Beyond Color: The New Carbon Ink

Abstract

For thousands of years, carbon ink has been used as a black color pigment for writing and painting purposes. However, recent discoveries of nanocarbon materials, including fullerenes, carbon nanotubes, graphene, and their various derivative forms, together with the advances in large-scale synthesis, are enabling a whole new generation of carbon inks that can serve as an intrinsically programmable materials platform for developing advanced functionalities far beyond color. The marriage between these multifunctional nanocarbon inks with modern printing technologies is facilitating and even transforming many applications, including flexible electronics, wearable and implantable sensors, actuators, and autonomous robotics. This review examines recent progress in the reborn field of carbon inks, highlighting their programmability and multifunctionality for applications in flexible electronics and stimuli-responsive devices. Current challenges and opportunities will also be discussed from a materials science perspective towards the advancement of carbon ink for new applications beyond color.

Graphical Abstract

For >30,000 years, carbon ink is mainly used as a black color pigment. However, the recent discovery of nanocarbon materials with remarkable physical and chemical properties has enabled new applications well beyond coloring. This article reviews recent advances in the development of nanocarbon inks and their emergent applications, highlighting the fundamental and technical challenges as well as opportunities.

1. Introduction

Initially discovered as a pigment to color an object’s surface for the production of images, text, or designs, carbon ink has played an important role in human civilization by helping preserve, spread, and exchange knowledge.^[1–3] One of the oldest known examples of carbon pigment can be traced back to ≈30,000 B.C. in the Paleolithic Age, in which charcoal was used for cave painting (Figure 1a). In ≈4,000 B.C., ink made from charcoal and carbon soot was applied for decorative purposes and soon after developed into the widely used Chinese solid ink sticks (Figure 1b,c). As the development of human society called for more portable and reliable pigments, water- and oil-based carbon ink, pencils, and electronic-grade ink powders continued to be developed (Figure 1d–f). However, coloring remained the main function for all of these traditional carbon inks. Fueled by recent material discoveries and advances in their mass production, nanocarbon materials, including fullerenes,^[4] carbon nanotubes (CNTs),^{[5, 6]} graphene,^[7] as well as some of their derivatives,^{[8, 9]} have been explored as fundamental building blocks for the formulation of new carbon inks. Because of their extraordinary electrical, optical, thermal, and mechanical properties, nanocarbons have shown the potential to significantly expand the applications of ink beyond its simple color function (Figure 1g–m).

Figure 1.

The evolution of carbon ink from a pigment with only a coloring function to multifunctional materials beyond color. a) A group of animals painted in a cave using charcoal as the pigment. Photo courtesy of Centre National de Préhistoire, France. Reproduced with permission. b) A pottery bowl with charcoal decorated with a human face pattern recovered from the Banpo archaeological site dating back to ≈4000 B.C. Photo courtesy of the National Museum of China. Reproduced with permission. c) Ancient Chinese inksticks dated to ≈200 B.C. Photo courtesy of Hubei Museum, China. Reproduced with permission. d) Quill pens for writing used during the 18^th to 19^th century. Photo courtesy of the Nation Museum of American History. Reproduced with permission. e) The opening of Genesis in the Gutenberg Bible (Biblia latina, Mainz, Johann Gutenberg, 1455), printed with oil-based carbon ink. Photo courtesy of the Otto Vollbehr Collection, Rare Book and Special Collections Division, Library of Congress. Reproduced with permission. f) The invention of pencil (traced to ≈1700) and the usage of graphite powder as the electric ink in ≈1980. g) A conductive electronic circuit printed from carbon ink. Reproduced with permission.^[219] Copyright 2007, The Royal Society of Chemistry. h) Early applications of carbon ink for sensing applications utilized the porous structures of nanocarbon to host the probe molecules, instead of directly using nanocarbon as probes.^{[220, 221]} i-g) Rigid (i) and flexible (g) thin film transistors fabricated from ink-jet printing CNT ink. Reproduced with permission.^{[222, 223]} Copyright 2007, American Institute of Physics; Copyright 2010, American Chemical Society. k) A gas sensor fabricated from a CNT solid ink. Reproduced with permission.^[179] Copyright 2012, Wiley-VCH. l) 3D printed scaffold from graphene ink. Reproduced with permission.^[224] Copyright 2015, American Chemical Society. m) Security ink made from fluorescent carbon dots. Reproduced with permission.^[43] Copyright 2016, Wiley-VCH.

Depending on structure, nanocarbons, most notably graphene and single-wall CNTs (SWCNTs), exhibit extraordinary properties, including high electrical conductivity,^[10] outstanding mechanical strength,^[11] structural flexibility,^{[11, 12]} and large surface area.^[13] The large surface area of nanocarbons further makes it possible to host a large number of functional groups for added chemical functionalities. Additionally, compared with other materials such as metal or silicon, nanocarbons are mechanically soft and flexible, making them suitable to serve as building blocks for flexible electronics and wearable devices.^{[12, 14]} Furthermore, thanks to the advances in carbon functionalization chemistry, the structure of nanocarbons can be modified to allow for systematic tailoring of their solubility, electrical, and optical properties for specific applications.

Nanocarbon ink is by nature programmable in terms of both function and form that it can be built into. On one hand, nanocarbon inks can be processed into almost any form of structure (e.g., dots, lines, percolative network and 3D hierarchical structures), particularly with the aid of modern additive manufacturing technologies (e.g., ink-jet, roll-to-roll, and 3D printing). Printing technologies using nanocarbon inks have advantages in feature resolution, device cost, and mass production compared with other device-fabrication processes like vacuum deposition and photolithography.^[15–17] On the other hand, the ink can be imbued with dynamic functionalities by choosing different types of nanocarbons with distinct chemical or physical properties. Indeed, formulating nanocarbons into ink is an effective way to preserve and integrate the extraordinary properties of nanocarbon into devices with desired responsive functionality, which can efficiently convert external perturbations into detectable electrical and optical signals as well as mechanical actuation.

Motivated by such programmability, nanocarbon inks are being investigated for applications such as real-time health-monitoring sensors,^[18–20] implantable medical devices^{[21, 22]} and autonomous robotics^{[23, 24]} to meet the increasing demand for flexible electronics and smart devices that can be used to detect and respond to different external perturbations. As a result, there has been increasing interests in utilizing nanocarbon inks as the fundamental building material for research and commercial innovations, as evidenced by the increasing number of publications and market expansion (Figure 2).

Figure 2.

Trends in research and applications based on nanocarbon ink. a) Trends in research on nanocarbon ink (using CNT and graphene ink, as examples) illustrated by the number of publications each year (Data from SciFinder on July 2020). b) The increasing commercial market size in China of conductive ink made from graphene and CNTs in 2014, 2019 and 2024 (prediction).^[225] The total market size presented by the area of the pie plot and indicated in black fonts.

Although nanocarbon inks feature many novel functionalities beyond color compared with traditional carbon inks, their basic formulation recipes remain largely unchanged. For example, traditional Chinese ink is made of charcoal or lamp black, water, and animal gelatin.^{[2, 3]} Likewise, modern nanocarbon inks are similarly made from three components, including functional nanocarbon materials, solvent, and additives (e.g., surfactants, binders, or viscosity modifiers). Nanocarbons are the key component that renders the ink with different functionalities and can be formulated into a liquid solution, paste, or solid ink. In general, bulk materials such as graphite and raw CNT powders are first mixed with surfactants^{[25, 26]} or polymers^[27] in aqueous solutions or organic solvents,^[28] followed by intense sonication to disperse the material and then centrifugation to remove the undispersed aggregates.

Nanocarbon ink can be amenable to modern printing technologies including inkjet, screen, spray, and 3D printing. In order to achieve the desired printing resolution and device performance, the combination of the different ink components must be adjusted to optimize the key parameters of the ink, such as loading, viscosity, and surface tension to meet the requirements of different printing techniques.^[29] As the dispersion,^{[30, 31]} surface functionalization^{[32, 33]} and solution processing of nanocarbons in different solvents^[30] as well as printing technologies^{[14, 16, 17]} have been recently reviewed by others, we refer the readers to these excellent articles for more detailed discussions on these topics. It is also important to note that although this review is focused on nanocarbons, inks formulated from many other emerging nanomaterials are generally compatible with nanocarbon inks and can be potentially printed using the same techniques. We direct the readers to review articles that already exist in those areas for further information.^{[14, 16, 17]}

This review aims to summarize the latest advances in the field of nanocarbon inks, with an emphasis on their use as programmable materials for smart devices that can respond to external stimuli. We will first introduce in Section 2 the fundamental properties that uniquely position nanocarbons for the development of a new generation of carbon inks. We note that in this review we have conceptually extended the definition of ink to a composite material that can be used to transfer/deliver nanocarbons on substrates. Therefore, the scope of our discussion is not constrained by the physical state of the ink, which may be in liquid, paste, or even solid form. Emergent applications, including examples in flexible electronics, sensors, and actuators, will be discussed in Section 3, categorized based on the electrical function and external perturbation (e.g., electrical, chemical and optical stimuli) to highlight the dynamic multifunctional nature of nanocarbon inks. In Section 4, we will discuss the current challenges in ink formulation and further advances required from the perspective of materials engineering. Some general design principles and specific techniques developed in nanocarbon solution processing will be closely examined to serve as a guide to enable inks with precise nanocarbon structures, high material loadings, and reduced additives. Approaches to tailor the ink-substrate interactions and build ordered structures in printed devices will also be discussed in this section. Finally, we conclude by summarizing the state of research on nanocarbon ink and highlight some new opportunities for future studies.

2. Nanocarbons with properties beyond color

Nanocarbons are a diverse family of nanomaterials with exceptional physical and chemical properties depending on their specific structure. These materials can be fully conjugated or non-conjugated (as well as partially conjugated). Conjugated nanocarbons include, most notably, quasi zero-dimensional fullerenes,^[4] one-dimensional CNTs,^{[5, 6]} and two-dimensional graphene.^[7] These carbon allotropes feature a fully conjugated π-electron system of bonded sp² carbon atoms. It is important to note that both CNTs and graphene can form stacked structures, which, in the case of CNTs, can be further subcategorized into single-wall CNTs (SWCNTs), double-wall CNTs, and multiwall CNTs. For graphene, the literature may refer indiscriminately to different structures as graphene, including single-layered, double-layered, and multilayer graphene, although the properties of these materials can be distinctly different. For clarity we will reserve the use of graphene to the single-layered structure and differentiate the different structures when possible. Depending on the structure, graphene and SWCNTs can possess properties to an extraordinary scale in terms of electrical conductivity, semiconducting properties,^[10] mechanical strength,^[11] and thermal transport (Table 1).^[34]

Table 1:

Physical properties of different nanocarbons. Data collected from references ^{[7, 8, 10, 11, 34, 41, 49, 51, 56, 60, 76, 83, 112, 231–238]}

		Historical Carbon Material		Beyond Color Nanocarbon Material
		Amorphous carbon51,231,232,233	Graphite16,234,235	Fullerenes236	Semiconducting SWCNTs10,34,56,76,112	Metallic-SWCNTs34,49	Graphene7,11,34,56,60,237	Graphene oxide83,238	Carbon dots8,41
Electrical	Electrical conductivity [S/m]	≈100	≈300 (┴) 3×105 (//)	≈1×10−3	3.1×106		1×108	Depending on the oxidation condition. ≈1×10−3	-
	Current density [A/cm2]	≈60	≈25	-	≈4×109		≈4×108	-	-
	On/Off current ratio	-	-	-	≈1×106	-	<10	-	-
	Subthreshold swing [mV/dec]	-	-	-	≈120	-	≈20	-	-
	Charge carrier mobility [cm 2 /V∙s]	≈10	≈10,000 (//)	-	> 79,000	-	≈10,000–15,000	-	-
Optical	Visible optical color	Black	Black	Black	Colorful for chirality enriched samples		Black	Black	Black
	Quantum yield [%]	-	-	-	< 1–2	-	-	-	≈20–95
	PL lifetime [ns]	-	-	-	< 1	-	-	-	≈4–5
Mechanical	Young’s modulus [Tpa]	≈0.1 (based on thin film)	-	-	≈1		≈1	≈0.2–0.3	-
	Tensile strength [Gpa]	-	-	-	≈10–60		≈130	-	-
Thermal	Thermal conductivity [W/m∙K]	≈0.2	≈1 (┴) ≈1×102 (//)	≈0.1	Principle direction: > 3000 Radial direction: ≈1.5		≈5000–5300	Depending on the oxidation condition, ≈1	-

Graphene is a two-dimensional honeycomb network of sp²-bonded carbon, while a SWCNT can be conceptually viewed as a seamless cylinder of graphene. The structure of a SWCNT, known as its chirality, can be uniquely indexed by a pair of integers (n,m). Both graphene and SWCNTs are atomically thin, featuring a surface area as large as ≈2600 m² g⁻¹, which is useful for achieving chemically functional surfaces and electrical percolation in printed structures.^[35] They are also lightweight (at a density less than ≈1.4 g cm⁻³) and can be synthesized from inexpensive precursors like hydrocarbons and alcohols.^[36] The combination of these extraordinary properties has motivated the use of nanocarbons in applications such as conductive ink and electronic devices.^{[12, 15, 16]} One of the most significant differences between graphene and SWCNTs is that graphene is a gapless semi-metal while SWCNTs can be metallic or semiconducting depending on their chirality.

