Advancing Single‐Walled Carbon Nanotubes toward Next‐Generation Electronic Materials: The Critical Roles of Purity and Processability

ABSTRACT

Single‐walled carbon nanotubes (SWNTs) have garnered significant attention due to their unique size‐ and structure‐dependent properties, making them highly promising for a wide range of applications. Among these properties, their exceptional electrical conductivity positions them as potential alternatives to traditional metal conductors. However, despite the outstanding conductivity of individual SWNTs (10⁵–10⁸ S m⁻¹), bulk SWNT materials do not exhibit a simple additive scaling of conductivity (<10⁵ S m⁻¹) due to various limiting factors. This discrepancy arises from the challenges associated with solution processing and purity, which are critical for translating the intrinsic conductivity of individual nanotubes into macroscopic assemblies. This review provides an overview of recent advancements in methodologies aimed at improving both the solution processability and electrical performance of bulk SWNT materials, with a particular emphasis on purification and sorting strategies. Additionally, we discuss the dual role of dispersants used during SWNT sorting, which facilitate tube de‐bundling but often remain on the nanotube surface as insulating residues, necessitating further processing to fully restore electrical performance. By consolidating recent insights, this review identifies the key mechanisms governing conductivity trade‐off and proposes practical pathways for translating the intrinsic performance of SWNTs into highly conductive bulk assemblies for future electronic applications.

Single‐walled carbon nanotubes (SWNTs) are promising for a wide range of applications. In particular, their exceptional electrical conductivity positions them as potential alternatives to traditional metal conductors. This review identifies the key mechanisms governing their conductivity within composites and proposes practical pathways for translating the intrinsic performance of SWNTs into highly conductive bulk assemblies for future electronic applications.

1. Introduction

Electronics dominate nearly every aspect of people's lives. Conductive materials, the foundation of electronics, are crucial for developing next‐generation devices. Electrical conductivity, defined as the ability of a material to transfer electrical current, is a standard metric used to evaluate the properties of conductors. Based on conductivity values, materials can be divided into three categories: conductors (conductivity over 10² S cm⁻¹), semiconductors (conductivity 10⁻⁸–10² S cm⁻¹), and insulators (conductivity less than 10⁻⁸ S cm⁻¹) [1].

Traditional conductors, including metals like copper, silver, aluminum, and steel, have been the dominant conductive materials for decades. They still hold an important position in modern electronics, such as wiring, power transmission lines, and solar panels. The processing techniques for these traditional conductors have matured over decades of development. However, as demand has increased drastically, natural resources have diminished, resulting in a significant rise in market prices [2]. In addition, metal conductors are prone to surface tarnishing and corrosion under extreme conditions, which shortens their service life and necessitates maintenance.

As the requirement for conductors has increased, new types of conductors, including carbon‐based materials [3], conductive polymers [4], metal oxides [5], superconductors [6], and nanomaterials [7], have emerged. Compared to others, carbon‐based materials (i.e., fullerenes, graphene, carbon black, carbon nanotubes) have attracted significant attention as next‐generation electronic building blocks due to their exceptional stability and mechanical properties. Carbon nanotubes (CNTs) have been a particularly attractive material since their discovery in 1991 by Iijima [8]. CNTs are a honeycomb lattice of sp²‐hybridized carbon atoms rolled into a hollow cylinder with a diameter in the nanometer range and a length that is on the micrometre scale. They are considered quasi‐1D crystals with translational periodicity along the tube axis. CNTs with one wall layer are single‐walled nanotubes (SWNTs), whereas a multi‐walled nanotube (MWNT) has several stacked layers (Figure 1b). The layered structure of MWNTs results in properties that are somewhat different from those of SWNTs, yet both have been employed in the preparation of conductive materials and nanocomposites. However, considering the vast body of work that has been done with both SWNTs and MWNTs, we sought to limit the scope of this review to exclusively focus on SWNTs. Additionally, SWNTs are more uniform in structure, allowing robust comparison between different studies.

FIGURE 1.

(a) Advantages and disadvantages of SWNTs. (b) Schematic illustration of the (7,5) SWNT chiral vector. (c) Schematic of the electronic DOS of m‐ and sc‐SWNTs. (d) Illustration of different contributions to conductance or resistance in dense networks of SWNTs. (e) Illustration of general methods toward the preparation of higher purity SWNTs with enhanced electrical conductivity.

Since the discovery of their unique electronic properties, SWNTs have shown great potential for constructing next‐generation electronic devices. Compared to traditional metal conductors, SWNTs are advantageous in several respects (Figure 1a):

• Density: SWNTs have a reported density range of 1.3–2.1 g cm^{− 3}, which varies depending on the synthesis method and the technique used for density measurement [9, 10, 11].

• Stability: SWNTs exhibit extraordinary environmental stability, which makes them more tolerant to extreme conditions, including ultra‐high temperatures [12, 13] and chemically corrosive conditions [14]. For example, results demonstrated by Tomkiewicz et al. showed that the failure time of SWNT materials is 108 times improved compared to copper wire in a saltwater environment under an applied electrical bias [15].

• Mechanical strength: The tensile strength of SWNTs ranges from 25 to ∼100 GPa, depending on chirality and defects [16, 17, 18].

• Carrier mobility: The charge mobility of SWNTs can be as good as metals [19], making them ideal materials for nanoscale devices.

• Market price: Opposite to the price trend for copper [2], the cost of manufacturing SWNTs at an industrial scale has decreased drastically and is predicted to continuously decrease in the future due to the growing interest in SWNT research [20].

However, significant technical challenges still exist in transferring the superior properties of individual SWNTs to highly conductive bulk SWNT materials, such as continuous wires, cables, and sheets. The limitations stem from several characteristics of commercial SWNT samples (Figure 1a):

• Low purity: As‐prepared SWNTs contain not only a broad distribution of different tube chirality, length, and diameter but also impurities such as catalyst particles and amorphous carbon.

• Poor solution processibility: The π‐electrons in the sp²‐hybridized carbon atoms of SWNTs result in strong π–π stacking interactions, leading to the formation of aggregated SWNT bundles. This aggregation significantly reduces their solubility in most common organic and aqueous solvents, posing challenges for their processing and application.

• Junction resistance: While conduction of electrons along a single tube is highly efficient, electron transfer between neighboring tubes occurs through a hopping or tunneling mechanism that is much less efficient, generating resistance.

Although individual SWNTs exhibit extraordinary conductivity, the conductivity of bulk SWNT materials does not simply scale with the sum of individual SWNT conductivities due to the above‐mentioned limitations. This may seem counterintuitive, but careful processing is required to transform a highly conductive material at the microscale into a similarly conductive bulk material at the macroscale. This review aims to provide insight into recent methodologies and potential mechanisms for enhancing both the solution processability and electrical conductivity of bulk SWNT assemblies, with a particular focus on purification and sorting strategies. Specifically, purification strategies target the removal of production‐derived impurities, while sorting/separation strategies further enhance the material purity by enriching nanotubes of desired chirality, diameter, or length. The discussion will begin with an overview of the conduction mechanisms of individual SWNTs and SWNT networks. It will then examine the factors influencing overall conductivity and explore sorting methods that could potentially enhance the material's conductivity.

2. SWNT Electrical Conductivity

2.1. Basic Properties of Individual SWNTs

A good understanding of the conduction mechanism in SWNTs is essential in order to achieve a desired goal. In general, the conductive properties of SWNTs, similar to water flowing through a pipe, are accomplished by electrons moving along a discrete tube surface. As mentioned in the introduction, the structure of a SWNT can be visualized as a single layer of graphite rolled into a cylindrical shape. As there are multiple ways to roll the graphene sheet, a wide variety of SWNTs with distinct structural parameters (i.e., chirality and diameter) and bandgaps can be formed.

SWNTs can be categorized according to their chirality, electronic type, and handedness. The chirality of SWNTs can be denoted based on the graphene lattice vectors. As shown in Figure 1b, the circumferential vector (C _h) is the chiral vector that connects two crystallographically equivalent sites on a graphene sheet (i.e., when the graphene sheet is rolled up to form a tube, atoms A and A’ become superimposed). The direction perpendicular to C _h is the tube axis. The SWNT geometry is defined by the equation C _h = n a₁ +m a₂ , where a₁ and a₂ are unit vectors of the hexagonal honeycomb lattice of graphene, and a pair of integers (n, m) represents the relative position of two atoms on a flat strip of graphene. The SWNT is metallic (m‐) when n–m = 3i (i is an integer) or when n = m (armchair), otherwise it is semiconducting (sc‐). The diameter of the SWNT can be calculated based on Equation (1):

$$d_t=\frac{\left|C_h\right|}{\pi }=\frac{a}{\pi }\times \sqrt{n^2+nm+m^2}\left(a=\sqrt{3}\times a_{c-c};a_{c-c}=1.42\mathrm{\AA }\right)$$ (1)

where a is the lattice constant, and a_c‐c is the C‐C bond length. SWNTs can be either right‐handed (P, or +) or left‐handed (M, or −). As the handedness does not affect the electronic properties of a SWNT, this review will not elaborate on the topic of handedness.

SWNTs are composed of continuously bonded carbon atoms, in which each carbon is connected to three other carbon atoms through σ bonds, and the remaining p‐orbital interacts with neighboring p‐orbitals to generate delocalized π (bonding) and π* (anti‐bonding) orbitals. The p‐orbital, perpendicular to the tube surface, can interact with other SWNT p‐orbitals, resulting in a π–π interaction that leads to the formation of SWNT bundles. The electronic behavior of SWNTs is also closely related to the energy level of π and π* bands. The calculated electronic density of states (DOS) of SWNTs shows characteristic peaks, known as van Hove singularities, which arise due to the 1D quantum confinement of the tube (Figure 1c). The Van Hove singularities in the valence and conduction bands are distinct for each (n, m) tube. When the valence and conduction bands are in contact, the tube exhibits metallic conduction, whereas sc‐ tubes have an energy gap between the two bands. Statistically, about 1/3 of nanotubes are m‐SWNT, and 2/3 are sc‐SWNT.

However, the DOS calculation does not account for SWNT curvature, which is induced by its cylindrical structure [21]. In this respect, only the armchair SWNT is truly m‐ type with a zero‐bandgap. Tubes with n–m = 3i have a very small bandgap and are referred to as quasi‐metallic. They can still be considered metallic at room temperature since their bandgap is negligible. Moreover, the curvature effect is more significant in tubes with small diameters, where the rehybridization between σ and π orbitals can affect their electronic structure. The energy gaps of these ultra‐narrow SWNTs can be more than 50% lower than predicted, making them exhibit unusual properties such as superconductivity [22]. However, this effect is not pronounced for tubes with a diameter greater than 1 nm.

Until now, all calculations assumed the SWNT structure was defect‐free. However, this assumption is unrealistic. Defects, such as vacancies, pentagons, and impurities, will significantly affect the local electronic properties [23]. Conduction within a single SWNT is highly chirality‐dependent and is a combination of ballistic and diffusive conduction [24]. Ballistic conduction happens when the tube is defect‐free, and the electron can travel without scattering. This process can be interrupted by impurities and tube defects to cause electron diffusion [25].

The dimensions of SWNTs also play a crucial role in defining their conductivity. According to theoretical predictions, armchair SWNTs with a large diameter exhibit a longer mean free path, which results in better conductivity [26]. Thus, longer SWNTs with larger diameters exhibit superior conductivity and, therefore, selective enrichment or production of m‐SWNTs with these characteristics is a reliable strategy for achieving higher conductivity.

However, this does not imply that sc‐SWNTs cannot be used in high‐performance electronics. Numerous studies have shown that sc‐SWNTs can also exhibit superior conductivity upon treatment with various moieties. For example, the introduction of palladium can make sc‐SWNTs behave like ballistic m‐tubes [25]. Moreover, doping semiconducting SWNTs with n‐ or p‐type molecular dopants can effectively tune their bandgap and align their energy levels with those of metal counterparts. Molecular dopants can form charge‐transfer complexes with SWNTs on either the inner or outer tube surface, significantly enhancing electronic conductivity while largely preserving structural integrity. This approach offers a promising route to further enhance the conductivity of sorted SWNT samples [27]. Notably, sc‐enriched samples exhibit greater sensitivity to molecular dopants, leading to the largest conductivity enhancement relative to both pristine and m‐enriched ones [28].

