Introduction

Carbon allotropes

Carbon materials consist of various allotropes. Carbon allotropes form when carbon atoms form covalent bonds between each other. There are several known carbon allotropes such as diamond, graphite and amorphous carbon. These are the traditional 3D bulk carbon materials. More recently, with the advent of advanced observation techniques, 2D graphene and 1D carbon nanotubes (CNTs) are identified as other, low-dimensional sp2 carbon allotropes.

To understand sp2 carbon, one can start with the electron orbital hybridization [1, 2]. Carbon’s atomic number is six. It has four valence electrons which occupy the 1s2, 2s2, 2p 1x , and 2p 1y atomic orbitals. With all four of the valence electrons contributing to covalent bonds, carbon is tetravalent and forms sp3 carbon, which we more commonly know as diamond. The bonding energy between two sp3 carbon atoms is 3.6 eV and contributes to diamonds’ great strength. Carbon can also form with sp2 hybridization in which three electrons form in-plane σ (covalent) bonds and the fourth electron forms a weak (non-covalent) inter-plane π bond. The π electrons are delocalized and provide graphite (and graphene) with its electrical conductivity. The inter-plane π bond has relatively low bonding energy of 31 meV [3] in comparison to the very strong in-plane σ bond (6.4 eV). This means that graphite is a layered structure formed from graphene layers that can easily slide apart from each other and it is this easy sliding action that allows us to use graphite for graphical purposes (e.g., a graphite pencil) and from which it derives its name.

Indeed, the individual graphite layers can be separated and ultimately yield a single layer, which we term graphene [1, 4]. The three equivalent sp2 σ orbitals are symmetrically angled with 120° in the XY plane (forming a characteristic hexagonal or honeycomb lattice), while non-hybrid 2p-orbitals are perpendicular to the XY plane. One can also take sections of a graphene sheet that can then be rolled up into a ball which we call a fullerene. In this case, some pentagon structures form amid the hexagons to allow for curvature much like a classic leather football. One can also roll up a graphene sheet to form a tube, and we call these structures single-walled carbon nanotubes. When several sheets are rolled up, we obtain multi-walled carbon nanotubes.

Graphene and carbon nanotubes

The properties vary between the various carbon allotropes, and thus each is of interest for a variety of applications. Indeed, so rich are their applications it would be too much to enter into in this review. Here, we focus only on the chemical vapor deposition (CVD) for the fabrication of graphene and CNTs since this growth route is the most promising for their manufacture in the semiconducting industry when considering both CNTs and graphene as important future materials in the electronics industry as well as for mass production. At first, we briefly look at the properties of graphene and CNTs that make them so promising and then we introduce the reader to CVD as a growth route.

Type and properties of graphene

In terms of morphology, graphene can be classified into planar graphene and vertical graphene. Planar graphene lies parallel to a substrate. With increasing layer numbers, planar graphene is usually divided into three types, namely mono-, bi-, and multi-layer graphene. In contrast, vertical graphene is free standing and anchor perpendicular to a substrate.

The enormous interest in graphene has been mostly driven by its unique electronic and electrical properties. Monolayer graphene has three key electronic properties. These have zero carrier density at the Dirac points and are pseudo-spin and relativistic carriers [1, 4]. These electronic properties provide a rich series of experiments in physics, such as modulation of carrier density with tuning gate voltage, quantum hall effect [5, 6], and band gap engineering in graphene ribbons [7, 8]. Thus, graphenes are potentially building blocks for transistors [9]. Bilayer graphene [10], especially with AB-Bernal stacking [1113], has tunable band gap depending on the gate voltage [14, 15]. Few-layer graphene is better suited for application in electrochemical energy systems [1620] as opposed to device use. Vertical (multi-layer) graphene [2123] has a curved and folded morphology in contrast planar graphene. Multi-layer graphene (graphite) is generally used as a catalyst support and as electrodes in energy systems [2432].

With the strong C–C bond in sp2 carbon, graphene yields the highest intrinsic strength of known materials [33, 34]. The graphene nano-platelets can blend with polymer for a composite [35] with reinforced mechanical strength and electrical conductivity. The thermal conductivity of graphene is also very high (~2000 W mK−1) and is the highest among all the carbon allotropes [1]. The optical conductance of graphene is independent of frequency over a wide range [36, 37].

An ideal graphene sheet comprises a perfect hexagon network similar to honeycomb. However, the pentagon and heptagon pairs are commonly observed in non-perfect graphene. The typical locations where such pairs are observed are in grain boundaries [38, 39], folded edges [40], and electron-irradiation-induced defects [41, 42]. The pentagon and heptagon pairs degrade the electronic conductance of graphene by introducing electron scattering sites [43, 44]. Moreover pentagon–heptagon pairs will also deform the graphene [45, 46]. Hence the production of single-crystal graphene via CVD approach is highly attractive to avoid the drawbacks of defects.

Moreover, the chemical properties of graphene can be modified by introducing functional groups [4752]. The chemical functionalization can disrupt pristine graphene with covalent bonds or van der Waals interactions of external groups or molecules [53, 54]. The intercalation of molecules between two graphene sheets is a means with which to obtain graphene and few-layer flakes. The large surface area of graphene-based complex materials (with functional particles) is also attractive in applications such as biological sensors [55], humidity sensors [56], antibacterial devices [57], complex anode material for Li-ion batteries [58], and catalyzed H2 generation [59].

Type and properties of CNTs

CNTs are fundamentally distinct as compared to graphene as they have curvature, viz. planar graphene has zero curvature, while carbon nanotube has a diameter (chirality)-dependent curvature [60]. The curvature influences the physical properties of CNTs such as electrical [6163], vibrational [64, 65], optical [66, 67], thermal [68, 69], and mechanical properties [70]. For ideal planar graphene, the electrical properties are dominated by the delocalized electrons in π orbital. For single-wall carbon nanotubes, the electrical properties differ because of the re-hybridization of πσ orbitals due to the curvature effect [7174]. When tube diameter is small (<2 nm), the chirality plays a role in the curvature effect and thus the physical properties [75].

In comparison to graphene, CNTs can be classified by wall numbers into three categories. They are single-walled nanotubes (SWNTs), double-walled nanotubes (DWNTs), and multi-walled nanotubes (MWNTs). SWNTs can be further divided into metallic, semi-metallic, and semiconducting according to their chiral angle. DWNTs have two layers of walls which is convenient for further functionalization [2, 76, 77] while preserving the straightness of the tube (since in the SWNT case the tube wall can collapse upon the introduction of defects/functionalization). MWNTs have various morphologies in which, for example, they can have parallel walls (concentric tubes within each other) [7880], a bamboo-like structure [81, 82], or a herringbone structure [83].

