Introduction

Recent scientific progress in the use of electronic devices for strain sensing applications has attracted significant attention in the fields of soft robotics (Fallahi et al. 2017; Zheng et al. 2015), electronic skins (Ho et al. 2016; Hong et al. 2016; Martinez et al. 2013), and wearable devices for human health monitoring (Amjadi et al. 2014; Choi et al. 2017; Trung et al. 2017). These applications require flexible strain sensors with high sensitivity and durable performance. In general, strain sensing range and sensitivity are two key parameters used to evaluate the quality of these sensors. Particularly, the sensitivity is quantified in the form of gauge factor (GF), and is defined as the ratio of relative electrical resistance change to applied strain. Conventional metallic strain gauges offer a low GF (around 2) and a limited sensing range (< 5% strain), while semiconductor-based sensors offer high GFs but a limited detection range, making neither of them suitable for wearable sensing applications (Zhang et al. 2017).

Elastomeric nanocomposites enhanced by conductive nanofillers are one of the most popular advanced materials for stretchable strain sensing applications. Polydimethylsiloxane (PDMS) (Chowdhury et al. 2018; Lu et al. 2012), polyurethane (Zhang et al. 2013), and natural rubber (Selvan et al. 2016) have been employed as elastomeric polymers in these nanocomposites. The addition of conductive nanofillers to these flexible matrices endows the base material with electrical properties which can be exploited to measure its deformation. Both capacitance-based and resistance-based strain sensing mechanisms have been reported in literature, though the electrical resistance-based method has been more commonly reported (Abshirini et al. 2018; Cao et al. 2015; Luo et al. 2018). The change in electrical resistance is believed to be induced by the reorganization of the electrically conductive network formed by nanofillers under mechanical stimulus (i.e., tension or compression). Researchers have explored a variety of nanofillers, including carbon black (Zheng et al. 2017), carbon nanofiber (Charara et al. 2019b; Chowdhury et al. 2016), graphene (Park et al. 2015), and carbon nanotubes (CNTs) (Ryu et al. 2015; Liu et al. 2012), to increase the conductivity of elastomers. Of these, multi-walled carbon nanotubes (MWNT) have been widely used to improve these polymers’ conductivity due to their excellent electrical characteristics at low concentration (Hu et al. 2008; Kang et al. 2006; Lipomi et al. 2011). Piezoresistive response in nanocomposite materials has been previously observed (Obitayo and Liu 2012). By measuring the electrical resistance change resulting from mechanical deformation in these materials, the applied strains can be calculated.

Numerous studies have been performed in recent years to produce a sensor with the capability of detecting strains up to 100%. However, in most of the cases, these sensors exhibit a non-linear sensing response over a wide strain range (Liu et al. 2015; Yamada et al. 2011) and a low GF at large strains (Cai et al. 2013; Cohen et al. 2012; Liu et al. 2015; Yamada et al. 2011). Recently, several approaches have been developed to fabricate sensors with high sensitivity and stretchability, such as micro/nanocracking-assisted conductive network resulting in sensors with high GFs but very low sensing range (Chen et al. 2016; Hoang et al. 2016; Kang et al. 2014) and the creation of porous nanocomposite materials (Li et al. 2017; Liu et al. 2016; Wu et al. 2017). However, the fabrication process to create such sensors is often highly complex and time consuming. Therefore, it is important to explore a novel strain sensor fabrication technique which allows for the facile manufacturing of highly stretchable, durable sensors with a wide sensing range.