In addition to intentional doping, unintentional doping can occur during SWNT handling and sorting. Procedures such as sonication in solvents [29] and pH variation [30] have been shown to induce doping effects, with acidic conditions leading to reduced electrical resistance. Since counterion‐SWNT complexation is inherently reversible, gradual oxidation (de‐doping) can occur upon exposure to air, particularly in n‐doped systems. While this reversibility has been advantageously exploited for pH sensing applications [31], it poses a challenge for the development of stable, high‐performance electronic devices. Strategies to enhance dopant stability using various doping methods have recently been reported and comprehensively reviewed [32, 33, 34].

It is widely acknowledged that individual sc‐SWNTs have a calculated conductivity of 10⁵ S m⁻¹, whereas the conductivity of m‐SWNT is 10⁸ S m⁻¹ [35, 36, 37]. The calculation is based on Equation (2):

$$\sigma =\frac{G\times L_{SWNT}}{\pi \times \left(r_{out}^2-r_{in}^2\right)}$$ (2)

where G is the measured conductance, L_SWNT , and $r_{\mathrm{out}}^2-r_{\mathrm{in}}^2$ represents the length and wall thickness of the SWNT. Nevertheless, these values should only be considered as a general reference, since tubes with different chirality, diameter, length, and extent of structural defects exhibit distinct conductivity.

2.2. SWNT Network Conductive Path and Conductivity

The electronic conduction mechanism of an individual SWNT is very different from that of a SWNT bundle or network (Figure 1d). Within a dense nanotube network, assuming that percolation is not a limiting factor, the movement of electrons occurs along the SWNTs, across tube‐tube junctions (tubes with the same chirality or with a different charity and distinct bandgap) [38, 39], via local scattering (induced by tube defects), and through field screening [40] (created by aggregated SWNT bundles). In this context, numerous situations can arise, rendering the electron transport behavior unpredictable with any specific analytical model [41].

Electron transport through inter‐tube junctions can be accomplished by hopping and tunneling. The contact resistance of a tube–tube and tube‐bundle junction is between 29 to 532 kΩ; [42] these values are proportional to the tube/bundle diameter, density, and the relative angle between tubes [43]. When the tubes are aligned parallel to each other (0° relative angle), the p‐orbitals, which contribute to the electron transport, have the strongest interaction, resulting in maximum electron transfer efficiency. For the aligned SWNT network, the average length of SWNTs plays a crucial role: short tube networks (length <100 µm) show an eightfold higher resistance than long tube networks (>300 µm) [44].

The conductivity (ρ) of SWNT network films can be directly measured using instruments such as four‐point probes or calculated through resistance (R) by Equation (3):

$$\rho =R\times \frac{A}{l}$$ (3)

where A is the surface area, and l is the film thickness.

In summary, theoretical calculations have provided useful guidance for experimental work on conductive SWNT‐based materials. It is now clear that improvement of conductivity within SWNT composites can be achieved through the following general directions (Figure 1e):

1. Selective production/enrichment of tubes with single chirality, diameter, and length.

2. De‐bundling SWNTs by dispersion in solvents.

3. Minimizing the insulating effect caused by impurities or dispersants.

Over the past two decades, a large amount of work has focused on purification and sorting of SWNTs, with the resulting materials showing improved performance in applications such as field‐effect transistors [45, 46, 47, 48], transparent flexible electrodes [49, 50, 51, 52], and sensors [53, 54, 55, 56, 57]. The purity of SWNTs can be controlled during the synthesis or through postprocessing, which is thoroughly discussed below.

3. Selective Synthesis

In general, the production of SWNTs through techniques, including high‐pressure carbon monoxide disproportionation (HiPco) [58], plasma‐torch growth [59], chemical vapor deposition (CVD) [60], and arc discharge [61], have enabled their commercial availability in bulk quantities. However, SWNTs synthesized using different methods exhibit significant variations in their properties [62]. The tube diameters, bandgap, length, and production cost vary with the preparation technique.

Among these methods, CVD has attracted the most attention due to its ability to directly grow well‐aligned, high‐quality SWNTs on a substrate, whereas other techniques often require post‐processing to enhance alignment and purity. In particular, the CoMoCAT catalyst enables the large‐scale synthesis of small‐diameter tubes. Despite its advantages, CVD still faces challenges in achieving precise structural control of SWNTs in large quantities. This method produces SWNTs by decomposing various carbon sources on transition metal catalysts supported on heated substrates. Further improvements in the controlled synthesis of specific SWNT structures can be achieved by optimizing the catalyst, substrate, temperature, and carbon source [63].

Catalysts with distinct chemical structures, particle sizes, surface morphology, and stability can result in the formation of different SWNT species [67, 68, 69]. To date, reported catalysts include cobalt (Co) [63, 64, 70, 71, 72, 73, 74, 75, 76, 77, 78], ruthenium (Ru) [70, 79], rhenium (Re) [66, 80, 81], iron (Fe) [82, 83], and alloy‐based catalysts [47, 84]. Most of the catalysts mentioned above either exhibit no obvious selectivity or primarily favor the growth of sc‐SWNTs. Notably, the Li group successfully demonstrated the use of W₆Co₇ as a catalyst to produce (12, 6) SWNTs with 92% purity (Figure 2a) [64]. Additionally, some researchers have explored catalyst modifications to enrich m‐SWNTs. For instance, Avetik et al. employed a noble gas ambient atmosphere (He/H₂/H₂O) during the thermal annealing of Fe nanoparticle catalysts, in combination with oxidative and reductive species, to increase the m‐SWNT fraction to a maximum of 91% [85]. The size and surface morphology of the catalyst have also proven critical for m‐SWNT enrichment. For example, Zheng et al. demonstrated the growth of >90% pure (12, 6) and >80% pure (8, 4) SWNTs by precisely controlling the symmetry of the catalyst surface [86]. Liu et al. demonstrated an effective template to achieve chirality control by exposing a sixfold symmetric face of a liquid “Trojan” catalyst (111), producing a highly enriched m‐SWNT sample with a purity of 90% (Figure 2b) [65]. Toshiyuki et al. demonstrated purity enhancement of m‐SWNTs from 50% to 90% after refining the catalyst particle size [87, 88]. Similarly, the SiO_x catalyst with a smaller diameter distribution was found to selectively enrich the growth of m‐SWNT (∼80%) [89]. Moreover, the kinetically controlled growth of m‐SWCTs by manipulating their binding energy with the catalyst and promoting their growth rate was demonstrated by Li et al. using CoRe₄ as the catalyst. This method significantly enriched m‐SWNT (>80%) and narrowed their diameter distribution (Figure 2c) [66].

FIGURE 2.

Strategies for selective synthesis of sc‐ and m‐SWNTs via catalyst engineering with significant work in (a) solid alloy catalysts [64], (b) solid “Trojan” catalysts [65], and (c) CoRe4 nanoparticles [66]. (d) Introduction of etching agents. (e) Total synthesis of SWNT with (e) buckybowl cap, (f) [n]cycloparaphenylene belt, and (g) short SWNT as reaction seeds.

The catalyst can either be supported by a substrate or exist as a floating catalyst (FC‐CVD) without support from the substrate. The advantage of FC‐CVD is that the as‐prepared SWNT can be directly collected with different morphologies, such as film [90], transparent electrode [91], aerogel [92], and yarns [93]. However, this method has a higher requirement for catalysts than traditional CVD. Frequently mentioned catalysts include Fe‐ [94, 95] and Co‐based [96] organometallic molecules, which decompose at relatively low temperatures to form nanoparticles with high catalytic activity. Since this review primarily focuses on purification and sorting methodologies, a detailed discussion of specific catalysts and their influence will not be included. Readers interested in a more in‐depth exploration of this topic are encouraged to refer to a recently published comprehensive review, which covers the fundamentals of catalyst design for the controlled synthesis of chiral SWNTs [97].

Selective production of sc‐/m‐SWNTs can also be achieved by either inhibiting the growth of the undesired type or selectively etching it away. Under an oxidizing environment, such as a mixture of methanol (MeOH) and ethanol (EtOH) [98], isopropanol (i‐PrOH) [99, 100], H₂O vapor [101], ZrO₂ [102], plasma [103], UV light [104], or a combination of different carbon sources [105], the growth of m‐SWNT is largely suppressed. Similarly, the yield and chirality selectivity of the FC‐CVD can be enhanced by the addition of oxidants such as H₂O [106], NH₃ [107], CO₂ [90], oxygen [108], and a combination of different oxidants [109]. On the contrary, selective enrichment of m‐SWNTs was found when using etchants such as SO₃ [110] and NO₂ [111] (Figure 2d). Moreover, some research indicates that the changing of precursor source and composition would create an etching effect during the production process and increase the content of m‐SWNTs (Figure 2d) [112, 113].

From an organic chemistry perspective, the synthesis of SWNTs with a specific chirality may theoretically be possible through total synthesis [114]. This approach requires the preparation of a seed, such as a buckybowl cap (Figure 2e) [115], [n]cycloparaphenylene aromatic belt (Figure 2f) [116], or an end‐functionalized short SWNT (Figure 2g) [117], followed by the stepwise elongation of the nanotube through the sequential addition of appropriate building blocks. The elongation process can be achieved entirely through organic synthesis via reactions such as the Diels–Alder reaction [118] or metal‐free CVD conditions [119, 120] (Figure 2e). Buckybowl caps can be synthesized either through the controlled opening of C₆₀ [121] or through organic synthesis [115, 122, 123]. The Amsharov group pioneered the controlled synthesis of m‐SWNTs by “programming” the precursor structure and initiating the growth of SWNTs with a single helicity from large‐diameter hemispherical buckybowls. This method successfully yielded (6, 6), (8, 8), (10, 10), (12, 12), and (9, 0) SWNTs in a one‐pot synthesis (Figure 2f) [115, 122, 123]. Additionally, short SWNTs with open ends, obtained by fragmenting longer SWNTs, have been used as “seeds” for the growth and cloning of ultra‐long SWNTs on quartz substrates, producing a broad range of SWNTs with various chirality (Figure 2g) [124, 125].

In summary, the controlled synthesis of SWNTs offers several advantages, particularly in producing nanotubes that are generally free from severe structural defects. However, while the controlled synthesis of mono‐chiral SWNTs is achievable, current methods struggle to balance purity, quality, and scalability [126]. The purity of synthesized SWNTs remains significantly lower than what can be achieved through post‐synthesis separation methods. Furthermore, compared to the synthesis of sc‐SWNTs, the enrichment of m‐SWNTs during synthesis has been less frequently reported [127]. While highly pure sc‐SWNTs are desirable for certain applications, the exceptional electrical conductivity and stability of m‐SWNTs make them equally valuable [128]. The lack of precise control over chirality, diameter, and length during synthesis further complicates these challenges, making post‐synthesis purification and separation indispensable. In the following section, we will explore purification techniques that effectively remove production impurities and enable the selective enrichment of SWNTs with specific structural characteristics.

4. Purification Strategies for Eliminating Production Impurities

Depending on the production method, the as‐prepared SWNT samples contain a great variety of undesired impurities, such as metal catalysts, amorphous carbon, different graphitic nanoparticles, and fullerenes, which makes it challenging for them to be directly used in practical applications.

Initially, the purification of as‐prepared SWNTs is accomplished through physical methods, where the raw material is subjected to thermal annealing [132], filtration [129], chromatography [133], and extraction [134] to remove the production impurities [135]. These methods are relatively gentle, as they exploit differences in size, aspect ratio, and density between SWNTs and impurities to achieve isolation, minimizing damage to the tube structure. As shown in Figure 3a, after the removal of production impurities, larger impurities, such as aggregates and graphitic sheets, can be removed efficiently. However, other impurities such as metal catalysts and amorphous carbon remain unseparated.

FIGURE 3.

(a) TEM and SEM images of raw (left) and purified (right) SWNTs. Reproduced with permission [129]. Copyright 1998, Elsevier Science B.V. (b) The enhanced performance of transparent conductors produced using predominantly m‐SWNTs compared to unsorted SWNTs. Reproduced with permission [130]. Copyright 2008, American Chemical Society. (c) A plot illustrating the electrical conductivity of composites as a function of SWNT concentration corresponding to SWNT percolation. (d) Conductivity as a function of surface concentration for length‐sorted SWNTs. Reproduced with permission [131]. Copyright 2008, American Chemical Society. (e) Methods and dispersants for solubilizing SWNT.