The large scientific and industrial interests in CNTs have been driven by the amazing electronic properties [84]. Depending on their electronic properties, SWNTs can be divided into metallic and semiconducting types [2]. Firstly, the metallic CNTs show ballistic transport behavior, i.e., electrons pass along the tube with little or no scattering. In other words, almost no heat is dissipated inside it, viz. CNT is capable of conducting large current while avoiding an excessive temperature increase. This conductivity behavior is attractive in integrated circuits in that transparent electrodes can be formed with metallic CNT networks [85]. Semiconducting SWNTs transport electrons in a diffusive manner. However, experiments show a high mobility in semiconducting SWNTs. Indeed, the electronic property of SWCNTs depends on their diameters and helicity [86]. This leads to exciting application potential in devices such as diodes, field-effect transistors [87, 88], field emitters, and memories [89]. CNTs can also be mixed into composite materials for energy applications such as electrodes for fuel cells [90], lithium/sodium ion batteries [91, 92], and super capacitors [93, 94] and used as a catalyst for photo-electricity transformation [95, 96] and as a supporting matrix [97].

As previously mentioned, the electrical properties of SWNTs are strongly related to their chirality. One can image SWCNT as a rolled-up graphene nanoribbon. Thus, the SWNT structure can be indexed with an integer pair (n, m), which serves as a rolling vector [98102]. When n is equal to m, the SWNTs have an armchair tube opening. They are metallic CNTs (no band gap) [103, 104]. Armchair SWNTs remain metallic regardless of tube diameter. When (nm)/3 is a non-zero integral, the tube edges are zigzag. They are semi-metals with a small bandgap [99, 105, 106]. Each of their bandgaps is proportional to the inverse of the diameter squared [99, 107109]. When (nm)/3 is any other value (non-integer), the SWNTs have a chiral angle, depending on (n, m). Each of these semiconducting SWNTs has a bandgap that is proportional to the inverse of the tube diameter [110112]. The representative values of the band gap are listed in Table 1.

Table 1 The relation between the type, the bandgap, and SWNT index is listed. The small band gap dependent on the diameter shows a curvature effect for semi-metallic and semiconducting SWNTs, viz. the band gap increases with decreasing tube diameter

Perfect CNTs have hexagon lattice structure. However, like graphene, the introduction of pentagon and heptagon pairs will alter the tube helicity and therefore its electronic structure [99, 117122]. One advantage of the introduction of a pentagon and heptagon pair is the creation of on-tube tuning of tube chirality/helicity. For example, a semiconducting/metallic junction (8, 0)/(7, 1) can form when introducing pentagon–heptagon pairs along the (8, 0) axial direction [122]. A semiconducting heterojunction (10, 0)/(9, 1) can form when inserting a pentagon–heptagon pair on a (10, 0) tube axis [122]. A complicated configuration of pentagon and heptagon pairs can build up a metal–metal junction such as (12, 0)/(6, 6) and (9, 0)/(6, 3) as well [123].

Another attractive property of CNTs is their mechanical properties [124, 125]. The tensile strength of the stiffest CNTs is approximately fifty times larger than that of steel, while their Young’s modulus is five times larger than that of steel. CNT can tolerate 30 % induced strain so that structural enhancement can be achieved when using CNT in polymer composites [124, 126129], metal powders [130], and ceramic matrices [131]. Besides, functionalization of MWNT with silane molecules enables the better solubility (chemical compatibility) for polymer matrix composites [132]. Also, CNT can be used as robust microscope probes [133] and nanoscale tweezers [134]. The thermal conductivity of MWNTs (from arc-evaporation production) exceeds 3000 W mK−1 [135]; however, MWNTs from catalytic CVD methods show a significant drop in thermal conductivity down to 200 W mK−1 [136]. Probably, the presence of catalysts impurities (usually metal particles) degrades the thermal conductivity by creating scattering sites.

The chemistry of CNT is more rich than that of graphene since it has a different chemistry on its inside and outside. This is one advantage of CNT over graphene, in which both sides are symmetric and show no difference. Moreover, one can take the advantage of the nanoscale space inside a SWCNT [137]. The filling of heterogeneous substances into CNTs can also be achieved [138140]. Some fillings hold potential in magnetic material engineering [80, 141]. The inside of a CNT can also be used to adsorb gas molecules, such as hydrogen [142145], nitrogen monoxide [146], nitrogen, and oxygen [147] as well as Xe [148]. The tube interior can work as a nano-channel of fluidics to let water [149] or solutions flow [150] through. The filling of CNT with water can work as a valve to switch on/off the transport of gas molecules [151] and a hydroelectric power generator [152]. With water filling, the phase transformation can be observed inside the tube, for example, ordered water [153] and ice crystals [154]. CNT with encapsulated catalyst particles can function as a nano-reactor for ethanol production [155], Pd-catalyzed hydrogenation [156], Pt-catalyzed fuel cell [157], and inner smaller CNT growth [158]. The encapsulated catalysts exhibit a confinement effect [159] from the tube diameter. The confinement effect can enhance the capacitance of encapsulated MnO2 [160] and the electron spin of encapsulated transition metals [161] such as Gd [162]. The filling of DNA molecule inside a tube enables biological application [163]. Moreover, a single-molecule biological activity (with retinal molecule inside a carbon nanotube) can be imaged dynamically in response to light exposure [164].

The outside of a CNT has the merit of a large surface-to-volume ratio, which provides a reactive surface. The chemistry of CNT exteriors can be tuned with covalent or non-covalent functionalization. Covalent functionalization aims to solubilize the CNTs by attaching hydrophilic species to them [165]. This aggressive process though can disrupt the structure of the tube walls. In contrast, non-covalent functionalization binds molecules gently to CNTs via Van der Waals forces without degrading the tube [166168]. When the biological molecules such as protein and DNA are used, biological functionalization can be obtained [169]. Functionalization of carbon nanotube can work as a drug delivery platform [170]. As above, the functionalization occurs at/over the wall structure. One can also dope CNT by replacing C in the wall with other elements such as B and N [171173], and BN nanotube was individually synthesized when C is totally replaced by both B and N [174176].

Chemical vapor deposition

CVD is a sophisticated synthesis technique for both industry production and laboratory research. CVD is applied in many areas, such as crystal growth, thin-film coating, and fiber and powder production. CVD is also well developed for the synthesis of graphene and carbon nanotubes. The principles of CVD [4] for sp2 carbon include two steps, firstly thermal decomposition of carbon feedstock and thereafter re-assembly of carbon radicals into sp2 carbon nanostructures.

