1 Introduction

With the rapid material development, carbon nanomaterial-based composites such as carbon nanotubes (CNTs) [1,2,3,4,5,6,7], graphene [8,9,10,11,12], graphite [12,13,14,15,16], carbon nanofibers [17,18,19], and carbon black (CB) [12, 20,21,22] have gained immense attention owing to their physical, mechanical, electrical, and thermal properties, which render them suitable for various applications such as sensors [6, 22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43] and energy harvesting devices [44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80]. CNTs exhibit remarkably high specific stiffness and strength, and hence are considered as potential candidates for developing composites. CNTs also exhibit high electrical and thermal conductivities. These properties along with their high aspect ratio, one-dimensional (1D) honeycomb lattice, and low density [81, 82] make CNTs extensively studied materials for the development of various composite and smart materials.

Graphene is a monolayer of carbon atoms tightly packed into a two-dimensional (2D) flat structure and is the basic building block of graphite materials. The geometrical (2D) and electronic effects of graphene on the field-emission properties of open-ended CNTs affect their mechanical and electrical properties, rendering them suitable for the formation of composite materials [83, 84]. In addition to these properties, graphene shows the lowest energy because of the overlap of 2\( p_{z} \) orbitals of carbon atoms, and hence imparts anisotropic properties to its composite materials due to the differences between the in-plane and out-of-plane bonding of carbon atoms and their dimensional (3D) geometrical properties. These unique properties and the cost-effectiveness of graphene-based composites make them potential filling agents [85]. Apart from these composites, various carbon nanomaterials such as carbon nanofibers and CB, which exhibit unique geometries and outstanding physical, electrical, and thermal properties, are also used as fillers for making composite materials [84, 86].

This review focuses on the progress in the development of carbon nanocomposites for sensor and power source applications, illustrating the expansion in the range of their applications. Herein, we have discussed the working principle, structural designs, and output performance (sensitivity, sensing range, etc.) of carbon nanocomposite-based sensors. We have also discussed the working mechanism and the improvement in the design and output performance of carbon nanocomposite-based thermoelectric and triboelectric generators, which can harvest multimode energies from the surrounding environment and mechanical stimuli. Finally, the perspectives and challenges for the development of carbon nanocomposites are discussed to expand their application range.

2 Carbon Nanomaterial-Based Composite Mechanical Sensors

2.1 Working Mechanism of Carbon Nanomaterial-Based Sensors

Carbon nanomaterials exhibit outstanding properties such as highly mechanical and electrical properties, which are suitable for sensing applications. First, there are various methods to see the sensing capability of the carbon material itself. The conductance of carbon nanomaterials (CNTs, graphene, etc.) can be altered dramatically by introducing mechanical strain using atomic force microscopy. This is the reason this strain causes structural changes in carbon nanomaterials such as the change in chirality, thus altering their electrical properties [87]. Hence, the incorporation of carbon nanomaterials into polymers is an effective approach for the development of highly sensitive and large-measurement range mechanical sensors with simple device structures. Zhang et al. and Yi investigated the piezoresistivity of conductor-filled polymer composites [88]. They found that the resistance of the composites decreased with an increase in the uniaxial pressure following the concept of “conducting paths”, i.e., piezoresistivity. In particular, this piezoresistive effect as an important working mechanism of mechanical sensors can be induced by the tunneling effect of the randomly distributed carbon nanomaterials in the polymer matrix. The application of a mechanical stimulus alters the tunneling distance between the adjacent CNTs, resulting in the breakup and formation of conductive networks and an increase in the inter-filler distances in the polymer matrix. This mechanism has been validated both experimentally and theoretically [41, 89].

2.2 Design of Piezoresistive Carbon Nanomaterial Sensor

Various methods have been used to optimize the design and enhance the performance of carbon nanomaterial-based mechanical sensors. Especially, the selection of carbon nanomaterials strongly affects the sensor performance. There have been several reports on the detection of mechanical force using CNT composites [2,3,4, 6, 7, 12, 22, 90, 91]. Darren et al. developed transparent elastic films of CNT strain sensors with high conductivity. These films could repeatedly measure high strain (\( {{\varepsilon }} \ge 150{\% }) \) [92]. Eun et al. developed sandwich-like stacked piezoresistive strain sensors with stretchable (100%), transparent (62%), ultrasensitive (gage factor of 62), and patchable properties [93]. The incorporation of graphene, which exhibits high electrical conductivity and excellent mechanical properties with a large aspect ratio, into these nanocomposites can lead to the development of sensors with good electrical and mechanical responses [15, 94]. Wang et al. reported a graphene-polydimethylsiloxane (PDMS) nanocomposite as a strain sensor. The sensor exhibited a gage factor (GF) of 233 at the graphene concentration of 8.33 vol%, which was measured within the strain range of 2% [94]. Reduced graphene oxide (rGO)-polymer nanocomposites with a strong interface interaction between the conducting filler and the matrix can improve the electrical response of sensors [95]. In addition to the carbon nanomaterials mentioned above, CB has gained immense attention for sensor applications due to its excellent electrical properties and low cost [21, 96, 97]. Wang et al. reported a mathematical model for piezoresistivity and analyzed the changes in the effective conductive paths of CB and silicone rubber [98]. Carbon nanofibers, which are readily available and are synthesized using chemical vapor deposition (CVD), have also been used extensively for the preparation of smart materials [18, 19, 86, 99, 100]. Ferreira et al. reported the electrical and piezoresistive responses of epoxy-carbon nanofiber composites. The composites showed a GF value of 9.8 and were thermally stable up to 75 °C [19]. In addition, many studies have also been carried out on hybrid carbon material-based composite sensors using two or more carbon nanomaterials. First, Fig. 1a shows hybrid filler systems consisting of CNTs (0.5–1 wt%) and CB (0.5–4 wt%). It shows the conductive network structure consisting of CNTs and CB is changed by stretching hybrid nanocomposite, explaining CNT–CNT contact and CB bridging effect [22]. Secondly, hybrid nanocomposites composed of CNTs grown on graphene nanoplatelets (GNPs) show synergy effects, which can be used to monitor the onset of irreversibly permanent deformation in the composites shown in Fig. 1b [34]. Thirdly, Fig. 1c also shows the piezoresistive responses of hybrid nanocomposites consisting of few-layer thermally rGO (TRGO) and multiwall CNTs (MWCNTs) mixed with natural rubber [35]. The piezoresistive response of such elastomeric composite materials can be controlled by tuning their relative hybrid filler contents. In addition to the combination of carbon nanomaterials, various carbon nanomaterial-metal piezoresistive strain sensors such as CNT-Al2O3 nanocomposites (Fig. 1d) have been developed for inducing the hybrid synergy effect [101].

