Research Article www.acsami.org
Cite This: ACS Appl. Mater. Interfaces 2017, 9, 41078-41086
Precise Engineering of Conductive Pathway by Frictional DirectWriting for Ultrasensitive Flexible Strain Sensors Zhikang Zeng,†,‡ Yan Yu,*,† Yongming Song,† Ni Tang,† Lei Ye,† and Jianfeng Zang*,†,‡ †
School of Optical and Electronic Information and ‡Innovation Institute, Huazhong University of Science and Technology, 1037 Luoyu Road, 430074 Wuhan, China
ACS Appl. Mater. Interfaces 2017.9:41078-41086. Downloaded from pubs.acs.org by MIDWESTERN UNIV on 01/24/19. For personal use only.
S Supporting Information *
ABSTRACT: Highly sensitive strain sensors that can detect small strain are in high demand in the fields of displays, robotics, fatigue detection, body monitoring, in vitro diagnostics, and advanced therapies. However, resistive-type sensors that are composed of electrically conductive sensing films coupled with flexible substrates suffer from the limits that their gauge factors (GFs) at small strains (e.g., 0.1−1%) are not high. Herein, through frictional direct-writing of graphite rod on the composite paper substrates, we produced strain sensors with extremely high gauge factor at small strains. The sensors exhibited a gauge factor of 9720 at a small strain of 0.9%, minimum strain detection up to 0.05%, strain resolution of 0.05%, response time of 40 ms, and high stability (>5000 bending−unbending cycles). Compared with the literature results so far, our sensors hold the highest GF value at small strains. Such high sensitivities are due to the precise control of narrow twodimensional percolative conductive pathway, which means the content of conductive graphite sheets is close to the conductive percolation threshold. The strain sensors have a rapid response to microdeformation changes and can monitor various structural changes, including human motion, through facilitative and effective installation of device designs. KEYWORDS: strain sensor, small strain, graphite, percolation threshold, high sensitivity
■
INTRODUCTION
detection, body monitoring, in vitro diagnostics, and advanced therapies.16−23 Generally, resistive-type sensors are typically composed of electrically conductive sensing films coupled with flexible substrates. When the sensors are stretched, microstructural changes in the sensing films lead to change in the electrical resistance as a function of the applied strain. After the release of strain, re-establishment of the sensing films to their original states recovers the electrical resistance of sensors. For these strain sensors, the gauge factor (GF) relates to the change in the electrical resistance to the applied strain through
Nowadays, flexible electronics is one of the fastest growing areas of the electronics industry.1−15 Among these emerging areas, flexible strain sensors that can measure human motions and monitor personalized health are attracting a great deal of attention owing to their unique characteristics, such as wearability, light weight, high flexibility, etc.16−19 Consequently, in recent years, there has been notable research effort devoted to the development of flexible strain sensors to fulfill the requirements of future technology. Important considerations for the development of the flexible strain sensors are the choice of sensing materials and device structures to accomplish the requirements of ultrahigh sensitivity, flexibility, durability, fast response, stability, linearity, etc. Gauge factor (GF), reflecting the sensitivity and often evaluated by the relative change in resistance versus applied strain, is vital for detecting the microstrain (less than 1%).20−39 These sensory systems could be useful in diverse applications needing ultrahigh displacement sensitivity, such as healthcare monitoring (especially weak motions including breathing, phonation, facial expression changes, blink, and pulse), human−computer interaction, electronic skins, and particularly, microstrain detection in artificial vessels implanted for aortic dissection therapy. Thus, highly sensitive strain sensors that can detect small strain are in high demand in the fields of the displays, robotics, fatigue © 2017 American Chemical Society
GF = (ΔR /R 0)/ε
(1)
where ΔR/R0 is the normalized change in electrical resistance and ε is the mechanical strain. According to the classical theory of percolation conductivity, in a two-dimensional (2D) plane model, two-dimensional percolative metal−insulator transition at the critical area fraction, Φc, will happen with an increase in the conductive particles.40,41 Generally, conductive networks near the percolation threshold are now of potential interest as sensitive strain gauges because they may exhibit a high GF value. Received: September 24, 2017 Accepted: November 2, 2017 Published: November 2, 2017 41078
DOI: 10.1021/acsami.7b14501 ACS Appl. Mater. Interfaces 2017, 9, 41078−41086
Research Article
ACS Applied Materials & Interfaces
Figure 1. Fabrication process of our graphite-based strain sensor: (a) schematic of frictional direct-writing method. (b) Scanning electron microscopy (SEM) image of flexible substrate (poly(ethylene terephthalate) (PET)−Al2O3 adhesive layer). (c) The growth of the graphite sheets on the substrate surface (optical images). (d) The resistance change trend as a function of writing times.
0.05% strain, with obvious resistance change recorded. Such high sensitivities due to the content of conductive graphite sheets are close to the conductive percolation threshold. The strain sensors have a rapid response to microdeformation changes and can be used to monitor various structural changes, including human motion, through facilitative and effective installation of device designs.
