adlayer flakes or multilayer graphene can grow on a metal with a low C solubility by surface mediation via three possible mechanisms, schematically shown in Figure 3 —synchronous growth, ^{Reference Sun, Raji, Zhu, Xiang, Yan, Kittrel, Samuel and Tour16,Reference Liu, Gong, Wilt, Sakidja and Wu17} on-top growth, ^{Reference Yan, Peng, Zhou, Li and Liu18–Reference Ta, Perello, Duong, Han, Gorantla, Nguyen, Bachmatiuk, Rotkin, Lee and Rummeli21} and underneath growth. ^{Reference Ta, Perello, Duong, Han, Gorantla, Nguyen, Bachmatiuk, Rotkin, Lee and Rummeli21–Reference Wu, Zhai, Li and Yang30} In synchronous growth, all layers nucleate and grow simultaneously. In the case of on-top growth, graphene grows layer by layer on top of previous layers or multiple layers grow from the same nucleation site, although the upper ones grow more slowly. For underneath growth, the lower layers grow more slowly. In this case, C for the growth of the bottom layers is mainly supplied by catalytic decomposition of the C precursor on the catalyst surface that is not covered with graphene (see Figure 3d). for new layer growth the decomposed C species, mainly C atoms, diffuse across the graphene edges ^{Reference Nie, Walter, Bartelt, Starodub, Bostwick, Rotenberg and McCarty22–Reference Li, Chou, Zhong, Liu, Dolocan, Zhang, Zhou, Ruoff, Chen and Cai24} or through the substrate into the interface between existing graphene layers and the underneath substrate to occur. ^{Reference Fang, Hsu, Song, Birdwell, Amani, Dubey, Dresselhaus, Palacios and Kong26–Reference Hao, Wang, Liu, Chen, Wang, Tan, Nie, Suk, Jiang, Liang, Xiao, Ye, Dean, Yakobson, McCarty, Kim, Hone, Colombo and Ruoff28} It has also been theoretically proposed that C may penetrate graphene layers, but to the best of our knowledge, there has been no experimental proof of this phenomenon. ^{Reference Wu, Zhai, Li and Yang30}

Figure 3.

Schematics of multilayer graphene growth models. (a) Synchronous growth—all layers nucleate and grow simultaneously. (b) On-top growth—graphene grows layer by layer on top of previous layers or multiple layers grow from the same nucleation site, although the upper ones grow more slowly, and (c) underneath growth—the lower layers grow more slowly. The upper diagrams are perspective views and the lower ones are cross-sectional views. The numbers in (b) and (c) represent the growth sequence of graphene layers. (d) Schematics of possible paths for C reaching the bottom layers.

Both hydrogen and oxygen play important roles in multilayer graphene growth. Hydrogen may passivate a graphene edge to prevent its bonding to the metal surface so as to lower the barrier for C atoms diffusing across it. ^{Reference Zhang, Wang, Xin, Yakobson and Ding29} Oxygen is also crucial in helping hydrocarbon precursors to completely dehydrogenate in order to provide C atoms. ^{Reference Hao, Wang, Liu, Chen, Wang, Tan, Nie, Suk, Jiang, Liang, Xiao, Ye, Dean, Yakobson, McCarty, Kim, Hone, Colombo and Ruoff28} A high purity Cu substrate with a smooth surface and C precursors with low partial pressure are preferred for monolayer growth with almost no adlayers. Otherwise, multilayer growth may result. ^{Reference Liu, Zhou, Cheng, Chen, Lin, Qu, Bai, Ivanov, Liu, Huang and Duan31–Reference Lee and Ye33}

It should be noted that in contrast to the unambiguous identification of the underneath growth sequences, there is still lack of direct experimental proof for the synchronous growth and on-top growth models, and these need to be further investigated.

On the other hand, the number of layers of graphene grown by the DSP process can be controlled kinetically. For example, C solubility in the substrate can be tuned by temperature or metal alloy composition; ^{Reference Wu, Chou, Ji, Wu, Chen, Jiang, Hao, Kang, Ren, Piner and Ruoff34,Reference Addou, Dahal, Sutter and Batzill35} the amount of dissolved C can be controlled by the thickness of the substrate; ^{Reference Kim, Zhao, Jang, Lee, Kim, Kim, Ahn, Kim, Choi and Hong8} and C precipitation can also be controlled by its concentration gradient from the surface to the inner bulk and the cooling conditions. ^{Reference Yu, Lian, Siriponglert, Li, Chen and Pei6,Reference Weatherup, Dlubak and Hofmann36,Reference Xiao, Wu, Lv, Wu and Qian37}