Metal catalysts are particularly difficult to remove because they are either tightly attached to the ends of SWNTs or trapped inside the tubes. Amorphous carbon, on the other hand, is less stable than SWNTs under chemical treatment due to the higher prevalence of dangling bonds and defects [136]. Taking these considerations into account, a chemical oxidation procedure was developed. The as‐prepared SWNTs undergo gas‐phase oxidation using air, O₂, Cl₂ [137], or humid air, followed by refluxing in a concentrated acid solution (e.g., HCl, HNO₃). Finally, repeated washing with H₂O yields purified SWNTs [138, 139, 140]. During these steps, amorphous carbon is first removed through oxidation, as it reacts more quickly than SWNTs in the presence of oxidants. Metal catalysts are then eliminated by dissolving them in concentrated acid solutions. Compared to liquid oxidants, gas‐phase oxidation is less effective due to limited contact, making concentrated acid treatment necessary to remove remaining amorphous carbon. However, a major drawback of this procedure is the unavoidable damage to the intrinsic structure of SWNTs. For instance, the end‐caps of SWNTs [141], which contain pentagonal rings and are less stable, are prone to oxidative opening, which is followed by uncontrollable sidewall oxidation. Modifications of this procedure, such as ultrasonic and microwave‐assisted acid treatments, help remove impurities while minimizing damage to the tubes [142].

After removing production impurities, SWNT mixtures still contain a diverse range of nanotubes varying in chirality, diameter, and length. Due to their similar properties, precise isolation methods are required. The following sections provide a comprehensive description of post‐synthesis purification techniques for efficiently separating SWNTs based on bandgap, chirality, diameter, and length.

5. Sorting Strategies to Enrich Specific Chirality, Length, and Diameter

5.1. Importance of Purity and Sorting Methods

As discussed in Section 2 above, theoretical studies indicate that charge transfer between tubes with distinct band gaps is orders of magnitude lower than that between tubes with a narrow bandgap distribution. This has also been confirmed experimentally. In 2003, Chattopadhyay et al. reported that the surfactant octadecylamine (ODA) efficiently precipitated m‐SWNTs at high concentrations, yielding an m‐SWNT‐enriched sample with 4.5 times higher conductivity (0.78 S cm⁻¹) than its sc‐SWNT counterpart (0.17 S cm⁻¹) [143]. Later studies found that SWNT‐based transparent electrodes enriched with m‐SWNTs via density gradient ultracentrifugation (DGU) exhibited enhanced conductivity and improved transmittance compared to unsorted SWNTs (Figure 3b) [130]. Notably, m‐SWNTs with diameters ranging from 1.4 to 1.6 nm are particularly well‐suited for building transparent electrodes. Their optical transmittance is maximal around 550 nm, aligning with the wavelength range where human vision is most sensitive and solar radiation is most intense.

The conductivity is also largely affected by tube diameter: tubes with a larger diameter exhibit higher conductivity than smaller diameter tubes. Experimental results obtained by Yanagi et al. show that m‐SWNTs with different diameters (1.34, 1.0, and 0.84 nm) exhibit diameter‐dependent surface resistance (65 Ω sq⁻¹ for 1.34 nm, 450 Ω sq⁻¹ for 1.0 nm, 1 kΩ sq⁻¹ for 0.84 nm) [144]. Similarly, Tyler et al. found that diameter‐refined m‐SWNT films with larger‐diameter tubes (1.6 nm) had half the resistance of smaller‐diameter tubes (1.25 nm) and seven times lower resistance than unsorted films [145].

The length of SWNTs is another crucial factor influencing conductivity, primarily by affecting the percolation threshold in a SWNT network (Figure 3c). As the SWNT concentration increases, the percolation network becomes denser, and conductivity improves exponentially until it reaches a plateau [146]. This critical concentration is referred to as the percolation threshold. Both theoretical and experimental studies show that, at a given surface concentration, tubes longer than 200 nm form more transparent and conductive networks, highlighting the importance of tube length distribution on conductivity (Figure 3d) [131].

These experimental results align with theoretical predictions, emphasizing the need to selectively enrich large‐diameter, long m‐SWNTs for high‐conductivity materials. Purifying SWNTs remains challenging due to their highly conjugated nature, insolubility, and tendency to form bundles in most solvents. While ultrasonication in solvents like N‐methyl‐2‐pyrrolidone (NMP) and N, N‐dimethylformamide (DMF) can temporarily disperse them, the solutions are typically low in concentration and unstable, with SWNTs precipitating over time. Moreover, high‐energy sonication can shorten SWNTs and introduce defects, degrading their intrinsic properties (Figure 3e) [147].

Various strategies have been developed to improve the solution processability and purity of SWNTs [62, 148, 149, 150, 151]. The established methods fall into three general categories: covalent, non‐covalent, and endohedral filling. Unlike covalent modification, non‐covalent functionalization preserves the tube's conjugation and intrinsic properties. Endohedral filling modifies the inner tube space, altering the nature of the SWNTs. Although the efficiency and scalability of these methods vary, many have achieved notable success, with some enabling SWNT sorting in a single dispersion step (Figure 5). However, highly purified SWNTs often require further sorting using techniques such as DGU, chromatography, or aqueous two‐phase extraction (ATPE) to achieve the desired level of purity.

FIGURE 5.

(a) Summary of general strategies for optimizing biomolecule‐SWNT dispersion. (b) Chemical structure of the A, T, C, G base pairs and their robust complementary interactions. (c) Single‐point scanning mutation to evolve recognition sequences for m‐SWNT, where sequence C enhances recognition of (7, 7) [204]. (d) Selection process for efficiently screening a sequence library of 10¹⁸ ssDNA sequences on SWNT surfaces [208].

As discussed in the previous sections, tube diameter is related to chiral vectors, and length can also be enhanced using similar techniques with modifications. Thus, all three parameters, diameter, length, and chirality, will be discussed, with a focus on separation techniques. Handedness‐based separation is not included, as it doesn't impact electronic properties.

5.2. Covalent Functionalization

Covalent surface functionalization involves attaching chemical groups to the SWNT surface through covalent bonds. Although this method typically requires harsh reaction conditions, it offers permanent modifications, ensuring reproducibility. Theoretical studies suggest that the energy and chemical reactivity of individual SWNTs depend on their chirality and diameter [152, 153, 154, 155]. Several studies have demonstrated that covalent functionalization can be used to selectively purify SWNTs. However, most of the covalent functionalization methods irreversibly disrupt the π‐conjugation of the tube surface. Setaro et al. reported a non‐destructive [2+1] cycloaddition approach that enables functionalization of SWNT on a gram scale [156]. This method reconstructs the extended π‐network, thereby preserving the exceptional quantum optoelectronic properties of SWNTs. However, recent progress in this area has been limited. As discussed in previous sections, surface defects significantly alter the intrinsic electronic properties of SWNTs, making covalent functionalization unsuitable for many applications. Given these considerations, this review will not explore this topic further. Readers interested in a more detailed discussion can refer to a recently published review [157].

5.3. Non‐Covalent Functionalization

Non‐covalent surface functionalization preserves the tube's conjugation, unlike covalent methods. Molecules interact with the SWNT surface through van der Waals forces, π–π stacking, and hydrophobic interactions. Small molecules like surfactants and polycyclic aromatic compounds adsorb onto the surface, while long‐chain molecules such as polymers and DNA wrap around the tubes, forming stable supramolecular assemblies. This enables SWNTs to disperse in organic solvents and aqueous media, broadening their applications.

5.3.1. Surfactants

A surfactant is a molecule that mediates between two incompatible phases or materials. It typically has a polar/hydrophilic “head” and a nonpolar/hydrophobic “tail”. Based on the charge state of the head, surfactants are categorized into ionic and non‐ionic types. Ionic surfactants are further divided into anionic, cationic, and amphoteric (zwitterionic) categories (Figure 4).

FIGURE 4.

Categorization and summary of commonly employed surfactants.

Anionic surfactants, characterized by a negatively charged head group, are most widely used. Examples include sodium dodecylsulfonate (SDS) [158, 159, 160, 161, 162, 163], sodium dodecylbenzenesulfonate (SDBS) [164, 165, 166, 167, 168, 169], sodium deoxycholate (DOC) [170, 171], and sodium cholate (SC) (Figure 4) [171]. SC and DOC are also referred to as bile salts, as they are produced in the liver and stored in the gall bladder, playing an important role in digestion. In contrast, cationic surfactants possess a positively charged head group, with common examples being cetylpyridinium chloride (CPC) and dodecyl‐, tetradecyl‐, or cetyl‐trimethylammonium bromide (DTAB/TTAB/CTAB) (Figure 4) [172, 173]. Amphoteric surfactants, which contain negative and positive charges on their head group, exhibit pH‐sensitive behavior. They act as anionic surfactants under basic conditions, cationic surfactants under acidic conditions, and resemble non‐ionic surfactants under neutral conditions. Non‐ionic surfactants, such as Brij [171], Triton X‐series [171, 174, 175, 176, 177], Pluronic [171, 178, 179], polyvinyl pyrrolidone (PVP) [171, 178], and Tween 60/80 [171, 180, 181, 182], are superior as they are less affected by pH variations (Figure 4). Surfactants interact with the tube surface in different morphologies, such as cylindrical micelles, hemi‐micellar assemblies, and via random absorption, depending on the tube diameter, surfactant structure, and their relative concentration [160].

A tremendous amount of work has been done on surfactant‐SWNT dispersion optimization [183]. Variation of parameters such as concentration [158, 171, 175, 184, 185, 186], temperature [185], sonication conditions [168], pH [187, 188], and the inclusion of co‐surfactants [172, 189] all showed enhancement of dispersion concentration and long‐term stability [158]. Surfactants with varying binding affinities [159, 162, 167, 173, 190] and surface coverage [165, 176, 191] can produce high‐concentration SWNT dispersions in aqueous media, but their selectivity for specific SWNT types is limited. Therefore, surfactants are typically used as an initial step in SWNT sorting, mainly to enhance solution processability. Subsequent separation steps, such as DGU, ATPE, and chromatography, are required to enhance the purity of specific SWNT species, and these methods will be discussed below.

5.3.2. Biomolecules

Biomolecules are another distinct class of adsorbents capable of effectively dispersing SWNTs in aqueous solutions. The most widely studied biomolecules are DNA [192, 193], oligopeptides [194, 195, 196], and proteins (Figure 5a) [197, 198]. Non‐covalent functionalization with biomolecules not only stabilizes SWNTs in aqueous media but also imparts unique functional properties, such as biocompatibility and targeted binding capabilities.

Single‐stranded DNA (ssDNA) is a highly promising surfactant for wrapping SWNTs, as it contains both a hydrophilic backbone and hydrophobic nitrogenous bases—adenine (A), cytosine (C), guanine (G), and thymine (T) (Figure 5b). In early studies, the ability of ssDNA to selectively disperse pure or enriched single‐chirality SWNTs was quite limited, often requiring additional chromatographic separation steps to enhance selectivity [199, 200]. Later research revealed that the base composition and sequence of ssDNA can critically influence the strength and specificity of its interaction with SWNTs, as well as the conformation and packing density of ssDNA molecules on the nanotube surface [201, 202]. Initial sequence screening highlighted that sequences with alternating G and T repeats, d(GT)_n (where n = 10–40), demonstrated superior separation performance compared to other homogeneous DNA sequences and base combinations [203]. Tu et al. proposed that certain ssDNA sequences can form intricate 3D folding structures on the SWNT surface, creating “recognition sequences” that enable chirality‐dependent sorting of SWNTs through chromatographic separation (Figure 5c) [204, 205].

Molecular dynamics (MD) simulations have been widely employed to explore the structural intricacies of DNA‐wrapped SWNTs. However, despite these efforts, identifying effective resolving sequences for chiral sorting has proven to be a costly, time‐intensive, and inefficient process, significantly hindering progress in the field. The absence of systematic strategies or predictive models has further exacerbated these challenges, making it difficult to optimize sequence discovery. To address these limitations, recent advancements have turned to machine learning (ML) techniques, which have shown remarkable potential in enhancing the chiral sorting efficiency. By analyzing extensive datasets and uncovering previously hidden patterns, ML approaches have successfully identified resolving sequences capable of isolating 22 pure single‐chirality species across a broad range of chirality with high enantiomeric purity [206]. These methods have dramatically improved both the number of resolving sequences (from 10² to 10³ sequences) and the success rate of finding them (from ∼10% to over 90%). Furthermore, ML has provided critical insights into the molecular recognition mechanisms governing DNA‐SWNT interactions, enabling the development of more efficient and targeted sorting strategies [207].

In addition to sequence‐dependence, studies also showed that the length of ssDNA is crucial for achieving length‐sorting of SWNTs [209]. For example, d(GT)₃₀‐wrapped CoMoCAT SWNTs with an average length of 500 nm [210], while d(GT)₂₀₀ yielded significantly longer tubes (1–4 µm) [211]. Leveraging the power of ML, subsequent research has further scaled up the sorting sequence library to 10^{1 8} sequences, allowing for the systematic selection of high‐affinity long DNA sequences (Figure 5d) [208]. Beyond designing ssDNA for specific sorting purposes, Zheng's group developed a length‐based SWNT separation approach using PEG‐induced micro‐cluster formation. By gradually increasing PEG concentration, shorter SWNTs required a higher PEG concentration to cluster than longer SWNTs, enabling length sorting through this crowding effect [212].