Fig. 1
figure 1

Hybrid effects of carbon nanomaterials in polymer matrix: a Schematic illustration of the network structure consisting of CNTs and CB before and after stretching for PVDF nanocomposites containing CNT contents below and above ΦCNTs (0.53 wt%). Reproduced with Permission [22]. Copyright 2016, American Chemical Society. b SEM micrographs of the carbon fillers (CNTs, GNPs, and CNT-GNP hybrid nanocomposite). Reproduced with Permission [34]. Copyright 2013, Elsevier. c\( \Delta {\text{R}}/{\text{R}}_{0} \varepsilon \) as a measure of piezoresistive sensitivity for the examined nanocomposites. Reproduced with Permission. [35] Copyright 2017, Elsevier. d The preparation process of the multiscale glass fabric/epoxy/CNT-Al2O3 composites. Reproduced with Permission [101]. Copyright 2014, Elsevier

Full size image

Apart from carbon nanomaterial (CNTs, graphene, CB, carbon nanofiber etc.)-based composite sensors, various carbon nanocomposite sensors with different matrices [materials with very high stiffness (cement) to very soft materials (Ecoflex)] have also been developed. Han et al. investigated the mechanical, electrically conductive, and piezoresistive behaviors of cement/CNT-CB composites showing high sensing compressive strengths (0–10 MPa) and low sensing strains (0–0.35 mε) (Fig. 2a) [37]. Figure 2b shows the experimental and numerical data reported for epoxy-CNT composite sensors. At the tensile strain of 6000 μ\( {{\varepsilon }} \), these sensors show a GF of 117 and a conductivity of \( 10^{6} \,{\text{S/m}} \) [102]. Figure 2c shows the piezoresistive behavior of vinyl ester/carbon nanocomposites under axial tension and compression. These composites exhibit cyclic compressive strain (0–0.5%) and tensile strain (0–0.25%) sensing capabilities and a GF of 2.6 [39]. Figure 2d shows flexible composites prepared by incorporating zero-dimensional (0D) CB and 1D CNTs into thermoplastic polyurethane (TPU). CNT/TPU composites show a GF of 6.8 and a wide sensing range (0–ca. 135% strain). Under 20% strain, CB/TPU composites exhibit higher sensitivity (with a GF of 10.8) than CNT/TPU composites. These composites exhibit a strain sensing range of 0–ca. 90% [40]. Figure 2e shows a polydimethylsiloxane (PDMS)-MWCNT nanocomposite foam sensor. PDMS is flexible and its foam structures are advantageous for pressure sensing. Ecoflex, which is a very soft polymer material and possesses mechanical properties comparable to those of human skin, is widely used for preparing carbon nanocomposites. As can be observed from Fig. 2f, CNT/Ecoflex nanocomposites can withstand a strain of 650% and can still maintain their conductivity. They show GF values of 0.65 and 48 over the stain ranges of 0–400% and 400–700%, respectively [31]. Therefore, carbon nanomaterial-based strain sensors (composed of materials with different stiffness values) find various applications such as in structure, facture, and tactile sensing systems.

Fig. 2
figure 2

Effect of carbon nanocomposites on various properties of matrix. a SEM images of CNT/NCB composite fillers reinforced cement-based materials with multifunctionality. Reproduced with Permission [37]. Copyright 2015, Elsevier. b Influence of \( {{\theta }}_{\text{max} } \) on sensor sensitivity of CNT/Epoxy composites. Reproduced with Permission [102]. Copyright 2010, Elsevier. c Compressive stress and normalized change of electrical resistance vs. strain for a MWCNT/VER composite. Reproduced with Permission. [39] Copyright 2012, Elsevier. d The relative resistance changes as a function of tensile strain for CB/TPU-8 and CNTs/TPU-4 composites. Reproduced with Permission. [40] Copyright 2017, Elsevier. e Sensing performance for dynamic load of MWCNT/PDMS nanocomposite foam. Reproduced with Permission. [25] Copyright 2017, Elsevier. f Image of a strain sensor at the initial length and stretched to 700% strain and relative change in resistance vs strain on wrinkled CNTs on Ecoflex substrate. Reproduced with Permission. [31] Copyright 2016, John Wiley and Sons

Full size image

Figure 3a shows multi-point and multi-directional strain mapping sensor systems based on MWCNT-silicone elastomer nanocomposites and their anisotropic electrical impedance tomography [6]. Such sensor systems show robust contact conditions and require low-cost fabrication processes. Carbon nanocomposite sensors use the mechanism of resistivity change in order to detect normal and shear forces. Nanocomposites with patterns similar to human skin show high sensitivity (15.1 kPa−1) and multiple-sensing capabilities (Fig. 3b) [41]. Such patterned nanocomposites and novel design sensor present a robust technology platform to further improve the sensitivity and response time of conventional composite elastomers for various sensor applications [103,104,105]. Pang et al. developed a sensor made up of graphene porous networks (GPNs) and PDMS. This sensor exhibited a wide pressure sensing range (2000 kPa and 25% strain) and high sensitivity (GF of 8.5), as shown in Fig. 3c [42]. Such sensors can be used for monitoring or even recognizing the walking state, determining the degree of finger bending, and monitoring the wrist blood pressure. To overcome the limitations of CNT geometry, CNT fiber-based strain sensors have been developed (Fig. 3d). This sensor can measure strains greater than 900% with high sensitivity and exhibit fast response and good durability [43].