The carbon feedstock varies according to the hydrocarbon selected. Examples include gaseous methane, ethylene, and acetylene, as well as solvents such as methanol, ethanol, isopropanol, and acetone and aromatic hydrocarbons. Usually, catalytic metals are used during CVD to enhance decomposition of the feedstock and perhaps also to aid sp2 carbon assembly.

Catalyst choices are often transition metals, for example, Fe, Co, and Ni for CNT growth and Cu and Ni for graphene CVD growth. However, the non-catalytic substrates are emerging for sp2 carbon growth and also, for example, Si/SiO2 wafers.

In terms of the heating source, there are two types of CVD, conventional thermal CVD and plasma-enhanced CVD. In a thermal CVD system, the heat generated from a resistance heating element is transferred to the target substrate and feedstock gas.

In a plasma-enhanced CVD oven, a plasma source is added to assist the thermal cracking of the carbon feedstock. Plasma CVD has certain benefits, such as that it decreases the growth temperature and thus, for example, minimizes catalyst particle agglomeration which is important for CNT growth. However, in terms of synthetic graphene by plasma CVD it is very desirable for growing vertical graphene but has limited potential for nano-crystalline planar graphene (thus far).

In terms of gas pressure during CVD synthesis, one can work at atmospheric pressure or at low pressure. Generally, low-pressure CVD facilitates rapid growth due to the larger probability of carbon species at the substrate. In contrast, ambient-pressure CVD is economic as it avoids the need for a complicated vacuum system but has lower growth kinetics.

CVD growth of graphene

There are many routes with which to fabricate graphene. These include mechanical exfoliation [177], reduced graphene oxide [178180], (electro-) chemical exfoliation [181, 182], ribbon formation by unzipping CNTs [183, 184], bottom-up synthesis from aromatic molecular precursors [185], epitaxial growth over silicon carbide [186, 187], segregation upon transition metals on cooling [188191], organic molecule adsorption with post hydrogen annealing [192], e-beam irradiation-driven graphene growth [193], and CVD [16, 23, 194196]. When synthesizing graphene for electronic devices, graphene is generally required to form over a large area. Moreover, graphene should ideally be crystalline with negligible defects. Balancing the production cost and the above prerequisites, CVD is the most promising route to produce large-area device-grade graphene. Indeed, graphene synthesis by CVD is one of the most popular techniques because of its potential to be scaled up, and it is already well established both in industry and in research labs.

Substrate pretreatment

Prior to graphene synthesis by CVD, substrate preparation is important. Substrates for graphene growth can be classified into two categories. They are metals and non-metals. When using metals such as Cu, the pretreatments before CVD include general surface cleaning via organic solvent soaking [197], partial Cu surface etching with strong HNO3 [198] or weak acetic acid [199], smoothing the surface with electrical or chemical mechanical polishing [200, 201], passivating the surface with oxygen [199] or argon exposure [202], surface rearrangement and grain size increase with H2 annealing [4], liquefied Cu above its melting point [203, 204], and thorough removal of carbon contamination by air sintering [195, 205].

Without any pretreatment, second-layer flakes form probably beneath a full coverage of monolayer graphene from trapped carbon species [206]. With pretreatments [200, 201, 207], carbon diffusion is suppressed because there is a lack of carbon-trapping sites (e.g., Cu grooves and contamination from the supplier processing).

Oxygen, in particular, plays an important role in fabricating both single-crystal graphene [199, 202] and large-area strict monolayer graphene film [195, 205]. Oxidation treatment is vital to remove surface organic contaminant, as shown in Fig. 1.

Fig. 1
figure 1

Influence of oxidation pretreatment on the graphene formation over Cu foil. CVD-grown graphene without pretreatment: a Raman mapping of full width half maximum of 2D mode, b SEM images of graphene transferred on Si/SiOx, and c AFM images of the same graphene. And graphene growth with oxidation pretreatment: e Raman mapping of FWHM 2D, f SEM images of graphene transferred on Si/SiOx, and g AFM images of the graphene. The SEM images of Cu foil before CVD growth: d without pretreatment, graphene flakes form h with oxidation pretreatment, Cu surface remains clean without graphene flake formation. Reprinted (adapted) with permission from Ref [195]. Copyright (2015) American Chemical Society

In terms of Ni supports, the optimization of the thickness of Ni film upon a support (e.g., Mo foil or Si/SiO2 wafer) is crucial to control the number of graphene layers and the homogeneity of the graphene layers by controlling the C solubility which is high in Ni [188, 190, 208].

When employing non-metal substrates, for example, silicon oxide or silicon nitride, usually cleaning in organic solvents is used initially [209]. Oxygen treatment (air burning 800 °C for 1 h) [210] is also implemented, and this has been shown to enhance the nucleation of graphene.

Thermal chemical vapor deposition

Thermal CVD typically results in planar graphene with controlled layer numbers such as mono-, bi-, and multi-layer graphene. In a thermal CVD procedure, Ni and Cu are the most popular substrates to grow graphene [4] though a few works use Ru [211], Co [212], Fe [213, 214], and Ag [215]. Polycrystalline Ni films (on Si/SiO2 support) facilitate the formation of thin continuous graphene films (layer number varies from 1 to 10) at ambient pressure [216, 217] and also at low pressure [218]. Single-crystal Ni (111) is good for the efficient growth of monolayer graphene (coverage ca. 90 %), while Ni polycrystalline foil has a lower yield (72 %) [4]. Among all the substrates, Cu currently attracts the most attention because it is far easier to grow monolayer graphene due to the low carbon solubility found in Cu. In general, the polycrystalline graphene grain sizes found with Cu are larger than those found in Ni [4]. In polycrystalline Cu, monolayer graphene is usually argued to grow in a self-limited manner [194]. A further advantage of Cu as a substrate is that it can be scaled up relatively easily for large-area graphene fabrication such that even 30-inch graphene transparent films transferred in a roll-to-roll approach have been demonstrated [219]. In an effort to remove the need to etch copper away from synthetic graphene after synthesis, the use of thin Cu films over Si/SiO2 [220] can be advised as it sublimes during the reaction such that at the end only graphene remains [221].

Another approach to avoid transfer (which can damage and contaminate the graphene) is to directly synthesize graphene over a non-metal substrate with no metals present. Non-metallic substrates include quartz [222], sapphire [223], Si/SiO2 [210], Si3N4 film [224], MgO crystals [225], and hexagonal BN [226228]. This form of CVD growth is slower than when using metals; however, at times there are advantages, for example, the electron mobility of graphene grown over Si3N4 was shown to exceed that of transferred graphene grown over Cu [224].

