Highly Conductive Polydimethylsiloxane/Carbon Nanofiber Composites for Flexible Sensor Applications

Transcription

1 FULL PAPER Flexible Sensors Highly Conductive Polydimethylsiloxane/Carbon Nanofiber Composites for Flexible Sensor Applications Shoieb A. Chowdhury, Mrinal C. Saha, Steven Patterson, Thomas Robison, and Yingtao Liu* A polydimethylsiloxane (PDMS)/carbon nanofiber (CNF) nanocomposite with piezoresistive sensing function is presented. Excellent electrical conductivity is achieved by dispersing the CNFs into PDMS. A facile, low cost, and scalable fabrication procedure allows the sensors to be made in different shapes. The piezoresistive sensors show repeatable response up to 30% tensile strain. In addition, the characterization of sensing mechanism using an in situ mechanical testing system within a scanning electron microscope reveals the reorganization of CNF network by varying fiber alignment and interfiber distance in the nanocomposites under tensile load. To validate the wearable sensing capability, nanocomposite straps are employed to monitor the finger motions under various bending speed and holding time. Two different shapes of compressive sensors, including cylinder and truncated cone, are tested in compression strain as low as 3%, with gauge factors of 18.3 and 6.3, respectively. In addition, the sensing capability is independent of the applied strain rates and is highly repeatable in 1000 cycles under compression. Finally, the developed nanocomposites are made into sensor arrays for pressure sensing ranging from 35 to 690 kpa. The versatility and ease of fabrication of the reported nanocomposites can bridge the current challenge between performance and applicability of flexible sensors. 1. Introduction Sensors with high flexibility are of immense scientific interests due to their applications in a wide range of areas, such as soft robotics, [1] artificial skins, [2] wearable devices, [3] and structural health monitoring. [4] Conventional metallic and semiconductor based sensors do not satisfy the requirement of such applications due to their brittle nature and lack of conformability in various shapes. [5] Commercially available metallic strain gauges only work up to 5% maximum strain with low sensitivity. [6] To satisfy these demands, new materials that are both flexible and responsive to external loads are needed. Depending on their sensing mechanisms, these materials can be classified differently, while piezoresistive [7] and capacitive [8 10] based sensors S. A. Chowdhury, Prof. M. C. Saha, Prof. Y. Liu School of Mechanical and Aerospace Engineering University of Oklahoma 865 Asp Ave. Rm 212, Norman, OK 73019, USA yingtao@ou.edu S. Patterson, Dr. T. Robison National Security Campus operated by Honeywell Botts Rd., Kansas City, MO 64147, USA The ORCID identification number(s) for the author(s) of this article can be found under DOI: /admt are the most popular methods. Electrically conductive nanomaterials, such as carbon black, [11] planar graphene, [12] carbon nanotube (CNT), [13] carbon nanofiber (CNF), [14] gold nanowires, [15] conductive polymer, [16] silver nanowires, [17] and silver flakes, [18] have been widely studied for the piezoresistive sensing applications both in stand-alone form and by dispersing into flexible polymers. Several review papers have summarized the state of the art of current research on highly flexible nanocomposites. [19 21] When used in standalone form, such as CNT bucky papers, these nanoparticles can provide superior electrical conductivity, but are lacking in flexibility. [22] The more practical approach is to disperse nanoparticles within a flexible polymer, such as PDMS, [23,24] polyimide, [25] and polyurethane. [26] It has been reported that a small amount of CNTs or CNFs dispersed in polymer can significantly enhance the electrical conductivity by orders. Among those, CNF is two orders of magnitude cheaper than its counterparts CNT and graphene. [14] Multiple key parameters, including electrical conductivity, sensitivity, sensing range, durability, and manufacturing cost and complexity, can impact on the performance of nanocomposite sensors. Advanced manufacturing technologies, such as etching and photolithography, have been employed to create critical microstructures in