Although significant progress has been made in sorting SWNTs with well‐designed ss‐DNA sequences, the efficiency is still low, and the synthesis of ss‐DNA is costly. Genomic DNA offers a cost‐effective alternative to synthetic DNA for SWNT sorting [213]. Kim et al. showed that salmon genomic DNA can disperse and selectively enrich (6, 5) SWNTs, though selectivity was impacted by the m‐ tube content in the raw sample [214, 215]. Progress in this area has been limited, as the random base sequences in genomic DNA lack the reproducibility of synthetic ssDNA.

Peptides and proteins, which interact with SWNTs through hydrophobic and π–π interactions, are also effective dispersants. Their dispersibility depends not only on sequence but also on their 2D and 3D structures [216]. Proteins rich in basic residues, such as lysozyme (LSZ) and histone (HST), disperse SWNTs more efficiently at intrinsic pH compared to other common proteins. In contrast, bovine serum albumin (BSA) performs better under acidic/basic conditions. Notably, all these proteins showed a preference for m‐SWNTs [198]. β‐lactoglobulin (β‐LG) has been found to selectively disperse SWNTs with specific diameters (D < 4 nm and D > 30 nm) due to the presence of two possible binding sites that correspond to specific tube curvatures, which in turn relate to different tube diameters [217].

Similar to the challenges in optimizing DNA sequences for improved SWNT chiral sorting, designing peptide and protein structures to achieve specific sorting requirements is both costly and inefficient. However, advances in computational protein design, guided by selection rules, have enabled the creation of tailored structures for selective SWNT interactions [218]. Additionally, MD studies showcased that flanking residues play a critical role in chirality‐based separation. The results indicate that threonine, when flanked by positively charged lysine, preferentially stabilizes m‐SWNTs, whereas glycine residues interact with both sc‐ and m‐species [219]. These findings offer valuable insights into the interactions between SWNTs and proteins/peptides, paving the way for future molecular designs tailored to specific requirements.

In summary, biomolecules can interact with SWNTs, facilitating their de‐bundling and solubilization in aqueous media. Recent studies have shown promising results in using ML to screen ssDNA, protein, and peptide sequences for sorting SWNTs by chirality and length. However, this approach remains time‐consuming and cost‐inefficient. Combining ssDNA with other sorting techniques is highly desirable for achieving efficient SWNT isolation.

5.3.3. Small Molecules

Small molecules, such as polycyclic aromatic compounds, can interact with SWNTs through π–π stacking. These compounds typically consist of two key components: an aromatic moiety and a hydrophilic moiety. The aromatic moiety binds to the nanotube surface via π–π interactions, while the hydrophilic moiety ensures solubility in aqueous solvents [220]. As shown in Figure 6a,b, small molecule dispersants include pyrene derivatives [221, 222, 223, 224, 225], thienothiophene derivatives [226], N‐annulated perylene derivatives [225], perylene diimide derivatives [221, 224], cyclic aromatic compounds [227], and natural/synthetic fluorophores/dyes have been demonstrated to disperse SWNTs in organic media. Notably, fluorophores and dye molecules, such as Purpurin [50], Rhodamine B [228], and riboflavin analogues [229], not only solubilize SWNTs, but also impart intriguing optical properties to them, enabling various sensing applications [228].

FIGURE 6.

The non‐covalent functionalization of SWNTs with (a) aromatic molecules, (b) fluorophore/dye, and (c‐d) molecular tweezers (images reproduced based on refs. [230] and [231]). The aromatic regions interact with the nanotube surface through π–π stacking to form stable dispersions while preserving the intrinsic structure of SWNTs.

While most reported small molecules lack significant selectivity toward specific SWNT species, there are notable exceptions demonstrating SWNT sorting or enrichment. Li et al. identified an aromatic cyclic Schiff base that exhibits selectivity toward certain sc‐SWNTs, including (7, 5), (8, 6), (12, 1), and (9, 7) (Figure 6a) [227]. Purpurin, a natural anthraquinone dye, has been shown to efficiently exfoliate SWNTs in aqueous NaOH solutions, selectively enriching m‐SWNTs and enabling the fabrication of highly conductive buckypaper with a conductivity of 3, 726 S cm⁻¹ [50]. Additionally, Ishimaru et al. demonstrated that commercially available riboflavin tetrabutyrate can selectively disperse sc‐SWNTs, with sorting efficiency being concentration‐dependent. At lower concentrations, it effectively enriched (8, 6) and (8, 7) SWNTs (Figure 6b) [229].

Perylene derivatives have also demonstrated selectivity in certain cases. Ernst et al. reported a perylene‐imido‐diester compound with hydrophobic and hydrophilic tails that solubilizes and individualizes SWNTs in water, enriching specific sc‐SWNTs such as (9, 5), (10, 3), and (11, 1) [232]. El‐Refaey et al. introduced a π‐molecular tweezer, a U‐shaped molecule with two aromatic perylene diimide groups connected by a dimethyl phenyl group and short alkyl chains, that efficiently sorts SWNTs with diameters around 0.9 nm ((9, 4) and (7, 6)) at a 90% sorting rate (Figure 6c) [230]. When the ends of the U‐shaped molecule were closed to form interlocked macrocycles around SWNTs, the diameter‐selective properties further improved [233, 234, 235, 236, 237, 238, 239]. A related approach, involving mechanically interlocked nanotubes, utilizes ring‐opening metathesis polymerization to form a polymer decorated with multiple U‐shaped structures. These molecules were subsequently ring‐closed around SWNTs using ring‐closing metathesis. The resulting hybrids featured highly individualized SWNTs and demonstrated significantly enhanced mechanical properties compared to the matrix polymer (Figure 6d) [231].

Metallo‐porphyrins, such as Zn‐ [240], Fe‐ [241], Co‐ [241], and Ni‐based porphyrins [242], have also been reported to disperse SWNTs. Although this method lacks selectivity, the porphyrins enhance electronic performance through charge transfer between the metal‐centered molecules and SWNTs [242]. Additionally, other π−π conjugated systems, including small graphene fragments and graphene oxides, have proven effective in dispersing SWNTs [243, 244, 245].

5.3.4. Polymers

Conjugated polymers (CPs), either with a fully conjugated backbone or conjugated blocks connected by nonconjugated linkers, can wrap around SWNTs, and form stable assemblies. Similar to polycyclic aromatic compounds, these polymers interact with SWNTs through π–π stacking, but the interaction is much stronger due to their larger delocalized π‐orbitals. This strong binding affinity results in dispersions that remain stable upon long‐term storage [246]. Additionally, polymer structures can be tailored for selective binding to specific chirality SWNTs, making this method more versatile than others [150, 247]

The selective enrichment of SWNTs using CPs in organic solvents began in 2007 when Nicholas and co‐workers isolated high‐purity (7, 5) SWNTs using poly(9, 9‐dioctylfluorenyl‐2, 7‐diyl) (PFO), starting with CoMoCAT tubes having a narrow diameter distribution [248]. Since then, research on CP‐SWNT interactions has expanded significantly, with some reports achieving sc‐SWNT purities exceeding 99.9% [249]. The preparation of high‐purity CP‐SWNT solutions involves four key steps:

1. Sonication: Sonication of raw SWNTs with the CP in a solvent, forming a mixture of CP‐SWNT complexes, excess CP, and bundled/uncomplexed SWNTs.

2. Centrifugation: Separates bundled and uncomplexed SWNTs by sedimentation.

3. Filtration & Washing: The supernatant from step 2 is collected, filtered through an ultrafiltration membrane, and the residue is thoroughly washed with solvent to remove excess polymer.

4. Re‐dispersion: The residue from step 3 is redispersed in an organic solvent via sonication to obtain the final dispersion.

Over the past several decades, extensive research has explored a wide range of CPs and optimized their structure, dispersing conditions to enhance SWNT dispersibility (Figure 7a). In general, CPs can be broadly categorized based on their building blocks, such as PFO, polythiophenes, and polycarbazoles (PCz) (Figure 7c). Despite their structural differences, these polymers generally exhibit a much higher selectivity for sc‐SWNTs than m‐SWNTs. This selectivity stems from variations in the interactions between SWNTs and polymers. Metallic tubes are believed to have a higher density of free surface electrons, leading to stronger charge transfer interactions with polymers. These interactions promote the bundling of m‐SWNTs in nonpolar organic solvents, ultimately causing their precipitation during ultracentrifugation.

FIGURE 7.

(a) Summary of general strategies toward better polymer‐SWNT dispersion. Polymer backbone rigidity alters the dispersed tube diameter. Modification through alternating backbone rigidity using (b‐c) monomers with varying backbone rigidity enables selective enrichment of SWNTs with specific diameters, and (d) hybrid conjugated/nonconjugated polymer architectures with tunable backbone flexibility (scheme redrawn based on ref. [256, 257]) enhance dispersion yield, selectivity, and stability. Polymers with (e) electron‐poor building blocks enhance sc‐SWNT selectivity, while (f) electron‐rich building blocks are more selective toward m‐SWNTs. (g) Preparation of a polymer library with different types of reactions to afford different linkers. Structures redrawn based on ref. [261]. (h) Polymers with extended alkyl sidechains lead to selectivity toward larger diameter tubes. Scheme redrawn based on ref. [262]. (i) Biaxially extended conjugation exhibits improved wrapping ability to yield high‐purity sc‐SWNTs. Chemical structure redrawn based on ref. [46]. (j) Polymer with electron‐poor and electron‐rich sidechains enables selective sorting of certain SWNTs, chemical structures redrawn based on refs. [263, 264].

The interaction between CPs and SWNTs is strongly influenced by the relative orientation of the polymer chain and the nanotube framework, which can be modified by varying the backbone and sidechain structure of the polymer [250]. With respect to the polymer backbone, homopolymers exhibit distinctly different selectivity when compared to copolymers [251]. For example, PFO and PCz preferentially interact with sc‐SWNTs of smaller diameters, whereas their copolymers with anthracene [252], xanthenoxanthene [253], and pyrene [254] demonstrate enhanced selectivity for larger‐diameter SWNTs. This shift occurs because these comonomers increase the rigidity and planarity of the polymer backbone [255]. (Figure 7b) However, extreme cases lead to the opposite effect: a highly rigid backbone becomes too planar to effectively wrap around SWNTs. Hwang et al. introduced pyrene units to modify the backbone kink angle and found that a slightly kinked polymer achieves better dispersions of sc‐SWNTs than either a highly kinked or entirely linear backbone [254]. Flexible polymers incorporating non‐conjugated linkers, such as alkyl chains or oligo(ethylene glycol), between conjugated units have been shown to provide better coverage of the tube surface, thereby improving sorting yields (Figure 7d) [256, 257]. Beyond linkers, block copolymers consisting of CPs with long non‐conjugated chains, such as polyisoprene [258, 259] and polystyrene [260], attached to both polymer ends have shown remarkable improvements in selectivity and dispersion stability. The non‐conjugated chains act as barriers, preventing SWNT reaggregation.

Comonomers with electron‐donating/withdrawing properties can modify the backbone's electronics, thereby enhancing selectivity toward specific SWNT species. Several reported electron‐donating comonomers, including benzothiadiazole (BT) [265, 266, 267, 268], pyridine (Py) [263], bipyridine (BPy) [269, 270], diketopyrrolopyrroles (DPP) [254, 271], dithiafulvene [272], and tetrathiafulvalene [272], have been shown to improve dispersibility and selectivity toward sc‐SWNTs to varying degrees (Figure 7e). Conversely, electron‐deficient conjugated π‐systems tend to shift dispersion selectivity toward m‐SWNTs. Studies from the Adronov group demonstrated that polymer backbones incorporating electron‐withdrawing nitro‐functionalities [273], electron‐poor cationic groups (Figure 7f) [274], or significant dipole moments [275, 276] result in dispersions that are enriched in m‐SWNTs. Further exploration through the synthesis of new CPs with tailored structures to achieve specific backbone flexibility is laborious and time‐consuming. With this in mind, Just et al. have shown that introducing small molecules or oligomers as CP chaperones can provide a similar effect. These chaperones facilitate more efficient polymer folding around SWNTs, improving dispersion and selectivity without requiring complex polymer synthesis [277].