Plasma chemical vapor deposition

Plasma CVD leads to the formation of vertical graphene and also planar graphene. With plasma-enhanced CVD, one can grow vertical graphene on various substrates including metals and non-metals [22, 23]. Moreover, PECVD can speed up the growth rate of planar graphene on non-metal supports by 15 min to achieve full coverage, [229] as compared to thermal CVD by at least 4 h [210]. In addition, PECVD can decrease the growth temperature significantly, for example, at 600 °C [230] one can generate high-quality graphene, while thermal CVD usually requires temperatures of 1100 °C or above [222].

Carbon feedstock

The carbon feedstock for graphene growth is typically a hydrocarbon. They can be categorized into gas, liquid, and solid. The most frequently applied is gaseous methane (CH4) [194, 196, 231, 232] in Cu-based CVD. Also, ethylene (C2H4) [233] and acetylene (C2H2) [212, 225] have been used successfully for growing graphene. Liquid carbon source commonly seen in laboratory CVD growth of graphene includes alcohols [234240] (for example methanol, ethanol, isopropanol), acetone [234], large hydrocarbons [241243] (pentane, hexane), and aromatic hydrocarbons [235, 244246]. Solid carbon sources have also been explored such as carbon doping into bulk metals [247, 248], amorphous carbon [249251], self-assembly monolayer [252], camphor [253256], DNA [257], insect protein [258], flower petal [259], PMMA [260266], PS polymer [267], and waste plastics [268].

The carbon feedstock for carbon nanotube growth is similar to that for graphene. They are methane [269, 270], ethylene [271275], acetylene [276278], alcohol [80, 279285], and aromatic hydrocarbons [286288]. Carbon monoxide [289] has also been used to grow single-walled carbon nanotubes.

Growth mechanisms and dynamics

Thermodynamics describes the possibility of a reaction. It can be evaluated through the activation energy (Ea). The reaction process initializes when one satisfies certain conditions to overcome the reaction energy barrier (e.g., by increasing temperature, introducing a catalyst, etc.). Kinetics, however, describes the reaction rate. The kinetics can be evaluated using kinetic constants, for example, the growth rate of graphene (such as increase rate in size, density, or coverage of graphene flake over Cu substrates). The kinetic constants can be obtained from statistics of the experimental data. To obtain thermodynamic information such as activation energies, one can plot the so-called Arrhenius plot which is the logarithm of a kinetic constant against the reciprocal of temperature.

Thermodynamics and activation energy

Now, we turn to the parameters that affect the Ea such as gas pressure, type of feedstock, type of substrate, oxygen incorporation, plasma enhancement, and catalyst. The activation energies are in the range of 1.5–3.7 eV in low-pressure Cu–methane CVD [199, 290292]. When increasing the pressure to ambient atmosphere, Ea increases by ca. 3 eV [292].

With respect to the influence of carbon feedstock choice, when using ethane (C2H4) as an alternative carbon feedstock in low-pressure CVD, activation energies are similar in the range from 2.4 to 3.1 eV [293, 294]. When using small-molecule solvents (ethanol, acetone, and isopropanol), Ea decreases to 1.4–1.6 eV [192]. When using large aromatic hydrocarbons (C22H14 and C24H12), the Ea drops to 1.46 and 1.87 eV, respectively [246].

In terms of substrate choice, when using Co and Ru substrates (under vacuum), the activation energies show no significant difference to that found for Cu (1.4–2.9 eV) [294, 295]. However, when switching to non-catalytic oxides, large increases in Ea occur such as 4.75 eV for SiO2 [210] and 6.75 eV for Si3N4 [224]. This highlights the importance of a catalyst in decreasing the reaction energy barrier.

A tiny amount of oxygen adsorbed on a Cu surface can reduce the Ea [199]. Oxygen incorporated with H2 in the reactions can lead to the formation of OH radicals which decrease the energy barrier. Compared to thermal CVD, plasma-enhanced CVD (PECVD) significantly reduces the Ea to 1.03 eV [296]. In addition, metal promoters such as Ni [290] reduce the Ea to 1.7 eV from 3.7 eV. Indeed, Ni foil [290] (in the upstream of target substrates) has been shown to enhance the decomposition of the carbon feedstock and thereafter accelerate graphene growth. We list the Ea values in Table 2 for an overview.

Table 2 Thermodynamic activation energy (Ea) is listed from different works in the literature. The growth conditions are also included

We now look at the growth kinetics of graphene. The kinetic constants can be accelerated by increasing reaction temperature, reactant concentrations, and catalyst activity. Firstly, an increase in CVD temperature can accelerate graphene growth. Indeed at temperatures below 1000 °C, full graphene coverage does not occur and pores remain in the film even after 150 min of CVD growth [291]. With temperatures in the range from 720 to 800 °C, graphene growth saturates to 50–80 % coverage [297]. However, at the elevated temperature of 1000 °C full coverage can be achieved within 10 min of CVD reaction time [291, 297]. Secondly, the growth kinetics can be enhanced by an increase in reactive carbon species. Reactive species concentrations increase either with an increase in the methane partial pressure [298] or an increase in reactive carbon species. The carbon species concentration can also be enlarged by an increase in temperature [299, 300]. The amount of carbon (larger than the equilibrium level for nucleation) is attributed only to the carbon source choice [291].

Thirdly, the catalytic activity of Cu and Ni can improve the growth kinetics. Indeed, CVD on Cu or Ni substrates typically requires < 1 h to achieve full coverage graphene growth [194, 196, 216, 218]. However, the CVD duration over a non-catalytic ceramic oxides requires several hours or days [210, 222, 224].

Furthermore, metal vapor is an efficient means to speed up the growth kinetics [196, 301]. Specifically, Cu vapor trapping facilitates graphene full coverage growth in 10 s [195, 196], whereas 15 min CVD growth times are required in a non-trapping setup with the same growth conditions [196]. Methane decomposition into reactive carbon species is probably enhanced by the trapped Cu vapor.

To better understand the thermodynamic and kinetic aspects of graphene growth, in situ characterization techniques are being developed, e.g., in situ SEM [302, 303], LEEM [304, 305], and XPS [306].