Fig. 3
figure 3

Strengthening sensor properties of carbon nanomaterial based composites. a Schematic of electrical impedance tomography and concept of multi-dimensional strain measurement of MWCNT-silicone elastomer nanocomposites. Reproduced with Permission. [6] Copyright 2017, Springer Nature. b Conductive composite elastomers with interlocked microdome arrays. Reproduced with Permission [41]. Copyright 2014, American Chemical Society. c Relative resistance variation of the GPN-PDMS composite with applied static compressive loading. Reproduced with Permission. [42] Copyright 2016, American Chemical Society. d Relative change in resistance versus strain for unsupported CNT fibers, CNT fibers on an unstrained Ecoflex substrate, and CNT fibers on an Ecoflex substrate prestrained by 100%; the strain ranged from 0 to 450% strain (inset). Reproduced with Permission. [43] Copyright 2015, American Chemical Society

Full size image

3 Carbon Nanocomposite Based TE Generators

3.1 TE Materials and Energy Conversion Mechanism of TE Power Generation

Thermoelectric (TE) materials are extensively studied due to applications in TE generators (TEGs). TEGs provide clear form of energy by converting waste heat or solar heat directly into electrical power and reduce greenhouse gas emissions. They do not involve the use of moving parts and working fluids and do not produce noise [44]. In order to develop high-performance TEGs, TE materials with high Seebeck coefficients, high electrical conductivities, and low thermal conductivities must be used. Carbon nanomaterial/polymer composites with optimized carbon nanomaterial and polymer ratio fulfill these requirements and offer the advantages of both the high electrical conductivity of carbon materials and low thermal conductivity of polymers.

In 1821, Thomas Johann Seebeck discovered that a compass needle is deflected by two different metals which form a closed loop electrically in series and thermally in parallel when exposed to a temperature gradient. This phenomenon is known as the Seebeck effect and is the working principle of TEGs. Figure 4a shows the schematic of the Seebeck effect. If there is a temperature difference between the top and bottom junctions of an assembly of two semiconductors, the electron energy levels in the n- and p-type semiconductors are shifted to different degrees and a potential difference between the junctions creates an electrical current and a magnetic field (this is the reason for the deflection of the needle). Although a single p–n junction can act as a TEG, many p–n junctions are connected in series to achieve high TE performance, as shown in Fig. 4b.

Fig. 4
figure 4

Reproduced with Permission [44]. Copyright 2008, AAAS

a Schematic of Seebeck effect and b thermoelectric generator with many p–n junctions in series.

Full size image

3.2 TE Performance

The performance of TE materials and TEGs is quantified by their Seebeck coefficients, power factors, and the figure of merit. The Seebeck coefficient S of a TE material is defined as

$$ S = - \frac{\Delta V}{\Delta T}\left( {{\text{V\,K}}^{ - 1} } \right) $$
(1)

where \( \Delta {\text{V}} \) and \( \Delta {\text{T}} \) are the differences in the potential and temperature of the top and bottom junctions, respectively.

The power factor PF of a material is defined as

$$ {\text{PF}} = {\text{S}}^{2} {{\sigma }}\;\left( {{\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} } \right) $$
(2)

where \( {{\sigma }} \) is the electrical conductivity of the material. The PF of TEGs is a measure of their performance and efficiency in terms of electronic transport. The figure of merit ZT (dimensionless) of a material is defined as

$$ {\text{ZT}} = \frac{{{\text{S}}^{2} {{\sigma T}}}}{k} $$
(3)

where T and k are the absolute temperature and thermal conductivity, respectively. The figure of merit of a material is the measure of its electronic transport (\( {\text{S}}^{2} {{\sigma }} \)) and thermal transport (k) at temperature T. The figure of merit of TEGs is useful for determining their TE power efficiency in terms of electrical and thermal transports.

3.3 Carbon Nanocomposite Based TE Materials

The quantities S, \( {{\sigma }}, \) and k are strongly interrelated in a such way that an increase in \( {\sigma } \) results in a decrease in S and an increase in k. Hence, it is difficult to simultaneously improve both the PF = \( {\text{S}}^{2} {{\sigma }} \) and ZT = \( {\text{S}}^{2} {{\sigma T}}/k \) of TEGs. The addition of carbon nanocomposites to TE materials makes it possible to control these interrelated quantities (S, \( {{\sigma }}, \) and k) quasi-independently [106]. In carbon nanomaterial/polymer based TEGs, various internal interfaces are formed between the carbon nanomaterial and the polymer. These interfaces act as scattering centers for phonons and reduce the lattice thermal conductivity of the TEG more than its electrical conductivity, as shown in Fig. 5.

Fig. 5
figure 5

Reproduced with Permission [45]. Copyright 2011, American Chemical Society

A schematic of ordered PEDOT: PSS on nanotubes and junctions as electrically conductive path and scattering center of phonon.

Full size image

Furthermore, low-dimensional carbon particles such as 1D CNTs, 2D graphene and graphite, and 0D CB induce the quantum confinement and carrier-energy filtering effects, resulting in the ordering of the polymer chains, as shown in Fig. 6a for CNTs and Fig. 6b for graphene. As a result, the increase in S is higher than the decrease in \( {{\sigma }} \). For this reason, combinations between carbon nanomaterial and polymer have been studied to enhance the PF and ZT of TEGs.