Controlling the stacking order of multilayer graphene is crucial as well. However, current understanding of their formation is superficial, and results in the literature are contradictory. Ta et al. suggested that graphene flakes growing on top of the first full layer are non-AB-stacked, while those grown by the underneath growth mechanism are highly oriented AB-stacked. ^{Reference Ta, Perello, Duong, Han, Gorantla, Nguyen, Bachmatiuk, Rotkin, Lee and Rummeli21} On the other hand, most of the other work reported a high degree of AB-stacked order comes from the suggested on-top growth mode, ^{Reference Yan, Peng, Zhou, Li and Liu18–Reference Han, Lee and Yeo20} while bilayer graphene through the underneath growth model shows non-AB-stacked order. ^{Reference Nie, Wu, Xing, Yu, Bao, Pei and McCarty23,Reference Fang, Hsu, Caudillo, Song, Birdwell, Zakar, Kalbac, Dubey, Palacios, Dresselhaus, Araujo and Kong25}

Engineering of large-area graphene single crystals

Normally, graphene nuclei on metal surfaces are not aligned. These misaligned nuclei grow and finally merge together, resulting in a polycrystalline graphene film with grain boundaries containing pentagonal and heptagonal carbon rings. ^{Reference Li, Cai, Colombo and Ruoff13,Reference Huang, Ruiz-Vargas, van der Zande, Whitney, Levendorf, Kevek, Garg, Alden, Hustedt, Zhu, Park, McEuen and Muller38} To improve its crystallinity, one straightforward strategy is to reduce the nucleation density and achieve a large-area single crystal from one nucleus. The nucleation density can be reduced using a Cu substrate with high purity and a smooth surface, a low carbon precursor partial pressure, and a high growth temperature. ^{Reference Li, Magnuson, Venugopal, An, Suk, Han, Borysiak, Cai, Velamakanni, Zhu, Fu, Vogel, Voelkl, Colombo and Ruoff39,Reference Yan, Lin, Peng, Sun, Zhu, Li, Xiang, Samuel, Kittrell and Tour40} Oxygen can also significantly suppress graphene nucleation while increasing graphene growth rate. ^{Reference Hao, Bharathi, Wang, Liu, Chen, Nie, Wang, Chou, Tan, Fallahazad, Ramanarayan, Magnuson, Tutuc, Yakobson, McCarty, Zhang, Kim, Hone, Colombo and Ruoff41–Reference Guo, Jing, Xiao, Zhou, Lin and Wang43} Both centimeter-sized graphene single crystals ( Figure 4a) ^{Reference Hao, Bharathi, Wang, Liu, Chen, Nie, Wang, Chou, Tan, Fallahazad, Ramanarayan, Magnuson, Tutuc, Yakobson, McCarty, Zhang, Kim, Hone, Colombo and Ruoff41} and rapid graphene lateral growth of 60 μm/s ^{Reference Xu, Zhang, Qiu, Zhuang, Zhang, Wang, Liao, Song, Qiao, Gao, Hu, Liao, Liao, Yu, Wang, Ding, Peng and Liu44} have been achieved with the assistance of oxygen. A Cu substrate alloyed with Ni also showed rapid growth. ^{Reference Wu, Zhang, Yuan, Xue, Lu, Liu, Wang, Wang, Ding, Yu, Xie and Jiang45} By tuning C solubility in the alloy, an appropriate amount of C can be dissolved in the near surface region and then quickly attached to the graphene nuclei. This is much faster than C supplied from the gas phase followed by the adsorption-dehydrogenation-diffusion process. ^{Reference Li, Colombo and Ruoff2} By locally feeding the carbon precursor, a ∼1.5-in. (3.8-cm) graphene single crystal was grown on a Cu₈₅Ni₁₅ substrate (Figure 4b). ^{Reference Wu, Zhang, Yuan, Xue, Lu, Liu, Wang, Wang, Ding, Yu, Xie and Jiang45}

Figure 4.

Growth of large-area graphene single crystals from individual nuclei. (a) Centimeter-sized graphene domains (pale area) on Cu (orange area). Reproduced with permission from Reference Hao, Bharathi, Wang, Liu, Chen, Nie, Wang, Chou, Tan, Fallahazad, Ramanarayan, Magnuson, Tutuc, Yakobson, McCarty, Zhang, Kim, Hone, Colombo and RuoffReference 41. © 2013 AAAS. (b) Left: schematic of the local gas feed; right: optical image of the ∼1.5-in. (3.8-cm) graphene single crystal (pale area) grown on Cu₈₅Ni₁₅ at 1100°C in 150 min. Reproduced with permission from Reference Wu, Zhang, Yuan, Xue, Lu, Liu, Wang, Wang, Ding, Yu, Xie and JiangReference 45. © 2015 Nature Publishing Group.