Most reported nanocomposite strain sensors were manufactured by molding and casting (Amjadi et al. 2016; Obitayo and Liu 2012). Recently, various additive manufacturing methods have been used for fabricating conductive polymers. Fused deposition modeling, based on the melting of thermoplastic filaments to manufacture a desired 3D model, has been widely used for the printing of conductive thermoplastic polymers (Costa et al. 2014; Gnanasekaran et al. 2017; Kumar et al. 2018; Kwok et al. 2017; Leigh et al. 2012; Wei et al. 2015; Yu et al. 2017). Other printing techniques have been used for manufacturing conductive polymers, including inkjet printing-based deposition of ink droplets (Michelis et al. 2015), aerosol jet printing (Agarwala et al. 2018; Goh et al. 2018a), and direct writing (Kim et al. 2012; Vatani et al. 2013). Recently, extrusion-based 3D printing technology has been reported as a rapidly expanding approach to fabricate customizable structures and prototyping of elastomers, such as PDMS, and related elastomeric nanocomposites (Charara et al. 2019a; Guo et al. 2017; Hwang et al. 2018; Truby et al. 2018). Due to its thixotropic flow properties, PDMS can be 3D printed under ambient conditions and used to create relatively thin multi-material devices, such as a bionic ear (Mannoor et al. 2013). Although the printed PDMS is able to maintain its geometric fidelity after curing, the low elastic modulus of the pre- and post-cured PDMS and its deformation under gravity prior to curing still restricts the 3D geometries that can be fabricated due to its inability to maintain shape fidelity directly post deposition. Hinton et al. reported a 3D printing method for PDMS in a hydrophilic support bath, aiming to enable true freeform fabrication of complex structures (Hinton et al. 2016). Dispersion of nanoparticles in PDMS can significantly improve materials’ elastic modulus and viscosity, leading to 3D printable and highly elastic nanocomposites. Carbon-based nanoparticles, including graphene and CNTs, have been dispersed in elastomers and 3D printed for electronic and biomedical applications (Jakus et al. 2015; Wang et al. 2017). Most previous work has focused on nanocomposite characteristics, 3D printability, and potential applications. However, there is a lack of studies exploring the effects of 3D printing and curing parameters on the nanocomposite’s properties and sensing response due to the various nano- and micro-structures formed within the material during the manufacturing process (Goh et al. 2018b).

In this study, an electrically conductive nanocomposite is 3D printed on a pristine elastomer substrate to fabricate highly stretchable and wearable strain sensors. To obtain optimal piezoresistive sensing function, nanocomposites are printed varying processing parameters, including needle diameter, curing temperature, and MWNT concentrations. All the printed materials are tested under cyclic tensile load to identify their sensitivity and linear sensing response range. The optimized sensors are further tested to validate their reliability and repeatability under various cyclic loads and loading rates. The piezoresistive sensing mechanism was demonstrated by comparing the MWNT conductive network using in situ micro-mechanical testing in a scanning electron microscope (SEM). Finally, the fabricated sensors are used as wearable sensors to detect the motions on a human hand and fingers.

Experimental

Materials

Unless otherwise stated, all the materials in this study were used as received. SYLGARD 184 PDMS base elastomer (part A) and curing agent (part B) were purchased from Dow Corning. MWNTs with an average aspect ratio of 100 and diameter of 50–90 nm were purchased from Sigma-Aldrich. Tetrahydrofuran (THF) purchased from Sigma-Aldrich was used as the solvent.

Preparation and 3D printing of nanocomposites

Pristine PDMS sheets were first manufactured using an aluminum mold. The PDMS prepolymer was prepared by mixing the base elastomer and curing agent at the 10:1 part A to part B weight ratio, as recommended by the manufacturer. The mixed prepolymer was degassed in a vacuum desiccator at room temperature for 30 min and cured in an oven at 65 °C for 4 h. The manufactured pristine PDMS sheets were used as substrates during the 3D printing process.

The electrically conductive nanocomposites were prepared by dispersing MWNT in the PDMS base elastomer using the solvent-assisted ultrasonication method. MWNT and PDMS base elastomer were added into 30 ml THF and mixed for 5 min using a magnetic stirrer at 350 RPM. After mixing, a 750 watt probe sonicator was used to sonicate the PDMS base elastomer/THF/MWNT solution for 30 min. The mixture was kept on a hot plate at 65 °C and stirred at 100 RPM overnight to evaporate all the THF solvent. Finally, the PDMS curing agent was added to the prepared nanocomposites and mixed for 3 min before loading into a 3 mL syringe and 3D printing on the pristine PDMS substrate. The schematic of the material preparation process is shown in Fig. 1a. Nanocomposites with various MWNT concentrations were printed and characterized to identify the optimal nanocomposite formulation with the highest piezoresistive sensing function.