Fig. 6
figure 6

Ordering of a polymer on a surface of carbon particle and \( {\pi } - {\pi } \) interaction between carbon materials and polymer: a CNTs and polymer; TEM images of CNTs coated by PANI. Reproduced with Permission [46]. Copyright 2010, John Wiley and Sons. Schematics of ordered PANI on CNTs and \( {{\pi }} - {{\pi }} \) interaction between CNTs and PANI. Reproduced with Permission [107]. Copyright 2012, Royal Society of Chemistry. b graphene and polymer; TEM images of rGO coated by PEDOT. Reproduced with Permission [48]. Copyright 2013, Royal Society of Chemistry. Schematics of ordered EDOT on rGO and \( {{\pi }} - {{\pi }} \) interaction between rGO and PANI. Reproduced with Permission [49]. Copyright 2014, Royal Society of Chemistry

Full size image

0D CB has been widely used as a dopant for TE materials owing to its dispersion ability and high conductive properties. The thermoelectric behaviors of CB/polyethylene (PE) [108], and CB/poly (4-methyl pentene-1)/PE [109] composites have been investigated. Although the resistance change and heat flow (by temperature) of such composites have been investigated, no study has been carried out to measure their Seebeck coefficients, power factors, and figures of merit. To the best of our knowledge, the TE performances of CB/polymer composites were first reported by Casey et al. [110]. They prepared CB/poly (3, 4-ethylenedioxythiophene): poly (4-styrenesulfonate) (PEDOT: PSS) composites and achieved a PF of 0.96 μ\( {\text{W\, m}}^{ - 1} {\text{K}}^{ - 2} \). They used dimethyl sulfoxide (DMSO) to improve the electrical conductivity of PEDOT: PSS. In addition, the effect of DMSO on the TE performance of CB/PEDOT:PSS was investigated by Wang et al. [111]. The addition of DMSO as a second dopant (a first dopant is CB) slightly increased both the S (~ 5.15 to ~ 6.5 μ\( {\text{V\,K}}^{ - 1} \)) and PF (from ~ 0.016 and ~ 0.075 μ\( {\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} \)) at PEDOT: PSS = 30 wt%.

In CNT/polymer composites, the 1D CNTs arrange in the polymer chains on their surfaces in the axial direction and thus impart good conducting properties. Depending on the number of tube-wall layers, CNTs are classified as single-wall (SWCNTs), double-wall (DWCNTs), and MWCNTs. SWCNTs show better electrical conductivity and dispersion properties than DWCNTs and MWCNTs. These properties are advantageous for achieving high TE performances. Yao et al. combined SWCNTs with polyaniline (PANI) [112] and achieved improved Seebeck coefficient and electrical conductivity because of the increase in the carrier mobility caused by the ordered (by \( \pi - \pi \) interactions) PANI chains along the surface of the SWNCTs. On the other hand, the thermal conductivity of the composite did not increase significantly because of the reduced lattice thermal conductivity by the phonon scattering at many SWCNTs-PANI interfaces. Hence, the S, \( {{\sigma }}, \) and k of the composite could be controlled quasi-dependently, thus enhancing its TE performance (S = 40 μ\( {\text{V\,K}}^{ - 1} \), PF = 20 μ\( {\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} \), ZT = 0.004). The PF could be further enhanced (S = 65 μ\( {\text{V\,K}}^{ - 1} \), PF = 176 μ\( {\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} \) [113] and then, S = 38.9 μ\( {\text{V\,K}}^{ - 1} \), PF = 217 μ\( {\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} \) [50]) by optimizing the doping level (camphor sulfonic acid, CSA) and microstructure of PANI. MWCNTs are also often used for forming composites with PANI because of their low cost as compared to that of SWCNTs. Meng et al. blended MWCNTs with PANI and obtained the following TE performance parameters: S = 28.6 μ\( {\text{V\,K}}^{ - 1} \), PF = 5.04 μ\( {\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} \), ZT = 0.003. Although the TE performance of this composite was better than those of pure PANI (S = 2.74 μ\( {\text{V\,K}}^{ - 1} \)) and MWCNTs (S = 12.2 μ\( {\text{V\,K}}^{ - 1} \)), it was inferior to that of SWCNTs/PANI composites. This is why MWCNTs are inferior to SWCNTs in terms of properties and dispersion ability [46]. The enhanced TE performance of the MWCNTs/PANI composite could be attributed to the size-dependent energy filtering effect of the MWCNTs enwrapped by the PANI layer. Table 1 lists the TE properties, fabrication, and cost of various CNT/polymer composites [45, 90, 107, 114,115,116,117,118,119,120].

Table 1 A summary of thermoelectric performances for carbon particle/polymer composites
Full size table

Like CNTs, graphene and graphite are also utilized for making composites owing to their outstanding properties. Various types of chemically treated graphene materials such as GNPs, graphene nanosheets (GNs), and rGO have been used for developing TEGs. Kim et al. reported the composites of PEDOT: PSS and rGO with many functionalities [126]. PEDOT: PSS polymer chains were intercalated inside the stacks of rGO and arranged in rGO surface by \( {{\pi }} - {{\pi }} \) interactions, resulting in an increase in the carrier mobility of the composite and the formation of many internal interfaces between rGO and PEDOT. (TE performance: S = 58.77 μ\( {\text{V\,K}}^{ - 1} , \) PF = 11.09 μ\( {\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} , \) and ZT = 0.021). Zhang et al. used \( {\text{C}}_{60} \), which possess a large Seebeck coefficient (~ 2000 μ\( {\text{V\,K}}^{ - 1} ) \) and low thermal conductivity (~ 0.16 \( {\text{W\,m}}^{ - 1} {\text{K}}^{ - 1} \)) [135] as a dopant to further enhance the TE performance of rGO/PEDOT:PSS composites [127]. The addition of \( {\text{C}}_{60} \) prevented the restacking and agglomeration of rGO [135].\( {\text{C}}_{60} \) also facilitated the formation of interfaces between rGO and PEDOT and the ordering of the PEDOT chains on the rGO surface, which in turn improves the TE performance of the composite (S = 22.9 μ\( {\text{V\,K}}^{ - 1} , \) PF = 32.4 μ\( {\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} , \) ZT = 0.067). In addition to these carbon nanomaterials, graphite has also been used for preparing TE composites due to its ready availability as compared to graphene. However, the PF of graphite-based TE materials [131, 132] is lower than that of graphene-based TE materials [49, 122, 123]. This is the reason that graphite shows low electrical conductivity and poor dispersion ability. Table 1 lists the TE performances of graphene/polymer and graphite/polymer composites.