Morphologies of grown graphene crystals depend on both the growth conditions and the substrate crystallographic orientation. ^{Reference Geng, Wang and Yu46} Typically, they are compact hexagons with sharp edges ( Figure 5a) or four- or six-lobed shapes with protruding branches or dendritic edges (Figure 5b), whose growth involves two mechanisms. The sharp-edge hexagons, arising from the kinetic Wulff construction (which gives a shape of the crystal with its minimized surface Gibbs free energy) have zigzag edges and grow by an edge-attachment-limited growth mechanism. That is, graphene growth is limited by the rate of C incorporation into its lattice rather than C supply by diffusion from the surrounding environment, a diffusion (mass transport)-limited process, which results in the protruding branches and the dendritic edges. ^{Reference Hao, Bharathi, Wang, Liu, Chen, Nie, Wang, Chou, Tan, Fallahazad, Ramanarayan, Magnuson, Tutuc, Yakobson, McCarty, Zhang, Kim, Hone, Colombo and Ruoff41,Reference Artyukhov, Liu and Yakobson47,Reference Jacobberger and Arnold48}

Figure 5.

Scanning electron microscope images of graphene single crystals with (a) sharp edges and (b) dendritic edges. Reproduced with permission from Reference Hao, Bharathi, Wang, Liu, Chen, Nie, Wang, Chou, Tan, Fallahazad, Ramanarayan, Magnuson, Tutuc, Yakobson, McCarty, Zhang, Kim, Hone, Colombo and RuoffReference 41. © 2013 AAAS.

Another strategy is to grow aligned graphene nuclei, which then grow and merge seamlessly to form a large single crystal. The advantage of this strategy is the much faster growth rate when many graphene nuclei grow simultaneously versus growth from a single nucleus. For example, because the lattice mismatch between graphene and Cu(111) is small (4%), with appropriate conditions, compact hexagonal graphene nuclei can epitaxially grow on Cu(111) with edge alignment parallel to 〈101〉. ^{Reference Nguyen, Shin, Duong, Kim, Perello, Lim, Yuan, Ding, Jeong, Shin, Lee, Chae, Vu, Lee and Lee49} With the development of a technique to produce a large-area Cu(111) single-crystal foil, a graphene single crystal with size up to 5 cm × 50 cm has been achieved, demonstrating a promising solution for the preparation of large-area graphene single crystals ( Figure 6 ). ^{Reference Xu, Zhang, Dong, Yi, Niu, Wu, Lin, Yin, Li, Zhou, Wang, Sun, Duan, Gao, Jiang, Wu, Peng, Ruoff, Liu, Yu, Wang, Ding and Liu50}

Figure 6.

Synthesis of large-area graphene single crystals by seamless stitching of aligned graphene domains. (a) Schematic of the experimental design for the continuous production of a single-crystal Cu(111) foil with a hot temperature zone at the central area of a furnace tube. The temperature gradient provides the driving force to move Cu grain boundaries from the low-temperature region to the high-temperature region. In practice, the continuous motion of grain boundaries can be realized by continuously moving the substrate. (b) The schematic evolution of the Cu(111) grain growth, which starts from a single nucleus at the tapered end of the foil. (c) Photograph of a 5 cm × 50 cm single-crystal Cu(111) foil. (d) Representative low-energy electron diffraction patterns from six different regions as marked with the number in (c) show aligned orientation. (e) Optical image of graphene domains on the Cu(111) surface. The parallel dotted lines indicate all the graphene domains have aligned orientation, which ensures a seamless stitching of these domains to form a single crystal. Reproduced with permission from Reference Xu, Zhang, Dong, Yi, Niu, Wu, Lin, Yin, Li, Zhou, Wang, Sun, Duan, Gao, Jiang, Wu, Peng, Ruoff, Liu, Yu, Wang, Ding and LiuReference 50. © 2017 Elsevier.

Low-temperature growth

Considering the high growth temperatures (˃1000°C), which are normally required in the previously mentioned conventional route to obtain high-quality graphene, low-temperature growth is desired to reduce energy consumption and the associated cost, improve system safety, and more importantly, integrate with present semiconductor manufacturing technologies, or grow graphene directly on flexible substrates.

