Atomic precision manufacturing of carbon nanotube—a perspective

When the size of a material decreases to nanoscale, many different phenomena occur, and surface atoms become more important in affecting or sometimes determining the properties of this material. Starting from this simple 'size effect', the past few decades have witnessed a blooming growth in nanoscience and nanotechnology. Extensive progress has been achieved in developing fabrication techniques, investigating novel properties, as well as concept-proofing various applications [1].

Carbon nanomaterials have played an irreplaceable role in this 'small' field. One crucial reason is that carbon is one of the earliest materials that can be prepared into atomic-thin layers. In such an extreme situation, all the carbon atoms in the hexagonal lattice are on the surface, so the atomic precision becomes unprecedentedly important. Three main types of these carbon nanomaterials consisting of all-surface atoms are fullerene, carbon nanotube (CNT) and graphene, which are also well known to be representatives for zero-dimensional (0D), one-dimensional (1D) and two-dimensional (2D) materials [2–4].

While in all three cases that the atomic arrangement influences properties of a carbon nanomaterial, CNT seems to be the most extreme one. In a single-walled CNT, two factors exclusively determine the electronic structure of this CNT. The first is the rolling angle of the carbon hexagonal lattice and the second is the number of periodicity used to form the seamless tube. The former factor is called chiral angle that is between 0° and 30°, and the latter is associated with tube diameter (shown in figure 1). Alternatively, two integers, n and m, can be used the strictly define the structure of single-walled CNT. This pair of numbers (n,m) is referred as chiral index, or chirality, in the nanotube community. Among all chirality combinations, 1/3 of the CNTs are metallic and the other 2/3 are semiconducting. Because of this structure-property correlation, 'chirality control' or 'chirality-specific synthesis' have been an ultimate target for the production of CNT in the past 30 years.

Zoom In Zoom Out Reset image size

There have been two main approaches to obtain atomically precise CNTs: post-synthesis sorting and direct chemical vapor deposition (CVD) synthesis. The former is developed about 10 years earlier than the latter.

Roughly between 2004 and 2014, while the whole community realizing direct CVD synthesis of atomically precise CNT is highly challenging, researchers working on CNT dispersion noticed that there may be another way. They observed that different types of CNT behave differently in a solution, and then with the aid of some separation techniques, solution containing a specific type of CNT can be obtained. Accordingly, density gradient ultracentrifugation [5], deoxyribonucleic acid (DNA) assisted separation [6], gel-chromatography separation [7] were quickly established. All these methods can produce CNT solutions with certain types of CNT and/or even single-chirality CNT [8], and recent optimizations even allowed to obtain single chirality CNTs up to sub-milligram scale [9]. One very important and eye-catching fact that researchers begin to realize is the color of CNT—CNTs are not always black. Solutions of specific type of CNTs are actually colorful when CNTs are pure in their atomic structures [10] (figure 2). With the development of these separation techniques, more recently there have been efforts making aligned films from CNT solutions [11–13], and these aligned, high-density, chirality-separated films can be used to build high-performance electronics [14, 15].

Zoom In Zoom Out Reset image size

The long-awaited breakthrough in direct CVD synthesis of chirality-specific CNT appeared in 2014 [16]. Researchers developed a novel Co–W catalyst having a high melting temperature, this solid catalyst produces specifically (12,6) single-walled CNTs with a purity over 92%. By tuning the catalyst pretreatment and reaction conditions, other types of CNTs are also produced at high selectivity (figure 3). For example, (16,0) CNTs are obtained over 80% purity and (14,4) CNTs are obtained with over 97% purity [17, 18]. The (14,4) CNT is of particular interests as it is semiconducting and may be used directly as the channel material of a field effect transistor. The mechanism behind this series of growth selectivity is believed to be the nucleation on the specific facet of this Co–W solid catalyst, meaning certain atomic arranged CNT seeds are preferably formed. This model is proposed based on a systematic density function theory calculation and verified by transmission electron microscopy (TEM) characterization of the catalyst-CNT interface. Overall, this revolution in catalyst design led to the breakthrough of atomic-precise synthesis of CNTs. In the same year, one other progress from a different approach is also reported [19]. A specially designed molecule, which can later be transformed into a certain CNT cap, is used as the precursor for CNT growth. With the help of additional growth, this seed can elongate into an atomically precise (6,6) CNT.