CPs are synthesized through a variety of polycondensation reactions, including Suzuki coupling, copper(I)‐catalyzed azide‐alkyne cycloaddition (CuAAC), aza‐Wittig coupling, and Sonogashira cross‐coupling, with each introducing different linkers in the polymer backbone (Figure 7g) [261]. For example, the widely used Suzuki coupling directly forms a C─C bond between monomers, whereas other reactions introduce linkers such as 1, 2, 3‐triazole, azomethine, vinylene, or ethynylene. Studies by Gerstel et al. suggest that polymers with 1, 2, 3‐triazole linkers exhibit selectivity toward high chiral angle SWNTs (θ > 20°) [278]. Meanwhile, polymers with azomethine linkers demonstrate unprecedented selectivity for dispersing (8, 7), (7, 6), and (9, 5) SWNT species, primarily due to structural and conformational matching between the polymer backbone and the nanotube surface [261].

Although the polymer backbone primarily governs interactions with SWNTs, research indicates that polymer sidechains also play a crucial role in dictating polymer conformation and selectivity. Increasing sidechain length has been shown to enhance both dispersibility and diameter selectivity. For example, regioregular poly(3‐alkylthiophene)s (rr‐P3ATs) selectively enrich CoMoCAT SWNTs with diameters below 0.76 nm. Extending the alkyl sidechains not only improves the nanotube concentration within dispersions, but also promotes the dispersion of larger‐diameter SWNTs (Figure 7h) [261, 279, 280, 281]. Alternatively, branched sidechains can introduce steric hindrance, altering backbone conformation (e.g., dihedral angle) and thereby disrupting polymer‐SWNT interactions [282, 283].

Beyond sidechain length and structure, sidechain functionality also influences polymer‐tube interactions [46] and selectivity [264]. For instance, sidechain‐modified CPs with biaxially extended conjugation not only expand the sidechain pattern but also rotate sidechain moieties proximal to the backbone, enhancing sc‐SWNT dispersibility (Figure 7i) [46]. Polymers with polar sidechains improve selectivity toward sc‐SWNTs [284], while electron‐poor functionality on sidechains enhances the selectivity toward m‐ species (Figure 7j) [263, 264].

The solvent also plays a crucial role in CP‐SWNT dispersion. Modifying solvent properties such as viscosity [285] and dielectric constant [267] allows fine‐tuning of polymer solubility [286], dispersion selectivity, and efficiency. Polymers show distinct solubility and conformation when dissolving in various media. By combining different CPs and solvents, researchers can achieve selective solubilization of specific SWNT species, tailoring their distribution in solution. For instance, the rigid polymers PFO‐BT and PFO have limited solubility in toluene; however, they exhibit the highest sc‐SWNT selectivity when used to disperse SWNTs in toluene [286]. Moreover, Ji et al. demonstrated that in a highly viscous solvent such as tetralin, PFO predominantly adopts the β‐phase conformation, enabling selective dispersion of (10, 9) SWNTs with a 1.3 nm diameter [287]. Combining toluene and tetralin, Dzienia et al. found that PFO‐BT exhibited strong enrichment of (7, 3) SWNTs [288].

In addition to the above‐mentioned factors, other parameters such as polymer molecular weight [289], concentration [285], and the molar ratio of SWNTs to polymers [290] also significantly impact the efficiency of dispersion. While modifying polymer structures and dispersion conditions has enabled enrichment of certain SWNT species, this method still presents several notable drawbacks that limit further progress. For example, there is an inherent trade‐off between selectivity and dispersion concentration. Additionally, the process of preparing CP‐SWNT dispersions is highly time‐consuming, which hinders its scalability for industrial use. The nanotube concentration in the final dispersion remains relatively low, and the polymer is difficult to recycle. Overcoming these limitations will require significant advancements.

5.3.5. Endohedral Functionalization

Endohedral filling of SWNTs is an effective method for tuning their properties while maintaining their structural integrity [291, 292]. In such host‐guest complexes, the SWNT acts as a host, and some commonly reported guests include various gas molecules [293, 294], solvents (i.e., H₂O [295] and n‐hexane [296]), small organic molecules [297], fullerenes [298, 299, 300], metallocene complexes [301, 302, 303, 304, 305], and polyoxometalate (POM) clusters [298, 306, 307].

For small molecules, such as solvents and metal complexes, the endohedral filling will either alter the tube density [301, 303, 304, 305, 308] or act as a dopant [309, 310, 311]. So far, the use of the endohedral filler to sort SWNTs by their chirality is rarely mentioned. However, since the filler can alter the density of the tube and enlarge the density differences between SWNTs, the filled SWNTs are easier to isolate by DGU. When acting as a dopant, the filler promotes the charge transfer and even alters the conduction band of the SWNT [309, 310, 311].

Larger molecules can help separate tubes with various diameters through endohedral filling as the filling factor and efficiency are largely related to the tube diameter [313]. Bis(cyclopentadienyl) cobalt is observed to fill only tubes of one specific diameter (∼0.92 nm), while for wider nanotubes, the molecule desorbs easily during the filling process [302]. Endohedral functionalization with POM clusters enabled clean isolation of sc‐SWNT with a large diameter after extracting with CPs (Figure 8) [306, 312].

FIGURE 8.

POM clusters (∼1 nm) were filled in the inner cavities of SWNTs, and ∼98% pure s‐SWNTs (∼1.3–1.4 nm diameters) were selectively extracted. Redrawn based on ref. [312].

Similar to the previously mentioned non‐covalent functionalization techniques, the endohedral functionalization method suffers from low yield and poor selectivity. As a result, it often needs to be combined with other techniques to isolate highly purified SWNT samples.

5.3.6. DGU

DGU is extensively used in biology as it allows the separation of subcellular components with distinct buoyant densities. Hersam and co‐workers adopted this method and demonstrated its efficiency in separating SWNTs with different diameters [199, 314]. In the initial finding, ssDNA‐wrapped SWNTs were separated by a linear gradient prepared by aqueous diluted iodixanol. During ultra‐centrifugation, the use of structure‐discriminating surfactants enables the engineering of density differences among SWNT with different diameters and electronic type/chirality. This allows the nanotubes to migrate to regions that match their densities. This method demonstrated high selectivity, enabling the enhanced and precise separation of (6, 5) and (7, 5) chiral SWNTs, whose diameters differ by only 0.7 Å.

The buoyant density of SWNTs in aqueous solution is influenced by several factors, including the intrinsic density of the nanotube, surface functionalization, hydration layers, and internal filling. As discussed in the previous section on surfactants, dispersing SWNTs with surfactants results in the formation of a thick encapsulating layer of surfactant molecules. This additional mass and volume significantly contribute to the overall buoyant density of the surfactant–SWNT complex during DGU. The density of the surfactant–SWNT complex is highly tunable, allowing for precise control over DGU separation resolution and efficiency. The packing density of surfactant molecules on the nanotube surface is strongly dependent on both surfactant concentration and SWNT chirality. For example, when dispersing SWNTs in SDS, increasing the SDS concentration leads to a gradual saturation of the nanotube surface's capacity to bind additional surfactants. DGU separations performed at different SDS loadings reveal that surfactant structures on m‐SWNT surfaces reach saturation at lower SDS concentrations than those on semiconducting tubes, enabling more effective and selective separations (Figure 9a) [315]. The density of the surfactant–SWNT complex can also be tuned by selecting different surfactants or by employing combinations of surfactants, providing additional flexibility in optimizing DGU separation resolution and selectivity. A comparative study of three bile salts, deoxycholate (DOC), taurodeoxycholate (TDOC), and sodium cholate (SC), revealed that DOC forms a relatively thicker bound layer on SWNTs compared to the others, whereas TDOC exhibits a higher surface adsorption density. These findings provide valuable insight into surfactant selection for tuning the properties of surfactant–SWNT complexes in DGU processes [323]. The usage of co‐surfactants has an even larger effect on optimizing the diameter separation. A 3:2 mixture of SDS:SC permits significantly improved isolation of m‐SWNTs, while a 1:4 SDS:SC ratio gave better enrichment of sc‐SWNTs (>97%) with sharp diameter distributions [314]. This ratio has also been demonstrated on raw SWNTs with larger tube diameters (arc‐discharge tubes with diameter >1.2 nm) and showed efficient isolation of large diameter (∼1.6 nm) sc‐SWNTs (>99%) [324]. The density difference can also be tailored to target a particular species by choosing the appropriate combination and ratio of competing surfactants. The DOC and SDS co‐surfactant system showed an isolation of 97% pure (6, 5) SWNTs [325]. Theoretical studies indicate the binding affinities of DOC‐SWNT and SDS‐SWNT are diameter/chirality‐dependent, which allows the SDS to gradually replace the DOC throughout the separation time to achieve good separation resolution [326]. Other co‐surfactant systems with SDS, such as perylene bisimide surfactant [327] and polymethyl(1–butyric acidyl)silane (BA–PMS) [328], also gave promising results in separating m‐ and sc‐SWNTs.

FIGURE 9.

(a) As surfactant (SDS) concentration is increased, SDS adsorbed on m‐SWNTs becomes limited in its ability to reorganize before SDS adsorbed on sc‐ species. Reproduced with permission [315]. Copyright 2010, American Chemical Society. (b) Separations were conducted with different Pluronic block copolymers. Reproduced with permission [179]. Copyright 2010, American Chemical Society. (c) Spontaneous partition of SWNTs in PEG/dextran two‐phase system, redrawn based on ref. [316]. (d) Fine control of pH enables the separation of select nanotubes. Redrawn based on ref. [317]. (e) Schematics illustrating at sufficiently acidic pH, m‐ tubes acquire a positive net surface charge while sc‐ species remain anionic. (f) Anionic hydrogel beads enable isolation of sc‐SWNTs (>99.94% ± 0.04% purity). Both schemes redrawn based on ref. [318]. (g) Schematic diagram of m‐/sc‐ SWNT separation using a column filled with agarose gel beads. Redrawn based on ref. [319]. (h) Single‐surfactant multicolumn GC. Reproduced based on ref. [320]. (i) Monochiral SWNT fractions of sc‐SWNTs with a purity of 61–95% separated by GPC under pH control. Reproduced with permission [321]. Copyright 2014, American Chemical Society. (j) HPLC system using SDS and seven cholate derivatives as eluent. Redrawn based on ref. [322].

Apart from ionic surfactants, Pluronic block‐co‐polymers with both hydrophobic polypropylene (PPO) blocks and hydrophilic polyethylene oxide (PEO) blocks were also used to disperse SWNTs for DGU separation. Results indicate Pluronics with shorter PPO chain lengths (Pluronic F68) lead to higher purity sc‐SWNTs (>99%). In contrast, X‐shaped Tetronic block copolymers (Tetronic 1107) display better separation of m‐SWNTs (74%) [179]. Under basic conditions, the isolation efficiency of m‐SWNTs with Tetronic 1107 was further enhanced to ∼99% (Figure 9b) [329].

Apart from surfactants, endohedral water filling of SWNTs is also strongly diameter‐dependent, suggesting that water filling plays a significant role in diameter‐based separation of SWNTs during DGU [308]. The calculated densities of empty, water‐filled, and water+surfactant‐filled SWNTs showed that filling the cavity is essential for achieving a density that increases with tube diameter [330, 331, 332]. These fundamental studies offer valuable guidelines and insights for optimizing DGU sample preparation and tuning experimental parameters, ultimately enabling more precise separation of SWNTs by DGU.

Aside from the buoyant density of SWNTs, the DGU medium and gradient also play a crucial role in separation resolution and efficiency. Iodixanol is the most commonly used medium for DGU, and studies have shown that its separation efficiency can be fine‐tuned by adjusting parameters such as concentration [333], density gradient [334, 335], and pH [336]. These optimizations enable reliable separation of sc‐ and m‐SWNTs, and, in some cases, even isolation of single‐chirality species. Ghosh et al. reported that tailored nonlinear density gradients can significantly improve DGU separations, allowing an enrichment of ten different (n, m) species in a single step [337]. Even though iodixanol is the most employed gradient medium, certain drawbacks were also reported, such as being expensive, hard to remove from the isolated SWNT samples, and the density gradient becoming very steep after a long centrifugation time, which reduced the sorting resolution. Usage of cheaper media, such as sucrose, was also demonstrated, resulting in a purity of m‐ and sc‐SWNTs of 69% and 95%, respectively [338]. When caesium chloride (CsCl) was used as a gradient medium, excellent diameter selectivity was achieved, allowing isolation of (11, 10) SWNTs [339].