Non-conjugated (or partially conjugated) nanocarbons include nano-diamond,^[37] graphene oxides (GOs),^[38] and carbon dots.^{[8, 9]} Although some locally-conjugated-areas may exist, the overall structure of these systems is characterized by sp³ bonding. Particularly, in GOs and carbon dots there exist various functional groups (e.g., hydroxyl, carboxyl, or sulfonate groups) or heteroatoms (e.g., oxygen, nitrogen and sulfur), making these nanomaterials poorly conductive or even insulating. However, non-conjugated nanocarbons may have other properties useful for ink applications.

Due to the high contents of hydroxyl and carboxyl groups, GOs feature superior water solubility, allowing for the formulation of aqueous ink at high concentrations without using surfactants or other stabilizers. More notably, since the viscoelastic properties of GO solutions vary considerably depending on concentration, the ink made from GOs can be easily tailored for different printing techniques. For example, ink with a low concentration of GOs can be inkjet printed into continuous thin films, while highly-concentrated GO solutions behave like a gel with higher elastic modulus, making it possible to build complicated hierarchical structures by 3D printing. Once printed, the GOs can be subsequently reduced by thermal annealing or photo and chemical reduction to yield reduced GOs (rGOs) with improved electrical conductivity.^[39] Although the electrical conductivity and mechanical properties of rGOs are generally inferior than that of pristine graphene, GOs still represent a useful strategy to fabricate certain types of devices, potentially in a large-scale and low-cost manner.^[40]

Carbon dots exhibit similar properties as GOs in terms of water solubility, chemical functionality, and electrical conductivity. Meanwhile, carbon dots also display tunable photoluminescence (PL), which makes them suitable as fluorescent ink for applications in biosensing,^[41] information encryption,^[42] and anticounterfeiting.^[43] We note that many new types of nanocarbons, including carbon nano-horns,^[44] carbon cones,^[45] and linear carbon chains (e.g., carbynes),^[46] are also discovered and synthesized with interesting electrical and optical properties. However, these nanocarbons have yet to be integrated into ink applications.

The variety of nanocarbon properties makes them particularly suitable for advanced ink applications. Unlike traditional ink materials, such as carbon soot or charcoal, which have little functionality beyond color, nanocarbons (e.g., graphene and SWCNTs) can exhibit exceptional electrical, mechanical, optical, and thermal properties. In this section, we will review the properties of different nanocarbon materials, with a focus on graphene and SWCNTs in their idealized forms (i.e., single nanotube or single-layer crystalline structure). We note in many cases the extraordinary properties observed in individual nanocarbons have yet to be realized in printed structures. However, these fundamental properties set the upper limits of the expected performance and a critical review of these limits will help highlight the vast potential of nanocarbon inks.

2.1. Electrical conductance of SWCNTs and graphene

SWCNTs and graphene have shown extraordinary electrical transport properties. An individual SWCNT can reach the ballistic conductance limit of 4e²/h over the μm length scale, where e is the electron charge, and h is Planck’s constant.^[10] Ballistic conductance occurs when the mean free path of the electron surpasses the length scale of the conductor. In this condition, the contribution of electron scattering to resistivity becomes negligible and the device resistance is limited only by the contact resistance.^[10] Besides this high conductance, SWCNTs are blessed with large current-carrying capacities, far surpassing those of traditional metals, such as copper and gold.^{[10, 47, 48]} For example, metallic SWCNTs can reach a current density on the order of ≈10⁹ A cm⁻², which is 8-orders of magnitude higher than traditional carbon ink materials, like amorphous carbon, and about 1-order of magnitude larger than gold nanowires.^[49–51] Graphene similarly exhibits high electrical conductivity with a current density on the order of ≈10⁸ A cm⁻², the upper limit set by the intrinsic electron-phonon scattering in graphene.^[52] This high conductivity and current density make graphene and CNTs attractive candidates for efficient and small-footprint printed electronics^{[53, 54]} and transparent conductive films.^[55]

2.2. Semiconducting properties

An important advantage of SWCNTs over graphene as well as traditional carbon ink materials is the fact that they can be semiconductors, featuring high carrier mobility (on the order of 79 000 cm² V⁻¹ s⁻¹ for individual SWCNTs)^{[56, 57]} and tunable bandgaps from ≈0 to ≈2 eV depending on the nanotube chirality.^[58] Like bulk semiconductors, electrical transport in a CNT semiconductor can be electrically gated by the field effect, with an ON-state conductance on the order of ≈7000 μA μm⁻¹ and ON/OFF ratios approaching ≈10⁶.^[10] These properties, together with the high carrier mobility, make SWCNTs suitable for high performance field-effect transistors (FETs).

As a semimetal,^[38] graphene can be considered as a zero-bandgap semiconductor, wherein the conduction and valance bands meet at the Dirac point.^[59] A negative gate bias pushes the Fermi level above the Dirac point, leading to electron conduction while a positive gate voltage increases hole conduction, allowing graphene to behave as an ambipolar transistor for sensor applications. Crystalline single-layered graphene can display an ON-state current of ≈2700 μA μm⁻¹, although its ON/OFF ratio is intrinsically low.^[60] To open a band gap and increase the ON/OFF current ratio, modifications through chemical functionalization,^[61] geometric constraints,^[62] bilayers and heterojunctions,^[63] and substrate interactions^[64] have been explored. Such devices often exhibit low threshold swings (down to 20 mV dec⁻¹) on the order of that seen in traditional semiconductors such as silicon.^[65]

The electrical properties of semiconducting SWCNTs and graphene make them attractive as energy efficient low power logic devices. However, device processing and materials inhomogeneity have made it difficult to achieve industrial figures of merit, particularly in printed thin film devices^[66] where film thickness, junction resistance, and heterogeneity can rapidly degrade performance. However, promising advances are being made in this field, and will be reviewed in Section 4.1.

2.3. Optical properties

Carbon materials, including naturally abundant charcoals and artificially synthesized CNTs, are typically black in color due to their broad-band absorption and high refraction of visible light. This optical property of various carbon materials has been well known and utilized for thousands of years to make black inks for writing and decorative purposes. Although most nanocarbons appear to be equally black to the naked eye, these materials actually feature different degrees of “blackness.” For example, conventional black paint and carbon black show a blackness of ≈90–95%, with ≈5–10% of the incident photons reflected,^[67] while CNTs (in the form of a CNT forest) can exhibit a blackness of more than ≈99.995% over the ultraviolet to terahertz region, with only ≈0.005% of the incident light being reflected (Figure 3a).^[68] Commercially available Ventablack^® paints made from CNTs have also shown a blackness higher than ≈99.8%.^[69] The inclusion of CNTs as pigments for ultra-black inks can open avenues for many useful applications, including low reflective shielding, high performance lenses, and astronomy radiation detectors.^[68]

Figure 3.

Optical properties of nanocarbons. a) Reflectance of various forms of CNTs with other black materials over the visible, near-IR, and mid-IR wavelength range. Reproduced with permission.^[68] Copyright 2019, American Chemical Society. b) Photographs of a subset of chirality-purified SWCNTs. Reproduced with permission.^[73] Copyright 2017, American Chemical Society. c) Schematic of the density of electronic sates of a semiconducting SWCNT. Reproduced with permission.^[74] Copyright 2002, American Association for the Advancement of Science. d) Absorbance spectra of chirality-enriched SWCNT species. The E₂₂ absorption peaks (labeled by the dashed black line) are responsible for the colors in the visible, as shown in b, whereas the E₁₁ absorption peaks fall in the shortwave infrared, opening opportunities for applications such as bioimaging and IR photodetectors. Reproduced with permission.^[226] Copyright 2009, Nature Publishing group. 2D PL maps of e) pristine (6,5)-SWCNTs and (f) OCC-tailored (6,5)-SWCNTs (p-nitroaryl OCCs). Reproduced with permission.^[77] Copyright 2013, Nature Publishing Group. g) Optical photograph of aqueous solutions of dispersed carbon dots excited at different wavelengths (in nanometers), as indicated. Reproduced with permission.^[9] Copyright 2006, American Chemical Society.

Although the synthetic raw materials are black, single-chirality pure SWCNTs are actually colorful (Figure 3b).^{[58, 70, 71]} For example, the semiconducting (6,5)-SWCNT solution appears purple, the semiconducting (7,6)-SWCNT is greenish, while metallic (7,4)-SWCNT is yellow. The color of metallic SWCNTs, unlike conventional metals such as gold, mainly depends on their discrete excitonic transitions between the van Hove singularities, the frequency of which is related to the electron density and nanotube structure.^{[71, 72]} The color of semiconducting SWCNTs originates from the chirality-dependent, direct bandgap transitions.^{[58, 70, 73]} Incident light is maximally absorbed at the nanotube’s chirality-dependent E₁₁, E₂₂, and E₃₃ excitonic transitions (Figure 3c), displaying characteristic absorption peaks (Figure 3d) and color for each chirality-enriched SWCNT solution.

