Low-Cost and Scalable Nanomanufacturing Processes for Obtaining Carbon Nanotube-Based Devices

Abstract

The increasing demand for materials with enhanced properties and high-performance devices has driven substantial research into nanomanufacturing, particularly using carbon nanotubes (CNTs), because of their exceptional properties and high sensitivity to chemical doping. In this way, this work summarizes nanomanufacturing methods for CNT-based devices developed in Brazil, covering the complete cycle from nanocomposite production to functional device assembly across cellulosic, polymeric, and metallic matrix systems. For cellulosic matrices, vacuum filtration enables the production of buckypaper, which is subsequently assembled into chemiresistive, thermoresistive, and thermoelectric devices. For polymeric matrices, 3D printing combined with surface functionalization techniques (spray coating, inverted immersion, and direct immersion) produces piezoresistive robotic sensors, metal-free thermal sensors, and biomedical scaffolds for tissue engineering. For metallic matrices, electrodeposition can produce Cu-CNT-coated aluminum comparable to traditional copper power transmission cables, while arc welding produces stainless steel composites with properties comparable to commercial high-grade steels. These devices have commercial and industrial applications, with low-cost and scalable production methods in comparison with conventional materials. Characterization results demonstrate that CNT integration into diverse matrices successfully bridges nanoscale properties to macroscopic functional devices. Current challenges include uniform CNT dispersion and structural defect control, laboratory to industry scale transition, and long-term device stability under environmental conditions. Future perspectives encompass lab-on-chip systems, wearable devices, 3D-printed smart structures, Internet of Things integration, and machine learning-enhanced analytics.

1. Introduction

The increasing demand for materials with enhanced properties, miniaturized sensors, and high-performance devices has driven substantial research into nanomanufacturing, from nanostructuring for nanocomposite production to functional device assemblies. This approach enables the development of sensors, coatings, reinforcements, and components with superior performance compared to conventional materials [1,2,3]. Among the diverse families of nanomaterials used in such applications, carbon-based structures have attracted both academic and industrial attention due to carbon’s ability to form multiple allotropes through different hybridization states (sp, sp², and sp³) [4], enabling the synthesis of materials with vastly different properties, such as fullerenes [5], graphene [6], and carbon nanotubes (CNTs) [7]. The literature indicates some advantages for CNT-based nanocomposites in substrate adhesion, mechanical properties, and electrical properties [7,8,9].

Discovered by Iijima in 1991 [10], CNTs are one-dimensional cylindrical structures of sp²-hybridized carbon atoms arranged in hexagonal lattices, classified as single-walled, double-walled, or multi-walled based on their number of cylindrical shells [11,12]. Their unique structure imparts exceptional mechanical, electrical, and thermal properties, such as tensile strengths up to 100 GPa, conductivities of ${10}^3$ to ${10}^8$ S/m, and thermal conductivities of 2000 to 7500 W/m·K, along with high aspect ratios, large surface area, low density, and chemical stability [13,14,15], making them ideal candidates for sensing applications and nanocomposite reinforcement [7,16].

The nanomanufacturing of CNT-based devices requires incorporating the nanotubes into matrices that enable practical handling and device assembly. To overcome the handling challenges posed by the nanoscale dimensions, over the past decades, researchers have developed strategies to incorporate CNTs and other carbon-based nanomaterials into nanocomposites using matrices such as polymers, cellulose, ceramics, and metals [17,18]. This integration, while facing challenges such as agglomeration due to van der Waals interactions [19,20], is facilitated by covalent functionalizations such as the introduction of carboxylic acid groups (-COOH) through wall oxidations via acid treatment, which enhance dispersibility in polar solvents and interfacial adhesion through hydrogen bonding [21,22]. Recent works also point to the usage of hybrid nanocomposites, using carbon black or graphene with the CNTs to improve the integration and performance of the materials [23,24,25].

Several nanomanufacturing approaches have been developed to produce CNT-based devices using different matrix systems. In cellulosic matrices, vacuum filtration enables the fabrication of buckypapers (BPs), which are thin, flexible films formed by entangled CNT networks dispersed on cellulosic paper [26,27]. Devices usually employ them fixed onto solid substrates such as Bakelite with copper terminals and conductive silver ink contacts [28,29,30,31]. Chemiresistive sensors based on buckypapers have demonstrated capability for detecting adulterants in beverages [28] and ether-amine in mining industry wastewater [29], while Pinheiro et al. [30,31] have demonstrated thermoresistive and thermoelectric properties of buckypaper-based devices.

In polymeric matrices, additive manufacturing through 3D printing enables the production of on-demand complex and customizable geometries [32,33]. Combined with coating and impregnation techniques such as inverted immersion, direct immersion, and spray coating, this approach produces nanostructured devices with tailored properties [34,35,36,37]. Piezoresistive sensors produced by spray coating on poly(acrylonitrile-butadiene-styrene) (ABS) substrates have shown potential for robotic haptic feedback systems [34,38], while thermoresistive sensors based on poly(latic acid) (PLA) and CNT nanocomposites offer sustainable alternatives to conventional metal-based thermal sensors [35]. Furthermore, biomedical scaffolds produced by inverted immersion coating of PLA filaments demonstrate the versatility of these techniques for tissue engineering applications [36,37].

In metallic matrices, electrodeposition and arc welding techniques enable the production of CNT-reinforced coatings and high-performance welds [39,40,41]. Pulsed electrodeposition of Cu-CNT coatings on aluminum substrates has achieved significant improvements in electrical conductivity [39], while direct current electrodeposition on aluminum wires achieved conductivity of 72.96% of the International Annealed Copper Standard (IACS) [40]. Arc welding using nanostructured flux-cored wires has produced stainless steel composites with enhanced mechanical properties and cavitation erosion resistance suitable for high-demand applications, such as hydroelectric turbine repairs [41].