Layer-by-layer assembly (LBL) of carbon materials/polymer composites has also been investigated to enhance their TE performance. Chol et al. fabricated a composite composed of graphene/PANI/DWCNTs/PANI stacked quad-layers to enhance its TE performance [133]. At 40 quad-layers, the Seebeck coefficient and PF of the composite were optimized (S = 130 μ\( {\text{V\,K}}^{ - 1} , \) PF = 1825 μ\( {\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} ) \). The TE performance of this quad-layer composite could be further enhanced by using PEODT: PSS as the stabilizer when graphene and DWCNTs were used as the carbon nanomaterial [134]. At 80 quad-layers of graphene-PEDOT: PSS/PANI/DWCNTs-PEDOT: PSS/PANI, the composite shows a Seebeck coefficient of 120 μ\( {\text{V\,K}}^{ - 1} {\text{and}} \) a PF of 2710 μ\( {\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} \) was also measured.

Carbon nanocomposites, which offer the advantages of both the carbon materials and polymers, are used as TE materials. Carbon nanomaterials exhibit high electrical conductivity and low dimensions to order the polymer chains on their surface. Polymers on the other hand, exhibit low thermal conductivity, facile processability, and flexibility, which are suitable for the fabrication of TEGs. TEGs with various nanostructures can be easily fabricated by selecting various carbon materials with ordered polymer chains and fabricating unique layered structures. The design of TEGs can be optimized to enhance their TE performance by controlling the amounts of the carbon material and polymer used.

3.4 Carbon Nanocomposite Based TEG

Carbon nanocomposites are widely used for preparing TEGs because their polymer matrix exhibits light-weight and low cost, as compared to solid-state inorganic TE materials. Carbon nanocomposite based TEGs are advantageous over conventional electric power generators since they do not require any moving parts and hazardous working fluids, do not produce noise, and are environmentally friendly as they do not involve any toxic constituents [136]. TEGs can use waste heat from biomass, the sun and earth, which are easily available, and convert it into electricity for space, automobile, and building applications [137].

Wang et al. produces TEGs consisting of only p-type TE materials [50]. As shown in Fig. 7a, they fabricated four p-type SWCNTs/PANI composite legs on a film and connected them electrically in series using a silver paste and gold wires with a power density of 10.4 μ\( {\text{W\,cm}}^{ - 2} {\text{K}}^{ - 1} \). Hong et al. developed a facile and cost-effective spray-printing method for the fabrication of p-type legs. In this method, p-type CNT/P3HTs were spray-printed, patterned on a substrate (as shown in Fig. 7b), and connected using dispenser-printed silver electrode lines [51]. This method yielded the maximum output power of 32.7nW, and the following TE performance parameters were obtained: S = 97 \( \pm \;11 \upmu {\text{V\,K}}^{ - 1} , \) PF = 325 \( \pm \;101 \upmu {\text{W\,m}}^{ - 1} {\text{K}}^{ - 2} \). Usually, the temperature difference is applied along the planar direction of TEGs. Suemori et al. applied the temperature difference along the thickness direction of a TEG consisting of p-type SWCNTs/PS composites [52]. Placing gold electrodes at the top and bottom surfaces of SWCNTs/PS composites and applying the temperature gradient along the thickness direction of the composites as shown in Fig. 7c, a power density of 55mW\( {\text{m}}^{ - 2} \) can be achieved at \( \Delta {\text{T}} = 70 {\text{K}} \). Unlike TEGs consisting of only p-type TE materials, Fang et al. fabricated TEGs consisting of both p- and n-type TE materials [53]. PEDOT: PSS (p-type) and SWCNTs (n-type) were deposited alternately on a film, interconnected by silver electrodes, and then rolled. During rolling, the in-plane temperature gradient was applied to the 100 p–n legs TEG (Fig. 7d). The power density of these legs was 0.13 μ\( {\text{Wcm}}^{ - 2} \) at \( \Delta {\text{T}} \) = 45 K. In order to fabricate continuous p–n leg patterns and large-area TEGs, Zhang et al. used a roll-to-roll (R2R) printing method as shown in Fig. 7E [54]. The PEDOT: PSS (p-type) and graphene (n-type) strips were printed sequentially and alternatively on a flexible plastic film by a printing roller and were then connected by a connector strip. An output power of 0.24 μ\( {\text{W\,m}}^{ - 2} \) was generated at \( \Delta {\text{T}} = 10\, {\text{K}} \).

Fig. 7
figure 7

TE generator and fabrication method: a TE generator consisting of four p-type SWCNTs/PANI composite legs. Reproduced with Permission [50]. Copyright 2016, John Wiley and Sons. b TE generator consisting of spray-printed p-type CNT/P3HT composites. Reproduced with Permission [51]. Copyright 2015, Royal Society of Chemistry. c TE generator consisting of spatula-printed p-type SWCNTs/PANI composites. Reproduced with Permission [52]. Copyright 2013, AIP Publishing. d Rolled TE generator consisting of p–n legs (PEDOT: PSS–SWCNT). Reproduced with Permission [53]. Copyright 2017, John Wiley and Sons. e TE generator consisting of p–n legs (PEDOT: PSS–graphene) printed by roll-to-roll method. Reproduced with Permission [54]. Copyright 2016, Manufacturing Letters. f TE generator consisting of layer structured p–n legs (CNTs/PVDF–PVDF (insulator) − CNTs/PVDF). Reproduced with Permission [55]. Copyright 2012, American Chemical Society

Full size image

Unlike TEGs in which TE materials are connected electrically in series by electrodes, Hewitt et al. prepared layer-structured TEGs by using an insulating PDVF layer for fabricating a p–n junction as shown in Fig. 7f [55]. The p-type CNTs/PVDF and n-type CNTs/PVDF layers were alternated between the PVDF insulation layers and formed a p–n junction. A maximum power of 137 nW was generated by 72 p–n junctions.

Various efforts have been made to optimize the performance of TEGs by using various carbon nanomaterials and unique fabrication methods. The TEG performance is affected by TE materials and their arrangement (only p-types or p–n legs) and size (number of legs). In addition, various fabrication methods such as spray printing, roll-to-roll, and LBL have been investigated for facile and large-scale production of TEGs with enhanced efficiency.