Zoom In Zoom Out Reset image size

Following these breakthroughs, another class of catalyst, carbides, are also found to produce certain type of CNTs with high selectivity [20, 21]. For example, a selective growth of (2m, m) CNTs are achieved using Mo₂C and WC catalyst. The purity of (12,6) CNTs obtained by this method reached 90% and the purity of (8,4) CNTs reached 80%. Current atomically precise CNTs from direct growth are still less pure than solution-sorted samples and their formation mechanism is still not fully clear [22], but certainly they possess great advantages in terms of cleanness and crystal quality. Unlike those prepared by solution sorting which are wrapped with various surfactants or DNA, these CNTs obtained from CVD have much cleaner surfaces. Also, as-grown CNTs undergo no damaging processes like sonication, and therefore the high crystal quality is preserved. In addition, CNTs can grow into horizontal arrays without post-alignment [23]. All these factors make direct CVD growth, though very challenging, a preferred approach for the production of atomically precise CNTs, particularly for the application in nanoelectronics.

The challenges for atomic-precision synthesis for CNTs not only lie in the production itself but also in the atomic-precision structure characterization. But fortunately, many different approaches have been established to identify the structure index, i.e. (n,m), of a CNT.

Among these different methods, TEM is one of the most straightforward approaches. TEM or scanning TEM imaging can tell the atomic arrangement directly if the resolution is high enough. When high resolution is not available, nano-beam electron diffraction is also capable of identifying the (n,m) pair of a CNT. Although slight tilting or bending of the CNT will affect the chirality assignment, recent advances in low-loss electron energy loss spectroscopy (EELS) further allows to reveal the electronic state at a local area [24]. Therefore, despite the complicated specimen preparation process, TEM is usually recognized as the most straightforward and precise method for atomic-scale structural identification.

Optical characterizations are one group of powerful tools that can distinguish the atomic structure of CNTs. These tools mostly include Raman spectroscopy, optical absorption spectroscopy, photoluminescence (PL) spectroscopy and Rayleigh scattering spectroscopy. Raman spectroscopy was well adapted for CNT research owing to its high sensitivity. Because of the resonant effect, Raman spectroscopy is the earliest spectroscopic technique capable of detecting single CNT. One main limitation of above optical approaches is their tube bias, meaning they are usually unable to detect all types of CNT in an equal efficiency. For example, a Raman spectrum largely pronounces the signal of those CNTs who are 'resonant', and metallic CNTs are all blind in a PL spectrum due to the absence of an optical band gap. Therefore, there have been arguments that some inconsistency could occur between optical and microscopic assignments [25]. However, Rayleigh scattering emerges to be an effective and non-bias optical approach to detect more CNTs. A very recent work even shows that Rayleigh scattering can distinguish the handedness of a CNT [26], providing almost the similar level of precision as TEM.

A more systematic summary can be found in a recent review [27]. Atomic precision identification of CNT needs to be cross-checked by multiple methods. Each single method may lead to a serious error and sometimes an inaccurate conclusion [25]. Further development for quick atomic-precision structure assessment is still highly demanded.

The production of CNT is an extreme manufacturing process. It requires a high reaction temperature and usually a specially designed reaction equipment. Currently, there are several 'manufacturing' techniques that can produce CNTs at near-industrial scale, e.g. tons per year or more. There are also techniques in laboratory, as described in the previous section, to produce CNT with 'atomic precision', or at least 'close-to-atomic precision'. However, 'atomic precision manufacturing' is still not possible.

Figure 4(a) describes three representative direct synthesis approaches to produce CNTs, with an original radar chart showing their productivity, cost, and CNT controllability (particularly the ability of achieving atomic precision). Among the three CVD methods listed here, fluidized bed is able to manufacture CNTs on a large scale and at a low cost, and it is already very successful in producing multi-walled CNTs for battery additives [28]. However, there is little control over the produced single-walled CNT in this method; the product is mixed with many chiralities so the degree of atomic precision is not high. In a wafer supported growth, however, CNTs can be nicely controlled with close-to-atomic precision as described previously. CNTs can also be grown directly into high-density horizontal arrays. Still, the productivity in such a process is extremely low, probably less than mg per batch. As the third approach, aerosol CVD gives a compromised productivity and controllability between fluidized bed and wafer supported growth. There is exciting progress that it produces CNT films with intrinsic colors—meaning CNTs with very small structural variation [29]. This is the first example that close-to-atomic precision CNTs are manufactured into a macroscopic form by direct synthesis (figure 4(b)).