These approaches offer a cost advantage due to the simplicity of the methods and the limited quantity of raw material required. CNTs’ cost depends on several factors, such as the number of walls, length, and purity; prices can reach as high as USD 1000/g for high-quality and pure CNTs, while lower-quality nanotubes range between USD 100 and 200/kg [17,42]. The CNTs used in these studies cost approximately USD 17/g, making these methods cost-effective by using small amounts of CNTs, typically only a few dozen milligrams [28,29,30,31,34,35,36,37].

In addition, 3D printing benefits from inexpensive materials. For instance, 1 kg spools of ABS or PLA are usually priced below USD 20. Similarly, BPs make use of filter papers, which are also available for under USD 20. In the case of metallic matrices, while electrodeposition and arc welding processes might incur higher overall costs, nanomanufacturing offers a cost advantage over conventional materials. For example, nanostructured wires for arc welding are priced at approximately USD 64.3/kg, compared to commercial wires, which cost around USD 112.8/kg [39,40,41].

Thus, this review presents low-cost and scalable nanomanufacturing processes to obtain CNT-based devices across cellulosic, polymeric, and metallic matrix systems. These processes demonstrate examples of functional devices with their characterization and performance results. For this, Section 2 examines buckypaper-based sensors produced by vacuum filtration in cellulosic matrices. Section 3 explores additive manufacturing combined with coating techniques for polymeric matrix devices. Section 4 addresses electrodeposition and arc welding for metallic matrix devices, including Cu-CNT coatings and stainless steel-CNT welding. Finally, Section 5 presents conclusions and discusses current challenges and future perspectives, including Internet of Things integration, machine learning-enhanced analytics, energy applications, and biomedical devices.

2. Cellulosic Matrix Devices: Buckypapers

Buckypapers (BPs) are micrometer-thick films of entangled CNT networks held together by van der Waals forces, exhibiting a porous structure with high surface area [26,27]. These flexible and lightweight macroscopic films enable the practical manipulation of nanoscale materials for integration into functional devices. The dimensionality of CNTs represents a central challenge for device fabrication, and BP preparation emerges as an effective strategy to overcome this difficulty. The vacuum filtration technique is the most common method for BP preparation, offering a simple, low-cost, and versatile processing route with adjustable parameters such as nanomaterial type, dispersing medium, membrane characteristics, solution concentration, and vacuum pressure [43,44,45].

The production of buckypapers follows the procedure illustrated in Figure 1. First, functionalized MWCNTs-COOH (99.80% purity, external diameter 10–30 nm, length 1–10 $\mathsf{\mu }$m) are dispersed in isopropyl alcohol (1.0 g/L) using ultrasonic agitation at 40 kHz for 60 min at room temperature [29,30]. In the covalent functionalization process, the use of strong acids (H₂SO₄ or HNO₃) and ultrasonic bath treatment forms carboxylic acid groups (-COOH) through the oxidation of the nanotube surface, breaking the $\pi$ bonds and improving dispersion in polar solvents without the addition of surfactants [29]. The dispersion is then vacuum filtered through cellulose filter paper (grammage: 80 g/m², nominal thickness: 205 $\mathsf{\mu }$m, pore size: 14 $\mathsf{\mu }$m) using a Büchner funnel attached to a Kitasato flask connected to a vacuum pump (nominal vacuum of 2.66 Pa) [28,29]. The film is subsequently dried at 100 °C for 1 h to ensure complete removal of humidity and solvent. The final BP exhibits approximately 175 $\mathsf{\mu }$m total thickness, with the MWCNT-COOH layer (approximately 67 $\mathsf{\mu }$m) permeating about 40% of the filter paper [28,31]. The carboxylic acid groups establish hydrogen bonding with cellulose hydroxyls due to strong polar interactions, transforming the paper from an electrical insulator into a conductive material [29,31].

For sensor device fabrication, BP samples are cut from the regions between the center and edge of the film with dimensions of approximately 2.0 cm × 0.5 cm (active area of 1.0 cm²) [28,29]. The samples are fixed onto solid substrates such as Bakelite or glass coverslips using polyvinyl acetate (PVA) glue or commercial silver conductive ink. Copper electrodes are connected to the BP using conductive silver paint, with electrode separation of 1.5–1.6 cm depending on the application [28,31]. For chemiresistive sensor applications, the BP samples are configured for two-point electrical resistance measurements, where analytes are applied directly to the sensor surface using a micropipette. L. Ferreira et al. [28] used chemiresistive BPs to identify adulterants in Port wine samples, while D. Ferreira et al. [29] used them to detect ether-amine, an expensive substance used for ore flotation, in industrial wastewater. For thermal sensors, Pinheiro et al. [30,31] developed both coplanar and sandwich configurations. In the coplanar arrangement, the BP tape is connected to copper electrodes that serve as contacts for resistance measurement, while additional electrodes on the opposite side conduct heat from an external resistor. The sandwich configuration consists of BP placed between copper-based hot and cold sides (active area of 16 cm², separated by 1.5 cm) on a Bakelite substrate, with an aluminum heat sink establishing the temperature gradient for thermoelectric measurements [31].