4 Triboelectric Nanogenerators (TENGs)

The recent technological advances have increased the demand of wearable devices and portable electronics worldwide. Most of these devices use a consumable battery, which must be replaced periodically. The high battery replacement cost makes these devices expensive. Hence, efforts are being made to develop permanent self-powered devices to prevent pollution from the after-treatment of the battery and reduce the processing costs. Triboelectric generators convert mechanical energy into electrical energy. They store energy using triboelectrification, which is common in our daily life. They exhibit high efficiency at low frequencies and multiple working modes, and hence are used for various sensing applications. Carbon nanomaterials (such as CNTs, graphene, etc.) exhibit outstanding mechanical and electrical properties. These properties of carbon nanomaterials are exploited to enhance the performance of TENGs.

4.1 Triboelectric Nanogenerator Mechanism

The working mechanism of TENGs is as follows. Triboelectrification, the main phenomenon of TENGs, occurs due to the difference in the fermi levels of different two materials. During friction process, one side gives electrons, while the other side receives them so that the surface becomes positively and negatively charged. During the separation of the charged surface, the negatively charged material transfers a positive charge to the electrode in contact with it to maintain the electrical equilibrium and sends the electrons to the wire that connects the electrode and TENG exterior. Conversely, as the two surfaces approach each other, the electrons come back into the electrode, which is bonded to the negatively charged material, with the charges on the two surfaces being parallel. The repetition of this mechanism transforms the mechanical energy into electrical energy. Apart from the contact-separation mode, the contact-sliding, single-electrode, and freestanding triboelectric-layer modes are the fundamental working modes of TENGs.

4.2 Material Selections and Device Structural Designs for Enhancing Performance

Owing to their flexible and stretchable characteristics, TENGs have gained immense attention for the fabrication of wearable and portable electronic devices. Since the resistance of conventional electrodes (e.g. Cu, Al, etc.) is affected by repetitive folding or cyclic deformation, they are not suitable for application in wearable devices. Carbon materials exhibit outstanding mechanical and electrical properties, and hence function as flexible electrodes. In addition, treated carbon materials are expected to exhibit outstanding performance as triboelectric friction layers. Especially, CNTs are not only used as flexible conductive materials due to their excellent mechanical and electrical properties but also serve as a TENG friction layers due to their surface roughness.

Park et al. prepared TENGs using spray coated SWNT films as the electrode and friction layer (Fig. 8a) [56]. Stretchable energy harvesting electronic skins are fabricated using porous PDMS, PDMS spacer, and SWNT films. The negatively charged PDMS surface generated by the UV/Ozone surface treatment results in additional surface charge through friction with the bottom SWNT electrode. An output voltage of 48 V is obtained at a low pressure of 1 kPa and energy harvesting is possible by finger tapping, bending, twisting, lateral straining, and sound-driven vibrations. Wang et al. reported TENGs using a friction layer of aligned CNTs on the PDMS surface Wang et al. (Figure 8b) [57]. When 40 μm CNTs were arranged on the PDMS surface, the sheet resistance as low as 280 Ω/sq2 and super hydrophobicity was achieved. The CNTs rendered the PDMS layer an effective dielectric electron donating layer. Hence, the PDMS layer positive triboelectric layer. The output voltage, current density, and power density of the TENG were 150 V, 60 mAm−2, and 4.62 W m−2, respectively. The TENG was well stretched and no change was observed in its performance after 12,000 cycles, indicating its high cyclic stability.

Fig. 8
figure 8

CNT based TENGs: a spray coated SWNT film based TENG. Reproduced with Permission [56]. Copyright 2014, John Wiley and Sons. b Aligned CNT structure based TENG. Reproduced with Permission [57]. Copyright 2016, RSC Pub. c Fingertip-inspired CNT-PDMS based TENG. Reproduced with Permission [58]. Copyright 2017, Elsevier. d Plasma etched PDMS-CNT based TENG. Reproduced with Permission [59]. Copyright 2017, Royal Society of Chemistry

Full size image

Composites consisting of CNTs dispersed in a polymer matrix can also be used as TENG friction layers. Chen et al. reported a fingertip-inspired freestanding TENG using double spiral CNT-PDMS (DS CNT-PDMS) electrodes prepared by dispersing CNTs in PDMS, transferring the dispersion in a spiral mold, and then curing it (Fig. 8c) [58]. The TENG was divided into three parts: fingerprint, epidermis, and dermis. The DS CNT-PDMS electrodes mimicking fingerprints generated energy with sliding. The porous CNT-PDMS structure mimicking the dermis served as a pressure sensor. The pure PDMS layer, mimicking epidermis, acted as the substrate for the DS CNT-PDMS electrodes and prevented their interference with the porous CNT-PDMS. Kim et al. carried out fluorocarbon plasma-etching on the surface of a CNT-PDMS composite to increase its surface area by exposing the CNTs embedded in the PDMS. They carried out the chemically fluorination of PDMS and CNTs to increase the charge density of the composites (Fig. 8d) [59]. The use of CNTs improved the mechanical properties such as durability of the composite. At the CNT concentration of 4 wt% and after plasma etching for 60 s, an output voltage of 77.8 V was achieved (area of the specimen = 20 mm × 30 mm). This is 248.7% higher than that of pure PDMS and 106.5% higher than that of un-etched PDMS-CNT. At 3 MΩ, an output power of 1.98 mW (3.29 W m−2) was obtained. When attached to human body, this device could distinguish the movement of the person with the speed to be charged.

The high electrical conductivity, outstanding mechanical properties, and large aspect ratio of CNTs enhance the stretchability and performance of TENGs. Hence, TENGs exhibit multiple working modes and versatility.