Fig. 1
figure 1

Schematics of the a nanocomposite material synthesis and b strain sensor fabrication processes

Full size image

A Tronxy X5S system was modified to 3D print the conductive nanocomposites by installing a geared syringe pump (Abshirini et al. 2018). Plastic syringes capped with needles of various diameters were used to print the nanocomposites on pristine PMDS substrates. The prepared MWNT/PDMS nanocomposites were printed into an electrically conductive pattern of two connected 30 mm long parallel lines. The designed 3D model was converted into G-code using the Repetier and Slic3r open source software. The extrusion pressure is adjusted based on the volumetric flow rate required depending on the filament diameter (in this case the syringe diameter) and the nozzle diameter (in this case the needle diameter). For this work, printhead speed of 1 mm/s is used to deposit the nanocomposite. The printed conductive pattern was cured at various temperatures (25 °C, 65 °C, and 130 °C) in an oven to explore the curing temperature effects on the obtained piezoresistive sensing function. A layer of pristine PDMS was coated on the printed nanocomposites and cured at 130 °C for 20 min before mechanical testing. The schematic of the strain sensor fabrication process is illustrated in Fig. 1b. The modified 3D printer, the printing process, and the final shape of the fabricated strain sensor are shown in Fig. 2c. The dimensions of the final sensors are 50 × 30 × 1.8 mm. The PDMS substrate and sealing layer are both 0.9 mm.

Fig. 2
figure 2

a Modified 3D printer, b depositing nanocomposite on the PDMS substrate, c final shape of the printed strain sensor on pristine PDMS substrate (ruler for scale shown in millimeter and centimeter)

Full size image

Piezoresistive characterization of the nanocomposite

Three parameters, including MWNT concentration, needle diameters, and curing temperature were optimized to obtain the highest piezoresistive sensing function of the developed strain sensors. Cyclic tensile tests were performed for each set of the fabricated strain sensors using an Instron single column uniaxial test machine at maximum strains of 10, 15, 20, 25, and 30%. The electrical resistance of the tested samples was continuously recorded using an Agilent 34401A multimeter during the mechanical tests for all samples. The experimental setup is shown in Fig. 3. The load/unload procedure was repeated for six cycles to ensure the repeatability of the sensing response in each test. The normalized electrical resistance change was calculated to quantify the sensing behavior of the 3D-printed sensors using Eq. 1.

Fig. 3
figure 3

Experimental setup of piezoresistive sensing tests under cyclic tensile load

Full size image
$${\text{Normalized resistance change}} = 100 \times \frac{{R - R_{\text{o}} }}{{R_{\text{o}} }}(\% )$$
(1)

where Ro is the initial electrical resistance of the sensor and R is the real-time measured electrical resistance. In addition, GF was calculated to quantify the piezoresistive sensitivity of the 3D-printed sensors using Eq. 2.

$${\text{GF}} = \frac{{\left( {\frac{{R - R_{0} }}{{R_{0} }}} \right)}}{\varepsilon }$$
(2)

where ε (mm/mm) is the applied strain on the sample.

The effect of the MWNT concentration on the piezoresistive behavior of the printed sensors was evaluated by testing nanocomposites with four different MWNT concentrations: 1.5 wt%, 2 wt%, 2.5 wt%, and 3 wt%. All nanocomposite sensors were printed using a needle of 0.41 mm diameter and cured at 130 °C. The optimal nanocomposite with the highest piezoresistive sensitivity was identified using the cyclic mechanical tests discussed above. The nanocomposite with the selected CNT content was used to print sensors using four needle diameters of 0.26, 0.41, 0.71 and 1.19 mm, cured at 130 °C, and tested under the same cyclic load conditions as discussed above to explore the effect of needle diameter on the properties of the printed nanocomposite sensors. Finally, the nanocomposites printed with the optimal formulation and needle size were cured at three different temperatures of 25 °C, 65 °C, and 130 °C to investigate the effect of curing temperature on their piezoresistive sensing function.

Nanocomposite sensors printed using all the optimized parameters and formulations were tested to evaluate their sensing reliability and repeatability. Uniaxial tensile loads at the rate of 1 mm/min were applied to identify the maximum strain that the sample carried before fracture. In addition, cyclic tensile loads were applied to find the maximum strain range in which the sensor showed a linear piezoresistive response. Moreover, the load rate dependency of the printed sensors was investigated under cyclic tensile load at loading rates varying from 5 to 200 mm/min, at a constant 10% maximum strain. The piezoresistive sensing response was recorded throughout all the experiments. The long-term sensing and fatigue behavior of the printed sensors was tested under tensile loads at a loading rate of 3 mm/min and 10% max strain for 400 cycles. Finally, the stress and electrical resistance relaxation behaviors of the sensors were investigated by holding the sample at 10% strain load for 8000 s while recording electrical resistance and stress data.