Reaction mechanisms over Cu

Cu is probably the most often used substrate since, because of its low carbon solubility, monolayer graphene with full coverage can easily be obtained with monolayer ratios exceeding 95 %, [194] and strictly (100 %) monolayer graphene [195, 309, 310]. A surface-mediated self-limited mechanism [194] is proposed such that once the Cu surface is fully covered, the catalytic process of Cu ceases and hence so does graphene growth. However, the surface-mediated mechanism cannot account for the secondary formation of graphene. For example, second-layer (additional) patches emerge in full coverage monolayer graphene film [195, 196, 309], and even agglomerate into parallel lines [311] and achieve full coverage for bilayer graphene [310].

It is argued that part of the reason for this is carbon diffusion into/out of Cu grain boundaries. For one case, the outside of Cu pocket is found to grow bilayer graphene [310]. However, with a tungsten foil inside the Cu pocket, monolayer graphene was achieved with tungsten argued to suppress carbon diffusion from the inside to outside of the pocket [310]. In another case, pulsed methane doses [309] as opposed to a continuous flow were introduced to deplete carbon for graphene growth to avoid carbon accumulation at Cu grain boundaries (and thereafter C diffusion to form second-layer flakes) Adsorbed solvent on the Cu substrate can also account for the additional graphene flake formation [192].

Some researchers propose that the carbon can diffuse through an initial graphene layer, nucleate, and form second-layer flakes underneath full coverage monolayer film [312314]. They discuss second-layer graphene growing above the full first layer of graphene [13, 315, 316]. In short, the origin of second-layer flakes is still not fully understood.

Liquid Cu substrates can play an important role in the growth of single-crystal monolayer graphene. Cu foils liquefy at elevated temperatures above its melting point (1080 °C). Monolayer single-crystal graphene (either round or hexagonal shaped) were fabricated over liquid Cu [203, 204, 317]. The grain boundary free Cu surface suppresses carbon diffusion into the Cu bulk and thereafter avoids the formation of second-layer flakes from carbon extrusion (from accumulated sites in the Cu bulk, e.g., grain boundaries). The advantage of liquid Cu also applies to other metals such as In [204] and Ga [318]. In contrast, solid Cu suffers from the evolution of polycrystalline graphene [319].

The geometry and quality of graphene flakes are probably dependent on Cu grain orientation (in solid Cu). Cu(101) and Cu (110) facilitate four-lobe graphene domain formation [320]. Moreover, graphene ribbon can be selectively grown on Cu twin-crystal striped surfaces like the (001) facet [321] by controlling the methane partial pressure, whereas Cu (111) facets produce higher quality monolayer graphene with a higher growth rate than Cu (100) facets [322]. The shape of graphene flakes can also be influenced by the gas conditions. Initially, square-shaped graphene domain can be controlled by growth conditions, e.g., methane partial pressure [323], regardless of the Cu lattice orientation. Generally, a high hydrogen flow ratio (in comparison to methane pressure) facilitates hexagonal graphene [206, 316], while low hydrogen partial pressure facilitates square graphene domains. Moreover, annealing with high-partial pressure H2 single gas can etch the square-shaped graphene flakes into hexagons [206].

Reaction mechanism over Ni

Ni as a substrate for graphene growth exhibits carbon segregation and precipitation [216, 218, 324]. Initially, in a CVD reaction, carbon atoms dissolve [325] into the Ni bulk forming a carbide [326]. Carbon segregation and precipitation from meso-stable Ni2C [327] occur upon cooling. However, graphene grown on pure Ni foil is inhomogeneous in layer numbers, ranging randomly from 1 to 10 layers.

Therefore, one needs to engineer the substrate to achieve homogeneous graphene layer formation. One example is to employ Ni thin films to decrease the quantity of carbon that can diffuse into the Ni [216, 328]. Furthermore, mild oxidants (CO2 and H2O) can be mixed with methane to grow thin few-layer graphene with diminishing carbon diffusion [324] because Ni sub-oxide not only increases catalytic activity but also leads to the suppression of carbon diffusion. Graphene growth over NiOx with CO2 enhancement is shown in Fig. 2.

Fig. 2
figure 2

Graphene growth on NiOx substrate with CO2 enhancement. a Schematic of CVD growth of graphene over NiOx surface. Middle is the energy state versus the reaction pathway of methane dissociation. b Raman spectra of graphene grown at 700 and 500 °C with/without CO2 incorporation. With CO2, the D mode decreases. c,d STEM image and EDX mapping of C, Ni, and O in a cross section of graphene/NiOx interface. e XPS of the surface of graphene over NiOx. There is direct observation of oxygen at top surface of the Ni substrates. Reprinted (adapted) with permission from Ref [324]. Copyright (2014) American Chemical Society

Alternatively, when using a binary alloy (Ni/Mo) [188, 190], one can produce strictly monolayer graphene. Figure 3 shows a monolayer graphene growth on an Ni–Mo alloy. Similarly to Ni/Mo alloy, one can control the graphene layers from monolayer full film to a bilayer full membrane by tuning the component ratio of Cu/Ni alloy [189].

Fig. 3
figure 3

Monolayer graphene growth from Ni–Mo binary alloy. a Schematic of growth mechanism: surface decomposition of carbon feedstock, carbon atoms’ dissolution into the bulk, Mo2C formation, and controlled carbon precipitation into graphene. b Photograph of full coverage graphene film transferred onto Si/SiOx substrates. c Optical micrograph of the transferred full coverage graphene, inset is typical Raman spectrum that confirms the monolayer graphene feature. d,e Discontinuous graphene film and graphene flakes with slow cooling rate after CVD. Adapted by permission from Macmillan Publishers Ltd: [Nature Communications] [188] copyright (2011). fi TEM characterization of the graphene monolayer membrane. jo AFM height images of the transferred graphene on Si/SiOx. From left to right are the evolution of monolayer full film, discontinuous film, and isolated flakes, according to various cooling temperature windows. Reprinted (adapted) with permission from Ref [190]. Copyright (2013) American Chemical Society

Reaction mechanism over non-metals

Graphene growth can also be achieved over non-metal substrates, for example Al2O3 [223], MgO [225], and SiO2 [210, 222] ). In comparison to Cu CVD, the growth over oxides generally requires higher temperatures >1100 °C, [210, 222]. In addition, surface oxygen [210], metal vapor [209, 329], and plasma [229] are necessary to promote graphene growth.

Graphene growth over sapphire (Al2O3) involves epitaxial growth by van de Waals interactions [223]. The unit cell of the grown graphene has a 30° rotation compared with that of sapphire substrate. High-resolution TEM (see Fig. 4) proves the parallel stacking of graphene over MgO crystals [225], which also confirms epitaxial growth. In addition, graphene was successfully grown over non-metal crystalline substrates such as graphite [330, 331], SiC [187], BN [226, 228], and graphene on a Cu support [13].