Graphene, which is a 2D carbon nanomaterial, exhibits excellent mechanical properties with a Young’s modulus of 1 TPa and an electron mobility of 10,000–50,000 cm2 V−1 s−1, and hence is used for fabricating flexible thin electrodes by chemical vapor deposition (CVD). Lee et al. reported transparent flexible TENGs using CVD-grown graphene (Fig. 9a) [138]. A single graphene layer showed an output voltage and a current density of 5 V and 500 nA, respectively. However, the higher the number of graphene layer, the lower the output voltage and current due to the weakening of the electronic interactions between the graphene layers and the puckering effect. Song et al. reported a TENG consisting of different materials on the top (graphene sheet) and bottom indium tin oxide (ITO) electrodes (Fig. 9b) [139]. Although the sheet resistance of ITO is lower than that of monolayer graphene, the resistance of ITO increases rapidly when strain is applied. Therefore, graphene was used as the top electrode, where strain occurred repeatedly but ITO was used as the bottom electrode in which there is no strain under pressure. The high conductivities of the electrodes contributed to the high performance of the TENG. At 200 MΩ, the output voltage was 104 V and the output power density was 60.7 μW cm−2. The TENG retained its performance even after 700 cycles.

Fig. 9
figure 9

Graphene based TENGs: a graphene sheet based TENG. Reproduced with Permission [138]. Copyright 2014, John Wiley and Sons. b Graphene-ITO electrodes based TENG. Reproduced with Permission [139]. Copyright 2016, IOP Publishing. c A flexible and transparent graphene based TENG. Reproduced with Permission [140]. Copyright 2016, IEEE. d Wearable e-textile based TENG. Reproduced with Permission [141]. Copyright 2016, American Chemical Society. e Graphene electrode and porous PDMS based TENG. Reproduced with Permission [60]. Copyright 2016, Springer Nature. F Conformal graphene based TENG. Reproduced with Permission [61]. Copyright 2016, Elsevier. g Power generator using water droplet. Reproduced with Permission [62]. Copyright 2016, American Chemical Society. h PEDOT: PSS coated graphene based TENG. Reproduced with Permission [63]. Copyright 2017, American Chemical Society. i Aligned graphene sheets based TENG. Reproduced with Permission [64]. Copyright 2017, Elsevier. j Crumpled graphene based TENG. Reproduced with Permission [65]. Copyright 2017, Elsevier. k rGONRs based single-electrode TENG. Reproduced with Permission [66]. Copyright 2016, Springer Nature. l GO based single-electrode TENG. Reproduced with Permission [67]. Copyright 2017, American Chemical Society

Full size image

Shankaregowda et al. also reported a flexible and transparent TENG using graphene (Fig. 9c) [140]. The CVD-grown graphene used by them provided high conductivity and optical transmittance (97%) to the TENG. On the PDMS surface, a thin plasma-treated layer (fluorocarbon polymer) was applied to improve the output performance. The maximum output voltage and current of 650 V and 12 μA, respectively were obtained at 4.3 Hz pressure. It showed an effective electrical power of 3.28 mW at 60 MΩ.

Wu et al. reported a wearable TENG with high compatibility with smart electronic textiles (e-textiles) and clothing (Fig. 9d) [141]. To make e-textiles, silver nanowire (AgNW)-coated textile was cured after coating with GO and was then reduced with hydrazine hydrate to make a textile/AgNW/graphene composite. With an increase in the frequency, the output current decreased with decreasing charging and discharging. The textile/AgNW/graphene core–shell structure increased the conductivity of the e-textiles. Its low sheet resistance of 20 Ω/sq2 provided superb conductivity, transparency, foldability, flexibility and stretchability to the TENG.

He et al. used graphene as the top and bottom electrodes and as the TENG friction layer (Fig. 9e) [60]. PDMS with 500 nm-sized ZnO spheres was cured and then etched with hydrochloric acid to prepare nest-like porous PDMS. Owing to its well-organized pores, the porous PDMS showed good transmittance and this TENG showed an output voltage and current of 271 V and 7.8 μA, respectively (current density of 1.95 μA cm−2). The peak output power of the TENG reached 0.39 mW at the load resistance of 9.01 MΩ. Chu et al. fabricated a conformal single-electrode TENG using a thin graphene (0.3 nm) electrode (Fig. 9F) [61]. PET (< 0.9 μm) was the portion of the substrate bonded to the human body and PDMS (1.5 μm) acted as the electrification layer generating electricity by contacting various clothes or human body. The plasma treatment produced nanostructures on the surface of PDMS to increase the surface area, and the fluorinated surface increased the peak power by 130 μW with increasing the difference in the electron affinity.

Kwak et al. fabricated an energy generator through triboelectrification between flowing droplets and graphene (Fig. 9g) [62]. A negative charge accumulated on the top side of the graphene as the water droplet moved. The triboelectrification-induced pseudocapacitance generated between the water droplet and graphene on PTFE produced an output power of 1.9 μW. When graphene was transferred onto the PTFE surface by the wet transfer method, triboelectrification between PTFE and deionized (DI) water occurred and the surface of PTFE became negatively charged. This surface attracted holes from the graphene. Therefore, the top side of PTFE and the bottom side of graphene formed an electron double layer, while the excluded electrons on graphene formed an electron double layer (EDL) with the water droplet. The strong negative triboelectric potential of the PTFE surface caused positive and negative charge accumulation at the top and bottom of the graphene layer, respectively.

Yang et al. prepared a wearable TENG with low resistance and high current density (2.4 μA cm−2) and output power (12 μW) by coating PEDOT:PSS (PH1000) on CVD-grown graphene (Fig. 9h) [63]. This TENG exhibited flexibility, stability, portability, and human compatibility, and hence could be used as a human joint sensor. Xia et al. fabricated a TENG using aligned graphene sheets (AGS) (Fig. 9i) [64]. An AGS/PDMS composite was fabricated by uniformly dispersing graphene sheets in PDMS followed by spin coating. The AGS/PDMS showed high capacitance and low dielectric loss due to the dispersion of graphene sheets. At 15 MΩ, an output power density of 4.8 W/m2 was achieved, while the output current and voltage were 21 μA and 530 V, respectively. The parallel structure of the layers and the small gap between the two sheets in the adjacent layers resulted in the maximum breakdown voltage and minimum dielectric loss of the TENG.