nanocomposites, resulting in novel pressure and strain sensors, though the manufacturing cost and scalability are often criticized. [24,27] Recently, graphene/ PDMS nanocomposites have shown good sensitivity with gauge factor of 61.3 and detectability of up to 20% tensile strain, but manufacturing requires freezing at 50 C and heat treatment at 750 C. [28] In terms of sensor performance, obtaining an optimum combination of sensitivity and sensing range is challenging. High gauge factors in the order of 1000 were obtained using a woven mesh configuration of graphene/polymer but this material only achieved up to 6% strain. [29] A graphene film was made in an easy single step process, and showed gauge factor of 1000 but its failure strain was only at 2% strain. [30] In general, highly flexible nanocomposites result in the trade-off of piezoresistive sensitivity. For example, Qin et al. reported that flexible nanocomposites using reduced graphene oxide and polyimide had the maximum stretching and compression sensing capability up to 50% strain with low gauge factors of and 1, respectively. [25] Majority of the works on sensor (1 of 10)

2 arrays for pressure sensing have a narrow operating range, mainly in the low-pressure range with a detection limit of few kpa. [9,31] A capacitive-based pressure sensor array developed by Woo et al. was able to achieve a pressure range of kpa, but required micropatterning and soft lithography. [10] Sensors with interlocked microstructure have shown both high sensitivity and deformability. However, the electrical resistance of the sensor was more than 100 MΩ, which is impractical for many applications. [27] Porous CNT/PDMS nanocomposites have been reported, but were limited in durability due to degrading and fracture of the porous polymer matrix. [32] Therefore, it is necessary to develop a new sensing material that can bridge the current gap in those requirements. Although significant efforts have been reported on developing novel materials with high piezoresistive sensitivity and high flexibility, the sensing mechanism has not been well explained. Previous published journals were able to model and simulate the piezoresistive sensing behavior. [33 35] To our best knowledge, the piezoresistive sensing mechanism has not been well experimentally characterized, particularly at the microscale to show single fiber realignment under external applied load conditions. In this study, we present a nanocomposite made of tailormade PDMS enhanced by CNFs that shows piezoresistive properties. The fabrication procedure is facile and of low cost, as it involves simply dispersing CNFs into PDMS using solventbased ultrasonication and cast molding. The resultant nanocomposite showed electrical conductivity in the order of 10 2 S m 1 starting from 8 wt% CNF content. Sensors in various shapes were fabricated and showed multiple sensing functions in both tension and compression loads. For tensile sensing, stable and repeatable sensing response was achieved for multiple cycles up to 30% strain. Both positive and negative piezoresistive behaviors (resistance increases and decreases, respectively, with tensile strain) have been reported for composites with high aspect ratio nanomaterials as CNT and CNFs. [36] To study the fundamental sensing mechanism of such materials under tensile strain, in situ micromechanical tests were carried out within scanning electron microscope (SEM) and showed both rotation of long nanofibers and interfiber distance variation that led to the formation of new conductive networks and change overall and resulted in the piezoresistive sensing function in the developed nanocomposites. To achieve higher sensitivity, different shapes of sensing units were fabricated for compressive load sensing applications. Cylindrical and truncated cone shape units showed gauge factors of 18.3 and 6.3, respectively, with minimum detectable compression of 3% strain. Reliability testing revealed stable response for up to 1000 cycles and no dependency on strain rate. Bending deformation via human finger movement was successfully detected using a glove with sensors attached. Two sensor arrays were fabricated to detect differential pressure ranging between 35 and 690 kpa. All these results demonstrate the promising applications as flexible strain and pressure sensors using the developed nanocomposite. 