A unique optical property for semiconducting SWCNTs is their bandgap photoluminescence (PL). Figure 3e is an excitation-emission map of a chirality-enriched (6,5)-SWCNT solution, which shows a peak emission at ≈980 nm (E₁₁) upon photoexcitation at ≈565 nm (E₂₂). Depending on the chirality, nanotube PL features sharp emission with peak wavelengths tunable over the shortwave infrared (≈0.9–3 μm).^{[70, 74]} This unique optical property has enabled the formulation of SWCNTs as fluorescent paints for the fabrication of diagnostic biochemical sensors^[20] as well as artistic paintings.^[75] Apart from the chirality, the SWCNT PL wavelength and intensity are sensitive to the surrounding environment and surface defects.^[76] Particularly, covalent bonding of organic functional groups to the nanotube sidewall is found to create quantum defects or organic color-centers (OCCs) at which excitons can be trapped and radiatively recombine to emit single photons even at room temperature (Figure 3f). These defect-induced PL occurs at longer wavelength (E₁₁⁻) than the intrinsic nanotube E₁₁ emission, featuring bright PL in the shortwave infrared and intriguing photophysics.^{[76, 77]} OCC-tailored SWCNTs have demonstrated the early promise for deep tissue bioimaging, sensing, and room-temperature quantum light sources.^{[76, 78]}

Carbon dots are another family of nanocarbons that show interesting optical properties originating from their small size.^{[8, 9, 41]} Carbon dots can be synthesized from hydrothermal reactions and show high solubility in water, biocompatibility, and size-tunable PL in the visible wavelength range (Figure 3g).^[8] The PL quantum yields of nitrogen-doped carbon dots can be higher than 90%,^[79] making them a strong candidate material for fluorescent inks useful in applications such as sensing,^[80] information protection,^[42] and anti-counterfeiting.^{[43, 81]}

2.4. Mechanical properties

The covalent, conjugated sp² bonding network gives both graphene and CNTs exceptional mechanical properties. Graphene is the strongest 2D material known.^[82] Nanoindentation studies of graphene by atomic force microscopy measure a Young’s modulus as high as ≈1 TPa, a tensile strength of ≈130 GPa, and an elongation up to break of ≈25%.^[11] These values suggest that graphene can be ≈260-times stronger and 5-times stiffer than A36 structural steel. Graphene is also lightweight, with an areal density of ≈0.77 mg∙m⁻²; a single sheet of graphene the size of a football field, if attainable, would weigh less than 1 g.^[13]

The mechanical properties of GOs at the atomic level are notably lower than those of pristine graphene, presumably due to the existence of many structural defects. For example, the Young’s modulus of monolayer GO was found to be ≈ 200 Gpa, which is only one fifth of that of pristine monolayer graphene.^[83] Nevertheless, the Young’s modulus of GOs is still higher than many conventional materials. Together with their high solubility as well as chemical tunability due to different functional groups, GOs are increasingly used as filler materials in polymer nanocomposites for the mechanical enhancement.

The tensile strengths of individual SWCNTs were recently determined to range from ≈25–66 GPa, depending on the chiral structure, with the highest tensile strengths found in smaller-diameter, near-armchair nanotubes.^[84] The CNTs are relatively soft in the radial direction, with a Young’s modulus decreasing with diameter.^[85] In part due to the combination of both high strength and mechanical flexibility, CNTs and graphene are intensively investigated for flexible devices^[29] or composite materials.^{[86, 87]}

2.5. Surface area and porosity of nanocarbons

One of the most significant differences between nanocarbons and graphite is the drastically increased surface area. The specific surface area of natural graphite is only ≈0.6 m² g⁻¹,^[88] while the carbon soot/black features an increased surface area ranging from ≈50 to ≈400 m² g⁻¹ depending on different synthetic methods.^[89] At the single layer limit, the surface area of graphene, GOs and SWCNTs can exceed ≈2600 m² g⁻¹.^[35] Such high surface area is particularly important for hosting a high density of molecular probes for sensing applications, achieving electrical percolation in printed structures, as well as applications in energy storage, catalysis and environmental remediation.

Notably, metal organic framework (MOF) derived porous nanocarbons can achieve a surface area exceeding 5000 m² g⁻¹.^[90–93] Since the structure of MOFs is highly designable and tunable, the MOF derived porous nanocarbons feature controllable pore sizes and surface areas. Although the applications of these porous carbon materials are still mainly focused on the field of energy storage and catalysis,^[91] considering their similar physical and chemical nature with other nanocarbons such as GOs and carbon dots, commonly used ink formulation strategies should be applicable to make functional inks from these MOFs derived nanocarbons.

2.6. Chemical Properties

Chemical functionalization provides a potentially powerful route to further control the structure and properties of nanocarbons for both added functionalities and ink formulations. CNTs in particular have been extensively investigated in this area. The sidewalls of CNTs can be covalently functionalized, such as by reacting them with diazonium salts in water or in chlorosulfonic acid.^{[76, 94]} Such functionalization has been used to tailor the physical and chemical properties of CNTs, including the solubility. Lately, work from our groups and many others have shown a small number of controlled functional groups can have a drastic positive impact on their optical and electronic properties.^{[76, 95]}

However, in most cases, covalent functionalization of the SWCNT sidewalls typically come at the sacrifice of the material’s electrical and optical properties due to the disruption of the conjugated pi-electron structure, thereby limiting its device applications. Double-wall CNTs provide a solution for addressing this tradeoff of chemical functionality and electrical properties in CNT ink formulations. Double-wall CNTs consist of exactly two concentric SWCNTs where the inner tube is uniformly wrapped by the outer tube.^[96] By outer wall-selective covalent functionalization,^{[95, 97, 98]} we can add chemical functionality to the surface while maintaining the structural integrity and thus the electronic properties of the inner-tube, creating a tube-in-a-tube structure.^[98] This strategy allows for high degrees of covalent surface functionalization to afford water solubility^[95] and enable chemically gated field-effect chemical sensors.^[98] Chemical gating significantly simplifies the device configuration of FET sensors, requiring only two physical electrodes (i.e., source and drain, without the need of a gate electrode) to achieve the same sensitivity while drastically enhancing the chemical selectivity.^{[98, 99]} Although the advantage of chemical gating has only been demonstrated with thin films, this device concept can be readily applied to printed devices.

Optical lithography can also be used as a simple, post-processing method of modifying the nanocarbon structure to impart or remove chemical functionality in printed thin films. For example, Ng et al. demonstrated that it is possible to selectively remove the chemical functional groups from CNTs by irradiating the film with a laser. By doing so, highly conductive patterns can be directly written into an otherwise chemically functionalized and electrically insulating thin film at micron-scale resolution.^[100] An analogous process can also be used to add chemical functionality to SWCNT thin films, wherein the film is selectively functionalized under irradiation with high spatial resolution using light-activated chemistry.^[101] The ability to covalently attach and reversibly remove functional groups with the aid of light provides a tool for modifying the printed nanocarbon devices, or enables direct writing of functional circuits.

3. Emergent Applications

The remarkable physical properties of nanocarbons have motivated their formulation into inks for direct printing into a variety of functional devices. In these applications, the nanocarbon material generally serves as the functionality-enabling element, or transducer, that converts external stimuli into detectable signals of another form. Figure 4 summarizes how different types of nanocarbons can be used to transduce stimuli into responsive signals. For example, actuators printed from graphene or CNTs and polymer inks can transduce external heat/light changes into a shape deformation.^[102] Interestingly, from a transducer perspective, electronic devices such as thin-film FETs may also be viewed as converting electrical inputs (e.g., a gate voltage) into an output source-drain electrical signal.^[103] Here we examine recent progress in developing such sensor and actuator applications based on the type of stimuli, including electrical, chemical, mechanical, heat, and light, as examples of the applications of nanocarbon ink.

Figure 4.

Schematic illustration showing how different types of nanocarbons can transduce different stimuli to detectable signals.

3.1. Electrical stimuli and flexible electronics

CNTs and graphene inks are uniquely positioned for printed electronics such as FETs and mechanical actuators. FETs can electrically switch ON and OFF in response to an applied electrical field while mechanical actuators convert an applied current into mechanical work. In a FET, a voltage is applied to a gate electrode, which switches the output source-drain current to a high or low state, forming the basis of binary logic used in computation. The efficiency of the transduction of the applied voltage to the output current determines the power efficiency, speed, and output current of the device. The attractive electrical properties of nanocarbons have enabled high performance transistors using traditional semiconductor manufacturing processes.^[66] However, problems with uniformity, structure-dependent electrical properties, and reproducibility have thus far limited the role of nanocarbons in printed electronics.^[66] Printing can significantly lower fabrication costs (Figure 5a), but at the expense of the device performance as additives and surfactants compromise the electrical transport properties of the nanocarbons.^{[104, 105]} Some recent work has sought to address these issues, with the aim of meeting several performance goals of the electronics industry, including high on-state conductance, low gate voltages, and low subthreshold swings.

Figure 5.

Nanocarbon inks enable fully printed high performance field effect transistors. a) Highly purified semiconducting CNTs can be printed into homogeneous films with high reproducibility and tunability. Reproduced with permission.^[103] Copyright 2017, Wiley-VCH. b) CNT inks can be used to print both the semiconducting channel, and the metallic contacts. Metallic CNTs enable low contact resistance compared to metal nanoparticle ink formulations like gold and silver. This enables a large electrical current and more efficient device operation. Reproduced with permission.^[109] Copyright 2016, American Chemical Society. c) The responsiveness of printed CNT field effect transistors, measured by the subthreshold swing and gate voltage, can be improved by using high purity semiconducting CNTs of a single chirality. Reproduced with permission.^[103] Copyright 2017, American Chemical Society. d) Responsiveness can further be improved by using an ultrathin and high-quality dielectric such as HfO₂. Reproduced with permission.^[118] Copyright 2020, Elsevier Ltd.

Specifically, one industrial goal is to increase the on-state conductance (normalized by channel width) to > 0.1 mS μm⁻¹,^[66] as a higher ON-conductance can reduce the device power consumption for more efficient operation and less heat generation.^[106] We note that ON-state conductance is strongly dependent on channel length: in the work cited above, a channel length of 120 nm was used which was free of CNT junctions. At a given source-drain voltage, the current in an FET is limited by the total resistance of the device constituents,^[107] which in the case of nanocarbon-ink-based materials derives from the contact resistance and resistance at the junctions between CNTs or graphene flakes in the printed films. One approach to improve the ON-state conductance is to reduce the contact resistance.^{[108, 109]} In the case of semiconducting SWCNTs, contact resistance arises from the Schottky barrier, which is created from the mismatch in the work function of the electrode material (such as Ag and Au metals) and the semiconductor’s conduction band.^[109] For graphene, a semimetal, the origin of the contact resistance is different, but also presents a problem and a limiting factor in total device conductance.^[110] To solve the problem, metallic CNT ink has been developed and used to print electrodes that show significantly reduced contact resistance. Franklin and co-workers^[109] demonstrated that the contact resistance can be reduced from ≈80 kΩ μm to 16.8 kΩ μm using metallic CNT contacts printed both above and below to sandwich the semiconducting SWCNT channel as opposed to contacts printed from silver ink on top of the channel (Figure 5b). This improvement was attributed to more efficient charge transfer between the metallic nanotubes and the semiconducting channel, as well as a high surface area contact due to the strong adhesive forces and close packing of the metallic and semiconducting CNTs at the interface. Double layer electrodes composed of a layer of metallic CNTs and a layer of Ag printed on top have also recently been demonstrated,^[111] combining both the high conductivity of metal contacts with the favorable mechanical and charge transfer properties of metallic CNTs. Such a hybrid metallic CNT/Ag electrode achieved a 60% reduction in contact resistance compared with those using only Ag. Considering their increasing commercial availability, relatively low cost, mechanical flexibility, as well as compatibility with printing technologies, semiconducting and metallic CNT inks may hold the potential to realize all-carbon printed electronic devices with the combination of high device performance and flexibility.^[112]

The second source of resistance in printed nanocarbon FETs, which can cap device conductance, is the junction resistance. Junction resistance arises from electron transfer between nanotubes with different chiralities and from the trapping of contaminants at such junctions.^[107] While chirality-based junction resistance can be reduced using nanotube purification techniques to achieve single chirality materials, as will be discussed later,^[103] junction resistance arising from contaminants trapped at junctions can be difficult to address, especially in ink formulations where additives are present. In this regard, engineering the additives so they are easily removed during post-printing treatments can be a useful approach to decrease the junction resistance, which will be discussed in Section 5.3.

An alternative approach to reduce the effect of junction resistance is to minimize the FET channel length, thereby minimizing the number of junctions through which charge carriers must pass. However, this approach is more challenging for fully printed devices due to the limited resolution in printing and alignment of multiple layers (e.g., > 30 μm for typical inkjet printers).^[113] A notable way to circumvent this issue is to use a vertical device geometry, where the channel length is defined by the thickness of the printed semiconducting SWCNT film, not the horizontal gap between the two printed electrodes as in the traditional device architectures.^[103] This technique was demonstrated with pure (6,5)-SWCNT semiconductors dispersed in tetrahydrofuran with polytetrafluorethylene and terpineol and printed with an aerosol jet printer (Figure 5c). The films thickness (i.e. the vertical channel length) was scaled down to 50 nm, far below the channel length resolution of a traditional device architecture. This device demonstrated an ON-state current density of > 25 A∙cm⁻² and a sub-threshold swing of < 150 mV dec⁻¹. However, because the printed nanotubes are deposited along the substrate rather than the vertical channel direction, the electrical transport still has to occur through many tube-tube junctions along the thickness, posing a major challenge and opportunity for further improvement using this vertical device architecture in printed electronics.