Fig. 4
figure 4

Graphene grown on MgO substrate. a,b TEM images of few-layer graphene on MgO lattice. c,d Graphene nano-flake on MgO surface with short exposure of carbon feedstock. e Magnified region from panel c highlights the graphene structure. f Cross-sectional view of graphene on MgO surface. g Raman spectrum of few-layer graphene after purification. h TEM of the purified graphene. Inset is the contrast profile of graphene interlayers. i Scheme of graphene growth on MgO: nucleation at the step edge. j HRTEM of magnified region of panel c. k Schematic of a nanographene flake illustrating the large number of edge defects. l Fast Fourier transform images from magnified region in panel h. Reprinted (adapted) with permission from Ref [225]. Copyright (2010) American Chemical Society

The growth mechanisms over non-metals are not yet clear. The graphene growth is argued to involve a surface deposition mechanism with a very slow deposition rate [210]. A vapor–solid–solid mechanism is also argued [210, 222]. SiO2 nanoparticles were suggested to nucleate graphene growth [210] probably through an intermediate of SiC formation [281]. Without any treatment of oxide substrates, the grain size of graphene is small, around 300 nm [210]. However, with a pretreatment of air annealing, the graphene size exceeds 10 microns [222]. The growth kinetics is probably limited with slow carbon attachment at graphene edges [222]. Si3N4 is also a candidate substrate for graphene growth [224].

The growth mechanism of vertical graphene is shown in Fig 5. The defects of initial graphene sheets supply the nucleation sites. Then, the graphene continues to grow with surface diffusion of active carbon species.

Fig. 5
figure 5

Vertical graphene (VG) grown from PECVD. a Low-resolution TEM images of VG including graphene top, bottom, and substrates, b high-resolution TEM image of VG bottom. The mismatches are probably nucleation sites. c HRTEM of VG stems on carbon onions. d HRTEM of the carbon onion. e Scheme of a proposed mechanism: nucleation at defects, growth at surface and edges. f HRTEM of a seamless edge of a graphene closure. g Schematic of the vertical graphene growth at its closure. h TEM and atomic model of a growing VG sheet with many curved steps and terraces. Reprinted (adapted) with permission from Ref [23]. Copyright (2014) American Chemical Society

Transfer protocol

Graphene grown on metals requires a transfer step for further characterization or device fabrication. Firstly, the graphene side is (spin-) coated with a supporting film, usually a polymer such as PMMA [216, 332, 333]. Secondly, the polymer/graphene/Cu stack is put afloat on an etching agent, such as FeCl3 [333] and ammonium persulfate (APS) [334, 335]. Thirdly, the polymer/graphene is cleaned in distilled water after thorough etching of the metal. Then, the polymer/graphene is fished onto a target substrate. Next, acetone is used to dissolve the polymer (e.g., PMMA) away. Finally, the transfer process is completed with high-vacuum annealing to minimize surface contamination. When the annealing temperature exceeds 300 °C, the electric property improves with negligible polymer residues [335]. In addition, the graphene on the undesired side of a substrate or foil can be pre-etched away with nitric acid [335] or APS [334] since graphene can grow on both sides of a foil/substrate. We list the transfer protocol and the chemicals for each step in Table 3.

Table 3 Transfer protocol for graphene over Cu

Alternative approaches for graphene transfer also exist. Bubbles can be used to intercalate into the interface between the graphene and the substrate and thus separate them. The bubbles are usually generated in an electrochemical reaction [338]. To recycle Cu foil, an electrochemical cell is used in a controlled fashion to dissolve a very thin layer of Cu (cathode) underneath the graphene [338, 339]. In this mild etching, hydrogen bubbles from the electrolysis of water can assist in peeling off the PMMA/graphene from Cu.

A bubble method was successfully developed that does not require an electrochemical cell and so can be applied to non-conductive substrates [337]. The process relies on the copious production of gaseous O2 which intercalates in between the substrate–graphene interface and can thus serve as a peeling agent. The schematic is shown in Fig. 6.

Fig. 6
figure 6

A bubble approach to transfer graphene from arbitrary substrates. a TEM micrographs of graphene transferred to a grid from Cu, Ni–Mo, Al2O3, and SiC substrates. b Raman spectra of transferred graphene. c Raman mapping of G/D mode of graphene transferred from Cu foil. d Photographs of key steps of the transfer: coating PMMA, immersion into bubble agents, and delaminating PMMA/graphene with Cu foil. e Schematic of the complete transfer route. Reprinted (adapted) with permission from Ref [337]. Copyright (2014) Royal Society of Chemistry

Mechanical transfer can be achieved by peeling off sticky epoxy/graphene from a substrate [341]. In a similar manner, thermal release tapes in a roll-to-roll dry transfer route can separate graphene from a support [219].

Alternative polymers such as polycarbonate (PC) [336] can be used to diminish polymer residue. PC polymer dissolves more effectively in chloroform than the PMMA/acetone combination. This can avoid the need for high-vacuum annealing. The use of fluoropolymers (CYTOP model) instead of PMMA can achieve fluorine doping in graphene [342]. This single step largely increases the electrical conductivity (with higher carrier density) but preserves the high transparency of the graphene. When fishing a polymer/graphene stack out of water onto a substrate, isopropanol is a better medium than water to guarantee better adhesion over the Si/SiO2 substrate [343]. However, without a polymer support, graphene can be directly fished onto a net upon Cu removal [340].

In terms of graphene over In or Ga, HCl solution can be used as an etching agent [204, 318]. For graphene on Ni, the etching agent can vary from mild hydrochloric acid [216], acetic acid [18, 19], O2 bubbles [337], and nitric acid [18]. In the case of oxide-based supports, HCl was employed to remove MgO nanocrystals [16]. KOH can etch SiO2 substrates (such as quartz or amorphous SiO2 film) [344, 345].

CVD growth of carbon nanotubes

The primary approaches to synthesize CNTs are arc-discharge evaporation of graphite, laser ablation, and CVD. In particular, CVD is well developed in industry and easier to scale up than arc-discharge and laser evaporation methods. Hence this section is focused on CVD growth of CNTs. The CVD approaches discussed here are divided into two sections, catalyst CVD and catalyst-free CVD [2, 346].

Substrate pretreatment

Prior to the CVD growth of CNT, the substrates often requires pretreatment. Air annealing at 400–750 °C is widely used. This process can oxidize the metal particles (catalyst particles) which minimizes Oswald ripening at the elevated temperatures used in the reaction. In addition, organic contaminants can be removed [347]. Moreover, long air annealing times can reduce substrate roughness [348]. Indeed, substrates with a rough surface can halt CNT growth or lead to twists the growth direction and is sometimes implemented for horizontally aligned CNT growth.