2. Results and Discussions 2.1. Nanocomposite Property: Morphology and Conductivity High aspect ratio nanomaterials, like CNTs and CNFs, have high surface area, which creates dominant wall-to-wall van der Waals attractive force. As a result, the CNFs synthesized using chemical vapor deposition (CVD) method usually come as highly bundled agglomerates and require additional external energy to overcome the force and break the agglomeration. [37] It is essential to uniformly disperse the CNFs into the polymer matrix to obtain desired mechanical and electrical properties. Both conductivity and piezoresistive sensitivity stem from conductive paths created by CNFs throughout the matrix, which may not be achievable without proper dispersion. There can be order of magnitude differences in electrical conductivity of the overall composite between well and poorly dispersed nanofillers as reported in literature. [38] Without proper dispersion, more CNF content will be required, which will drive up the cost of manufacturing. To verify the feasibility of fabrication method, SEM micrographs were taken at different length scales for materials with 8 wt% CNF content. Figure 1 shows representative SEM images showing uniform dispersion of long and previously bundled CNFs in the PDMS matrix, with almost no agglomeration. These images indicate that nanocomposites Figure 1. Representative SEM micrographs at different length scales of nanocomposite with 8 wt% CNF content showing uniform dispersion of CNFs in PDMS (2 of 10)

3 Figure 2. Volume conductivity of prepared nanocomposites as a function of CNF content. produced by such solvent-based ultrasonication technique are suitable for this application. In addition to dispersion state, electrical conductivity (σ) and resistivity (ρ) are key material properties that need to be quantified to select the correct material. Figure 2 shows the variation in volume conductivity of the prepared nanocomposite specimens with different CNF weight concentration. As expected, electrical conductivity increases with increased CNF content as more conductive networks are created by the dispersed CNFs. An improvement in conductivity of almost 10 orders of magnitude is achieved when CNF content is changed from 0.5 to 8 wt%. For nanocomposites with 8 and 12 wt% CNF content, electrical conductivity is on the order of 10 2 S m 1, which proves that excellent electrical conductivity has been achieved. Additionally, this level of conductivity validated the success of this dispersion procedure as higher electrical conductivity correlates to higher uniformity of dispersion. Resistivity of the materials decreases with increased CNF content, as the inverse of conductivity. Due to their superior electrical properties, composites with a minimum of 8 wt% CNF were used for mechanical deformation detection and sensor arrays fabrication Piezoresistive Sensing: Stretching Deformation Stretching is one of the predominant modes of deformation that wearable sensors must naturally withstand and detect. To understand and quantify the stretching sensing capability, rectangular nanocomposite units (75 mm 10 mm 2.5 mm) were used. To quantify the piezoresistive behavior, relative resistance change ΔR/R o (where R o is the initial resistance at the unstretched condition) was recorded while applying quasi-static stretching at different strains in a cyclic manner. CNF/PDMS units were subjected to cyclic stretching (from unstretched condition) up to three different levels of tensile strain: 10%, 20%, and 30%. Figure 3a shows the change in relative resistance at various tensile strains for five cycles. It is important to first note the stability of the response produced under cyclic tensile strain, with almost no variation between the first and final cycle, indicating excellent repeatability of the sensing units. Second, the sensing response is in synchronization with applied cyclic stretching as shown in Figure 3b, with the relative resistance change following loading and unloading. The maximum change in relative resistance varies up to 30%, 50%, and 60% as levels of maximum stretching are increased as 10%, 20%, and 30%, respectively. Finally, relative resistance change for a single stretch and release cycle at