While not contingent on nanocarbon ink properties, it is worth noting that some recent works using metal nanoparticle inks in combination with hydrophobic/hydrophilic surface patterning^[114] or immiscible solvents to create sub-micron electrode gaps on semiconducting CNT films.^[115] This type of hybrid approaches provides a potential path forward to all-inkjet printed FETs with ultrashort channels.

Printed nanocarbon transistors have also been improved through various methods to reduce the operating gate voltage and subthreshold swing. In order to compete with current commercial FETs, these parameters are expected to be < 2 V and < 120 mV dec⁻¹, respectively.^[66] A promising approach to achieving low gate voltages in fully printed nanocarbon FETs is through the use of an ionic gel gate material, which uses an electrolytic double layer at the semiconductor surface to gate the source-drain current, and can be more efficient than traditional metallic/dielectric layered gates. Such a gate material can be processed as an ink and printed directly on the nanocarbon channel and electrode materials. For example, Bao and co-workers recently demonstrated fully printed and flexible transistors using a poly(vinylidene fluoride-co-hexafluoropropylene) gate dielectric, which exhibited 0.2 mA cm⁻¹ ON-state currents with an operating gate voltage of < 1 V.^[116]

For graphene FETs, an approach for improving the gating performance is to use printed two-dimensional heterojunctions as the gate dielectric. A two-dimensional heterojunction refers to the interface between graphene and another two-dimensional material, in this case an insulator. For example, hexagonal boron nitride inks have been printed on top of printed graphene channels, providing a thin, flexible, environmentally stable, and pinhole-free dielectric layer. The resulting devices had exceptional mobility of ≈91 cm² V⁻¹ s⁻¹ and could be operated with a gate voltage of < 5 V.^[117] Similarly, coupling a high quality HfO₂ dielectric (deposited by atomic layer deposition) with inkjet-printed CNT channels has enabled a subthreshold swing of 92 mV dec⁻¹, which is well within the industrial requirement (Figure 5d).^[118]

Nanocarbons can also be used to transduce electrical stimuli into mechanical actuation. While it is possible to directly convert an electrical current to a mechanical output in a graphene flake^[119] or in twisted fibers of pristine CNTs,^{[120, 121]} the translation of these phenomena to fully printed nanocarbons is not currently feasible. Instead, printed actuators currently use either nanocarbon/polymer composite ink or multilayer structures with nanocarbon conductors printed on polymer or paper films, wherein the polymer does the mechanical work in response to joule heating of the nanocarbons.^{[102, 122]} As the heat generated by joule heating is directly proportional to the current squared, metallic CNTs are attractive for this application, compared to other conductors, because of their high current-carrying capacity. Additionally, CNTs and graphene have high thermal conductivity, allowing the heat to quickly spread throughout the matrix.^[122]

Since these devices are made from macroscale, uniform nanocarbon films, printing is a useful fabrication strategy. A variety of printing techniques can be used to translate this general actuation mechanism into working devices. A particularly useful technique is inkjet printing, thanks to its ability to tune the thickness of the printed layer using multiple printing passes or through the ink composition. Such an approach was demonstrated by Wang et al. and used to fabricate self-folding origami structures on polymer films.^[123] In this case, the ink was composed of oxidized SWCNTs dispersed in a water/ethanol solution. A voltage applied to inkjet-printed regions heated the underlying substrate, inducing it to bend into three-dimensional structures. The induced mechanical force was as high as 0.088 N, which is 60-times the weight of the origami structure. By applying a voltage in cycles, the origami structure can be electrically triggered to “walk” 1 cm per step. Multiple printing steps were used to tune the thickness and therefore resistance of the CNT film, enabling the device to operate at small driving voltage compared to a thin film.

An additional printing method is spray coating, which is capable of achieving uniform nanocarbon coatings for mechanical actuator device fabrication on a large scale.^[102] For example, Sachyani et al.^[102] showed that ink made from 1 wt% CNTs in water with a commercial polymer dispersant can be spray coated down to a thickness of ≈1 μm on polyamide films. Under an applied voltage of 100 V, the CNT-polyamide bilayer demonstrates a mechanical response but only by a few millimeters. However, by using thicker layers of CNTs with higher conductance and sandwiching the polyamide film between the printed CNT layer and a polyurethane acrylate layer, the mechanical response of this system was drastically improved, demonstrating a deflection of up to 300 degrees. This improvement was attributed to the larger thermal expansion of the polyurethane, which exerted an additional force on the polyamide film.

3.2. Chemical stimuli

Nanocarbon inks can also be used to create systems that are responsive to various chemical stimuli, such as biomolecules and ions, eliciting an electrical or photonic signal due to their intrinsic electrical and optical properties. This responsiveness can be used in chemical sensing, environmental monitoring, and disease diagnosis. For such applications, nanocarbon inks can be incorporated into four basic types of devices, including electrochemical sensors, field effect sensors, chemiresistors, and optical sensors (Figure 6a), all of which can be fabricated by scalably printing these multifunctional inks at low cost.

Figure 6.

Nanocarbon inks can be used to print devices responsive to chemical stimulus. a) Printed nanocarbon chemical sensors have achieved excellent sensitivity in recent works, demonstrating their potential for use in biomarker-base diagnostics. Nanocarbon inks can be used to print a variety of sensing platforms, namely electrochemical, field effect, chemiresistive, and optical sensors. Note the range of sensitivity is based on published reports, not necessarily reflecting the fundamental limit for each technique. b) Nanocarbon inks, particularly those made from graphene and its derivatives, are suitable for electrochemical detectors due to their high surface area, electrical conductivity, and electrochemically active sites. c) Semiconducting carbon nanotube and graphene inks have been widely used in field effect sensors, where their electrical transport can be sensitively modulated by chemical interactions between probes and target molecules such as single stranded DNA shown here. d) Carbon nanotube inks have been useful in making chemiresistive gas sensors, due to their high surface to volume ratio, and electrical conductivity which is sensitive to surface interactions, particularly with gas molecules. e) The bright and environmentally sensitive fluorescence of graphene quantum dots make them useful in printed fluorescence sensors and are commonly composited with molecular imprint polymers for enhanced selectivity

Electrochemical sensors harness the high surface area and electrical transport properties of nanocarbons. Graphene and its derivatives are the most commonly used nanocarbons in electrochemical sensors thanks to their electrochemically active edges (Figure 6b).^[124] Electrochemical sensors are composed of at least 3 electrodes (a working electrode, counter electrode, and reference electrode), with the signal transduction occurring from the oxidation or reduction of electrochemically active target molecules at the working electrode surface which can be printed from CNTs or graphene inks. This redox reaction generates an electrical signal, either a current or voltage spike, depending on the operating mode of the sensor.

In some cases, nanocarbon inks are used to decorate a current collector where the nanocarbons act as anchors for probes, bringing the target molecules close to the current collector surface, thereby improving the selectivity and efficiency of the device. For example, Mollarasouli et al.^[125] achieved 0.5 pg mL⁻¹ sensitivity for the heart failure biomarker tyrosine kinase by printing an ink composed of antibody-functionalized graphene quantum dots on an commercial carbon electrode. This sensitivity is sufficient for the detection of heart failure in a clinical setting and requires less time than traditional diagnosis using enzyme-linked immunosorbent assays, demonstrating the benefit of mass produced, sensitive, and portable sensors printed from nanocarbon inks.

Alternatively, the current collector can be printed from nanocarbon inks. In this case, the nanocarbon current collector can be further functionalized, not just added as anchor sites on an existing current collector. An example is demonstrated by attaching enzymes to nanocarbon surfaces for small molecule sensing.^[126] Integrating enzymes on electrochemical biosensors is an attractive detection method, as enzymatic reactions can produce electroactive intermediates or products that can be subsequently oxidized or reduced at the electrode surface. Claussen and co-workers recently demonstrated this with inkjet-printed graphene films, which were subsequently laser annealed and coated with platinum nanoparticles, which resulted in increased surface area and improved electrochemical activity, respectively.^[126] Phosphotriesterase enzymes were then covalently attached to the graphene. These enzymes convert paraoxon, the target molecule, into p-nitrophenol, which can be detected at a redox potential of 0.95 V. This device exhibited a detection limit of 3 nM and excellent selectivity, demonstrated by the lack of current response to a panel of interferent molecules similar in structure to the target. An additional sensing option is the attachment of antibodies to a printed nanocarbon electrochemical sensor.^[127] In this work, graphene-nitrocellulose ink was aerosol jet printed to form an electrode, followed by linkage with antibodies for interferon gamma (IFN-γ) and interleukin-10 (IL-10) cytokine monitoring. The devices showed detection limits of 25 and 46 pg mL⁻¹ for IFN-γ and IL-10, respectively, which are sufficient for use in clinical detection.

Field effect sensors utilize the spatial proximity of a target probe interaction to a semiconducting channel to electrostatically gate the current flowing through the semiconductor (Figure 6c).^[128] This effect can be measured either as a shift in the threshold voltage of the semiconductor channel (determined by sweeping the gate voltage) or as a change in the source drain conductance with the gate voltage set near the threshold value. Nanocarbon inks can be used in two different ways in field effect sensors: firstly, inks made from semiconducting CNTs or graphene sheets can be printed to act as the electrically conductive channel and subsequently functionalized with probes such as antibodies, single stranded DNA, or aptamers; secondly, nanocarbons can be used to decorate an existing semiconducting channel to act solely as chemically functional anchors for probes. This latter case was recently demonstrated with a pentacene semiconducting layer, with graphene oxide ink inkjet printed on the channel;^[129] thiolated single stranded DNA probes were then covalently linked to the GOs. Subsequent binding between the target DNA and the probe increased the ON-state conductance in the pentacene layer of the device, enabling detection of 100 fM DNA targets.

Printed devices are also attractive as chemiresistor sensors due to their simplicity. Although typically not as sensitive as FET sensors, chemiresistor sensors require only two electrodes instead of three that are typically necessary for field effect sensors. Chemiresistors require the target to interact with the active material of the device, thereby generating a change in the device resistance (Figure 6d). Printed nanocarbon chemiresistors have proven especially useful in gas sensing applications. For example, Wu et al. printed a chemiresistor from a graphene-tungsten oxide suspension and showed the sensor exhibited 10 ppm sensitivity to NH₃ with minimal baseline drift and a response time of only tens of seconds.^[130] Both drift and response time have been two of the key challenges in chemiresistive sensing. In another example, Ammu et al. demonstrated a chemiresistor made from CNTs that had been inkjet-printed from a surfactant suspension onto a paper substrate, which achieved 250 ppm sensitivity for nitrogen dioxide.^[131]

Due to their high solubility in water, GOs can easily form a stable colloid ink without the need of dispersants. Although GOs themselves have rarely been used as chemiresistors due to the poor conductivity, various post-printing methods have been developed to reduce GOs to rGOs for sensing purposes. For example, Robinson et al. used hydrazine hydrate vapor to reduce the thin continuous network made from GOs to yield rGO devices that can detect chemical-warfare agents and explosive down to the ppm level.^[132] Similarly, Dua et al. demonstrated low-cost inkjet printing to fabricate complex pattens on different flexible substrates. After being chemically reduced to rGOs by ascorbic acid, the printed patterns can be directly used as chemiresistors for organic vapor detection.^[133]

In addition to electrical-based sensors, nanocarbon inks can also be used to make photonic-based chemical sensors, converting molecular recognition events into light emission. This optical transduction of chemical binding is made possible by charge transfer or conformational changes induced by target-probe interactions near a nanocarbon fluorophore, leading to new or shifted emissions. For this type of sensing mechanism, the optical properties of nanocarbons and their tunable chemical functionality are particularly valuable.^{[76, 134]} Chemical tunability enables different probes to be attached to the nanocarbon surface, as well as the ability to tune the excitation and emission wavelengths by chemical modification.