Scanning electron microscopy (SEM), as shown in Figure 2a, reveals the BP microstructure across different applications. SEM micrographs obtained in secondary electron detection mode show that the cellulose fibers of the paper membrane form a support framework for the CNT film, with MWCNTs-COOH distributed on the surface and between micrometric cellulose fibers, forming a nanostructured sheet [28,31]. The morphology shows agglomerated nanotubes uniformly distributed in the BP, with retention and impregnation of CNTs into the cellulosic fibers due to surface interactions between them [31]. Cross-sectional analysis, performed after sample fracture following immersion in liquid nitrogen, confirms that the MWCNT layer permeates the paper structure due to the forces exerted on the nanotubes through the pores during vacuum filtration [28,29,46]. This integrated nanocomposite morphology demonstrates that functionalized CNTs have greater adherence to paper than pure CNTs since the carboxylic groups allow strong interaction with the hydroxyls of cellulosic fibers [31].

Raman spectroscopy, displayed in Figure 2b, provides structural information about CNT quality and CNT-analyte interactions in buckypaper sensors. All BP samples exhibit characteristic D and G bands of functionalized MWCNTs in the 1100–1450 cm⁻¹ and 1525–1700 cm⁻¹ regions, respectively [29,31]. Raman spectra are typically fitted using Lorentzian deconvolutions to identify characteristic bands and sub-bands, including satellite bands around the D band (${\mathrm{D}}^{″}$, ${\mathrm{D}}_{LA}$, ${\mathrm{D}}_{LO}$, ${\mathrm{D}}_L$, ${\mathrm{D}}_R$) that arise from structural changes due to chemical functionalization, and the ${\mathrm{D}}_{middle}$ sub-band (1450–1500 cm⁻¹) corresponding to amorphous carbon [29]. The G band splits into ${\mathrm{G}}_{outer}$ (1571–1576 cm⁻¹) and ${\mathrm{G}}_{inner}$ (1586–1603 cm⁻¹) sub-bands associated with the outermost and innermost tubes, respectively, along with the D’ sub-band as a defect-activated shoulder [29]. The ${\mathrm{G}}^{\prime }$ band (2500–3000 cm⁻¹) comprises ${\mathrm{G}}_{inner}^{\prime }$ and ${\mathrm{G}}_{outer}^{\prime }$ sub-bands related to diameter distributions. For BPs exposed to ether-amine, additional vibrational modes appear in the 2727–3306 cm⁻¹ region corresponding to C-H and N-H stretching vibrations [29]. Analysis of band shifts reveals n-type charge transfer from ether-amine to the CNTs’ outermost layers, evidenced by redshifts of up to 4 cm⁻¹ in ${\mathrm{G}}_{outer}^{\prime }$ [29]. The ${\mathrm{D}}_{middle}$ sub-band analysis shows that 100% ether-amine induces crystallinity loss with approximately 42% in comparison to as produced BP, indicating that although ether-amine acts as an electron donor, it introduces amorphous carbons that could block conduction channels [29].

The electrical characterization illustrated in Figure 2c demonstrates the sensing capabilities of buckypaper devices through the monitoring of electrical resistance as a function of time using two-point measurements with digital multimeters [28,29,31]. The response of the sensor is calculated from the change in normalized relative resistance according to

$$Response={\displaystyle \frac{R_f-R_0}{R_0}},$$

(1)

where $R_0$ is the initial electrical resistance without analyte and $R_f$ is the final resistance after analyte application, both measured at room temperature [28,29]. For chemiresistive sensing, the working principle is based on changes in electrical properties induced by charge transfer, changes in electron scattering, contact effects, and capacitance changes between analytes and CNTs during chemical interaction [28]. For Port wine adulteration detection, water addition produces positive normalized resistance responses (0.50 ± 0.04 for 5% water, 0.75 ± 0.16 for 10% water), while ethanol addition produces negative responses (−0.65 ± 0.10 for 5% alcohol, −0.21 ± 0.26 for 10% alcohol) [28]. This distinct behavior enables discrimination between adulterating agents, with response times of 10.0 ± 3.60 s for 5% ethanol and recovery times of up to 50.6 min for 10% water [28]. For ether-amine detection, responses increase dramatically with concentration: 446.25 ± 150.14% for 1%, 39,699.14 ± 9154.10% for 5%, 174,241.57 ± 130,005.32% for 10%, and up to 723,974% for 100% ether-amine [29]. Response times range from 3.62 ± 0.99 min (5% ether-amine) to 14.86 ± 1.32 min (10% ether-amine), with recovery times of 9.23 to 21.16 min depending on concentration [29]. For thermal sensing, BP exhibits thermoresistive behavior with electrical resistance decreasing during heating and sensitivity up to −10.05% at 322 K [31]. The thermoelectric configuration in the coplanar arrangement achieves a maximum thermovoltage of −1.2 mV and thermoelectric power of −0.09 mV/K from a temperature gradient of 19 K [31].

Principal Component Analysis (PCA, Figure 2d) validates sensor discrimination capability for buckypaper-based multisensor platforms. PCA represents an unsupervised linear exploratory technique that transforms multidimensional datasets into a reduced coordinate space of principal components (PCs), enabling the visualization of data variability and pattern recognition [28,29]. For wine adulteration detection, PCA processing generates principal components, with PC1 explaining 62.0% and PC2 explaining 23.6% of total variance (cumulative 85.6%), showing clear separation between unadulterated and adulterated samples in distinct quadrants of the biplot [28]. For ether-amine detection, PCA analysis (PC1: 71.43%, PC2: 20.40%, cumulative 91.84%) demonstrates distinct clustering for different concentrations of 1%, 5%, and 10% ether-amine [29]. These results validate BP sensors as low-cost multisensor platforms for beverage quality control and mining industry applications, offering accurate, portable, and accessible analytical methods as alternatives to complex techniques such as chromatography, colorimetry, and FTIR spectroscopy [29].