different strains is shown in Figure 3c e. Initially, as the material is stretched, the relative resistance change is negative, indicating negative piezoresistive behavior. Before the maximum stretch level is reached and unloading stage is started, a reverse behavior is observed with increasing resistance indicated by the second peak. This reverse behavior in relative resistance is quantified as 2.5%, 5.5%, and 10% (shown by two horizontal orange dashed lines in Figure 3c e) for strains of 10%, 20%, and 30%, respectively. These figures clearly indicate two distinct steps of piezoresistive response under stretching. A similar response showing negative piezoresistance under stretching or second peaks have been reported before for CNF and CNT based nanocomposite. [39 41] Toprakci et al. hypothesized that a similar twostage behavior for CNF/PDMS material is due to out-of-plane rotation of fibers that reduces the average distance between conductive fibers and creates more conductive paths leading to decrease in the resistance. In the later stage of stretching, the in-plane interfiber separation becomes more dominant, which increases the resistance. [39] Rathi et al. attributed similar negative piezoresistance, and eventual saturation of conductive networks, in CNT/PDMS nanocomposites to the effect of Poisson s ratio in the transverse direction, which produces in-plane CNT reorientations. [41] However, in both cases, no supporting evidence was provided to prove any of the hypotheses. It is well known that for nanofiller-based materials, piezoresistance is generated due to change in interfiber distance and number of conducting paths under external load. [42] To explore the sensing mechanism, the CNF/PDMS material was characterized using an in situ micromechanical testing stage in SEM at different levels. In situ SEM images were taken to observe the change in interactions between fibers under deformation as shown in Figure 4. Clear interactions between nanofibers were observed where both rotation of nanofibers and change in interfiber distance were visible. The rotation phenomenon, shown by magnified SEM images at unstretched and stretched conditions in Figure 4a,b, is due to alignment and reorientation of out of plane nanofibers. Reorientation of fibers reduces both the average distance between fibers and number of possible conductive paths that can be created throughout the matrix. Both phenomena correspond to the increase in conductivity, which produces negative piezoresistance at the first stage of stretching. As strain of the nanocomposites is increased, interfiber separation becomes more dominant. The increased interfiber distance reduces the amount of tunneling current flow and corresponds to positive piezoresistance at the second stage of stretching. The interplay of both CNF reorientation and separation dictates the overall sensing response under stretching, and the behavior should be similar for composite materials containing high aspect ratio nanofillers. These results experimentally explain the piezoresistive behavior of such materials under stretching not properly explored before (3 of 10)

4 Figure 3. Sensing response of CNF/PDMS under stretching. a) Relative resistance change with cyclic strain at maximum strains of 10%, 20%, and 30% for five cycles; b) loading and unloading cycles; c e) relative resistance change for a single strain cycle showing negative relative change in resistance and a final positive change in resistance (marked by horizontal dashed lines) for different maximum stretching level. and can provide experimental validation as well as guidance for improving existing theoretical models Compression Sensing To explore the sensing response of CNF/PDMS under compression, single sensor units with cylindrical (10 mm height 10 mm diameter) and truncated cone shapes (10 mm height 10 mm base diameter 5 mm top diameter) were used. Different geometric shapes can be used to tune and improve the sensitivity of response under compression. [43] Cyclic compression and release at maximum strains between 3% up to 25% and relative resistance change were measured throughout the test. The results are shown in Figure 5 for sensing units of both shapes. Resistance decreases monotonically with increase in compressive strain. Once the compression