Zoom In Zoom Out Reset image size

However, for CNT transistors and ultimately CNT-based integrated circuits, atomically precise, high-density, well-aligned CNTs still need to be prepared on wafers, and then the ability of simultaneously processing multiple wafers will be highly demanded for the manufacturing. In such a case, catalyst preparation, operation parameter, and reaction equipment need to be redesigned. Although no one knows at this point when this final target can be achieved, and what the final solution will be, the author imagines that two previous processes can be good references for the CNT community (figure 5).

Zoom In Zoom Out Reset image size

The first is low pressure CVD (LPCVD). It was proposed last century for silicon industry, and horizontal LPCVD furnaces are widely used for depositing various thin films on silicon wafers. The combination of low operation pressure and wafer carrier allows to process multiple (up to 100) wafers in a single batch. Although there are concerns that precursor concentration slightly decreases along the flow, this problem can be overcome by improving inject design or by applying a gradient to furnace temperature. In addition, comparing to the processes for silicon industry, the manufacturing process of CNTs could be even simpler—the precursor is simply hydrocarbon or alcohol, and the exhaust is largely unharmful gases. However, catalyst needs to be prepared separately before LPCVD. Except for the catalyst preparation, LPCVD can be a very developed solution for the manufacturing CNT on multiple wafers.

The second is metal organic CVD (MOCVD). It is mostly famous for the manufacturing of GaN on sapphire for LED applications. Although the CNT production usually employs a slightly higher reaction temperature than GaN production, there is no concern that many years' successful experience in building MOCVD equipment allows a large degree in customizing the equipment for different reactions. The current multiple-pocket wafer holder made of coated graphite can also ensure temperatures above 1000 °C. One potential strength of MOCVD for CNT production is the possibility of depositing catalyst prior to CNT growth in the same chamber. There are a variety of metal organic precursors commercially available, which can likely to allow the preparation of various transition metal catalyst. Organic W is also available, and commonly used in atomic layer deposition (ALD) and focused ion beam (FIB) fabrication. Therefore, one can foresee that it may be not impossible to form W-based chirality-selective catalyst by MOCVD. Following this strategy, chirality specific CNT may be produced on multiple-wafer scale by a single MOCVD equipment. If achieved, the atomic-precision manufacturing of CNT will likely become reality as MOCVD are now highly developed after the success of GaN and could be extended to many other materials [30].

At this point, both LPCVD and MOCVD production of CNT are not receiving enough attention. This is probably because the market of atomically precise single-walled CNT is still unclear, so most CNT researchers are still working at laboratory scales. However, the past 10 years have witnessed a quick progress in chirality-specific growth of CNT as well as the high-end application of CNTs for transistors, even computers [31, 32]. Therefore, this could be a good time to claim that the CNT field is now ready to move from 'fabrication' to 'manufacturing', particularly 'atomic precision manufacturing'. CNT research started from 1991; 2021 is the year CNT reaches its 30 s. In addition to CNT itself, various other crystal nanotubes are also recently achieved on a CNT template, resulting in a class of new material called one-dimensional van der Waals heterostructures [33, 34]. Likely, we can expect an exciting 30 s for CNT research. There are plenty of room for the atomic-precision manufacturing of CNT, and probably also its function-designable heterostructures.

Part of this work was supported by JSPS KAKENHI (Grant Nos. JP18H05329, JP19H02543, JP20H00220, and JP20KK0114), by JST, CREST Grant No. JPMJCR20B5, Japan. Part of the work was conducted at the Advanced Characterization Nanotechnology Platform of the University of Tokyo, supported by the 'Nanotechnology Platform' of the MEXT, Japan, Grant Nos. JPMXP09A20UT0063 and JPMXP09A21UT0050.