Catalyst chemical vapor deposition

The more common catalysts used in the CVD growth of CNT are Fe, Co, and Ni or related alloys [2, 346]. There are two types in catalyst systems in CNT production by CVD. They are the floating catalyst CVD and supported CVD. In the floating catalyst system, the catalysts are not supported and reside in the gas flow [349, 350]. In supported catalysis, nanoparticles are pre-deposited on supporting substrates and serve as the catalyst particle. According to the CVD conditions implemented, one can obtain three types of CNT orientations. They are vertically aligned CNT [351, 352], horizontally aligned CNT [353357], and randomly aligned CNT [358].

Catalyst-free chemical vapor deposition

Non-catalysts such as ceramic oxides have been shown to grow CNT [346]. In addition, carbon allotropes, such as fullerenes [271, 354], diamond surfaces [359], graphite [271], and CNT [360, 361] are successful systems to synthesize CNTs.

Growth mechanisms

Thermodynamics and activation energy

The thermodynamics (e.g., activation energy) to grow CNTs has been carefully studied over various conditions [362366]. Table 4 provides a comparative overview of the parameters that affect activation energy. In this review, thermal CVD and PECVD are compared. For thermal CVD, the Ea of CNT growth can be as high as 1.21 eV (Ni–C2H2) [364]. In contrast, PECVD significantly reduces the Ea, e.g., 0.56 eV [363] and 0.76 eV (Ni–C2H2) [364]. These values are obtained in the temperature range (520–700 °C). Plasma CVD has benefits at low temperatures. For example, one could grow device-quality CNTs at 200 °C [367] with the plasma-assisted CVD. This allows CNT vertical arrays to be directly grown on a plastic polyimide substrate.

Table 4 Activation energy of carbon nanotube growth over various substrates

Many factors account for the lower Ea in PECVD than a thermal CVD [364]. Firstly, the catalytic particles could be heated by ion bombardment and increase the carbon diffusivity and solubility within them. Secondly, the activated etching gases could more effectively remove excess carbon species from the front nickel surface, and thus an enhanced uptake of carbon occurs at the larger fresh area of the metallic catalyst particle. Both factors could provide a driving force to allow carbon diffusion through the catalyst particles.

In terms of the role of the catalyst metal choice, the highest Ea is 3.5 eV for Co–C2H2 couple [366]. With Ni and Fe employed, the Ea drops significantly to 2.3 eV [366] and 1.3 eV [365], respectively. Indeed, the same trend occurs when using C2H4 as the feedstock gas. They are Co (1.69 eV) [362] and Ni (1.4 eV) [366]. However, a complex alloy of Co/Cr/Al [362] significantly suppresses the Ea barrier to 0.84 eV.

We now turn to the influence of carbon feedstock. When mixing a promoting gas such as NH3, the Ea (1.21 eV) [364] for Ni–C2H2 (NH3) decreases to half of that found with Ni–C2H2 (2.3 eV) [366] at ambient pressure. Different feedstocks with Ni catalyst, such as CH4 (1.5 eV), C2H4 (1.4 eV), and C2H5OH (1.3 eV), lead to varying activation energies. Co catalyst particles show a similar trend when switching from C2H2 (3.5 eV) [366] to C2H4 (1.69 eV) [362] and C2H5OH (1.1 eV) [366].

In thermal CVD growth, Ea tends to be larger (1.1–3.5 eV) in low-temperature zones (<600–700 °C) and much smaller (0.1–0.7 eV) in high-temperature zones(>600–700 °C) [366]. The decrease of Ea probably depends on the liquefaction of the metal catalyst particles. At elevated temperatures, the nano-sized metal catalysts liquefy at lower temperatures than their bulk counterparts [270, 368371]. The activity of catalyst particles can be enhanced upon increasing the temperature. Indeed, molten particles continue adsorbing carbon atoms and elongating (growing) nanotubes after nucleation. However, as the liquefaction temperature increases beyond the synthesis temperature, the carbon concentration increases in the metal particles [372]. Solidification decreases the carbon diffusion kinetics and finally ceases CNT growth altogether.

Similarly, in PECVD the Ea varies between two temperature regimes. The Ea is larger (0.76 eV) [364] at high temperatures (520–700 °C) but smaller (0.23 eV) [373] at low temperatures (120–520 °C). Indeed, the Arrhenius plot has a strict linear fitting at the low temperature range [373]. A linear fit occurs only in the low portion in the high-temperature zone [364].

One can engineer the kinetics of CVD with predictions (rather than trial-and-error mode) [374]. For example, the CNT synthesis can be conducted with a controlled diameter [375378] or even reasonably defined chirality [379381]. Moreover, the kinetic process of CNT formation can be monitored with enriched in situ characterization techniques, such as digital camera [382], reflection optical spectroscopy ([383, 384]), adsorption optical spectroscopy [385], Raman spectroscopy [386389], thermogravimetric analysis [288], XRD [390], XPS [391], and environmental TEM [82, 392394].

The termination of CNT growth is controversial. Early work deduced that the growth ceases upon the deactivation of the catalyst particles. The deactivation includes the poisoning of catalysts by amorphous carbon encapsulation [395397], or by Oswald ripening/sub-diffusion [398]. This theory is elegantly supported by the water-assisted super growth of vertical CNT arrays [351, 399, 400]. Indeed, water incorporation at elevated temperature removes the amorphous carbon layer and exposes the fresh catalyst surface [398]. Thereafter, the catalysts re-activate for the CNT growth. Similarly, oxidants such as O2 and CO2 show an enhancement in SWNT growth [400]. Moreover, oxidized substrates may drive CNT growth by preventing amorphous carbon condensation [80]. But, one convincing argument rises up for the non-poisoning nature of amorphous carbon layer [401]. Here, amorphous free radical condensates over catalyst are proposed to serve as a carbon source rather than a deactivation system. Indeed, high-resolution TEM shows that the amorphous carbon covering catalyst can continue CNT growth [402]. Beyond these two arguments, vertical CNT arrays show a genuine saturated growth behavior [403, 404].