release cycle is started, resistance starts increasing and almost completely returns to the initial resistance in its uncompressed state upon complete unloading. Cyclic resistance change also shows that the response is very stable and repeatable with no fluctuations between cycles. The minimum detectable compressive strain by the PDMS/CNF sensing units producing repeatable results was 3%. Sensitivity in terms of gauge factor (defined as (ΔR/R o )/ε, where ε is the applied strain) for the different shaped sensing units under compressive strains was tested for maximum strains of 3%, 5%, 10%, and 25%. For the sensor units of cylindrical shapes, gauge factors of 18.3, 14.8, 8.2, and 3.5, respectively, were obtained for the mentioned maximum compression levels, as shown in Figure 5a. Gauge factors for the truncated cone shaped unit were 6.3, 8.0, 5.5, and 3.5, respectively, as shown in Figure 5b. The sensitivity results indicate that the cylindrical sensing unit is more sensitive at low compression and saturates faster with increased compressive strains. The truncated cone shape unit is more sensitive at higher strains and saturates slowly in comparison. This contrasting response can be exploited to employ different shapes to extend the sensing range when making sensor arrays with these sensing units without the need of any microstructural engineering. Robustness and durability are critical for the proposed sensors to have any practical applications. Figure 6a shows the response of the truncated cone shaped unit under cyclic compression loading unloading for five cycles at varying strain rates. Six different strains rates (5%, 10%, 20%, 50%, 100%, and 200% strain per minute) were applied, relating to almost two orders of magnitude variation in strain rate. CNF/PDMS showed almost no dependency in strain rate on the sensing, suggesting excellent robustness. A test of 1000 cyclic compression loading (4 of 10)

5 Figure 4. SEM images showing in situ visualization of nanofibers interaction under tensile strain A) at 20% and B) at 40%; a d) shows the magnified images of the red circles containing the interaction region; a,b) shows rotation of nanofibers; c,d) shows change in-plane interfiber distance where (i) is at unstretched and (ii) at stretched state, respectively. revealed very little drift and fluctuation in relative resistance change, which shows great durability of the CNF/PDMS sensor (Figure 6b). In terms of quantification, the magnitude drift of relative resistance change over 1000 cycles is less than 5% (fluctuates between 92% and 96% maximum change), as shown in Figure 6c. The typical response of single loading unloading cycle is shown in Figure 6d, and indicates that upon unloading the initial resistance is almost regained, pointing to the existence of very small hysteresis. Additionally, three distinct stages of sensitivity can be noticed as strain is increased to its maximum value. The results provide evidence for the potential use of the proposed PDMS/CNF nanocomposite Application as Wearable Sensors One of the major applications for the proposed piezoresistive sensor is wearable devices that can be used for detecting (5 of 10)

6 Figure 5. Sensing response of CNF/PDMS under cyclic compression release at different maximum strains for a) cylindrical shaped unit and b) truncated cone shaped unit. human joint movements. To demonstrate the capability of wearable sensors, a glove was prepared by attaching CNF/ PDMS strips onto its five fingers. The strips were specifically located on the glove to be attached with the proximal joints of human fingers and would undergo deformation with finger bending motion. Bending motion of proximal joints in human fingers is vital for actions like grabbing items and gesturing. An image of the glove at its relaxed state and with a bend of the index finger is shown in Figure S9 of the Supporting Information, exhibiting its conformability. Bending was performed on all five fingers at an angle of 90. Three types of finger bending were induced as a) quick bend and quick release b) quick bend, hold for five seconds, and then release, and c) quick bend, hold for fifteen seconds, and release. Normalized resistance changes Figure 6. Relative resistance change in compression showing a) effect of strain rate in cyclic response; b) representative cyclic response for 1000 cycles showing durability; c) fluctuation of relative resistance change magnitude over 1000 cycles; d) single cycle response showing little hysteresis (6 of 10)