Significant research has demonstrated that altering the parameters of the DGU process can improve separation efficiency. In particular, reducing the separation rate, lowering the temperature, and increasing the SWNT concentration all enhance efficiency [340]. The addition of electrolytes (e.g., NaCl) can improve diameter sorting by increasing the amount of SDS absorbed on the tube surface and amplifying the density differences between SWNTs [324]. Bhatt et al. demonstrated that iterative centrifugation cycles at low centripetal force not only remove amorphous carbon and catalyst nanoparticles but also enhance the buoyancy of surfactant‐encapsulated, unbundled, high aspect ratio SWNTs, thereby enabling their bulk isolation [341]. Following DGU, the fractionation step is both crucial and challenging, as closely spaced layers can be difficult to isolate. The efficiency of this step is optimized by an automated fractionation apparatus that precisely extracts layers, further enhancing separation efficiency [342].

When the dispersant exhibits selectivity toward specific species, DGU efficiency is further improved. For example, non‐covalent functionalization with PFO can selectively enrich sc‐SWNTs prior to DGU, resulting in a more efficient separation of monodisperse SWNTs compared to non‐selective dispersants [343]. Similarly, CPs with pyridine units can selectively enhance the dispersion of certain chiral SWNTs, further improving the DGU separation efficiency to yield (10, 8) and (12, 5) chirality in high purities of 92.3% and 95.6%, respectively [344]. Covalent functionalization, which exploits the relative differences between sc‐ and m‐SWNTs, can also aid separation during DGU [345]. For instance, bromination reactions are more selective for m‐tubes over sc‐tubes, and the attached Br atoms induce a significant density difference between reacted and pristine nanotubes [346]. Consequently, the choice of functional group and solute can substantially tune separation efficiency [268, 347]. Endohedral filling further enhances separation resolution by amplifying the density difference between filled and unfilled species. Fillers such as nickelocene [303, 304, 305] and perfluorooctane [301] efficiently improve the isolation of encapsulated SWNTs with uniform diameters.

In summary, when combined with various dispersants, DGU enables the separation of chiral pure SWNTs. Although improvements in efficiency and resolution now meet strict sorting requirements and allow for the isolation of highly pure SWNT samples, the scalability of this method for large‐scale production remains limited.

5.3.7. ATPE

ATPE relies on liquid‐liquid phase separation, where a uniform mixture splits into two distinct liquid phases with different component concentrations. Originally used for biological separations, ATPE was applied to SWNT sorting in 2013 by Zheng's group (Figure 9c) [316]. Initial demonstration used two immiscible aqueous phases, PEG (more hydrophobic) and dextran (less hydrophobic). The hydrophobicity of SWNTs depends on their diameter and chirality. In small‐diameter (<1 nm) ranges, tubes with larger diameters are more hydrophobic due to curvature effects. In larger diameter (>1.2 nm) ranges, sc‐SWNTs are more hydrophobic than m‐SWNTs. During phase separation, more hydrophilic SWNTs preferentially move to the hydrophilic dextran phase (bottom), leaving hydrophobic SWNTs in the PEG phase. Later, researchers further employed combinations of DNA with partition modulators to sort single‐chirality nanotube species from a synthetic mixture [348, 349, 350]. The use of DNA was subsequently replaced with surfactants, which are readily available, inexpensive, and easier to implement. The hydrophobicity differences among various (n, m) species are significantly enhanced by varying the concentration of co‐surfactant systems (DOC/SDS and SC/SDS), enabling the chirality and diameter separation of SWNTs to be achieved using experimentally accessible surfactant concentrations [351, 352, 353, 354]. The purity of sc‐SWNTs isolated using the ATPE method can be significantly improved by repeating ATPE cycles, achieving >99.5% purity after four cycles [355].

Critical to this method is the hydrophobicity difference between various SWNT species. Considering that SWNT hydrophobicity can be tuned using redox reactions [356], variation of pH [348, 357, 358], and the application of surfactants [359, 360, 361, 362, 363], a range of synergistic methods can be implemented to enhance separations. In particular, the combination of multiple techniques enables efficient extraction of single chirality species. For instance, Li et al. extracted large‐diameter monochiral tubes by first endohedrally filling them, and then using ATPE under well‐controlled pH conditions (Figure 9d) [317]. Cao et al. performed ATPE first to narrow the chirality distribution of a SWNT sample, then further sorted them with CPs to isolate five single‐chirality SWNTs [364].

In summary, ATPE is a scalable and efficient liquid‐phase technique for separating individual SWNTs from mixtures. It offers several advantages, including flexibility in targeting specific (n, m) species, rapid processing, low cost, and reliable scalability. However, further research is needed to refine the control parameters of this technique to enable large‐scale separation with consistent performance. A recent review of this method provides deeper insights into its development [365].

5.3.8. Chromatography

As observed with the DGU method, different SWNT species exhibit distinct physicochemical properties that result in differences in their interactions with both mobile and stationary phases, allowing chromatographic separation. Various chromatographic techniques, such as anion exchange chromatography (IEX), gel chromatography (GC), and size‐exclusion chromatography (SEC), are well established in the separation of macromolecular structures, each exhibiting different separation efficiencies that depend on specific interactions with the analyte.

5.3.8.1. IEX

IEX is a method that uses an ion exchange resin as the stationary phase, separating components based on their differing affinities for the ions in the mobile phase and the counterions on the resin. IEX has been used to purify ss‐DNA‐SWNT complexes, taking advantage of a highly stable monolayer of DNA that forms on the SWNT surface [200, 203, 366]. When DNA with distinct sequences is wrapped around SWNTs, the resulting complex exhibits different geometry, dielectric properties, hybrid length, and diameter [200, 367]. The hydrophobicity of the SWNTs and van der Waals interactions between DNA‐wrapped SWNTs and IEX resin play an important role in IEX‐based separation [215]. In a series of pioneering studies, Zheng and co‐workers elegantly showed that specific DNA sequences can be tailored to allow the separation of both single‐chirality sc‐SWNTs [205, 368] and m‐SWNTs [204] from a synthetic nanotube mixture.

In addition to the use of DNA, it has been shown that differences in redox properties between m‐ and sc‐SWNTs can allow their separation by IEX. Chan‐Park and co‐workers demonstrated that m‐SWNTs are oxidized faster at low pH, causing an inversion of zeta potential from negative to positive when dispersed using a non‐ionic surfactant (Triton X‐405). Under the same conditions, sc‐SWNTs retain their negative zeta potential (Figure 9e). When the mixture was passed through an anionic resin (Sepharose4B‐SANS agarose beads), eluting with the non‐ionic surfactant, the negatively charged sc‐SWNTs eluted first, as the m‐SWNTs were electrostatically attracted to the stationary phase. The m‐SWNTs were subsequently eluted using the anionic surfactant, SDBS. Using this procedure, the authors claimed a sc‐SWNT purity of > 99.9% (Figure 9f) [318].

5.3.8.2. GC

GC enables the large‐scale separation of m‐/sc SWNTs and even single‐chirality SWNTs with specific band gaps. Tanaka and co‐workers found that when SDS‐dispersed SWNTs were passed through a column containing agarose gel beads, the sc‐SWNTs were selectively adsorbed onto the gel (95% pure), allowing the mobile phase containing enriched m‐SWNTs (70% pure) to be eluted (Figure 9g) [319, 369]. The core principle of this separation is the interaction between the SWNTs and gels, which depends on the tube electronic properties and surfactant conformation on the tube wall [370, 371]. When dispersing in SDS, m‐SWNTs have denser SDS layers on the tube wall than sc‐ tubes as mentioned in previous sections. This leads to a selective absorption of sc‐SWNTs onto the column media as the interaction between tubes and hydrogel is stronger with less dense SDS layers. In this regard, the concentration of dispersants is found to have a critical influence on the m‐/sc‐separation efficacy [372, 373]. Thermodynamic analysis indicates that the adsorption‐free energy of sc‐SWNTs onto the gel is lower than that of m‐SWNTs, which accounts for the m‐/sc‐SWNT separation [374].

Under basic conditions or in the presence of salt, more SDS condensed on the SWNT wall through electrostatic screening by the counterion, which reduced the interaction between SWNT and hydrogel, and the adsorption is dissociative [375, 376]. This pH‐dependent adsorption aligns with observations with other co‐surfactant complexes. For instance, the interaction differences between m‐ and sc‐SWNTs with the gel are amplified when HCl is added to SDS/SC‐dispersed SWNTs [377]. The absorbability of sodium hyodeoxycholate (SHC)/SDS‐coated m‐SWNTs is significantly enhanced with the addition of NaOH [378].

In addition to pH, the separation efficiency can be tuned by the gel media composition [379, 380, 381, 382, 383] and concentration [384, 385]. To‐date, agarose, Sephacryl, dextran‐based gels, and some self‐synthesized gels have been found to selectively adsorb SWNTs. The kinetic model based on the chirality‐selective adsorption to specific hydrogels further guides the hydrogel design for the separation of single‐chirality sc‐SWNT [386].

The adsorption of nanotubes on gels can also be diameter‐dependent. For example, Liu et al. found that sc‐SWNTs adsorbed on a Sepharose 2B gel (bead‐formed agarose‐based gel matrix) can be eluted stepwise using DOC as the eluent [387]. Smaller‐diameter sc‐SWNTs were eluted first using lower‐concentration DOC solutions, while larger‐diameter tubes were subsequently eluted with higher‐concentration DOC solutions. The concentration‐dependent adsorbability was later refined to enable the isolation of m‐SNWTs [388]. Other surfactants, such as cholic acid (HC) [389], also showed concentration‐dependent diameter selectivity and small‐diameter (5, 4) SWNTs were selectively absorbed on the gel at a high HC concentration.

In addition to concentration, loading amount, and column temperature are also crucial parameters for diameter and chirality separation [390, 391, 392]. The interaction strength between the gel and SWNTs increases at higher temperatures, enabling the separation of high‐purity sc‐SWNTs with desirable mean diameters [393]. Yang et al. observed that SWNTs with chiral angles less than 20° were adsorbed onto the gel column at temperatures below 18°C, while those with chiral angles greater than 20° began to adsorb at higher temperatures (20°C) [394]. When a concentrated SWNT dispersion (1 mg mL⁻¹) is loaded onto the gel column, the separation yield of multiple single‐chirality species is increased by approximately six times to the milligram scale in one separation run [395]. While sonication may shorten the SWNTs, gentler pulsed ultrasonication allows the preparation of a low‐defect, long SWNT dispersion and further enables a length‐dependent selectivity by GC [396]. Single‐chirality SWNTs could also be separated using several vertically connected gel columns in series (Figure 9h) [320, 397]. Some other studies also achieved enhanced selectivity by optimizing the separation procedure. Tulevski et al. found that when the solution loaded on the column is pre‐purified by DGU, highly pure sc‐ fractions (99.9% pure) can be resolved using a single surfactant system, which makes it possible to carry out iterative separation by re‐loading eluted samples on the column multiple times, increasing purity [398].

5.3.8.3. SEC

While GC separates SWNTs based on the interaction between the SWNTs and the gel, which depends on electronic properties and conformational differences, SEC separates the SWNTs exclusively by size. In SEC, the mobility of SWNTs is inversely proportional to their tube length. Duesberg et al. initially employed SEC with controlled pore glass (3000 Å) to separate the SDS‐SWNT complex into two fractions with distinct length distributions (mean length of 1000 and 500 nm) [399]. Afterward, the length sorting efficiency was enhanced by changing column media, dispersants, and eluents. Similar to other gel separation techniques, the use of different surfactants or a combination of multiple surfactant solutions as eluent under controlled pH conditions allows a more precise separation in terms of both length and chirality (Figure 9i) [321, 372, 400]. Notably, Yomogida et al. developed an HPLC system with seven cholate derivatives that separate 11 high‐purity single‐chirality species automatically from one starting nanotube solution (Figure 9j) [322]. In addition to the use of surfactants, other functionalization methods, including covalent functionalization [401, 402], DNA‐wrapping [403, 404, 405, 406, 407], and CP‐wrapping [408], have also demonstrated impressive length and/or chirality separation. The separation efficiency can be further improved by packing multiple columns with different porosities. This approach significantly narrows the length distribution of each fraction to less than 100 nm, reducing length variation [404, 405]. The detailed separation techniques and advancements since the invention of the method are summarized in Table 1.

TABLE 1.

Summary of recent developments in SEC.