In situ characterization of the piezoresistive sensing mechanism

Understanding the piezoresistive sensing mechanism is critical for the optimization of 3D printed sensor behavior. In this paper, in situ micro-mechanical tensile tests were conducted to investigate the reorganization of the electrically conductive MWNT network in the nanocomposites under tensile load. A printed nanocomposite sensor was manufactured without a pristine PDMS sealing layer, clamped on a micro-mechanical testing stage (Gatan Microtest 200), and tested in an SEM (TESCAN VEGA II), while SEM images were captured at different strain increments, as shown in Fig. 4. The reorganization of the MWNT network in the nanocomposite was demonstrated by comparing the SEM images at 0% and 20% strain.

Fig. 4
figure 4

In situ SEM analysis of the strain sensor using a micro-mechanical tensile stage

Full size image

Human body motion recognition

The feasibility of using the developed strain sensors as wearable sensors was explored. Five strain sensors were printed and mounted on a glove to detect finger motions, and the sixth sensor was attached on the wrist area to detect wrist bending. Electrical resistance data were recorded throughout the test to validate the sensors’ performance as wearable electronic devices.

Results and discussion

Distribution of the nanocomposite pattern in the sensor

Figure 5a shows the SEM images of the cross-section of the strain sensor, detailing the MWNT distribution within the printed nanocomposite. The half oval shape with a width of 598 μm and a height of 228 μm was the conductive nanocomposite embedded between two layers of pristine PDMS. Due to surface wetting, the cured cross-section of the 3D printed nanocomposite sensor was in the shape of half oval instead of half circle, as outlined in Fig. 5a. Good dispersion of the MWNT in the PDMS matrix can be observed in the high-magnification SEM image in Fig. 5b. No MWNT agglomerates or fractured nanotubes were observed during SEM imaging. The width and thickness of the deposited pattern using different needles were measured from the SEM images captured from cross-sections and tabulated in Table 1.

Fig. 5
figure 5

SEM images of the nanocomposite: a printed conductive nanocomposite on pristine PDMS substrate, b distribution of MWNT in the PDMS matrix

Full size image
Table 1 Width and thickness of the deposited pattern after curing
Full size table

Optimization of MWNT concentration in nanocomposites

To study the effect of MWNT concentration on the piezoresistive sensing function of the conductive nanocomposite, sensors fabricated with four different MWNT formulations were tested via cyclic loading at max strains between 10 and 30% at a loading rate of 2 mm/min. The electrical resistance changes of nanocomposites with various MWNT concentrations are shown in Fig. 6. By reducing the MWNT loading, the piezoresistive sensitivity of the printed nanocomposites improved, as indicated by the increased slopes at lower MWNT contents in Fig. 6. The average GFs of nanocomposites with 3 wt%, 2.5 wt%, and 2 wt% MWNT were 1.15, 1.56, and 2.51, respectively. However, the average GF of nanocomposite with 1.5 wt% increased up to 12.98. Nanocomposites with 1 wt% MWNT were also printed and tested following the same testing procedure. However, their poor electrical conductivity did not offer a measurable piezoresistive sensing function. Since the piezoresistive sensing mechanism is due to the reorganization of the conductive network of MWNTs under external load, an appropriate amount of MWNTs can form the most effective micro-scale connections among MWNTs, resulting in the nanocomposites being most sensitive to external loads. Therefore, nanocomposites with 1.5 wt% MWNT are considered the optimal material for sensor development in this paper.