A promising development in optical sensing with nanocarbon inks is the use of molecularly imprinted polymers (MIPs).^[135] MIPs are structures which contain nanoscopic cavities that are designed to trap target molecules with high selectivity. A recent development in this area is the coupling of MIPS with graphene quantum dots (GQDs), which can be subsequently printed on a substrate (Figure 6e).^[136] The template molecules (i.e. structural analogs of the target molecule) is blended together into an MIP-GQDs ink. After printed, the template molecules are removed with a solvent, leaving behind the molecular cavities, some of which are in close proximity to the GQD fluorophores. As the target molecules fill the molecular cavities left by the template in a manner reminiscent of antibody-antigen bonding,^{[137, 138]} the GQD fluorescence shifts due to changes in the local dielectric environment caused by the binding event between the target and cavity in the polymer. This approach has been used to detect methamphetamine,^[136] atropine,^[139] and dopamine,^[140] as well as the multiplexed detection of cephalexin and ceftriaxone.^[141]

3.3. Mechanical stimuli

Printed devices from nanocarbon inks are also capable of reporting mechanical stimuli. Such devices work by transducing applied external force to an electrical or an optical signal through the nanocarbon. Nanocarbons are advantageous in this application for several reasons: their sensitivity can be tuned by changing the structure of the nanocarbon (for example, the CNT length or graphene flake size) or the thickness of the printed film. Such tunability is desirable for detecting applied strains across multiple orders of magnitude. Furthermore, inks can be used to make large scale uniform films, which is useful for mapping strain on macroscopic objects. Finally, films can potentially be made transparent for mapping strain on transparent surfaces.

Strain sensing can be achieved with a percolation network of metallic CNTs. We recently showed that the sensitivity of such strain sensors can be greatly enhanced using thin films made of short CNTs (average length ≈0.8 μm) which can exhibit up to ≈1000-fold increase in resistance under the external mechanical stretching.^[55] The CNTs form a percolation network on a stretchable polymer substrate, and when the substrate is strained, the percolation network breaks slightly, leading to a jump in the film resistance. By using short CNTs, the breakage of the percolation network can occur at much smaller strains than when using long nanotubes (≈3.2 μm vs. ≈0.8 μm), thereby increasing sensitivity.^[55] A different approach to achieving tunable strain sensitivity was recently achieved by varying the thickness of the printed nanotube layer.^[142] In this case, multiwall CNTs were spray coated from an aqueous dispersion onto a heated polymer substrate to form a percolation network. As the thickness was increased from 0.75 to 3.5 μm, the strain sensitivity could be varied by an order of magnitude. Printed structures made from graphene or rGOs behave similarly, as the application of strain causes more tortuous electron transfer between adjacent or overlapping graphene platelets, which display a “fish-scale” structure under strain.^{[143, 144]}

An alternative transduction method for strain sensing is based on the optical emissions of CNT films. Mechanical strain can induce shifts in the nanotube photoluminescence since the bandgap of the SWCNT is sensitive to applied stress. For example, an ink made from semiconducting SWCNTs dispersed by poly(9,9-di-n-octylfluorenyl-2,7-diyl) in toluene was spray coated on a flexible substrate and then covered with a second flexible substrate.^{[145, 146]} This bilayer “skin” is then transferred to a substrate such as a wall. Local strain induced in this bilayer skin structure by the substrate causes spectral shifts in the nanotube photoluminescence. Readings of the emissions across the whole skin structure can thus be used to create high-resolution strain maps of the underlying macroscopic surfaces.

3.4. Heat stimuli

Due to their excellent thermal conductivity, carbon-based nanomaterials, such as graphene and CNTs, have been widely used in programmable thermal responsive systems. It was found that the thermal expansion coefficient of packed films made of graphene or CNTs is one or two orders of magnitude lower than that of various polymers, and a simple CNT-polymer bilayer architecture can generate the mechanical motion of bending based on the asymmetric deformation of the two layers reacting to a temperature change.^{[147, 148]} Note that thermal energy can also be introduced in the form of light illumination (see Section 3.5) or electrical resistive heating.

As a result of these thermal properties, the combination of nanocarbon ink and materials such as hydrogel polymers provides the opportunity to create a light or heat controlled programmable material, which is promising for application in areas such as biomedicine and robotics.^{[149, 150]} For example, Javey and co-workers reported programmable hydrogel composites made with poly(N-isopropylacrylamide) (PNIPAM) and SWCNT ink.^[151] The hydrogel transitions from a hydrated to dehydrated state at temperature above the lower critical solution temperature (LCST) 32–33 °C, resulting in a drastic volume reduction. SWCNTs can be composited with PNIPAM by directly mixing sodium deoxycholate (DOC)-dispersed SWCNT ink with an aqueous solution of the NIPAM monomer, which is UV-polymerized after the composite hydrogel is deposited on low density polyethylene substrates to form a bilayer structure. When this bilayer structure is fabricated into the shape of an unfolded cube and placed in warm water (48 °C), the strain induced by the shrinkage of the hydrogel patches causes the folding of each hinge into a cube (Figure 7a). The box could also quickly unfold when placed in cold water (23 °C). The inclusion of SWCNT ink in the hydrogel was found to accelerate the polymer’s thermo-response by ≈5-fold (taking just 2.7 s for the hinge to reach a 90° angle), possibly due to the creation of a more porous structure for water diffusion.

Figure 7.

Recent advances in nanocarbon ink for thermally and light responsive devices. a) Programmable hydrogel composites made with PNIPAM and SWCNT inks. The incorporation of SWCNTs into the polymer matrix enables an efficient thermal-mechanical actuator in which the strain induced by the shrinking hydrogel leads to the cooperative folding of each hinge when placed in warm water. Reproduced with permission.^[151] Copyright 2011, American Chemical Society. b) CNT-coated polymer fibers as an infrared (IR) radiation gating textile. The triacetate cellulose bimorph fibers undergo mechanical actuation that changes the inter-fiber spacing as a function of humidity. The CNT coated fibers showed a relative change in IR transmittance of over 35% compared to below 10% for uncoated counterparts, which is mainly ascribed to the distance-dependent electromagnetic coupling between CNT-coated neighboring fibers. Reproduced with permission.^[152] Copyright 2019, American Association for the Advancement of Science. c) CNTs as the crucial component of a photoactuatable pen array for molecular printing. The pyramidal pens and backing layer were made of PDMS-CNT composites, which can locally expand upon light irradiation. When light is turned on, the local deformation moves the pen out of plane and into contact with the substrate so that ink material is transferred to the surface. Reproduced with permission.^[156] Copyright 2018, Wiley-VCH. d) Photoactuators with wavelength selectivity, made of a polymer-SWCNT bilayer structure. By utilizing SWCNT inks with different chirality distributions, chromatic actuators that are responsive to selected wavelength ranges are achieved. Reproduced with permission.^[148] Copyright 2014, Nature Publishing Group.

CNTs can also be used to regulate thermal energy and towards this aim has been incorporated into humidity responsive systems. For example, CNT ink was recently applied as a crucial coating material to enable dynamic gating of infrared radiation through a textile.^[152] By coating triacetate-cellulose bimorph fibers with a thin layer of CNTs, it became possible to effectively control infrared radiation through the fabric. Due to competing hydrophobic and hydrophilic effects, the triacetate-cellulose bimorph fibers undergo mechanical actuation that changes the inter-fiber spacing as a function of humidity. When hot and wet, the yarn collapses into a tight bundle, increasing the electromagnetic coupling between the CNT-coated fibers, which shifts the IR emissivity to spectrally overlap better with that of the human body and thus facilitate radiative cooling. When cold and dry, the reverse effect takes place. This mechanism enables the CNT-coated textile to increase the relative IR transmittance by 35%, while the uncoated control is less than 10% (Figure 7b). This study suggests the possibility of producing smart wearable devices with self-powered thermal management capabilities.

3.5. Light Stimuli

Compared with thermal actuation, photoactuation shows advantages such as remote control and high resolution in terms of both position and time.^[153–155] Nanocarbons can be incorporated through ink as the photon absorbers and photothermal heaters that drive the actuation.^[155] Huang et al. recently demonstrated photoactuated polymer pens for molecular printing (Figure 7c).^{[156, 157]} For this proof-of-concept demonstration, a viscous high concentration CNT ink is formulated by mixing covalently functionalized CNTs and polydimethylsiloxane (PDMS) monomer precursors and then casted on silicon molds for pyramidal pens. As light is shined on the pens, local photothermal effect due to light absorption by the CNTs brings the illuminated pens to contact with the substrate for molecules transfer and delivery. In this manner, molecular printing can be controlled by toggling light on and off.

Furthermore, wavelength selectivity in photoactuation can be achieved using SWCNTs of specific chiralities.^[148] Figure 7d shows the color and optical absorption spectra of SWCNT inks of different types, including HiPco, which is a mixture of different chiralities, as well as metallic species and the (6,5)-SWCNT. A polycarbonate/SWCNT bilayer structure formed by vacuum filtration of these ink solution over a polycarbonate filtration membrane was found to bend in response to light radiation. The results show a strong dependence on the wavelength of the light used, with 850–1,050 nm most effective for actuating the (6,5)-SWCNT, whose absorption peaks at 980 nm. Although the examples discussed in this section used nanocarbon inks through conventional solution processing rather than printing, these photoactuation effects can be readily utilized to design photoactuated and reconfigurable structures that are printable.^{[156, 158]}

4. Challenges and Perspectives

The rich physical and chemical properties of nanocarbon inks have enabled the development of myriad devices that can specifically respond to different external perturbations and can be further tailored to intended applications. Yet, further efforts are required to overcome key scientific and engineering challenges toward formulating nanocarbon inks with optimal performance on multiple length scales (Figure 8). From a material science and engineering perspective, formulating a nanocarbon ink for advanced device fabrications should consider 5 key metrics: 1) nanocarbon with electrical, optical, mechanical, and chemical properties optimal for specific applications; 2) high material loading of nanocarbon in the ink; 3) additives that are “clean”; 4) tailored ink-substrate interactions for ink deposition; and 5) controlled micro/nanostructures from printing for intended device performance.

Figure 8.

Making nanocarbon ink for advanced fabrication of programmable materials and devices calls for advances across multiple length scales.

The challenges of formulating high quality nanocarbon ink may be addressed by analyzing and optimizing these key parameters. In this section, we will review recent technical breakthroughs in nanocarbon solution processing and fundamental science that could potentially be translated into the next generation of high-performance ink, including new approaches to sorting and purifying CNTs, ways of imparting chemical functionality for sensing and optical applications; new approaches to achieve high materials loading with minimal or removable additives; advances in surface chemistry for improved adhesion to the substrates; and methods for building devices with highly ordered structures of nanocarbon (e.g., aligned CNTs).