Reaction mechanism over catalysts

The reaction mechanism over catalysts include carbon dissolution and nanotube precipitation [2, 346]. Depending on the catalyst particle location, the growth mechanisms are divided into tip growth mode (where particle rises to the top) and base growth mode (where catalytic particles anchor to the supporting substrate). Ni over SiO2 often follows tip growth because of a weak particle–substrate interaction [80, 363]. However, Co and Fe over SiO2 substrates [405, 406] and over Al2O3 [78] favor base growth where a strong interaction limits the motion of the catalyst particles. Indeed, over Al2O3 support, the catalyst particles were found either embedded inside the supporting film or anchored on its surface (see Fig. 7).

Fig. 7
figure 7

Carbon nanotube (CNT) growth on various supports. a TEM images of CNT with root on the surface of Al2O3 thin film while catalyst buried inside the Al2O3 film. b TEM images of CNT with root on α-alumina nano-platelet. A Co particle elongates inside the CNT. The outer walls of CNTs align themselves with the (110) lattice fringes of the corundum support. c TEM micrographs of bamboo-like CNT grown on a graphite support. The right panel highlights the graphitic cap terminating the outside of catalyst particle. Correlation of CVD grown CNT diameter with catalyst particle size (d) and number of CNT walls with respect to catalyst particle size (e). Reprinted (adapted) with permission from Ref [78]. Copyright (2010) American Chemical Society

The reaction to form MWNTs over Ni includes (1) decomposition of carbon feedstock on the front face of a catalyst particle; (2) carbon dissolution and supersaturation in the bulk metal particle; (3) carbon precipitation to the back face; and (4) carbon incorporation at the root of a growing CNT. During the CNT growth, the Ni particle is raised upward spontaneously. The driving force to continue growth dynamics originates from the gradient in chemical potentials [407409], particle temperature gradients (front face > back face), and carbon concentration gradients (front face > back face). Namely, carbon (monomers or dimers) diffuses from high to low temperature and from high to low concentration in Ni particles. In terms of SWNTs, a carbon dissolution and nanotube precipitation mechanism applies. But, the driving force relies solely on the gradient in carbon concentration since the temperature difference is negligible in a sub-10 nm particle [410, 411]. To obtain SWNTs, one needs to optimize a number of parameters [2, 289, 412, 413], such as catalyst choice, temperature, and feedstock.

Diameter control of SWNTs can be achieved by various strategies, for example, using sulfur-promoted aerosol-assisted CVD [375], or to confine the catalyst particle size with secondary non-catalytic metals (such as Mo, Pt, Ru, Rh, and W). Thus, binary alloys such as Co/Mo [412, 414],Fe/Mo [413, 415, 416], and Co/Pt [417] are often used confine SWNT growth to the base growth mode. Furthermore, water incorporation (or other mild oxidants) help prevent Oswald ripening of Co over Mo, therefore narrowing down the diameters of the resultant SWNT [414]. Surprisingly, some degree of chiral control of SWNTs via CVD is emerging such as (12,6) tubes from Co/W clusters [418], (9.8) tube from CoSO4/SiO2 [419] and from Co/TUD-1 [420], cloning of (7,6) and (7,5) tubes [360], and (6,5) tube over CoxMgO substrates [421] and from Co/Pt catalysts [417].

In an atomic view, a theory of screw dislocation has been postulated for chiral SWNT growth [422]. In this work, the SWNT grows with C monomer attachment along the exposed tube edge in a screw-like fashion. Surprisingly, the screw theory has been confirmed with experimental observations under a field emission microscopy [379].

Reaction mechanism over non-metals

Table 5 summarizes different aspects of CNT growth from non-catalyst, mainly non-metals. CNT growth from carbon allotropes usually arises from defects such as nano-pores, open edges, and defects. Indeed, the exposed open edges of CNT walls (such as bulky onions and stack mismatches [23], Fig. 5) between graphitic sheets can facilitate the attachment of carbon adatoms [379]. On the one hand, the open-end tube structure has a high-energy state, which can crack carbon, the feedstock. These carbon atoms could initialize the formation of MWNTs or SWNTs depending on the size of the nuclei site. On the other hand, the defect sites or particle encapsulates (similar to Ni particle at tip) [396] may continue supplying sites to template the elongation. When expanding to other non-carbon particles, the initial nucleation occurs by forming carbon nano-caps.

Table 5 Growth conditions and results of CVD-grown CNTs, support type, curvature radius, and diameter of CNTs, nucleation site, CNT type, and particle location

Other elements such as Si, Ge, and SiC can also serve as catalyst particles for the growth of CNT. The Ge particle is proposed to be a liquid during growth. In contrast, Si and SiC are argued to be solid during growth.

SiO2 nanoparticles have also been used to grow CNTs (Fig. 8). When SWNT grows out from nano-SiO2 particles, SiO2 is assumed to be either a liquid [428] or an amorphous solid [357, 429] during the reaction. However, these hypotheses were deduced with micro-meter-scale resolution methods, like XPS.

Fig. 8
figure 8

Carbon nanotube growth over SiO2 support without a catalyst. a An overview with SEM graph of dense CNT networks. b Low-resolution TEM images of the CNTs. c Optical absorption spectra of the CNT on SiO2 samples. There is a peak referring to SiC. d Raman spectra present an SiC peak. f An SiC particle encapsulates inside the CNT tip. (eh) HRTEM of the CNT with a cap tip or with opening tip. Schematic showing reduction of SiO2 catalyst particle to SiC and subsequent carbon tube/fiber growth (i). Reprinted (adapted) with permission from Ref [281]. Copyright (2009) American Chemical Society

Many metal oxides [357, 427433] such as Al2O3, TiO2, ZrO2, ZnO, and MgO have also been used for CNT growth.

Low catalytically active metals such as Au, Cu, Pt, Ag, Pd, Mn, Mo, Cr, Sn, Re, Mg, and Al [344, 436439] have also been used for CNT fabrication. Last but not least, alkali salts [442] such as NaCl and Na2CO3 have also been demonstrated for the growth of CNT.

Conclusions

In this work, an overview of the CVD aspects involved in the synthesis of graphene and CNTs have been presented for both thermal and plasma-enhanced CVD. The shear breadth of possibilities is highlighted by the number of variations available in terms of the presence of catalyst, synthesis parameters, feedstock options, etc., all of which affect the thermodynamics and kinetics of the reaction and hence the final outcome of the as-produced material. The review highlights the enormous power of CVD as a synthesis route for graphene and CNTs as well as its versatility. Despite some remaining technical issues such as grain boundary control in large-area graphene or total chiral control of carbon nanotubes, the sheer volume of knowledge now available and recent advances suggest the use of CVD as a means to fabricate graphene and CNTs are not only set to stay, but will likely soon overcome many of the remaining technical challenges and thus allow graphene and CNTs to truly deliver their full promise.