7 Figure 7. Normalized resistance signal detecting cyclic bending of human fingers. a) Slow bend and release; b) quick bend and slow bend. produced by three types of bending motions mentioned above for all five human fingers were recorded. Figures 7 and 8 show the response signal generated by sensor strips attached to each finger for multiple bending and release cycles. Distinguishable signal patterns (normalized resistance) are generated for the low movement (Figure 7a) and fast and slow movement (Figure 7b) of each finger. The signal patterns consist of decreasing resistance with bending and subsequent increase with finger release. Relative resistance change remains stable when the fingers are held at bent condition and returns to its initial value upon release (Figure 8a,b). These distinct, reproducible, and consistent signals generated by finger bending movement show the human motion detectability of the developed sensor. of and 345 kpa were applied at different locations of the sensor array with cylindrical sensing units (Figure 9a). The sensor array with truncated cone shaped units experience a higher maximum pressure, 690 kpa (Figure 8c). The 3D bar plots of relative resistance change indicate that the sensor array successfully exhibited changes in signal corresponding to the maximum and minimum applied pressures. A gradual shift in magnitude of normalized resistance change with applied pressure level for both sensor arrays (Figure 9b,d) proves that a wide range of pressures can be detected, with the minimum detectable pressure of 35 kpa. Additionally, the resolution of differentiable pressure can be tuned by employing different shaped sensing units Application as Sensor Array A large area deployable sensor array capable of detecting a wide range of pressures can be especially useful. Two types of sensor arrays were fabricated using 3 3 units of cylinder sensing units, and 4 4 truncated cone sensing units. As discussed in the previous section, cylindrical units were found to be more sensitive at smaller compression, whereas the truncated cone shaped units were more sensitive to larger compression. Exploiting that characteristic, differential pressures 3. Conclusion In this paper, a flexible and stretchable CNF/PDMS nanocomposite with piezoresistive sensing functions has been developed using a simple, scalable fabrication technique via solvent-based ultrasonication. Uniform dispersion of CNFs inside PDMS produced electrical conductivity on the order of 10 2 S m 1. Sensing units showed repeatable response and good sensitivity up to 30% maximum stretching. Exploration of piezoresistive sensing mechanism using in situ stretching and SEM imaging revealed (7 of 10)

8 Figure 8. Normalized resistance signal detecting cyclic bending of human fingers. a) Quick bend and slow bend; b) quick bend and long hold. Figure 9. Sensor array response made of cylinder shaped units a,b) and truncated cone shaped sensing units c,d) shows detection of differential applied pressure (8 of 10)

9 that both rotation and separation of CNFs provide contributions. Using different shaped sensing units for compression showed that sensitivity can be tuned for different compression regions. Durable and reliable piezoresistive sensing response was achieved for over 1000 loading unloading cycles and no hysteresis or dependence on strain rate was observed. Along with these promising results, the proposed sensor showed practical applications as wearable device and sensor arrays. Human joint motion detection was demonstrated via finger bending by using gloves attached with sensor strips. Compression sensor array results showed a wide range of detectable pressures, ranging from 35 to 690 kpa. A combination of low-cost fabrication, flexibility, reliable sensitive response, and a wide range of applications prove the versatility of the developed sensor, where only a subset of requirements are met by others of similar purpose. 