Two issues need to be resolved to achieve high-quality graphene growth at low temperatures—an efficient supply of active C species and an appropriate attachment of intermediate species to a graphene crystal to form an ordered structure. The efficiency of the C supply can be improved by using carbon precursors with a low decomposition or a high adsorption energy such that the adsorbed carbon precursors can have enough time to decompose before desorption. ^{Reference Choi, Li, Cui, Fan, Zhang, Zeng and Zhang51} Various carbon precursors other than methane have been explored, such as aromatic hydrocarbons (e.g., benzene and p-terphenyl), ^{Reference Choi, Li, Cui, Fan, Zhang, Zeng and Zhang51–Reference Lee, Lee, Jo, Lee, Lee, Park, Lee, Kim, Bong and Cho55} aliphatic gases with low decomposition energy (e.g., acetylene), ^{Reference Weatherup, Dlubak and Hofmann36,Reference Weatherup, Bayer, Blume, Ducati, Baehtz, Schloegl and Hofmann56} liquid oxygenated carbon (e.g., methanol and ethanol), ^{Reference Guermoune, Chari, Popescu, Sabri, Guillemette, Skulason, Szkopek and Siaj57} and polymers (e.g., poly (methyl methacrylate). ^{Reference Li, Wu, Wang, Fan, Zhang, Zhai, Zeng, Li, Yang and Hou52} Growth temperatures from 100–1000°C have been reported. However, it should be noted that when using liquid or solid carbon precursors, they may adsorb onto the inside wall of the reaction chamber. When running the system, even without deliberately introducing carbon precursors, the desorption of those adsorbates may interfere with subsequent graphene growth, as it has been demonstrated that even a trace amount of carbon may promote graphene growth. ^{Reference Qing, Jia, Li, Liu, Li, Peng, Deng, Zhang, Li, Ruoff and Li58} Besides, some of the studies require a high-temperature annealing process for the substrate before graphene growth at low temperature. ^{Reference Choi, Li, Cui, Fan, Zhang, Zeng and Zhang51–Reference Zhang, Lee, Piner, Kholmanov, Wu, Li, Ji and Ruoff53} The whole process is not a true low-temperature process because a high-temperature process is still involved. In addition, as just mentioned, the system may be contaminated by the liquid or solid carbon precursors, which also probably promote graphene growth during the high-temperature “annealing” process rather than during the expected low-temperature “growth” process. Another strategy is to use a substrate with a high catalytic activity such as Ni, which is more active than Cu in cracking the hydrocarbons. ^{Reference Weatherup, Dlubak and Hofmann36,Reference Weatherup, Bayer, Blume, Ducati, Baehtz, Schloegl and Hofmann56} Energy-assisted CVD techniques such as plasma-enhanced CVD can also help to decompose hydrocarbons. ^{Reference Boyd, Lin, Hsu, Teague, Chen, Lo, Chan, Su, Cheng, Chang, Wu and Yeh59}

To continue perfecting the graphene structure, a mixture of polycyclic aromatic and aliphatic hydrocarbons were used. It was suggested that the aliphatic carbon species could incorporate small carbon fragments into defect sites to decrease defect density. ^{Reference Lee, Lee, Jo, Lee, Lee, Park, Lee, Kim, Bong and Cho55} It was also reported that high-quality graphene can be achieved with a slow growth rate, which allowed the intermediate hydrocarbon species attached to graphene crystals to have sufficient time to reorganize or shed hydrogen atoms. If the growth rate is too fast, these disordered intermediates can be trapped in the crystal, leading to local disruptions of the lattice or coordination defects. ^{Reference Jacobberger, Levesque, Xu, Wu, Choubak, Desjardins, Martel and Arnold60}

Challenges and prospects

Despite the advances in graphene synthesis achieved in the last few years, there are still many challenges. Neither the uniformity nor the stacking order of multilayer graphene can be easily achieved by kinetic control unless one grows a large-area multilayer graphene single crystal, which is limited to the submillimeter scale. ^{Reference Li, Chou, Zhong, Liu, Dolocan, Zhang, Zhou, Ruoff, Chen and Cai24,Reference Hao, Wang, Liu, Chen, Wang, Tan, Nie, Suk, Jiang, Liang, Xiao, Ye, Dean, Yakobson, McCarty, Kim, Hone, Colombo and Ruoff28} Defect control is another challenge. Although tuning the graphene nucleation density, and hence the number of grain boundaries, can solve this problem to some extent, there is still a lack of knowledge on the formation and control of many other defects such as vacancies, a distorted/disordered lattice in the grains, wrinkles, and functional groups. Quality degradation and a cost increase because of a necessary transfer process are also issues, and direct growth on target substrates is always preferred.