Type	Column	Dispersant	Eluent	Efficiency	Refs.
SEC	Controlled pore glass	SDS	SDS	Remove production impurities	[399]
GPC	Waters 150‐C Plus GPC	Covalent functionalized	THF	m‐SWNTs and sc‐SWNTs with differing diameters	[402]
HPLC	Silica beads	DNA	Buffer (pH = 7)	Length separation (527 ± 32 nm, 192 ± 19 nm, and 75 ± 8 nm) and Chirality enrichment ((6, 5))	[404]
HPLC	PL aquagel‐OH column	Oxidized SWNT	Milli‐Q water	Length separation (509 nm, 310 nm, 203, and 146 nm)	[401]
HPLC	COSMOSIL CNT	DNA	Buffer (pH = 7)	Length separation (393 ± 87 nm, 199 ± 49 nm, 109 ± 25 nm, 52 ± 21 nm)	[405]
HPLC	CNT SEC‐2000, 1000, and 300	DNA	Buffer (pH = 8)	Separation of free oligo DNAs and DNA/SWNT hybrids.	[406]
SEC	Sephacryl S‐200	SDS	SDS (elute‐m‐SWNT)→ Sodium cholate hydrate (elute sc‐SWNT)	“Zig‐zag” enriched microbundles	[372]
HPLC	Sepax CNT model	DNA	Sodium thiocyanate (NaSCN)	Length separation (630, 27, and 70 nm)	[407]
HPLC	SEC‐CNT	SDC	SC/SDC/NaSCN+ SC	Length separation (360 and 130 nm)	[409]
GPC	Sephacryl S‐200	SDS	SDS with different pH	(n, m) separation of 15 different nanotube species with a purity of 16–93%	[400]
GPC	Sephacryl S‐200	SC	SDS elute m‐SWNT and SC elute sc‐SWNT	95% pure sc‐SWNT with average lengths of 1.2, 0., and 0.3 µm	[410]
GPC	Sephacryl S‐200	SDS	SDS with different pH	(n, m) separation of eight different SWNT species with a purity of 61–95%.	[321]
HPLC	COSMOSIL CNT 3000	DNA	Buffer (pH = 7)	Length separation (668, 432, 264, and 153 nm)	[403]
GPC	Toyopearl HW‐75 resin (Tosoh)	Polymer wrapped	Toluene	Remove excess polymer and isolate >99.7% sc‐SWNT with different length distribution (940 ± 385 nm, 590 ± 250 nm, 360 ± 200 nm, and 315 ± 170 nm)	[408]
HPLC	Sephacryl S‐200 HR gel	SC	SDS and seven cholate derivatives	20 specific chirality‐enriched nanotubes, including 11 high‐purity single‐chirality species was isolated	[322]

5.3.9. Other Techniques

Dielectrophoresis is a technique that exploits the mobility differences between m‐SWNTs and sc‐SWNTs in a medium under an external electric field. Generally, m‐SWNTs, which have a larger dipole moment, migrate more rapidly toward the electrodes than sc‐SWNTs [411, 412]. As this method is more effective for aligning SWNTs rather than purifying them, it will not be discussed further in this review [413, 414].

In addition to the methods mentioned above, techniques such as differential mobility analysis [415], three‐step cross‐flow filtration [416], electrochemical selective adsorption [417], capillary electrophoresis [418, 419], and chemically modified thin film silicone elastomer tapes [420] have demonstrated varying levels of efficacy in sorting SWNTs. The use of these techniques is either not as widely reported as others or lacks efficiency, with no significant development following their initial reports. Therefore, this review will not cover them in detail.

In summary, raw SWNT samples can be dispersed in organic or aqueous media through covalent functionalization, non‐covalent interactions, or endohedral filling techniques. During dispersion, certain dispersants can preferentially enrich specific chiral species or nanotubes within a narrow length or diameter distribution. In particular, some reports have achieved sc‐SWNT purities exceeding 99.9% with CP [249]. The highly purified samples allow further advances in applications, including transistors [46, 256], flexible electronics [282, 283, 421], and transparent electrodes [50]. Although these approaches provide certain selectivity, additional purification steps, including DGU, ATPE, and chromatography, remain essential to obtaining highly monodisperse samples. Among these methods, ATPE currently offers the highest efficiency and cost‐effectiveness; however, all techniques still require significant development to enable scalable purification and single‐chirality sorting.

5.4. Summary of Single Chirality Sorting

“The harder the climb, the better the view” is a principle that aptly describes the development of isolating single‐chirality SWNTs. As mentioned earlier, pure single chirality SWNTs hold immense potential for the fabrication of next‐generation electronic devices, attracting worldwide attention. Through years of dedicated effort and collaboration across research groups, over 67 distinct chirality‐purified/enriched SWNTs have been successfully isolated. In this section, we summarize the purified, enriched, and selectively synthesized (n, m) SWNTs with the highest reported purity and some important variations (Table 2 and Figure 10)

TABLE 2.

Summary of purified (P), enriched (E), and selectively synthesized (n, m) SWNTs.

P/E	m‐/sc‐	Chirality	Techniques	Purity [%]	Dispersants	Most recent advance/important variation
P	sc‐	(5, 4)	GC	97	HC	GC with tuning surfactant concentration [389]
P	m‐	(5, 5)	Total synthesis	—	—	Diels−Alder reaction [118]
			ATPE	—	SDS/DOC	pH‐driven ATPE [357]
P	sc‐	(6, 4)	DGU	99	SDS/SC	Orthogonal iterative DGU [334]
			GC	93.4	SDS/SC	Milligram scale separation [394, 395]
p	sc‐	(6, 5)	Selective synthesis	95.8	—	Trimetallic catalyst NiSnFe [422]
			GC	96	DOC/SDS/SC	Milligram scale separation [395]
			DNA	87.5	DNA	DNA sequence defined by ML [423]
			HPLC	98	SDS/SC	HPLC with mixed surfactant system [322]
P	m‐	(6, 6)	DNA	90	DNA	DNA sequence defined by ML [423]
			Total synthesis	—	—	[115, 123]
P	sc‐	(7, 3)	GC	93.5	DOC/SDS/SC	Milligram scale separation [394]
			HPLC	91	SDS/SC	HPLC with mixed surfactant system [322]
P	m‐	(7, 4)	DNA	70	DNA	DNA sequence defined by ML [423]
P	sc‐	(7, 5)	Selective synthesis	18.6	—	Fe2(SO4)3/SiO2 [424]
			GC	91	DOC/SDS/SC	Milligram scale separation [394]
			HPLC	90	SDS/SC	HPLC with mixed surfactant system [322]
P	sc‐	(7, 6)	Selective synthesis	15.8	—	Fe2(SO4)3/SiO2 [424]
			GPC	95	SDS	GPC with pH control [321]
			GC	90.1	DOC/SDS /SC	Milligram scale separation [394]
			HPLC	96	SDS/SC	HPLC with mixed surfactant system [322]
P	m‐	(7, 7)	Selective synthesis	—	—	Cloned nanotubes with chiral pure seed [124]
			DNA	—	DNA	[204]
			DGU	—	Covalent functionalized	SWNT covalent functionalized with 4‐hydroxyphenyl group for DGU separation [345]
P	m‐	(8, 2)	DGU	—	Covalent functionalized	SWNT covalent functionalized with 4‐hydroxyphenyl group for DGU separation [345]
p	sc‐	(8, 3)	GC	91.8	DOC/SDS /SC	Milligram scale separation [394]
			HPLC	98	SDS/SC	HPLC with mixed surfactant system [322]
P	sc‐	(8, 4)	Selective synthesis	80	—	Solid carbide catalysts by controlling the symmetries of the active catalyst surface [86]
			GC	91.5	DOC/SDS /SC	Milligram scale separation [394]
			HPLC	75	SDS/SC	HPLC with mixed surfactant system [322]
P	m‐	(8, 5)	Selective synthesis	17.5	—	Solid Mo2C catalysts [425]
			ATPE	—	DNA	ATPE of DNA wrapped SWNT [350]
P	sc‐	(8, 6)	GPC	88	SDS	GPC with pH control [321]
			HPLC	94	SDS/SC	HPLC with mixed surfactant system [322]
			ATPE	82	CPs	Iterative ATPE [364]
P	sc‐	(8, 7)	Selective synthesis	18	—	FC‐CVD with H2O [106]
			GPC	77	SDS	GPC with pH control [321]
			HPLC	94	SDS/SC	HPLC with mixed surfactant system [322]
			ATPE	65	CPs	Iterative ATPE [364]
P	m‐	(8, 8)	DGU	—	Covalent functionalized	SWNT covalent functionalized with 4‐hydroxyphenyl group for DGU separation [345]
			ATPE	—	CP+SDS/DOC	[353]
			Total synthesis	—	—	[123]
P	m‐	(9, 0)	Total synthesis	—	—	[115]
p	sc‐	(9, 1)	GC	96.2	DOC/SDS/SC	Milligram scale separation [394]
			HPLC	95	SDS/SC	HPLC with mixed surfactant system [322]
p	sc‐	(9, 2)	GC	88.9	DOC/SDS/SC	Milligram scale separation [394]
			HPLC	95	SDS/SC	HPLC with mixed surfactant system [322]
E	sc‐	(9, 3)	Selective synthesis	∼20	—	Fe–Ni‐based catalyst [426]
P	sc‐	(9, 4)	GPC	76	SDS	GPC with pH control [321]
			GC	92.7	DOC/SDS/SC	Milligram scale separation [394]
			HPLC	76	SDS/SC	HPLC with mixed surfactant system [322]
P	sc‐	(9, 5)	HPLC	95	SDS/SC	HPLC with mixed surfactant system [322]
			ATPE	83	CPs	Iterative ATPE [364]
P	m‐	(9, 6)	DGU	—	Covalent functionalized	SWNT covalent functionalized with 4‐hydroxyphenyl group for DGU separation [345]
			ATPE	—	DNA	ATPE of DNA wrapped SWNT [350]
			Selective synthesis	45	—	Solid “Trojan” catalysts [65]
P	sc‐	(9, 7)	HPLC	83	SDS/SC	HPLC with mixed surfactant system [322]
P	sc‐	(9, 8)	ATPE	84.2	SDS/SC	Control temperature and addition of Potassium thiocyanate [427]
			Selective synthesis	48.2	—	Control synthesis with CoSO4/SiO2 [74]
P	m‐	(9, 9)	ATPE	—	CP+SDS/DOC	[353]
P	sc‐	(10, 0)	HPLC	94	SDS/SC	HPLC with mixed surfactant system [322]
P	sc‐	(10, 2)	HPLC	41	SDS/SC	HPLC with mixed surfactant system [322]
			GC	91	DOC/SDS/SC	Milligram scale separation [395]
P	sc‐	(10, 3)	HPLC	69	SDS/SC	HPLC with mixed surfactant system [322]
			GC	91.5	DOC/SDS/SC	Milligram scale separation [394]
E	m‐	(10, 4)	Selective synthesis	∼22	—	Fe–Ni‐based catalyst [426]
			GC	—	SDS	[428]
P	sc‐	(10, 5)	CP	60	PFO‐based CP	[268]
			HPLC	44	SDS/SC	HPLC with mixed surfactant system [322]
			ATPE	81	CPs	Iterative ATPE [364]
E	m‐	(10, 7)	DGU	34‐46	DOC/SDS	[429]
P	sc‐	(10, 8)	DGU	92.3	PFO‐BPy	DGU with stepwise extraction [344]
P	sc‐	(10, 9)	CP	71	PFO	PFO in tetralin [287]
			Selective synthesis	75	—	Co (1 1 1) catalyst [430]
P	m‐	(10, 10)	ATPE	—	CP+SDS/DOC	[353]
			Total synthesis	—	—	[123]
P	sc‐	(11, 0)	GC	72.4	DOC/SDS/SC	Milligram scale separation [394]
P	sc‐	(11, 1)	GC	91.7	DOC/SDS/SC	Milligram scale separation [394]
P	sc‐	(11, 3)	HPLC	85	SDS/SC	HPLC with mixed surfactant system [322]
			ATPE	85	CPs	Iterative ATPE [364]
E	sc‐	(11, 7)	Selective synthesis	24	—	Cobalt disilicide (CoSi2) as catalysts [431]
			CP	—	supramolecular polymer	[432]
E	sc‐	(11, 9)	Selective synthesis	28	—	Fe–Ni‐based catalyst [426]
			ATPE	—	CP+SDS/DOC	[353]
E	sc‐	(11, 10)	DGU	—	SC/CsCl	Encapsulated C60 to the (11, 10) SWNTs obtained by CsCl sorting [339].
P	sc‐	(12, 1)	GC	82.5	DOC/SDS/SC	Milligram scale separation [394]
			HPLC	51	SDS/SC	HPLC with mixed surfactant system [322]
E	sc‐	(12, 4)	ATPE	—	CP+SDS/DOC	[353]
P	sc‐	(12, 5)	DGU	95.6	PFO‐BPy	DGU with stepwise extraction [344]
P		(12, 6)	Selective synthesis	92	—	W–Co alloy catalyst [64]
E	sc‐	(12, 7)	ATPE	—	CP+SDS/DOC	[353]
E	sc‐	(12, 7)	CP	—	PFO‐Py	CP with a hydrophobic backbone and hydrophilic side chains [290]
E	sc‐	(12, 10)	CP	—	CP	Conjugated dendritic oligothiophenes [433]
P	m‐	(12, 12)	Total synthesis	—	—	[123]
E	m‐	(13, 1)	Selective synthesis	43.7	—	Solid “Trojan” catalysts [65]
E	sc‐	(13, 2)	SEC	32	SDS	pH‐control SEC [400]
E	sc‐	(13, 5)	CP	—	PFO‐Py	[290, 434]
E	m‐	(13, 7)	ATPE	—	DOC/SC	Endohedral filling and pH‐driven ATPE [317].
E	sc‐	(13, 12)	Selective synthesis	14.7	—	Co catalysts [435]
P	sc‐	(14, 4)	Selective synthesis	97	—	W6Co7 catalyst with a high percentage of (1 0 10) planes [436]
P	sc‐	(14, 6)	ATPE	75	DOC/SC	Endohedral filling and pH‐driven ATPE [317]
E	sc‐	(14, 7)	DGU	68	SDS/SC	[324]
			CP	—	CP	Conjugated dendritic oligothiophenes [433]
E	sc‐	(15, 4)	CP	—	CP	PFO‐BT [266] and hydrogen bonded supramolecular polymer [432]
E	sc‐	(15, 5)	ATPE	49	DOC/SC	Endohedral filling and pH driven ATPE [317]
E	sc‐	(15, 14)	Selective synthesis	20.6	—	Co (1 1 1) catalyst [430]
E	sc‐	(16, 0)	Selective synthesis	79.2	—	W6Co7 catalysts containing plenty of (1 1 6) planes [77]
E	sc‐	(16, 2)	CP	—	CP	Conjugated dendritic oligothiophenes [433]
E	sc‐	(16, 3)	ATPE	59	DOC/SC	Endohedral filling and pH driven ATPE [317]
E	sc‐	(16, 8)	Selective synthesis	74	—	Titanium carbide (TiC) nanoparticle catalyst [437]
E	sc‐	(16, 15)	Selective synthesis	35.5	—	Co (1 1 1) catalyst [430]
E	sc‐	(18, 17)	Selective synthesis	17.6	—	Co (1 1 1) catalyst [430]