4.1. Nanocarbon structural control

There have been many recent fundamental and technical studies in nanocarbons that have had a positive impact on printed functional devices. These include advances in the structural control of the nanocarbon material, such as the length of CNTs and thickness of graphene flakes.^[159–161] SWCNT length plays a key role in the previously discussed applications, impacting electrical transport, optical properties, and the transduction of mechanical stimuli. However, achieving a uniform and tunable length has remained elusive as most dispersal techniques involve high power sonication of the raw CNT materials, which reduces their average length. To overcome this challenge, our lab has recently developed a novel technique called superacid surfactant exchange (S2E) which can disperse and separate ultralong SWCNTs. Using this method, lengths of individual nanotubes reaching as high as ≈13 μm can be achieved as opposed to the < 1 μm lengths demonstrated by conventional dispersion processes that rely on sonication.^{[55, 159]} In the S2E process, the raw SWCNT materials are first dissolved in chlorosulfonic acid, which is subsequently added to aqueous solution of NaOH and DOC to neutralize the acid and stabilize the otherwise “naked” individual nanotubes (Figure 9a). S2E eliminates sonication, thereby avoiding cutting of SWCNTs and preserving the full lengths of as-synthesized nanotubes. Furthermore, we can reduce the length polydispersity and increase the average length (> 10 μm; Figure 9b) by length-dependent “self-sorting.” Although it is not fully understood, there is experimental evidence suggesting in the superacid the ultralong SWCNTs preferentially partition into the nematic phase while the shortest tubes stay in the isotropic phase. This length-dependent phase behavior is translated to the aqueous solution by S2E: the short nanotubes remain individually dispersed while the longer nanotubes precipitate as bundles, leading to self-sorting.^[160]

Figure 9.

New dispersion and sorting techniques in nanocarbon science enable novel and improved applications for printed functional devices. a) The nondestructive dispersion of SWCNTs can be achieved using the superacid-surfactant exchange (S2E) technique. Reproduced with permission.^[159] Copyright 2017, American Chemical Society. b) Ultralong SWCNTs can be obtained through the self-sorting process of iterative S2E. Reproduced with permission. Copyright 2018, Wiley-VCH. c) Density gradient ultracentrifugation (DGU) enables the isolation of single layer graphene sheets for high performance electrical devices. Reproduced with permission.^[161] Copyright 2009, American Chemical Society.

The ultralong SWCNTs significantly improve thin film properties compared to shorter nanotubes. For example, we have observed that thin films fabricated from long metallic CNTs with an average length of ≈3.2 μm can be as high as ≈3316 S cm⁻¹ at 85% optical transmittance and 100% tensile strain, while the short nanotube control (average length of ≈0.8 μm) only show a conductivity of ≈17 S cm⁻¹.^[55] The measured conductivity of the long tube film is greater than the conductance of ≈indium-tin oxide (ITO; >1000 S cm⁻¹), the prevailing material used in transparent conductive thin films.^{[55, 56]} Thin film transistors made from long, semiconducting SWCNTs also exhibit high carrier mobility (≈90 cm² V⁻¹ s⁻¹), which is comparable to state-of-the-art polycrystalline silicon semiconductors (≈100 cm² V⁻¹ s⁻¹) and surpasses nanocrystalline silicon (≈50 cm² V⁻¹ s⁻¹) and polymer semiconductors (< 1 cm² V⁻¹ s⁻¹).^[160] Furthermore, length-sorted ultralong nanotubes help improve the mechanical strength of the thin film network, which enables stretchable conductors with improved reproducibility and low hysteresis (retaining 96% of their conductivity with 1000 bending cycles), showing promise for printed flexible electronic devices.^{[55, 159]} Therefore, integrating length-sorted ultralong SWCNTs into functional ink could significantly improve the mechanical strength and electrical conductivity of various programmable devices.

A further advance in nanocarbon technology that could have significant impact on ink applications is the chirality-based sorting of SWCNTs, as discussed in Section 2. Chirality sorting can significantly improve FET performance with reduced subthreshold swings and threshold voltages.^[162] We direct the reader to recent reviews and books for advances on chirality sorting.^{[58, 70]}

Similarly, graphene sorting will be an enabler for high-quality printed graphene devices,^[163] as the electrical properties of graphene is distinctly different from multi-layer flakes. The use of high speed centrifugation in a variable density aqueous environment allows fractionation of the graphene flakes based on their thickness, enabling the isolation of monolayer graphene suspensions^{[161, 163]} which can subsequently be formulated into a high concentration ink for printed flexible circuits (Figure 9c).^[164]

4.2. Material loading

High concentration loading of nanocarbons is critical to making a good ink for high precision printing with minimized processing costs and batch-to-batch variation. However, conventional liquid exfoliation procedures can only generate limited amounts of nanocarbons dispersed in solution. For example, the concentration of surfactant-stabilized graphene solutions typically cannot exceed ≈0.3 mg mL⁻¹.^[25] Even sonication for more than a month can only improve graphene dispersions to ≈1.2 mg mL⁻¹.^[165] In addition, sonication is costly in terms of both energy and time, making this method unsuitable for large-scale production. Importantly, extensive sonication can cause significant structural damage to the nanocarbon, even shortening the nanotube length or breaking the graphene flakes into smaller pieces,^[166] which often deteriorate the performance of the resulting devices. However, new strategies that minimize or completely eliminate sonication have shown the promise to address this challenge, including: 1) pretreatment/activation of the nanocarbon, 2) dispersant/solvent engineering, and 3) selective extraction/condensation.

4.2.1. Pretreatment/activation of the nanocarbon

Pretreating or activating nanocarbon is shown to significantly enhance the solution concentration while also saving processing time. By soaking the bulk materials in certain solvents, such as water,^[167] ethanol,^[167] N-methyl-2-pyrrolidone (NMP),^[168] dimethyl sulfoxide and chlorosulfonic acid^[169] prior to the liquid exfoliation, the final concentration of the nanocarbon dispersion can be improved by ≈3–10-fold along with a decrease in the processing time by up to ≈90%. The underlying mechanism of this pretreatment approach is the spontaneous intercalation of the solvent molecules or ions between the bundled nanotubes or graphite layers, thereby overcoming the van der Waals interactions by disrupting the orderly packed structures to facilitate subsequent exfoliation (Figure 10a–c). For example, Coleman and co-workers show that NMP-pretreated graphite powder can be dispersed to a high concentration of ≈1 mg mL⁻¹ by just 1 min of tip-sonication, which is a significant improvement compared with the ≈90 mins required by tip-sonication without solvent pretreatment.^[168] In another demonstration, Alzakia et al. soaked graphite powder in ethanol for just 5 mins and found that the final concentration of the graphene dispersion increased by a factor of 10.^[167] In the case of SWCNT materials, Pasquali and co-workers demonstrated a high concentration (≈1.8 wt%) aqueous dispersion of SWCNTs can be obtained by pre-soaking the SWCNTs in chlorosulfonic acid, followed by mild sonication and dialysis (Figure 10d).^[169] The SWCNT aqueous dispersion is so concentrated that it forms a paste exhibiting a liquid crystalline phase, which can be further spun into highly ordered fibers.^[169]

Figure 10.

Achieving a high loading of nanocarbon in ink by solvent pretreatment of nanocarbon. a) Pretreatment by soaking the nanocarbon in certain solvents can induce intercalation of the ions or molecules, which reduces the van der Waals interaction between the nanocarbon, facilitating the subsequent exfoliation processes by sonication to yield high a concentration of nanocarbon in a short time. Reproduced with permission.^[167] Copyright 2020, American Chemical Society. b) Schematic illustration of the intercalation of solvent molecules into graphene layers. Reproduced with permission.^[170] Copyright 2018, Elsevier Ltd. c) Correlation between the graphene concentration and sonication time using pretreated (red) and pristine graphite (black). Pretreatment can alleviate the van der Waals interactions between the graphene layers, thus benefiting the dispersion procedures with higher concentration with reduce processing time. Reproduced with permission.^[227] Copyright 2015, American Chemical Society. d) Intercalation of SWCNTs by chlorosulfonic acid followed by mild sonication and dialysis can yield a surfactant-stabilized aqueous dispersion with concentration of up to 1.8 wt%. Notably, the concentration of the nanotube dispersion is so high that it looks like a paste. Reproduced with permission.^[169] Copyright 2018, American Chemical Society.

Apart from yielding high concentrations of nanocarbon dispersions, pretreatments can also preserve the flake size and tube length due to the reduced sonication time.^[170] Kim et al. obtained non-oxidized graphene flakes with a lateral size exceeding 30 μm by pretreating the graphite powder with dimethyl sulfoxide followed by bath-sonication with a reduced time of 45 min.^[170] These ultra-large graphene flakes when fabricated into a conductive thin film showed superior conductivity of up to ≈3.4×10⁵ S m⁻¹ due to the reduced grain boundaries.

4.2.2. Dispersant/solvent engineering

There are solvents/dispersants, including pyrene derivatives,^[171] N-cyclohexyl-2-pyrrolidone (CHP),^{[28, 172]} surfactant + sucrose,^[173] m-cresol,^[174] (DMSO) with alkaline metal^[175] and chlorosulfonic acid^{[176, 177]} that can disperse CNTs at various degrees (Figure 11). The pyrene derivatives effectively disperse graphene and SWCNTs due to the strong π-π stacking interactions between the nanocarbon and the aromatic planes of the pyrene derivatives. The concentration of graphene-pyrene dispersions is reported as high as ≈1.2 mg mL⁻¹ while the concentration of SWCNTs approaches ≈1.5 mg mL⁻¹. Pyrene-like molecules, such as naphthalene diimide (NDI) surfactants, were also found to efficiently disperse graphene up to ≈5 mg mL⁻¹.^[178] Compared with other materials, pyrene derivatives can stabilize graphene and SWCNTs with both a smaller amount of dispersant and greater stability. The dispersion remains stable at a wide range of pH and inorganic salt concentrations.^[171]

Figure 11.

Comparison of the concentration ranges of CNTs and graphene dispersed in various pure solvents or solvents with dispersants. Some of the chemical structures of the solvents and dispersants are shown at the bottom of the figure. Data collected from references.^{[25–29, 73, 161, 165, 171, 173–179, 228–230]}

Surfactants are widely used to disperse individualized SWCNTs but the concentration is typically too low for ink formulation. Leeds et al. found that adding sucrose to sodium dodecylbenzene sulfonate (SDBS)-dispersed SWCNTs in aqueous solution can drastically increase the concentration of SWCNTs (Figure 12a).^[173] The sucrose works by increasing the viscosity of the dispersion, locking the nanotubes in the viscous aqueous matrix and thereby preventing the nanotubes from rebundling. Significantly, even at ultrahigh concentrations (≈3.4 mg mL⁻¹), the majority of the SWCNTs remain as individual nanotubes as evidenced by the PL of the dispersion.

Figure 12.

Achieving a high loading of nanocarbon in ink by choosing specific molecules/solvent. a) Sucrose and SDBS can stabilize SWCNTs in aqueous solution up to ≈3.35 mg mL⁻¹. Reproduced with permission.^[173] Copyright 2012, Wiley-VCH. f) m-Cresol has been recently found to be a surprisingly good solvent to disperse CNTs with concentration up to ≈40 mg mL⁻¹ to form a nanocarbon paste. Reproduced with permission.^[174] Copyright 2018, National Academy of Sciences. CNTs can be mixed with certain solid selectors and compressed into a solid ink (c), which can be viewed as an extreme non-solvent version of ink with ultrahigh material loading. The solid ink can deliver the nanocarbon on the paper substrate for the fabrication of gas sensors (d). Reproduced from permission.^[179] Copyright 2012, Wiley-VCH.

There is virtually no organic solvent capable of dissolving CNTs. However, Huang and co-workers recently found m-cresol to be a surprisingly good solvent for CNTs, reaching concentrations of up to ≈40 mg mL⁻¹ to form multi-wall CNT paste (Figure 12b).^[174] As the concentration increases, the dispersion continuously transitions from dilute dispersion, thick paste, free standing gel, and dough-like materials. The exact mechanism that makes m-cresol an excellent solvent capable of dispersing CNTs is not yet fully understood, which may be an interesting open question.