Chen et al. reported a crumpled-graphene-based stretchable TENG (Fig. 9j) [65]. A CVD-grown graphene sheet was crumpled using a pre-strained tape to reduce its size and to enhance the effective friction area and roughness, which improved the performance of the TENG. When the pre-strain was 200%, a maximum power density of 1.5 μW cm−2 was achieved. The output voltage increased with an increase in the pre-strain. At the pre-strain of 250%, the output voltage was 15 V, which is 10 times higher than that obtained at 100%. The TENG could detect uniaxial tensile, stretching, compressing, and twisting forces.

Graphene oxide (GO) possesses various functional groups on its rough surface, and hence can be used as a TENG friction layer. The oxygen-containing functional groups on the surface of GO increase its negative charge due to the electronegativity of oxygen. Kaur et al. produced a single-electrode TENG using rGO nanoribbons (rGONRs) (Fig. 9k) [66]. The rGONRs were prepared by completely unzipping the MWCNTs treated with H2SO4 and KMnO4. The rGONRs were used as the friction layer with Al film after mixing it in PVDF. The rGONRs/PVDF films showed high surface roughness and had more functional groups than pristine rGONRs. The rough surface and functional groups of the rGONRs/PVDF films improved capabilities of their charge storage and transfer. Guo et al. used GO as the friction layer in a single-electrode TENG to increase its output voltage and prevent high leakage current (Fig. 9l) [67]. Owing to its high concentration of functional groups and high specific surface area, the GO layer enabled the sensitive force detection of the TENG and improved its antibacterial activity.

The high electrical conductivity of CVD-grown thin graphene sheets improves the flexibility, transparency, and performance on TENGs. The use of GO friction layers with high functional group density and large surface area enhances the performance of TENGs. Graphene-based thin TENGs show high efficiency and compatibility with human body and wearable devices.

4.3 Hybrid Carbon Nanomaterial Based TENGs

Apart from CNTs and graphene, various other carbon materials (with different dimensions) and carbon material hybrids are used for the fabrication of TENGs. When two or more carbon nanomaterials are used, their interactions compensate for weakness of network connection and improve their merits. Yi et al. designed an energy harvesting device with a stretchable TENG and stretchable supercapacitors using CB (Fig. 10a) [68]. CB was used as the top and bottom electrodes, as the friction layer at the bottom of the TENG, and as the stretchable electrode in the supercapacitors. Nano/microstructures were fabricated on the surface of the CB electrode as the bottom friction layer and silicon rubber as the top friction layer using a sandpaper to increase the contact area. During human movements, an open circuit voltage (VOC) of up to 400 V and a short circuit current density (\( \Delta \sigma_{\text{SC}} \)) of up to 97 μC m−2 was achieved. The TENG and supercapacitors were encapsulated with silicone rubber, and hence were waterproof and washable. Owing to its stretchability, the TENG could harvest energy from various deformations including pressing, stretching, bending, and twisting. Li et al. prepared a conductive elastomer by mixing CB and CNTs with Ecoflex to fabricate a TENG composed of repeating wavy and flat layers (Fig. 10b) [69].

Fig. 10
figure 10

Other and hybrid carbon materials based TENGs: a Carbon black based TENG. Reproduced with Permission [68]. Copyright 2018, American Chemical Society. b Carbon black and CNT based TENG. Reproduced with Permission [69]. Copyright 2017, John Wiley and Sons. c CNT and EGO based TENG. Reproduced with Permission [70]. Copyright 2017, Elsevier

Full size image

Energy harvesting could be achieved through pressure, stretching, and folding. At the external load of 10 MΩ, the maximum power density was around 0.1 W m−2. As device was stretched, the ΔQsc and Voc increased to 10 nC and 14 V, respectively. However, no improvement of performance was observed beyond 100% strain because the two layers were in contact with each other. Karumuthil et al. fabricated a piezo-tribo nanoenergy harvester using a ZnO nanorod, exfoliated graphene oxide, and MWCNTs dispersed in PDMS (Fig. 10c) [70]. A micro gap existed between the surface of PDMS and the copper structure, and triboelectricity was generated by pressing. When the device was continuously pressed and deformed, the positive and negative centers of the molecule were separated by internal deformation. The MWCNTs not only reduced the internal resistance of the device but facilitated the dispersion of ZnO and improved the stress distribution, thus enhancing its piezopotential. Carbon nanocomposites consisting of exfoliated graphene oxide (EGO) with one atom thickness show excellent piezoelectric properties and flexibility. Nanomaterials such as EGO and ZnO increase the dielectric constant, and hence the charge density of the PDMS surface. By exploiting the properties of carbon nanomaterials, the performance of TENGs can be improved. The optimal combination of various carbon nanomaterials improves the performance of TENGs through interactions between them.

5 Conclusions

Owing to their unique dimensional, electrical, thermal, and mechanical properties, carbon nanomaterial-based composites are suitable for sensor and energy harvesting applications. In this review, we have discussed the progress made in the development of sensors and energy harvesting devices over the past few years in terms of the selection of various carbon nanomaterials and matrices, unique structure design, and chemical treatments.

Carbon nanomaterial composite-based sensors exhibit high sensitivity and can detect multiple forces. The 3D mapping, structural design, chemical treatment, and optimization of carbon nanomaterials improve their sensor and light harvesting applications. Carbon nanomaterial based TEGs exhibit high ZT. The structural integration of various carbon nanomaterials like CB, CNTs, graphene, and rGO can improve the efficiency of thermoelectric systems.

Triboelectric carbon nanocomposite generators use the electrical and mechanical properties of carbon nanomaterials. These properties render these devices durable and improve their energy harvesting efficiency. However, their fundamental working mechanism has not been elucidated yet and should be investigated further in order to optimization their output performance and improve their energy conversion rate and efficiency.

Therefore, further investigations are required for optimizing the structural design of carbon nanocomposites to realize their high-performance and high-durability sensor and energy harvesting applications.