Fig. 6
figure 6

The effect of CNT concentration on the strain sensing response

Full size image

Optimization of nanocomposite printing parameters

The effect of the printing needle diameter was investigated by printing nanocomposites with 1.5 wt% of MWNT using four different needle diameters. As shown in Fig. 7a, the increased needle diameters resulted in wider nanocomposite patterns. The piezoresistive sensing function of these sensors is expected to be different, since the MWNT distributions and alignment in the nanocomposites can be affected by the shear flow generated during the printing process. The measured piezoresistive sensing functions of the four types of sensors tested under 10–30% strain are shown in Fig. 7b. It should be noted that the piezoresistive sensitivity was enhanced by reducing the needle diameter from 1.19 mm to 0.41 mm. The average GF is 13.01 for the sensors fabricated with needles of 0.41 mm diameter compared to 1.92 for the sensors printed using 1.19 mm diameter needles.

Fig. 7
figure 7

a Printed sensors of various widths due to the effect of needle diameter, b piezoresistive sensing function of sensors printed by various needle diameter

Full size image

The effect of needle diameter on the piezoresistive sensing function can be due to the alignment of the MWNT and formation of the conductive network within the printed nanocomposites. During the printing process, the flow rate of nanocomposites near the needle wall was lower than that near the center of the needle, resulting in the shear stress within the nanocomposite caused by the extrusion. The shear stress at a surface element parallel to the needle wall can be written as:

$$\tau \left( y \right) = \mu \frac{\partial u}{\partial y}$$
(3)

where μ is the dynamic viscosity of the flow, u is the flow velocity along the needle wall, and y is the distance away from the needle wall. Shear stress caused by shear flow can align MWNT nanoparticles along the printing direction, as shown in Fig. 8. Since the same amount of MWNT nanocomposites was printed due to the constant extrusion motor speed, the generated flow rate and shear stress using smaller needles were higher than those using larger needles, resulting in better alignment of MWNT within nanocomposites. Similar phenomena have been reported in literature when micronozzles were used to deposit nanofillers in nanocomposites (Compton and Lewis 2014; Farkash and Brandon 1994; Shofner et al. 2003). However, if the needle diameter is too small, the printed nanocomposite sensors may have discontinuous MWNT alignment and MWNT breakage during printing, resulting in reduced piezoresistive sensing function. The poor sensing behavior can be more significant under high tensile strains. This is evident by the non-linear response of the 0.26 mm needle size after 20% strain, which is likely due to poor and disrupted MWNT network in this sensor under high extension condition.

Fig. 8
figure 8

Schematic of nanotube alignment during the printing and realignment during the loading

Full size image

The curing temperature effect on the piezoresistive sensing function was characterized by printing sensors using the 1.5 wt% MWNT nanocomposite and a 0.41 mm needle, and immediately curing the printed nanocomposites at three different temperatures of 25 °C, 65 °C, and 130 °C. Figure 9 shows the piezoresistive response of these samples tested under 10–30% of tensile strain. The piezoresistive sensitivity of the printed sensors was reduced when low curing temperature was used. The average GFs of nanocomposites cured at 25 °C, 65 °C, and 130 °C were 1.8, 7.98, and 12.91, respectively, when tested under 10–30% of tensile strain. It should be noted that it only took 10 min to fully cure the printed nanocomposite at 130 °C, compared to 96 h when cured at 25 °C. Hence, it can be estimated that the alignment of the MWNTs was disrupted during the long curing process when a low curing temperature was employed. To take advantage of CNT alignment generated during the printing process, the microstructure of the conductive MWNT network should be “locked” quickly by curing the nanocomposites as quickly as possible. The ideal curing process should be the in situ curing of nanocomposites during the printing. Otherwise, the initial orientation of the MWNTs in the nanocomposites can change and reduce the piezoresistive sensing function of the printed nanocomposites.

Fig. 9
figure 9

Effect of curing temperature on the piezoresistive sensing function

Full size image

According to the obtained results, the best piezoresistive sensing function can be achieved using the 1.5 wt% MWNT printed on the pristine PDMS substrate using 0.41 mm needle and cured at 130 °C. The optimal nanocomposite formulation and manufacturing parameters were used to fabricate all the stretchable strain sensors that were characterized in the following sections.

Characterization of piezoresistive sensing function

To evaluate the maximum strain that the sensors carried prior to fracture, continuous uniaxial tensile load was applied at the load rate of 0.2 mm/min. Figure 10a shows the sensor mounted on the Instron uniaxial test machine before applying the load and the sensor at the moment before failure. The maximum strain the sensors carried was 146% tensile strain, which was close to the reported maximum strain that PDMS can sustain (elongation, 160%) (Cai et al. 2013). This high fracture strain allows the potential application of the developed sensor under high strain conditions.