3. 3D-Printed Polymeric Matrix Devices

Additive manufacturing, particularly fused deposition modeling (FDM) and fused filament fabrication (FFF), has emerged as a promising approach for fabricating CNT-based polymeric devices with complex geometries and tailored functionalities [35,47]. This technology enables the creation of three-dimensional objects from digital models, offering opportunities for rapid prototyping and on-demand manufacturing of customized parts for specific applications. The nanostructuring of polymeric substrates and filaments with CNTs through various coating techniques combined with 3D printing provides enhanced mechanical, electrical, and sensing properties while maintaining the design flexibility inherent to additive manufacturing.

Poly(lactic acid) (PLA) has gained particular attention in 3D printing applications due to its biodegradability, biocompatibility, renewable origin, and suitable melting point for FDM processing [35,36]. Similarly, poly(acrylonitrile-butadiene-styrene) (ABS) offers excellent mechanical properties and printability, making it suitable for structural and functional applications [34]. The combination of these thermoplastic polymers with CNTs through surface coating techniques such as spray coating, inverted immersion, and direct immersion enables the development of nanostructured devices with applications in pressure sensing, healthcare monitoring, biomedical engineering, and robotic systems.

3.1. ABS/CNT Piezoresistive Device

Spray coating is a post-printing functionalization method for depositing CNT-based nanocomposite coatings onto 3D-printed polymeric substrates, offering process simplicity, low material consumption, and applicability to complex geometries [34]. The production process consists of preparing the substrate and the coating ink. For the substrate, ABS sensing elements with specific geometrical features (25 mm length, 5 mm width at terminals, 7 mm width at central sensing area, and 0.8 mm thickness) are 3D-printed via FDM at 220 °C with 10 mm/s printing speed using a 0.2 mm nozzle [34]. The printing line orientation, parallel or perpendicular to the direction connecting the electrodes, induces intentional anisotropy in the electrical and mechanical properties of the substrate [47,48].

For the coating ink, an ABS/acetone solution is prepared by dissolving 5 g of ABS filament in 25 mL of acetone through ultrasonication at 40 kHz for 60 min. Separately, an acetone-DMF/MWCNT-COOH suspension is prepared by dispersing 20 mg of carboxylic acid-functionalized MWCNTs in a mixture of 10 mL acetone and 10 mL DMF, followed by ultrasonication at 42 kHz for 80 min [34,49]. DMF improves MWCNT dispersion stability, while acetone promotes superficial solubilization of the polymeric substrate during spray coating to enhance CNT–polymer adhesion [49,50].

The spray coating procedure, illustrated in Figure 3, is performed using an aerograph positioned approximately 10 cm from the substrate surface. The coating is applied in cycles: each one comprises CNT-COOH solution pulverization, 10 min of drying, and electrical resistance measurements to monitor the percolation process. After approximately five coating layers, the electrical resistance drops from above ${10}^{12}$ $\Omega$ to approximately ${10}^6$ $\Omega$, achieving electrical percolation at only 0.22 wt% CNT mass percentage [34]. After drying, the sensor terminals are covered with conductive silver ink, and the center is sealed with ABS/acetone solution [34].

FEG-SEM characterization, as shown in Figure 4a, reveals progressive spreading of the nanostructured film with an increasing layer number, with CNT agglomerates and random nanotube arrangement forming conduction channels over the ABS surface. Nanotubes appear fused to the substrate, indicating formation of a true nanocomposite layer with an estimated thickness of a few dozen CNT diameters, which should be below 1 $\mathsf{\mu }$m [34]. The Raman spectroscopy data in Figure 4b confirm the structural integration between CNTs and the ABS matrix. The spectrum exhibits overlapping fingerprints of both components, with blueshifts in the G-band (8 cm⁻¹) and 2D subbands (25–56 cm⁻¹) indicating compression and p-type doping from CNT–polymer interactions [34].

The piezoresistive response displayed in Figure 4c demonstrates an operational range from 140 to 363 kPa, with a maximum sensitivity of 0.75 ± 0.36% for samples with parallel printing lines. The anisotropic response allows distinction between directions of mechanical stress, with response times of around 1–6 s and recovery times of around 2–15 s, depending on printing orientation [34]. The spray coating method achieves functional performance at significantly lower CNT loadings compared to other piezoresistive nanocomposites reported in the literature, using 0.17 ± 0.19 mg per 62.24 ± 0.15 mg ABS substrate [34,48,51,52].

The operational range positions the ABS/CNT piezoresistive sensor for applications including flexible wearable motion monitors, robotic haptic feedback systems, and industrial inspection devices. As a practical application example, Pereira Junior et al. [38] demonstrated the implementation of this spray-coated sensor in a service robot wheel for terrain interaction detection. The wheel was 3D-printed using PLA and TPU filaments, with the sensor integrated into the wheel drum through a trigger pin and fixing component assembly. The on-board electronics, comprising an ESP32 microcontroller, an HX711 amplifier, and a Wheatstone bridge circuit, enabled wireless data transmission via Wi-Fi. During field tests, the piezoresistive response indicated wheel-ground contact and pressure variations from different soil types, demonstrating the potential for enhancing robot traction analysis and motion control in field operations [38].

3.2. PLA/CNT/Bioglass Biocompatible Scaffolds

Inverted immersion is a pre-printing functionalization method for coating polymeric filaments before 3D printing, enabling the incorporation of CNTs throughout the internal structure of the printed part [36]. The coating solution is prepared by solubilizing PLA (0.2 g) in acetone mixed with carboxylic acid-functionalized MWCNTs (0.007 g) under ultrasonication for 360 s. The inverted immersion procedure, illustrated in Figure 5, uses a funnel with an opening close to the filament diameter (1.75 mm). The coating solution is added at the funnel top, and the filament is pulled upward at 0.05 m/s until the entire length is coated. After drying at room temperature for 2 h, a second coating layer is applied to increase nanoparticle loading, followed by 5 h drying before use in 3D printing [36,37].