Finally, we note that solid ink can be viewed as an extreme case of ink where the usage of solvent is avoided (Figure 12c–d). The solvent-free nature of the solid ink makes ultrahigh loading of nanocarbon possible, with concentrations of more than ≈50% demonstrated with CNTs.^[179] Just like the “lead” in a pencil, these solid inks can also deliver nanocarbons by mechanical abrasion methods on different substrates such as paper, to fabricate responsive devices. For example, SWCNT-based flexible gas sensors have been demonstrated to detect toxic gases like ammonia with a detection limit as low as 0.36 ppm.^[179]

4.2.3. Selective extraction/condensation

Although the term “concentration” is perhaps one of the most frequently used parameters in the solution processing of nanocarbons, it should be noted there remains high uncertainty when evaluating the concentration of such solutions due to the lack of standards to clearly describe the composition (e.g., the percentage of individually dispersed SWCNTs). Strictly speaking, as graphene is defined as a single layered structure of sp² carbon, the “concentration of graphene,” if not specified, should only account for the concentration of the single-layer structures. However, the “concentration of graphene” in most publications actually encompasses a range of few layer graphene materials with different thicknesses and lateral sizes. Indeed, Coleman and co-workers showed that when the concentration of graphene is quoted as ≈1.2 mg mL⁻¹, the actual percentage of single layer graphene can be as low as ≈4% of the reported values.^[165]

The practical challenge of obtaining a solution with a high fraction of single-layer graphene also makes it difficult to determine the accurate “concentration.” To enrich the population of single-layered graphene in the dispersion, different methods including DGU and cascade centrifugation, have been used to sort graphene by its layer number.^{[161, 180]} Although these advanced sorting techniques are highly effective and represent the best control over the dispersion quality of graphene, the final concentration of the single-layered graphene after sorting can be less than ≈0.01 mg mL⁻¹, which significantly limits the integration of these valuable materials as inks in various applications. This fact is also true for CNTs and particularly, chirality enriched SWCNT solutions. Although different sorting methods including DGU,^[181] aqueous two phase extraction (ATPE),^[182] and column chromatography^[183] have been successfully applied to sort SWCNTs by their chirality, without additional post-concentrating treatment steps, the final concentration of these materials is typically lower than ≈0.1 mg mL⁻¹ (estimated based on the optical density and molar attenuation coefficient in the literatures^[184]). This is particularly true for the less populated species, such as armchair and zigzag CNTs. In this regard, methods that can selectively extract or concentrate the nanocarbon would be extremely useful.

Li et al. demonstrated a solvent exchange method that can concentrate the graphene by a factor of ≈20.^[185] Graphite flakes were first exfoliated into graphene in dimethylformamide (DMF) solvent, followed by distillation to exchange the DMF to terpineol, which has a higher boiling point (Figure 13a). The concentrating ratio is controlled by the volume difference between terpineol and DMF. In another demonstration, terpineol, when added into a dilute water/ethanol solution of ethyl cellulose (EC)-dispersed graphene flakes, can preferentially extract the graphene flakes into the immiscible terpineol phase with extraction factors of up to ≈10 (Figure 13b).^[186] We speculate that the same strategy could be applied to extract and concentrate single-layer graphene dispersions after DGU.

Figure 13.

Increasing the nanocarbon concentration in ink by selective extraction and condensation of the materials. a) Exchanging the solvent from DMF to terpineol by distillation can be used to concentrate graphene dispersions by ≈20-fold. Reproduced with permission.^[185] Copyright 2012, Elsevier Ltd. b) Extraction of graphene from the ethanol/water phase to the terpineol phase (highlighted by the red-dashed box) increases the graphene concentration by ≈10-times. Reproduced with permission.^[186] Copyright 2010, American Chemical Society. c) Schematic illustration of the workflow of the ATPE method to concentrate dilute chirality-enriched SWCNT fractions using a temperature swing. Such extraction can also be realized by tuning the surfactant (i.e., DOC) concentrations. Reproduced with permission.^[187] Copyright 2019, The Royal Society of Chemistry. d) A dual surfactant system of DOC and CTAB can be used to extract SWCNTs to the bottom phase (left) with an extraction factor of ≈10. The extraction is possibly due to the Coulombic attractions between the DOC and CTAB molecules, which remain wrapped around the nanotube (right). Reproduced with permission.^[159] Copyright 2017, American Chemical Society. e) Schematic illustration (left) and photo (right) of the reversible protonation/deprotonation of DOC-stabilized SWCNTs. The coagulate can be filtered out (f) and redispersed in H₂O with reduced volume to concentrate the SWCNTs. Reproduced with permission.^{[55, 159, 188]} Copyright 2018, Wiley-VCH; Copyright 2017, American Chemical Society; Copyright 2020, University of Maryland, College Park.

Selective extraction of dilute SWCNTs dispersion can be achieved by ATPE, which was originally adapted from biological science for chirality sorting of SWCNTs.^[182] ATPE is a versatile method that can potentially separate SWCNTs with any specific chirality.^{[73, 187]} However, the sorted chirality-enriched samples are usually quite diluted, posing a challenge to their formulation into ink. As the SWCNT partitioning between the two immiscible phases (i.e., polyethylene glycol (PEG) phase and dextran) in ATPE can be systematically controlled by parameters such as surfactant concentrations and temperature, it is possible to transfer the diluted fraction of SWCNTs from one phase to the other with reduced volume, so that the dilute SWCNT solution can be concentrated. This practice, as illustrated in Figure 13c, has been used to concentrate the SWCNT dispersion by a factor of ≈40.^[187]

We have recently discovered that cationic surfactant cetrimonium bromide (CTAB) and anionic surfactant DOC can form an aqueous two-phase system to selectively extract and concentrate SWCNTs.^[159] We observed when CTAB is added into the aqueous DOC-stabilized SWCNTs, the SWCNTs are preferentially extracted to a surfactant-enriched bottom phase (Figure 13d). This unique phenomenon may be driven by the Coulombic attractions between the two oppositely charged surfactants forming a more hydrophobic surfactant-rich phase that contains the SWCNTs. The extraction factor of this method is as high as ≈20. Compared with ATPE, this new extraction system requires no polymers, such as PEG or dextran, to induce the phase separation, eliminating the need of subsequent polymer removal for many applications.

A simple yet cleaner method to increase the concentration of SWCNTs exploits the reversible protonation/deprotonation effects of the DOC molecules.^{[55, 160]} DOC is a member of the bile salt surfactant family with pKa ≈6.7. Due to its slightly bent steroid rings that can effectively accommodate the nanotube curvature, DOC is perhaps the most potent surfactant for stabilizing SWCNTs.^[26] Our group found that when the pH of the DOC-stabilized SWCNTs was adjusted to below 6.5, which is lower than the pKa of DOC, the DOC-stabilized SWCNTs coagulate out from the solution due to the protonation of DOC to deoxycholate acid (Figure 13e). The coagulation can be readily separated from the rest of the solvent by filtration to form a grey to black solid depending on the amount of the SWCNTs within the material (Figure 13f).^[188] Since the individual SWCNTs are tightly encapsulated by the deoxycholate acid molecules, such protonation-induced coagulation does not cause re-bundling of the nanotubes. As a result, the SWCNT solution can be easily recovered by gently stirring the coagulation in H₂O as the pH is readjusted to ≈7–8 to deprotonate the deoxycholate acid molecules back to DOC.^[55] With this method, the SWCNT solution can be concentrated by a factor that is solely dependent, to the limit of the aqueous solubility of DOC, on the ratio between the initial solution volume and the final H₂O volume used for the re-dispersion.

4.3. Additives engineering and removal

Additives, including dispersants, viscosity modifiers, and binders are indispensable ingredients for making ink. Useful as they are during ink formulation, these additives are often unwanted in terms of the ultimate functionalities of the ink. For example, in printed transistors, the existence of a high fraction of nonconductive additives from the ink can block the electron pathways, thereby deteriorating the carrier mobility and device performance. This is also a critical issue for responsive devices such as electrochemical sensors, where the high resistance can cause a higher overvoltage and thus larger background current, reducing the signal-to-noise ratio and sensing sensitivity. Therefore, as we develop nanocarbon inks for various applications, we also need to find methods to cleanly remove excessive dispersants or new additive chemistries that will not significantly compromise the functionalities of the nanocarbon materials. In general, the additives (more specifically dispersants) used in ink are present in two forms: (1) additives that are physically absorbed on the surface of the nanocarbon, acting as stabilizers; and (2) free dispersants in solution that maintain the overall dynamic equilibrium with the physically absorbed dispersants. Different strategies are needed to reduce the amount of free and physically absorbed dispersants.

Reducing the amount of free dispersants from the nanocarbon ink can be achieved by physical techniques such as repeated ultrafiltration or dialysis. However, these methods suffer from relatively low yield and high material loss. Several high throughput solution-based techniques have been explored to selectively separate the functional nanocarbon material from the excess additives by tailoring the molecular interactions between the nanocarbons or through molecular crowding effects.^{[164, 189, 190]} For example, Hersam and co-workers showed that through the addition of the polymer/salt (ethyl cellulose/NaCl) to reduce the electrostatic repulsions between graphene flakes, a loose aggregate is formed while the majority of the dispersants remain in solution (Figure 14a).^[164] Since the individual graphene flakes are still encapsulated by the dispersant, the loose aggregate of graphene can then be separated from the excessive additives by mild centrifugation. Another example to reduce the amount of additive relies on the molecular crowding effect (Figure 14b).^[189] This entropic effect results in short-range attractive depletion forces between colloidal particles, which may consequently form aggregated clusters that can be more easily separated and re-dispersed as needed.^{[191, 192]} There are many different molecular crowding agents, ranging from polymers to small molecules (e.g., polyethylene terephthalate, polyvinylpyrrolidone, dextran, and NaSCN). For example, Zheng and co-workers used this molecular crowding effect to separate SWCNTs from residual polymers that are used for chirality sorting in the ATPE process (Figure 14b).^[190] The separated SWCNTs, in the form of loose aggregates can be re-dispersed in pure solvent such as water by mild sonication with a significant reduction of the amount of additives.

Figure 14.

Strategies to fabricate clean nanocarbon inks using different convertible, removable, and decomposable additives. a) The flocculation of graphene from excessive additives by tailoring the graphene interactions via the addition of salt. Reproduced with permission.^[164] Copyright 2013, American Chemical Society. b) A molecular-crowding strategy to selectively aggregate the SWCNTs from the excess amount of additives and dispersant. Reproduced with permission.^{[189, 190]} Copyright 2011, American Chemical Society. Copyright 2016, The Royal Society of Chemistry. c, d) Using convertible dispersants such as (c) graphene oxide or (d) GQDs to disperse different nanocarbon materials in aqueous solution. Reproduced with permission.^{[194, 195]} Copyright 2013&2014, American Chemical Society. e) Ammonium laurate is a cleaner dispersant for CNTs compared with SDS. Reproduced with permission.^[196] Copyright 2015, American Chemical Society. f, g) Removable/decomposable polymers as stabilizers for SWCNTs. Reproduced with permission.^{[199, 200]} Copyright 2010&2017, American Chemical Society.

Removing physically absorbed dispersants can be more difficult than removing their free counterparts in the solution. As discussed in Section 5.2, in order to disperse the nanocarbons with high concentration and colloidal stability, molecules featuring strong interactions with the nanocarbons are preferable. However, such strong interactions make the dispersant molecules tightly absorbed on the nanocarbon surfaces, which in turn makes them difficult to remove. Although repeated washing with a strong acid can be used to release physically absorbed dispersants, these harsh removal methods also induce irreversible aggregation of the nanocarbon materials and are typically incompatible with device fabrication.^[193] To solve these challenges, several strategies have been explored.