Fig. 10
figure 10

a Printed sensor under 0% and 146% tensile strains; b Piezoresistive sensing response of the flexible strain sensor from 10 to 70% tensile strain; c effect of loading rate on piezoresistive sensing behavior and details of the relative piezoresistive change under loading rate of 5 mm/min, 25 mm/min, 50 mm/min; d durability test of printed sensor for 400 cycles; e relaxation test showing the resistance and stress change under 10% strain for 8000 s; f FE results showing the strain distribution in the layers of the sensor at 40% strain

Full size image

The piezoresistive response of the printed sensors at different max strains was characterized to identify the linear strain sensing range. Cyclic tensile load tests at various max strains were conducted starting from 10% strain, incrementing by 5% up to 70% strain. The piezoresistive response of the sensor was measured continuously and is shown in Fig. 10b. The electrical resistance of the nanocomposites above 70% strain was beyond the working range of the multimeter used in this study, therefore, the sensing performance above 70% tensile strain is not reported in this paper. The sensing results in Fig. 10b shows that the sensor exhibits a linear response throughout the tested 10–70% strain range, with an average GF of 12.15, which is about six times higher than that of conventional metallic strain gauges (Yamada et al. 2011). The wide linear sensing range (at least up to 70% tensile strain) with a high GF demonstrated the potential strain sensing capability under large deformations.

To evaluate the load rate dependency of the printed sensors, cyclic tensile tests at 10% max strain were conducted under various loading rates. As shown in Fig. 10c, the piezoresistive response initially improved with the increased loading rate, indicating enhanced piezoresistive sensitivity at a higher loading rate, but the sensing behavior stabilized after the loading rate reached 50 mm/min. This demonstrated that the viscoelastic behavior of the material was affected at loading rates of 50 mm/min or lower, with the effect disappearing at loading rates of 50–200 mm/min. The detailed cyclic piezoresistive response of the sensor for the tests at loading rates of 5, 25, and 50 mm/min is also shown in Fig. 10c.

An important parameter to verify the robustness of a strain sensor is its long-term sensing response. The piezoresistive behavior of the sensor under a tensile fatigue test of 400 cycles loaded up to 10% strain is shown in Fig. 10d. The cyclic resistance change degraded gradually in the first 100 cycles and then became relatively stable. The reduction occurred in both the peaks and valleys in the initial cycles. The peak to valley resistance response was illustrated separately for three different points throughout the test: cycles 20–25 (start), 200–205 (middle), and 380–385 (end). The relative resistance that changed in the cycles of 20–25 was around 5.1% higher than that of cycles of 200–205 and 380–384. This result shows the acceptable long-term performance of the printed sensors.

To understand the relaxation behavior of the strain sensor, a 10% strain tensile load was applied and held for 8000 s. The sensor’s stress and resistance were recorded continuously by the Instron test machine load cell and multimeter, respectively. The normalized stress and resistance change (defined as the change in the real-time value from the initial values, divided by the initial value) throughout the experiment are depicted in Fig. 10e. Both stress and resistance degraded due to the viscoelastic behavior of the elastomer polymer in constant extension. However, the stress reduction was around 8% compared to 3.4% for the resistance, which revealed that the relaxation behavior of the stress was higher than that of the piezoresistance response, likely due to additional relaxation in the pristine PDMS layers, and not the conductive pattern.

Finite element analysis (FEA) simulations and experimental validations were conducted to analyze the difference between the overall applied strain on the sensor and the localized tensile strain carried by the printed nanocomposite pattern embedded in the sensors. Detailed FEA models and experimental validations of the FEA model using a video extensometer were discussed in the Supporting Material. Numerical results showing the longitudinal strain distribution in the substrate, nanocomposite pattern, and the top PDMS layer at the 40% applied tensile strain are depicted in Fig. 10f. The FEA simulation and experimental results showed that the true tensile strain carried by the printed nanocomposite was 6% smaller than the average strain applied to the sensor during mechanical testing for nanocomposite sensors with 1.5 wt% MWNT. Although this smaller strain can increase the calculated GF of the strain sensor, the difference in applied strain and actual strain on the nanocomposite was so small that the correction factor was not included in the experimental analysis to simplify the data processing procedures.