Vasconcelos et al. [36] employed inverted immersion to develop PLA-based biomedical scaffolds reinforced with CNTs and bioglass for osseous tissue engineering. The nanostructured PLA scaffolds were fabricated via FDM at 210 °C (first layer) and 180 °C (subsequent layers) with a 0.2 mm diameter extruder nozzle, ensuring that the CNTs deposited on the filament surface became incorporated throughout the entire polymeric matrix during extrusion. The scaffolds were designed with an interleaved honeycomb architecture forming hexagonal pores with inner spacing of 2 mm and 4 mm, which, according to the literature, favors angiogenesis and promotes good vascularization, attachment, proliferation, and cell differentiation [36]. Additionally, the scaffolds were thermally impregnated with bioglass particles on the surface by placing samples in a stove at 130 °C to increase bioactivity.

SEM characterization revealed the appearance of pores in the structure of the PLA/CNT filament, with an average pore size of 5.83 ± 3.20 $\mathsf{\mu }$m. The micrograhps, such as in Figure 6a, showed CNTs exposed before thermal impregnation, and the inner distance between filaments and layers ranged between 130 and 500 $\mathsf{\mu }$m for scaffolds with a 2 mm inner spacing and between 150 and 800 $\mathsf{\mu }$m for scaffolds with a 4 mm inner spacing, which falls within the optimal range for osteogenesis and angiogenesis [36]. The Raman spectroscopy results in Figure 6b confirm the successful interaction of CNTs with the PLA matrix, with characteristic D, G, and 2D bands, along with new peaks between 1400 and 2600 cm⁻¹ demonstrating interactive vibrational behavior [36,53,54,55].

The mechanical testing results displayed in Figure 6c demonstrated that PLA/CNT scaffolds showed a superior apparent compressive modulus of 0.58 ± 0.36 GPa and a compressive strength of 9.76 ± 6.28 MPa for scaffolds with an inner spacing of 4mm. These values fall within the range of trabecular bone properties, making them suitable for bone tissue engineering applications [36,56]. Cell viability assays using MC3T3 murine calvaria-derived preosteoblast cells demonstrated biocompatibility, with greater cell proliferation observed at 72 h among the scaffolds. These results position the PLA/CNT/bioglass scaffolds as promising biomaterials for bone repair applications in the vertebrae, ribs, skull, and joint bones [36,37].

The inverted immersion technique employed for these scaffolds was also investigated by Santos et al. [35] for thermal sensor applications. However, initial tests revealed that inverted immersion alone concentrated CNTs primarily in the interior structure, which was insufficient for surface-dependent thermal sensing phenomena. This limitation motivated the development of the direct immersion approach detailed in the next subsection.

3.3. PLA/CNT Thermoresistive and Thermoelectric Device

Direct immersion is a post-printing functionalization technique that involves submerging the entire 3D-printed structure into a CNT-containing solution after printing, creating functional surface layers essential for sensing applications [35]. The PLA/CNT-based ink is prepared by dissolving 200 mg of PLA filament in 5 mL of acetone at 60 °C, followed by ultrasonication at 40 kHz for 360 s. Separately, 7 mg of carboxylic acid-functionalized MWCNTs is dispersed in 1.5 mL of acetone under identical ultrasonication conditions. These solutions are mixed and homogenized by ultrasonication, resulting in a PLA/CNT-based ink with a concentration of approximately 31.85 mg/mL [35].

Santos et al. [35] developed a dual-coating protocol combining inverted immersion (pre-printing) with direct immersion (post-printing) for metal-free thermal sensors. First, a commercial PLA filament (1.75 mm diameter) was coated via inverted immersion with PLA/CNT-based ink using a funnel attached to one end of the filament, with two coating layers applied with 2 h intermediate drying and 5 h final drying. After 3D printing serpentine-shaped sensors on nanostructured filament via FFF (Stella 3 Lite printer, 0.2 mm nozzle, 215 °C first layer, 185 °C remaining layers, 100% infill), an additional surface coating was applied via direct immersion in the PLA/CNT-based ink, as illustrated in Figure 7. The sensors were dried for 5 h at room temperature to ensure complete solvent evaporation and optimal CNT–PLA adhesion [35].

SEM characterization, as shown in Figure 8a, reveals CNTs distributed throughout the trabecular microstructure of the PLA polymeric matrix, demonstrating successful incorporation and excellent adhesion of the nanotubes. Cross-sectional analysis shows a PLA/CNT coating with an average thickness of 130 ± 8.78 $\mathsf{\mu }$m surrounding the 3D-printed core, with a trabecular morphology that favors an increase in the contact area between the nanotubes and the PLA, leading to electrical improvements in the nanocomposite [35].

The Raman spectroscopy results displayed in Figure 8b confirm structural integration of CNTs into the PLA matrix. The spectrum exhibits characteristic vibrational modes of both PLA (C-CH₃, C-O-C, CH₃, C=O stretching) and CNTs (D, G, and 2D bands). The interaction between PLA and CNTs suppressed certain vibrational modes, including the C-CH₃ and C-O-C stretching modes. Blueshifts of 9 cm⁻¹ in $G_{outer}$ and 8 cm⁻¹ in 2D subbands indicate compressive deformations in the outermost walls of the MWCNTs, caused by the extrusion of the nanostructured filament during printing and the intertwining of polymer chains with MWCNT bundles, inducing p-type doping from CNT–polymer interaction [35].