The first strategy is to employ “convertible” additives that can be potentially converted into useful nanocarbons for the application. In this regard, nanocarbon-derived dispersants can provide a pathway for cleaner ink applications as these dispersant share similar structures as the functional nanocarbon materials.^{[194, 195]} Post-treatment methods, such as reduction or annealing, may also be used to convert these nanocarbon-based dispersants into conductive carbons that may synergistically enhance the device performance. For example, graphene oxide and graphene quantum dots (GQDs), two nanocarbon materials that feature high solubility in water, have been used as dispersants to stabilize different types of nanocarbon materials, including fullerenes, SWCNTs, and graphene (Figure 14c,d).^{[194, 195]} He et al. found that ≈100 mg of raw graphene powder could be stabilized by only 7.8 mg of GQDs (Figure 14d).^[195] In another example, Hersam and co-workers showed that nitrocellulose (NC)-dispersed graphene can be used as a conductive ink with highly tunable viscosity, making it compatible with different printing techniques. More importantly, post-printing annealing at ≈350 °C decomposes the residual NC into conductive sp² carbon that can synergistically contribute the overall conductivity of the graphene. As a result, the printed graphene patterns after annealing featured one of the highest electrical conductivities (≈40,000 S m⁻¹) reported for solution-processed nanocarbons.

Another method to make cleaner ink is through dispersant engineering. Dispersants that are rationally designed for easy removal would be extremely useful to make cleaner ink. Nilsson et al. reported the use of ammonium laurate as a novel type of surfactant to stabilize SWCNTs.^[196] Importantly, when depositing the SWCNTs on the substrate, the free ammonium laurates do not build up on the surface, as occurs with other common dispersants, such as sodium dodecyl sulfate (SDS), providing cleaner substrate surfaces with much less debris (Figure 14e). This is possibly due to the fact that the low ionic strength of the ammonium laurate solution prevents it from interacting with the substrate surface.

Significant progress has also been made for designing removable polymers that can be used to disperse and sort semiconducting SWCNTs. Polymer wrapping is one of the most effective methods for selectively and quickly isolating electronically homogenous semiconducting SWCNTs with relatively low cost.^[197] However, the process leaves behind significant polymer residues in the nanotube solution and due to the tight polymer wrapping, it is also generally difficult to totally remove the polymers from SWCNTs after the process. Arnold and co-workers found that even after several aggressive cycles of washing and ultracentrifugation to remove the polymers, 50 wt% of the polymers remained on the nanotube surfaces.^[198] These large amounts of polymers attached to the SWCNT surfaces, if left untreated, can significantly increase the junction resistance between nanotubes, increasing the overall sheet resistance of fabricated devices and decreasing the carrier mobility.

To this end, removable polymers have been designed to stabilize SWCNTs while also being easily removed by post treatment. Zheng et al. showed that a foldable oligomer (based on m-phenylene ethynylene) exists in an unfolded state and can be used to disperse SWCNTs in chloroform, but releases the SWCNTs in acetonitrile by changing to a folded state (Figure 14f).^[199] In another example, Gopalan and co-workers designed and synthesized a π-conjugated polymer based on polyimine to selectively extract semiconducting SWCNTs in toluene.^[200] The existence of imine bonds render the polymers degradable by acid so that the physically absorbed polymers on the sorted SWCNTs can be removed by rinsing with mild acidic solution (Figure 14g).

4.4. Tailoring the ink-substrate interaction

Tailoring the interaction between the ink and substrate is critical to control the morphology of the printed pattern, which can influence the performance of the resulting device. It is essential for the nanocarbon ink to wet the substrate so that a homogenous and continuous deposition of the functional materials can be achieved with the desired printing resolution. For proper wetting, the surface tension of the substrate should be ≈10 mN m⁻¹ higher than the ink based on empirical evidence.^[201] Figure 15a shows the surface tensions of commonly used solvents and substrates, with a line added to describe the ≈10 mN m⁻¹ correlation. For example, if the substrate is made of polycarbonate, then ink with a surface tension of less than 30 mN m⁻¹ (i.e. toluene or chloroform) is preferable to ensure that loaded functional nanocarbon can be homogenously delivered for the device fabrication. Failure to meet this requirement may result in poor wetting, causing ink-retraction and discontinuous deposition of the nanocarbon material, which can deteriorate the device performance.

Figure 15.

Tailoring the ink-substrate interaction for better nanocarbon deposition. a) Summary of the surface tensions of various solvents and substrates. The black line indicates where the surface tension of the solvent is ≈10 mN m⁻¹ less than that of the substrate. b) Design of surfactant-substrate interactions for homogenous deposition of SWCNTs. c) Atomic force microscopy (AFM) images of the deposition of SWCNT inks made with different surfactants. Reproduced with permission.^[202] Copyright 2014, American Chemical Society. d) AFM images showing the network morphologies of SWCNTs deposited on the substrate are highly dependent on the ink-substrate interactions: SC/DOC stabilized SWCNTs tend to form a uniform SWCNT layer while the SDS/SDBS stabilized SWCNTs tend to form large bundles. Reproduced with permission.^[203] Copyright 2013, American Chemical Society.

Apart from considering the intrinsic surface tension of the solvents and substrate, surface modification of the substrate is another useful strategy to ensure the homogenous deposition of the nanocarbon, especially for surfactant-stabilized materials. Amine-terminated molecules are often employed to enhance the interaction between surfactant-stabilized SWCNTs, which are usually negatively charged, with substrates. For example, Javey and co-workers examined the interaction between amine-terminated flexible substrates with SWCNTs stabilized by different commonly used surfactants, such as SDS, SDBS, and sodium cholate (SC; Figure 15b).^[202] Interestingly, while the steroid surfactant SC was able to induce the homogenous assembly of SWCNTs on the substrate, SDS- and SDBS-dispersed SWCNTs failed to achieve uniform deposition (Figure 15c). This is presumably because alkyl-chain based surfactants like SDS or SDBS will form a monolayer or bilayer structure that can prevent the adhesion of SWCNTs on the substrate due to the lack of static attraction force. In contrast, the steroid-based SC surfactant does not form this type of bilayer structure on the substrate, which allows the homogenous deposition of SWCNTs as a percolated network. The ink-substrate interactions can also determine the network morphologies of the deposited SWCNTs. For example, ink composed of SDS-dispersed SWCNTs tends to form large bundles while those made with DOC- or SC-stabilized SWCNTs can form uniform SWCNT networks (Figure 15d).^[203]

4.5. Building highly ordered microscale and nanoscale structures

Since ordered structures can significantly improve electrical, mechanical, and thermal properties,^[204] it is important to tailor the morphology of printed nanocarbon in order to achieved optimal device performance. Structural order of such materials needs to be controlled at both microscale and nanoscale to achieve long range ordering and fine resolution.

As drying is generally an unavoidable step after printing nanocarbon inks (excepting solid inks), tailoring the evaporation process can be a useful strategy to form ordered structures with required patterns.^{[205, 206]} Inspired by the coffee ring phenomenon, Lin and co-workers sandwiched the ink between two substrates with specific geometries that control the drying process (Figure 16a), allowing the solute to self-assemble into a series of ordered geometries (such as rings and lines, Figure 16b).^[205] This strategy has been used to build CNT line patterns on silicon wafers with spatial resolutions as high as ≈5 μm.^[207] Microscale self-organization of rGOs has also been reported using a similar approach.^[208]

Figure 16.

Building ordered CNT assemblies by controlling the ink’s solvent evaporation. a) Schematic illustration of the route to utilize controlled evaporation for the self-assembly of materials at the macroscale. Reproduced with permission.^[205] Copyright 2012, Wiley-VCH. b) Representative optical micrographs of the ordered stripes fabricated from controlled solvent evaporation. Reproduced with permission.^[206] Copyright 2009, Wiley-VCH. c) SEM (left) and atomic force microscopy (AFM; right) images showing the highly aligned SWCNTs structures made from floating evaporative self-assembly. Reproduced with permission.^[209] Copyright 2013, American Chemical Society. d-e) Representative AFM images of SWCNTs assembled on chemical templates. Reproduced with permission.^[216] Copyright 2006, National Academy of Sciences. Schematic illustration (f) and transmission electron microscopy image (g) of assembling SWCNT arrays with precise pitch by using DNA nanotrenches. Individual SWCNTs resting in the nanotrenches are indicated by red arrows. Preproduced with permission.^[217] Copyright 2020, American Association for the Advancement of Science.

Controlled evaporation can also be utilized to fabricate ordered structures down to the nanoscale. Gopalan and co-workers showed that polymer-wrapped SWCNTs dissolved in organic solvent can be deposited on solid substrates to form nanotube thin film with alignment of within ±14° by floating evaporative self-assembly (Figure 16c),^[209] which is achieved by the alignment of SWCNTs at the ink/water interface, followed by the deposition to the substrate during the de-pinning of ink/water/substrate and ink/air substrate interface.^[210] By first sorting the SWCNTs into electronic grade semiconducting materials, this method can be used to fabricate aligned semiconducting thin films for FETs with device performance exceeding Si.^{[211, 212]}

Further control of the printed structure down to the individual nanotube level can be achieved by combination with template-directed assembly.^[213–216] For example, Wang et al. show that chemically patterned hydrophilic templates on a hydrophobic surface can drive nanotubes in solution to assemble at the hydrophobic-hydrophilic interface, allowing the shape, orientation, and linkage of SWCNTs to be controlled at submicron-length scale (Figure 16d,e).^{[215, 216]} A recent breakthrough by Sun et al. represents one of the finest controls ever achieved in directed assembly of SWCNTs.^[217] By designing DNA parallel nanotrenches to host individual SWCNTs, the authors made highly aligned SWCNT arrays with a uniform pitch as small as ≈10.4 nm and with an angular deviation less than 2° (Figure 16f,g).^[217] Considering that these highly aligned CNT arrays can be fabricated with high assemble yield (> 95%) and large area (≈0.35 cm²),^{[217, 218]} this method may hold the potential to directly compete with state-of-the-art device fabrication approaches.

5. Conclusions and Outlook

Carbon ink is undergoing a renaissance due to the discovery of nanocarbon materials that have unlocked a host of remarkable physical properties which can offer printable functionalities well beyond color. Depending on their structures, nanocarbons can feature extraordinary chemical and physical properties, such as ultrahigh electrical conductivity, superior mechanical strength, and bright photoluminescence in the shortwave infrared. Driven by an in-depth understanding of the fundamental physics and chemistry, knowledge of solution processing, and the need for large scale cost-effective additive manufacturing of smart devices, the development of nanocarbon inks has enabled a versatile, programmable material platform with a wide range of emergent applications. For example, nanocarbon inks have made possible the fabrication of novel devices with revolutionary potential, such as flexible electronics and implantable health-monitoring sensors, and in a manner that is both facile and cost-effective. The continuous discovery of the new types of nanocarbons, such as carbon nano-horns,^[44] carbon cones,^[45] and carbynes^[46] with interesting electrical and optical properties may keep refueling and propelling the development of nanocarbon inks for the most advanced applications.

While such advances have been explored, there are still key engineering metrics and challenges that must be met for the formulation of nanocarbon inks that can fully harness the remarkable properties of nanocarbons in printed devices. For example, more efforts are needed to understand the correlation between device performance and ink properties (such as viscosity) in order to better guide the ink formulation. Post-printing treatments, such as high-temperature thermal annealing, are also routinely used to improve device performance. However, as polymer- or paper-based substrates are used to enable flexible devices, we must also develop post-printing treatment strategies that can operate at low temperatures to avoid damaging the substrate and device while still removing additives and densifying the functional nanocarbons as needed. Finally, although selective extraction methods discussed in Section 5.2.3 may be used to concentrate dilute solutions of mono-chirality-pure SWCNTs for applications requiring specific optical or electrical properties, the methods remain small-scale and relatively costly. Industrial-scale sorting and solution processing of mono-chirality pure SWCNTs at low cost is therefore a pressing need for obtaining superlative nanocarbon inks that meet the needs of many high-value applications. While these challenges exist and others are sure to emerge, the field of nanocarbon inks is well positioned to enable this intrinsically programmable material platform for innovative new applications with sophisticated functionality well beyond color.