FIGURE 10.

Summary of purified, enriched, and selectively synthesized (n, m) SWNTs.

6. Dispersant Removal

While the solubility and purity of SWNTs are significantly enhanced after sorting, the required surfactants and wrapping agents remain on the SWNT surface and act as insulators between tubes, largely suppressing their conductivity in thin films. In the previous sections, we discussed the electrical current transport within individual SWNTs and SWNT networks. When the SWNTs form a percolation network in the presence of other materials, the charge transfer between SWNTs at a small distance is facilitated through a hopping or tunneling mechanism. For the dispersant‐SWNT material, the conductive path is attributed to the percolation network and the electronic tunnelling across separate SWNTs [438]. When the loading of SWNTs is below the percolation concentration, composite materials may still be conductive, as the conduction may occur through a tunneling mechanism when tubes are close enough (distance <1.8 nm) but do not necessarily touch [439]. Charge flows through the shortest and least resistive pathways between tubes, avoiding indirect connections. At higher SWNT loading, SWNTs form direct contact with each other, and the resistance of the resulting composite decreases, following percolation theory. However, increasing SWNT concentrations is not always beneficial, as high concentrations can cause SWNTs to bundle/aggregate, thereby decreasing the conductivity [440]. A more fruitful approach may be to remove the dispersant that prevents close contact between nanotubes after formation of thin films.

Removal of surfactants has been accomplished by techniques such as acid oxidation, annealing, and solvent washing. Acid solutions, such as nitric acid [441, 442] and sulfuric acid [443], have been explored for eliminating residual surfactants and enhancing the conductivity of transparent conductive films (Figure 11a). Annealing, for example, in vacuum [444, 445] or oxidizing environments [446, 447], can desorb and/or decompose spurious adsorbates and surfactants from the surfaces of SWNTs and further enhance conductivity. Organic solvents, such as acetone, ethanol, dimethylacetamide, N‐cyclohexyl‐2‐pyrrolidone, dichloromethane, and acetonitrile, disrupt the surfactant‐SWNT interaction and allow for the facile isolation of purified SWNTs (Figure 11a) [448, 449].

FIGURE 11.

(a) Summary of general procedures to remove surfactants from the SWNT surface. (b) Supramolecular polymers (structures redrawn based on refs [432, 452, 453].) (c) Schematic illustration of the separation cycle of sc‐SWNTs using hydrogen‐bonding polymers (redrawn based on ref. [432].) (d) Degradable polymer backbone structures with various degradation mechanisms, including imine bonds [454, 455, 456], disilane bonds [457], photo‐degradable linkers [458, 459, 460, 461], and stimuli‐changeable conformation [462, 463]. Structures redrawn from corresponding refs. (e) Degradable polymer sidechain bearing photo‐ [464] and thermally‐ [465] degradable bonds. Structures redrawn from the respective refs. (f) CP bearing self‐immolative sidechains, which are stabilized with an end‐cap (“lock”) that is easily “unlocked” with an appropriate “key” to release highly conductive pure SWNT. Scheme redrawn based on ref. [466].

Generally, polymers are relatively harder to remove than small molecules due to their strong interactions with the tube surface. Polymer residues on the nanotubes hinder electron transfer, reducing the conductivity of SWNT‐based materials [282]. Rapid annealing and cooling cycles have been shown to thoroughly remove the polymers from the surface of sc‐SWNTs [450]. A thin layer of yttrium oxide can also oxidize the polymer coating on the tube surface, allowing the oxidized products to be removed by washing with dilute HCl. However, this method requires harsh oxidizing conditions, and multiple treatments are often needed to fully remove the polymer from thicker films [451]. To address this issue, researchers have explored degradable polymers and sidechains to enhance recyclability and minimize the insulating effects caused by residual polymers. Supramolecular polymers have demonstrated effective SWNT dispersion while allowing for subsequent disassembly into monomeric units upon exposure to a hydrogen‐bond disrupting agent (Figure 11c) [432, 452, 453]. CPs with chemically degradable bonds (i.e., imine bonds [454, 455, 456] and disilane bonds [457]) or photo‐degradable linkers [458, 459, 460, 461] (Figure 11d) can depolymerize into monomers upon application of certain stimuli, enabling the clean removal of polymers and yielding polymer‐free SWNTs. Other studies suggest that conformational changes in the CP backbone can disrupt polymer‐SWNT interactions, facilitating polymer removal [462, 463]. Once the polymer detaches, the unwrapped SWNTs precipitate out of the solution.

In addition to degradable backbones, researchers have explored removable sidechains to achieve similar effects. The tunneling resistance of a tube‐polymer‐tube junction is very sensitive to polymer separation layer thickness and also depends on the relative orientation between the SWNTs [438, 467]. Most CPs contain long sidechains to enhance their solubility, which largely contributes to the insulating layer thickness. Mechanisms for sidechain removal include photo‐cleavable bonds (Figure 11e) [464], thermally‐cleavable bonds (Figure 11e) [465], and self‐immolative sidechains [466] (Figure 11f).

Several key considerations must be taken into account when using degradable polymer backbones or sidechains for dispersing and releasing SWNTs:

1. Polymer stability: The polymer must be stable enough to withstand the dispersion process and storage period. For example, ultrasonication, commonly used during dispersion, can generate significant heat that may degrade sensitive polymers.

2. Dispersibility: The efficiency of dispersion is highly dependent on the structure of the polymer backbone and side chains, as previously discussed; therefore, polymer design should carefully consider dispersibility as a key factor.

3. Polymer removal efficiency: Harsh conditions, prolonged reaction times, or incomplete removal are undesirable and can limit the applicability of the resulting materials.

4. By‐product removal: By‐products that require additional steps or harsh conditions (e.g., extensive solvent washing or high temperatures) to be removed from the nanotube surface can also hinder downstream applications.

While the removal of polymer backbones or sidechains is well established in the literature as a proof‐of‐concept for efficiently isolating dispersant‐free SWNTs after specific treatments, most reported methods fail to meet all the aforementioned criteria. As a result, much of the research remains at the proof‐of‐concept stage, facing significant technical challenges that hinder practical application. Among these methods, the self‐immolative sidechains developed by Adronov and co‐workers offer distinct advantages over traditional stimulus‐responsive cleavage mechanisms. Due to protection of the end‐cap with tert‐butyl(dimethyl)silyl (TBS) groups, the polymer remains stable under ambient conditions, yet rapidly degrades into small molecules upon exposure to a specific trigger—tetrabutylammonium fluoride (TBAF). The sidechain cleavage by‐products can be easily removed from the nanotube surface through simple solvent washing (Figure 11g) [466]. Upon sidechain removal, the conductivity of the free‐standing SWNT film increased by approximately 60‐fold [466]. Building on this mechanism, Yu and Adronov further demonstrated the fabrication of thin‐film electrodes on Mylar substrates that are conductive, flexible, and transparent (Figure 12a) [52]. In addition to Mylar, elastomeric substrates (polydimethylsiloxane, PDMS) were used to develop highly sensitive pressure sensors with a sensitivity of 1,655 kPa^{− 1}, a broad dynamic range (0.003–70.1 kPa), and near‐instantaneous response (Figure 12b) [57]. These pioneering studies not only highlight the vast potential of SWNT‐based materials but also underscore the importance of the self‐immolative strategy for dispersants in developing the next generation of high‐performance, functional SWNT electronics [52, 468, 469].

FIGURE 12.

(a) Conductive, transparent, and flexible thin film fabricated on a Mylar substrate. subsequent removal of the self‐immolative side chain significantly reduces sheet resistance without compromising transparency. Reproduced with permission [52]. Copyright 2024, American Chemical Society. (b) Highly flexible and sensitive pressure sensor fabricated by using post‐TBAF‐treated CP‐SWNT film as the conductive layer on an elastomeric substrate for monitoring human motion. Reproduced with permission [57]. Copyright 2025 X. Yu and A. Adronov., published by Wiley‐VCH GmbH.

7. Summary and Outlook

In this review, we explored the electrical conductivity of SWNTs and the key challenges that hinder their widespread adoption as potential replacements for traditional metal conductors. We provided a theoretical foundation for understanding these limitations and discussed various practical methods developed over the past few decades to enhance the purity and solution processability of SWNTs.

A major research focus has been the selective enrichment of SWNTs based on their diameter, electronic type/chirality, and length. This review has summarized the progress in purifying and enriching (n, m) SWNTs, along with the highest purity levels achieved to date. Since the last comprehensive review [148], significant advancements have been made, not only in improving the purity of previously purified chiral tubes but also in isolating additional chiral species. Despite these advancements, large‐scale separation of specific nanotube types remains a significant challenge. Current techniques have only achieved enrichment at the milligram scale, which is still several orders of magnitude below the threshold required for industrial applications.

Beyond purity concerns, the use of dispersants, necessary for processing SWNTs in solution, introduces another critical limitation. These additives often act as insulating barriers, impeding the electrical performance of SWNT‐based materials. Although various strategies have been developed to overcome this issue, they typically involve a trade‐off between efficient dispersant removal and maintaining high dispersion quality and SWNT purity. Developing methods that can selectively enrich SWNTs while ensuring effective yet mild removal of dispersants remains a key challenge.

Looking ahead, further advancements in scalable separation techniques, coupled with innovations in dispersant removal strategies, will be essential for unlocking the full potential of SWNTs in electronic applications. Bridging the gap between laboratory‐scale separation and industrial‐scale production will require interdisciplinary efforts. Addressing these challenges will not only pave the way for SWNTs as next‐generation conductors but also expand their applicability across various high‐performance electronic and optoelectronic devices.

Author Contributions

Xiao Yu collected all references, wrote the original draft, and made all the Figures; Alex Adronov reviewed and edited the manuscript.

Funding

Natural Sciences and Engineering Research Council of Canada (NSERC) through the Discovery Grant program (RGPIN‐2023‐04824).

Conflicts of Interest

The authors declare no conflict of interest.

Data Availability Statement

All data are obtained from peer‐reviewed articles, books, and patents as reported in the references list. No other datasets have been used.