The thermoresistive performance in Figure 8c demonstrates negative temperature coefficient (NTC) behavior, with electrical resistance decreasing upon heating due to quantum hopping and tunneling mechanisms activated by thermal energy. Across the 33 °C to 45 °C range, the maximum response was −4.61 ± 0.32%, with response and recovery times of around 600 s and 560 s, respectively. The temperature coefficient of resistance (TCR) was −0.39 ± 0.01%$·$°C⁻¹, remarkably close to commercial Pt100 thermistors (0.3851%$·$°C⁻¹) [35,57]. Additionally, the sensors demonstrated thermoelectric properties via the Seebeck effect, generating up to 400 $\mathsf{\mu }$V from a temperature gradient of 12.6 K (thermoelectric power of 31.75 $\mathsf{\mu }$V/K). Positive thermoelectric power values suggest p-type charge carriers, consistent with the doping effects observed in Raman spectroscopy [35].

The dual-coating strategy addresses single-method limitations: inverted immersion alone concentrates CNTs internally, while direct immersion alone functionalizes only external surfaces. Together, these techniques enable production of nanocomposite structures with tailored properties throughout the entire volume and surface, positioning the 3D-printed PLA/CNT sensors as sustainable alternatives to conventional Pt100 thermistors and thermocouples for body temperature monitoring and thermoelectric energy harvesting applications [35].

4. Metallic Matrix Devices

The nanostructuring of metallic matrices with CNTs enables the development of materials with enhanced electrical conductivity, improved mechanical strength, and superior wear resistance [58,59]. Unlike polymeric and cellulosic matrices, where CNT incorporation produces sensing devices, metallic matrix nanocomposites focus on enhancing intrinsic material properties for applications in electrical conductors, protective coatings, and structural components exposed to erosive environments. Electrodeposition and arc welding represent effective techniques for producing CNT-reinforced metallic coatings and welds, offering cost-effectiveness, experimental simplicity, and scalability for large-scale production [39,41].

4.1. Cu-CNT Coatings

Electrodeposition is an electrochemical process for the controlled formation of metallic films on conductive surfaces. Both direct current (DC) and pulsed current electrodeposition have been employed for Cu-CNT coating fabrication. Pulsed electrodeposition provides significant advantages in controlling coating morphology and density through the adjustment of pulse width and frequency, resulting in smoother, denser, and nodule-free films [39]. The process involves substrate preparation through anodization, preparation of the electrolytic solution containing copper and dispersed CNTs, and the electrodeposition step itself.

For substrate preparation, aluminum sheets undergo cleaning and anodization to create a porous alumina layer that improves adhesion and coating structure. The anodization process uses H₂SO₄ with copper plates as counter-electrodes and the aluminum substrate as the working electrode, applying 10 V and 3 A for 120 min to create a porous Al₂O₃ layer with average pore diameters of 12.6 $\mathsf{\mu }$m [39].

The electrolytic solution is prepared by mixing pentahydrated copper sulfate (CuSO₄·5H₂O) in demineralized water, separately dispersing functionalized MWCNTs using dimethylformamide (DMF) and isopropyl alcohol via ultrasonication, and then combining the solutions [39]. The electrodeposition employs pulsed current with an 80% duty cycle and a current density of approximately 83.3 mA/cm², with the anodized substrate as the cathode, as illustrated in Figure 9.

Process parameter optimization is critical for achieving uniform coatings. Key parameters include anodization time, electrode-to-sample distance, MWCNT concentration in the electrolytic solution, and electrodeposition duration. Silva et al. [39] demonstrated that increasing the anodization time to 120 min and the MWCNT concentration to 1.0 mg/mL, with an electrode-to-sample distance of 1.5 cm and an electrodeposition time of 120 min, resulted in optimal coating growth and uniformity.

FEG-SEM characterization, as shown in Figure 10a, reveals the evolution of coating morphology with processing optimization. Initial conditions (stage 1) showed non-uniform Cu-MWCNT dispersion, with cluster formation attributed to agglomeration during electrodeposition. With optimized parameters (stage 3), micrographs consistently revealed uniform coating on the anodized substrate surface with robust MWCNT concentration in the pores [39]. EDS elemental mapping, displayed in Figure 10b, confirmed compositional transition from Cu-dominated to CNT-dominated surfaces, demonstrating effective nanocomposite incorporation [39].

The Raman spectroscopy data in Figure 10c provide insights into structural changes and doping effects. The D band exhibited satellite sub-bands (${\mathrm{D}}_l$, ${\mathrm{D}}_r$, ${\mathrm{D}}_{LO}$, ${\mathrm{D}}_{middle}$) associated with structural changes from functionalization and ultrasonication. The G band split into ${\mathrm{G}}_{outer}$ and ${\mathrm{G}}_{inner}$ sub-bands related to outer and inner tube diameters, along with a ${\mathrm{G}}_{BWF}$-like peak indicating plasmon-phonon coupling. Focusing on the outermost tube component, a blueshift of approximately 16 cm⁻¹ in ${\mathrm{G}}_{outer}^{\prime }$ for optimized samples compared to as-received MWCNTs demonstrates p-type doping induced by the pulsed electrodeposition process [39]. The Amorphous Carbon Degree (ACD) for as-received MWCNTs was approximately 12.97%, and processing maintained or improved nanotube crystallinity [39].

Surface electrical conductivity measurements using the four-point Kelvin probe technique (ASTM D257-07) demonstrated electronic transport enhancement. The control sample exhibited a surface electrical conductivity 337.62 S, while the optimized sample achieved 514.30 S, representing a 52.33% increase attributed to enhanced electron transport paths created by well-distributed Cu-MWCNT nanocomposites [39]. The pulsed electrodeposition technique creates efficient pathways for electron transport through uniform nanocomposite coverage and strong interfacial bonding.