4. Experimental Section Materials: The PDMS formulation (Gelest Inc.) was controlled by adjusting the constituent of materials: prepolymer, copolymer, silica filler, and platinum catalyst. In this paper, the prepolymer and silica filler ratio was maintained as 10:2. CNF nanoparticles with an average diameter of 100 nm was ordered from Applied Science Inc. (PR-24XT-LHT) and used as the conductive nanofillers. Tetrahydrofuran (THF, ACS stabilized) purchased from Macron Fine Chemicals was used as the solvent to disperse the CNF. The detailed characterization of PDMS polymer and CNF nanoparticles are reported in the Supporting Information. Preparation of Nanocomposite Sensing Units: CNFs were mixed with THF (1 g of CNF per 50 g of THF) by a mechanical mixer and dispersed via ultrasonication (Sonic Vibracell with Ti horn) at low amplitude for 4 h. PDMS prepolymer and silica filler were mixed separately via a mechanical mixer at 1500 rpm for 10 min and added into the CNF/ THF solution. Subsequently, 2 h of sonication was performed to obtain uniform dispersion of CNFs into PDMS. [35] The solvent was removed in an oil bath at 60 C with magnetic stirring followed by vacuum degassing (100 kpa). Copolymer and catalyst were mixed with the degassed viscous solution as the penultimate step. Finally, uncured nanocomposites were cast into mold and cured inside a hot press at 80 C for 3 h, followed by post-curing at 150 C for 2 h. Testing and Characterization: CNF dispersion in the fabricated nanocomposites was characterized in SEM (VEGA TESCAN at 20 kv) after sputter coating the samples with gold palladium alloy (about 2 nm thickness). Electrical conductivity was measured using the four-probe method with an Agilent millimeter by attaching four wires (Figure S6, Supporting Information) with silver paste on the material surface (for low resistivity samples). To measure the electrical conductivity of nanocomposites with low CNF loading, a resistivity adapter (Keithley 6105) was used. [44] In situ stretching of nanocomposite samples were carried out using a DEBEN microtester in displacement-controlled mode in an SEM, and images were taken at different strain level to observe the realignment and reorientation of CNFs. For continuous sensing response measurement in cyclic stretching and compression, an INSTRON uniaxial tester was used to mechanically deform the sample while resistance was recorded with an Agilent multimeter and RS-232 logger at 3 Hz sampling frequency. All tests were performed at 10% strain min 1. To demonstrate the effect of strain rate on the electrical response of the nanocomposite, mechanical deformation was performed at six different rates (5%, 10%, 20%, 50%, 100%, and 200% strain min 1 ). For stretching and compression sensing, samples with 12 and 8 wt% CNF loading, respectively, were used. All sensor units for compression sensing were sputter coated on top and bottom, and connected with copper wires to minimize the contact resistance. For finger bending detection, sensing units were attached to textile gloves using polymer epoxy. All the wearable sensor experiments were performed in compliance with the relevant laws and institutional guidelines of the University of Oklahoma, and have received appropriate Ethical Committee approval. The consent of all the participants in the wearable sensor experiments has been obtained. Sensor Array Fabrication: Sensor arrays were created with both cylinder and truncated cone shaped sensors. The sensors were arranged in a series of rows to produce each sensor array. Copper tape was used to interconnect each sensor due to its very high conductivity, ease of availability, and good adherence to PDMS surface. All the sensors were attached to electrodes with two parts silver epoxy that minimized contact resistance. Neat PDMS films were placed on the top and bottom of each sensor array during testing to provide support and act as insulation layers to prevent short circuits. Supporting Information Supporting Information is available from the Wiley Online Library or from the author. Acknowledgements The authors acknowledge the financial support from the Honeywell Federal Manufacturing & Technologies, LLC Kansas City Plant, and Dr. Thomas Robison is Program Manager. Conflict of Interest The authors declare no conflict of interest. Keywords elastomer, flexible sensor, nanocomposite, piezoresistivity, pressure sensor Received: August 28, 2018 Revised: September 24, 2018 Published online: October 25, 2018 [1] M. Totaro, A. Mondini, A. Bellacicca, P. Milani, L. Beccai, Soft Rob. 2017, 4, 400. [2] A. Chortos, Z. Bao, Mater. Today 2014, 17, 321. [3] J. H. Kim, J.