In situ characterization of the piezoresistive sensing mechanism

In situ micro-mechanical testing under SEM was conducted to illustrate the piezoresistive sensing mechanism by demonstrating the reorganization and rearrangement of MWNT conductive network within the printed nanocomposite. The SEM image of a relaxed nanocomposite sample (0% strain) is shown in Fig. 11. Although only a low percentage of the MWNTs was on the surface exposed to the SEM imaging, the realignment of MWNT was visible in several zones. After applying a 20% tensile strain on the sample, the rotation and sliding of MWNT were observed, and is shown in the magnified SEM images (P1–P5 zone) in Fig. 11. The zoomed SEM images with solid borders show the initial orientation of the MWNT (at 0% strain), while the pictures with dashed border show the reorganization of MWNT network under 20% tensile strain. The reduced gap between two adjacent nanotubes can be seen in zones P1 and P3. Additionally, zones P2–P4 show two separated nanotubes connecting after applying the load. The cumulative effect of the rearrangement of the MWNTs in the conductive network resulted in the reduction of the resistance of the conductive nanocomposite by increasing the tensile load. This piezoresistive sensing function of the strain sensor can be observed in Fig. 10c, d.

Fig. 11
figure 11

SEM images of the working area in the in situ micro-mechanical testing under SEM: SEM images for five different zones (P1–P5) comparing the initial unstrained state (show in the boxes with solid borders) and final loaded state at 20% strain (show in boxes with dashed borders)

Full size image

Human body motion recognition

To validate the application of the 3D-printed sensor as wearable sensors, a “smart” glove was prepared for human hand motion detection. Five sensors were attached to each finger on the glove to detect the finger motion, and one sensor was attached to the wrist area to monitor the bending motion of the wrist, as shown in Fig. 12a. Two tests were performed to validate the sensors. First, the fingers and wrist were folded and unfolded repeatedly, while the piezoresistive sensing data were recorded simultaneously. Relative electrical resistance changes between 22 and 55% were obtained during the tests, as shown in Fig. 12b. In the second test, there were 10 s holding period after the bending and relaxation. The recorded piezoresistive sensing results shown in Fig. 12c were able to capture the bending and holding of each motion. In particular, the plateau area in each cycle represented the holding time in both the bending and relaxation periods of the motion. The peaks of the plots showed the relaxation and the valleys showed the bending motion. These experiments demonstrated the potential of using the 3D-printed sensors as wearable sensors, with potential applications in robotics and biomedical sensors.

Fig. 12
figure 12

a Smart glove demonstrating the six possible individual sensing modes (five for the fingers and one for the wrist); b sensing response of repeated bending of human hand joints; c sensing response of hold and bending of hand joints

Full size image

Conclusion

Piezoresistive strain sensors consisting of PDMS/MWNT nanocomposites were 3D printed on a PDMS substrate and characterized for potential wearable sensor applications. The detailed nanocomposite formulation and 3D printing parameters were first optimized to obtain the sensors with the highest piezoresistive sensitivity. The strain sensing capability of the sensors was evaluated by testing the printed sensors under cyclic tensile loads. The experimental results showed that the nanocomposites with an average GF of 13.01 in the 10–30% strain range could be manufactured using a PDMS/MWNT formulation with 1.5 wt% MWNTs, printed by 0.41 mm diameter needles, and cured at 130 °C. The 3D printing process was able to enhance the alignment of MWNTs in the nanocomposites, leading to improved sensor performance. The developed sensors with optimal properties carried up to 146% tensile strain before fracture, demonstrating their high stretchability and flexibility. Experimental results showed a linear sensing response, with an average GF of 12.15 in the 10–70% strain range. Additionally, the highly repeatable sensing response was observed in the long-term 400-cycle tensile test. The long-term relaxation test showed the sensors only suffered a modest 3.4% resistance degradation after 8000 s under a tension load, due to the viscoelastic properties of the PDMS polymer. The application of the strain sensors in wearable electronics was investigated by attaching six sensors on a glove to detect bending in the human fingers and the wrist. The results indicated that the fabricated sensors were able to monitor the motion of the human body joints and can be used in the wearable sensing applications.