As an additional application, the electrodeposition technique has been extended to nanostructured cable fabrication. Rodrigues et al. [40] coated aluminum alloy 1350 wires using direct current electrodeposition with an electrophoretic solution containing copper sulfate and iodine-doped functionalized CNTs. The Cu/f-CNT coating was electrodeposited homogeneously on the aluminum substrate at 1.2 A, with copper films cultivated in (111) and (220) directions, favoring thin film formation with good crystallinity. Raman spectroscopy confirmed n-type doping from iodine, evidenced by redshifts in ${\mathrm{G}}_{inner}$ and ${\mathrm{G}}_{outer}$ sub-bands. The nanostructured wire achieved 72.96% IACS electrical conductivity, approximately 18% higher than conventional aluminum wire, with conductivity reduction of only 6.23% when heated from 25 °C to 50 °C [40]. These results demonstrate the potential of Cu-CNT-coated conductors for power transmission applications in extreme climate environments.

4.2. Stainless Steel-CNT Welding

Coatings based on metallic matrix nanocomposites reinforced with CNTs can balance hardness and ductility while improving the wear resistance of industrial equipment. Arc welding using nanostructured consumables enables the production of CNT-reinforced coatings with enhanced mechanical properties and cavitation erosion resistance, suitable for demanding applications such as hydroelectric turbine repairs [41,60,61].

Several approaches have been developed for preparing nanostructured wires for Gas Tungsten Arc Welding (GTAW). Reis et al. [60] employed electrodeposition to coat 316L stainless steel wire with Ni/CNTs using an electrolytic bath containing CNTs (2.0 g/L), cetyltrimethylammonium bromide (2.0 g/L), nickel chloride (0.2 mol/L), nickel sulfate (1.0 mol/L), and boric acid (0.5 mol/L), applying 1.0 A/dm² for 2 h to produce 1.0 wt.% Ni/CNT coating. This wire was used as filler metal on C-Mn steel via pulsed GTAW at 120 A with argon shielding at 10 L/min. Pinheiro et al. [61] dispersed CNT/Ti6Al4V/CaF₂ powders (mass ratio 1:40:4) at 40 kHz in isopropanol/H₂O₂ (40 mL:17 mL) for 20 min, dried at 140 °C for 3.5 h, packed the mixture into U-shaped low-carbon steel tapes, closed them tubularly, and drew them to 2.4 mm diameter wire. This wire was deposited on A131 grade D steel by GTAW at 160 A with argon at 12 L/min in a single pass at 5 cm/min.

Loayza et al. [41] developed a nanostructured flux-cored approach specifically for stainless steel applications. MWCNTs (5 wt.%) were ultrasonicated (55 kHz, 120 W) with 304L stainless steel powder (6.0 g) in 40 mL isopropanol for 15 min at 10 °C. Subsequently, 17 mL of 30% v/v H₂O₂ was added and homogenized for 25 min to perform chemical treatment that mitigates van der Waals forces and improves CNT dispersion. The mixture was dried at 130 °C for 1 h, and the CNT/304L nanocomposite was used as flux in a tubular wire with 304L steel sheath. This wire was deposited on 304L plates by pulsed GTAW (peak current 130 A, base current 80 A, peak time 1.5 s, base time 1 s, welding velocity 20 cm/min) with two-layer cladding and 30% pass overlap, as illustrated in Figure 11.

Comprehensive characterization of the stainless steel-CNT composite produced by Loayza et al. [41] demonstrates the effects of CNT incorporation on the microstructure and properties. SEM micrographs, as shown in Figure 12a, showed an improved distribution of chemically treated MWCNTs (black regions) into 304L SS particles (white regions) compared to as-received CNTs, attributed to the mitigation of van der Waals forces by H₂O₂ treatment. TEM analysis revealed that chemically treated nanotubes exhibited fewer defects, less wall exfoliation, and reduced deformations, evidencing good crystallinity of both outer and inner walls. The weld metal microstructure showed CNT segregation around dendrite contours and grain boundaries, with the composite exhibiting cellular-dendritic austenite morphology. The EDS elemental mapping in Figure 12b revealed enrichment of C, Cr, and Mn in interdendritic spaces, with M₇C₃ carbides precipitating around solidification structure boundaries, while M₂₃C₆ carbides formed in the liquid phase with a spherical morphology [41].

EBSD analysis provided detailed crystallographic information. The composite exhibited 64% grain refinement compared to conventional 304L SS (average grain size 98 ± 64 $\mathsf{\mu }$m versus 272 ± 173 $\mathsf{\mu }$m), attributed to CNTs acting as barriers inhibiting grain growth and providing favorable nucleation sites through the pinning effect [41]. CNT incorporation increased high-angle grain boundaries from 76.4% to 84.5%, with low-angle boundaries slightly increasing from 2.5% to 3.2%. The Kernel Average Misorientation (KAM) analysis showed increased elastic energy accumulation in the composite, particularly inside grains and at interdendritic region borders. The orientation distribution functions (ODF) revealed texture components characteristic of cold-rolled stainless steel, attributed to thermal anisotropy of CNTs aligning with preferential heat flow direction [41].

Raman spectroscopy of the nanostructured flux revealed significant structural changes from chemical treatment. The ${\mathrm{I}}_D$/${\mathrm{I}}_G$ ratio dropped from 1.12 to 0.68 after H₂O₂ treatment, indicating improved crystallinity. The ${\mathrm{D}}_{LO}$ sub-band disappeared in treated samples, showing exfoliation of defective outer walls. Analysis of ${\mathrm{I}}_{G_{outer}}$/${\mathrm{I}}_{G_{inner}}$ ratios showed values increasing from 0.33 to 1.06, indicating similar crystallinity between outer and inner walls after treatment. The Amorphous Carbon Degree (ACD) decreased dramatically from 76.5% to 17%, demonstrating effective purification through the removal of amorphous carbon and defective layers [41].