-Y. Hwang, H. R. Hwang, H. S. Kim, J. H. Lee, J.-W. Seo, U. S. Shin, S.-H. Lee, Sci. Rep. 2018, 8, [4] I. Kang, M. J. Schulz, J. H. Kim, V. Shanov, D. Shi, Smart Mater. Struct. 2006, 15, 737. [5] C. Lucarotti, C. M. Oddo, N. Vitiello, M. C. Carrozza, Sensors 2013, 13, [6] T. Yamada, Y. Hayamizu, Y. Yamamoto, Y. Yomogida, A. Izadi-Najafabadi, D. N. Futaba, K. Hata, Nat. Nanotechnol. 2011, 6, 296. [7] a) M.-Y. Cheng, C.-M. Tsao, Y.-Z. Lai, Y.-J. Yang, Sens. Actuators, A 2011, 166, 226; b) L. Pan, A. Chortos, G. Yu, Y. Wang, S. Isaacson, R. Allen, Y. Shi, R. Dauskardt, Z. Bao, Nat. Commun. 2014, 5, [8] S. Yao, Y. Zhu, Nanoscale 2014, 6, [9] S. C. Mannsfeld, B. C. Tee, R. M. Stoltenberg, C. V. H. Chen, S. Barman, B. V. Muir, A. N. Sokolov, C. Reese, Z. Bao, Nat. Mater. 2010, 9, 859. [10] S.-J. Woo, J.-H. Kong, D.-G. Kim, J.-M. Kim, J. Mater. Chem. C 2014, 2, (9 of 10)

10 [11] E. S. Bhagavatheswaran, M. Parsekar, A. Das, H. H. Le, S. Wiessner, K. W. Stöckelhuber, G. Schmaucks, G. Heinrich, J. Phys. Chem. C 2015, 119, [12] W. Yuan, J. Yang, K. Yang, H. Peng, F. Yin, ACS Appl. Mater. Interfaces 2018, 10, [13] Y. Zhang, C. J. Sheehan, J. Zhai, G. Zou, H. Luo, J. Xiong, Y. Zhu, Q. Jia, Adv. Mater. 2010, 22, [14] S. Wu, J. Zhang, R. B. Ladani, A. R. Ravindran, A. P. Mouritz, A. J. Kinloch, C. H. Wang, ACS Appl. Mater. Interfaces 2017, 9, [15] S. Gong, W. Schwalb, Y. Wang, Y. Chen, Y. Tang, J. Si, B. Shirinzadeh, W. Cheng, Nat. Commun. 2014, 5, [16] T. S. Hansen, K. West, O. Hassager, N. B. Larsen, Adv. Funct. Mater. 2007, 17, [17] S. Lee, S. Shin, S. Lee, J. Seo, J. Lee, S. Son, H. J. Cho, H. Algadi, S. Al-Sayari, D. E. Kim, Adv. Funct. Mater. 2015, 25, [18] N. Matsuhisa, M. Kaltenbrunner, T. Yokota, H. Jinno, K. Kuribara, T. Sekitani, T. Someya, Nat. Commun. 2015, 6, [19] T. Q. Trung, N. E. Lee, Adv. Mater. 2017, 29, [20] T. Q. Trung, N. E. Lee, Adv. Mater. 2016, 28, [21] S. Zhao, J. Li, D. Cao, G. Zhang, J. Li, K. Li, Y. Yang, W. Wang, Y. Jin, R. Sun, ACS Appl. Mater. Interfaces 2017, 9, [22] J. DeGraff, R. Liang, M. Q. Le, J.-F. Capsal, F. Ganet, P.-J. Cottinet, Mater. Des. 2017, 133, 47. [23] S. Wu, S. Peng, C. H. Wang, Sens. Actuators, A 2018, 279, 90. [24] B. Nie, X. Li, J. Shao, X. Li, H. Tian, D. Wang, Q. Zhang, B. Lu, ACS Appl. Mater. Interfaces 2017, 9, [25] Y. Qin, Q. Peng, Y. Ding, Z. Lin, C. Wang, Y. Li, F. Xu, J. Li, Y. Yuan, X. He, ACS Nano 2015, 9, [26] Q. Fan, Z. Qin, S. Gao, Y. Wu, J. Pionteck, E. Mäder, M. Zhu, Carbon 2012, 50, [27] J. Park, Y. Lee, J. Hong, Y. Lee, M. Ha, Y. Jung, H. Lim, S. Y. Kim, H. Ko, ACS Nano 2014, 8, [28] S. Wu, R. B. Ladani, J. Zhang, K. Ghorbani, X. Zhang, A. P. Mouritz, A. J. Kinloch, C. H. Wang, ACS Appl. Mater. Interfaces 2016, 8, [29] X. Li, R. Zhang, W. Yu, K. Wang, J. Wei, D. Wu, A. Cao, Z. Li, Y. Cheng, Q. Zheng, Sci. Rep. 2012, 2, 870. [30] X. Li, T. Yang, Y. Yang, J. Zhu, L. Li, F. E. Alam, X. Li, K. Wang, H. Cheng, C. T. Lin, Adv. Funct. Mater. 2016, 26, [31] H. B. Yao, J. Ge, C. F. Wang, X. Wang, W. Hu, Z. J. Zheng, Y. Ni, S. H. Yu, Adv. Mater. 2013, 25, [32] Y. A. Samad, Y. Li, S. M. Alhassan, K. Liao, ACS Appl. Mater. Interfaces 2015, 7, [33] X. Ren, G. D. Seidel, J. Intell. Mater. Syst. Struct. 2013, 24, [34] T. Theodosiou, D. Saravanos, Compos. Sci. Technol. 2010, 70, [35] X. Ren, G. D. Seidel, Compos. Interfaces 2013, 20, 693. [36] T. J. Kang, A. Choi, D.-H. Kim, K. Jin, D. K. Seo, D. H. Jeong, S.-H. Hong, Y. W. Park, Y. H. Kim, Smart Mater. Struct. 2011, 20, [37] Y. Geng, M. Y. Liu, J. Li, X. M. Shi, J. K. Kim, Composites, Part A 2008, 39, [38] Y. S. Song, J. R. Youn, Carbon 2005, 43, [39] H. A. Toprakci, S. K. Kalanadhabhatla, R. J. Spontak, T. K. Ghosh, Adv. Funct. Mater. 2013, 23, [40] T. Zhai, D. Li, G. Fei, H. Xia, Composites, Part A 2015, 72, 108. [41] P. Rathi, M. K. Jha, K. Hata, C. Subramaniam, ACS Omega 2017, 2, [42] L. Chen, G. Chen, L. Lu, Adv. Funct. Mater. 2007, 17, 898. [43] C. L. Choong, M. B. Shim, B. S. Lee, S. Jeon, D. S. Ko, T. H. Kang, J. Bae, S. H. Lee, K. E. Byun, J. Im, Adv. Mater. 2014, 26, [44] V. Mironov, J. Kim, M. Park, S. Lim, W. Cho, Polym. Test. 2007, 26, (10 of 10)