The mechanical testing results displayed in Figure 12c demonstrate substantial property improvements. Microhardness increased by 45% in the nanostructured composite (305 ± 15 HV_0.3) compared to conventional 304L SS (210 ± 12 HV_0.3), approaching values of commercial alloys like Stellite 21^®. Vibratory cavitation testing (ASTM G32-16) revealed 64% reduction in average erosion rate (from 1.49 ± 0.13 mg/h to 0.53 ± 0.07 mg/h) and 69% reduction in maximum erosion rate (from 2.23 ± 0.15 mg/h to 0.67 ± 0.08 mg/h), with cumulative mass loss reduced by 67% after 20 h. This cavitation resistance is comparable to commercial alloys such as Stellite 21^®, Cavitec^®, and Cavitalloy used for hydroelectric turbine repair. XRD analysis after cavitation testing showed preferential growth of the austenitic $\gamma$(111) phase, suggesting microstructural accommodation induced by cavitation erosion [41].

Economic viability analysis demonstrated that the fabrication of a 15 kg standard flux-cored coil with 1.5 wt.% MWCNTs costs approximately USD 964 (USD 64.3/kg) in laboratory conditions, representing a 43% cost reduction compared to commercial Stellite 6 wire at USD 112.8/kg [41]. This cost advantage, combined with comparable mechanical performance and cavitation resistance, positions stainless steel-CNT composites produced by arc welding as a cost-effective alternative to commercial high-performance alloys for industrial applications such as hydroelectric turbine repairs.

5. Conclusions and Future Perspectives

This review presented low-cost and scalable nanomanufacturing processes to obtain CNT-based devices, covering the complete cycle from nanocomposite production to functional device assembly across cellulosic, polymeric, and metallic matrix systems. For cellulosic matrices, vacuum filtration enabled the production of buckypapers that were subsequently assembled into chemiresistive sensors through fixation on solid substrates with copper terminals and conductive silver ink contacts, demonstrating excellent performance for detecting adulterants in beverages and industrial reagents such as ether-amines. For polymeric matrices, additive manufacturing via fused filament fabrication combined with surface functionalization techniques (spray coating, inverted immersion, and direct immersion) enabled the production of nanostructured devices with tailored properties, including flexible piezoresistive pressure sensors with potential for robotic haptic feedback systems and metal-free thermal sensors with temperature coefficient of resistance comparable to commercial Pt100 thermistors. For metallic matrices, pulsed electrodeposition of Cu-CNT coatings on aluminum substrates achieved an electrical conductivity of 72.96% IACS for power transmission applications, while arc welding using nanostructured flux-cored wires produced stainless steel composites with 64% grain refinement and enhanced cavitation resistance suitable for hydroelectric turbine repairs. These results demonstrate that the integration of CNTs into diverse matrices successfully bridges nanoscale properties to macroscopic functional devices through accessible nanomanufacturing routes.

Despite the promising results, several challenges remain for the widespread adoption of CNT-based nanomanufactured devices. Van der Waals interactions between nanotubes promote agglomeration, requiring functionalization strategies and optimized dispersion protocols to achieve homogeneous dispersion within the matrix materials. Chemical functionalization and processing conditions can introduce defects in the CNT structure, affecting their intrinsic properties. Balancing functionalization for improved dispersion while preserving nanotube crystallinity is both a challenge and an opportunity, as defects also could help tailor CNT properties. The matrix preparation is also important, such as common solvents for CNTs and polymers, as well as surface chemical treatment of metals, can improve the production process. Transitioning from laboratory-scale production to industrial-scale manufacturing while maintaining consistent quality and cost-effectiveness requires optimization of process parameters and development of automated production systems. Variations in CNT sources, functionalization degrees, and processing conditions can lead to inconsistent device performance. Establishing standardized protocols and quality control metrics is necessary for commercial applications. The operational stability of CNT-based devices under environmental conditions (humidity, temperature variations, mechanical stress) and during extended use periods requires further investigation and development of protective strategies.

By addressing these challenges, future developments in CNT-based nanomanufacturing encompass diverse applications, such as:

Lab-on-chip systems, electronic tongues, and electronic noses: The integration of multiple CNT-based sensors into compact platforms enables the manufacturing of portable devices for rapid detection and identification of chemical and biological substances, with applications in food quality control, environmental monitoring, and medical diagnostics.

Wearable devices based on flexible and biodegradable materials: The combination of CNTs with biocompatible polymers such as PLA and cellulose enables the development of flexible, sustainable wearable sensors for continuous health monitoring, including body temperature, pressure, and biochemical parameters.

Nanomanufacturing of 3D-printed parts with built-in sensing capabilities: Additive manufacturing combined with CNT functionalization techniques allows the fabrication of structural components with integrated sensing functions, eliminating the need for external sensor attachment and enabling smart structures for aerospace, automotive, and robotics applications.

Internet of Things integration for real-time monitoring: CNT-based sensors integrated with IoT platforms enable distributed sensor networks for real-time data acquisition and transmission, with applications in industrial process control, environmental monitoring, and personalized healthcare systems.

Machine learning and deep learning for sensor data processing: Advanced computational algorithms can enhance pattern recognition, analyte identification, and sensor calibration from CNT-based sensor arrays, improving discrimination capacity beyond conventional statistical methods such as PCA and enabling autonomous decision-making systems.