However, because the critical region where percolation occurs corresponds to a narrow range of Φ, it is difficult for the existing methods to correctly target Φc. Consequently, although GF is expected to diverge to very large values near Φc, it is difficult to find cases in the literature where GF is larger than 1000 under microstrain, e.g., 1%. (Typically, reported strain sensors that are based on percolative conductive networks on flexible substrates showed GF values lower than 1000 at 1% strain.20,22−39 Up to now, according to our literature review, the only report that had GF value higher than 1000 is ref 21, about 5000 at 1% strain.) Herein, we propose a low-cost frictional direct-writing method that uses composite films as flexible substrates and graphite rods as the feeding source of conductive phase. Through frictional direct-writing of graphite traces on the composite films substrates, sample with graphite conductive path can be prepared. The GF values of our samples are extremely high: at a small strain of 0.05%, the GF values could be higher than 3000, whereas at the small strain from 0.1 to 0.9%, the GF values could be up to 9720. Compared with reported strain sensors that are based on percolative conductive networks on flexible substrates, our sensors hold the highest GF value at a small strain (below 1%).20−39 These indicated that the sensor could typically detect the microdeformation of
■
MATERIALS AND METHODS
Materials. Composite papers (3M) were used as flexible substrates. Specification of the paper is 6000 mesh. Graphite rods drawing with ruler-guided in orthogonal direction were repeated for several times (Figure 1c) to form a stripe of uniform graphite flakes on the flexible substrates. The graphite traces were typically of 2 mm width and 10 mm length, or 1 mm width and 2 mm length. Then, the two ends of the rectangular copper (Cu) foil were used as contact electrodes to lead to the conductive line. Sensor Fabrication. Figure 1a describes the fabricating procedures of our graphite-based strain sensor. The device was fabricated by simple drawing with a graphite rod on a certain flexible composite film with a rough surface.42−44 The composite film consisted of a 75 μm thick poly(ethylene terephthalate) (PET) layer and a 15 μm thick Al2O3 adhesive layer (Figure 1a). The SEM images of the film surface is shown in Figure 1b, reflecting its surface with a certain degree of roughness. On the one hand, owing to the PET layer, the composite 41079
DOI: 10.1021/acsami.7b14501 ACS Appl. Mater. Interfaces 2017, 9, 41078−41086
Research Article
ACS Applied Materials & Interfaces
Figure 2. Main performances of the sensor: (a−c) the resistance−strain properties of three typical samples. (d) Comparison of the GFs at small strains in our sensors and those in literature results. (e) Bending and releasing of the sensor and the response time. (f) Durability doing 5000 cyclic strain tests. film had a superior mechanical flexibility. On the other hand, Al2O3 layer allowed the graphite sheets to be easily worn away from the graphite rod and attached to the flexible polymer surface. We drew graphite on the film trace by trace and detected the resistance of the graphite trace in real time (drawing pressure: ∼100 MPa, drawing speed: ∼0.01 m/s). With increase in the drawing times, the graphite sheet was gradually filled onto the flexible substrate (see Figure 1c and also see Figure S1). As shown in Figure 1d, the resistance reduced gradually and the percolation threshold appeared at period III, where one drawing led to the resistance dropping by 4 orders of magnitude. To obtain the samples with graphite contents close to the percolation threshold, we stopped drawing once the stupendous dropping of the resistance was observed. Characterizations. Field emission scanning electron microscopy (Sirion 200 and Nova NanoSEM 450) and optical microscopy (Zeiss Axiovert 200MAT) were employed to observe the morphology of sensor’s surface. The temperature changes and differences on sensors’ surfaces were detected by a microscopic thermal imager (FLIR T420). The electromechanical properties of the strain sensors were measured by sensitive current-measuring instrument (Keithley 6485) through the methods described in reported works.45
between 0.1 and 0.9%, the GF values were larger than 3580 and could be up to 9720. Compared with the reported resistive-type strain sensors that are based on tunneling model (with a highest GF value of 1000),23,32,35−39 our samples had much higher GF values at the strain below 1%, representing a high sensitivity to microstrain. To ensure that our highly sensitive strain sensor is a good application prospect, we discuss the dynamic performance of the samples. First, we focus on its response and recovery time to determine how quickly the strain sensors move toward a steady-state response. To examine the response time of our sensor to external forces, we applied 1 V voltage to it and detected the resistance change as the sensor was bent (max ε is about 0.5%) and released (Figure 2e). As presented in Figure 2e, the sudden fall and rise in the resistance change represent the bending and release of the sensor, respectively. It can be noted that the response time and the recovery time were both 40 ms (see Figure 2e), indicating that our sensors responded very quickly to strain. Second, the durability was examined by doing 5000 cyclic strain tests, and the sensor performance remained unchanged up to about 5000 loading and unloading cycles under 0−0.5% strain (see Figure 2f and also see Figure S5). These results present the strain sensors’ features of good durability and fast response to external forces.46,47 It is important to note that in the process of continuous preparation of samples, the graphite rod surface would be gradually smoother. Therefore, with the same graphite rod, the samples would have some differences. However, we drew graphite on the film trace by trace and detected both the resistance of the graphite trace in real time and the resistance after each trace. Also, we stopped drawing once a stupendous drop in resistance was observed. These could effectively guarantee the consistence performance of the sensors. So, interestingly, the performance of the samples was basically divided into three categories: (1) the initial resistance of the first few samples was usually between 0.4 and 0.7 MΩ, the GF was higher than 900 at 0.2% strain; (2) the initial resistance of
■
RESULTS AND DISCUSSION Main Performances of the Sensor. Figure 2a−c shows the three typical samples that we can obtain primarily (marked as sample 1 in Figure 2a, sample 2 Figure 2b, and sample 3 in Figure 2c). Before the metal−insulator transition happens due to increase in strain, the samples showed the peak of the GF values: 3230 for sample 1 at a small strain of 0.3%, 3580 for sample 2 at a small strain of 0.5%, 9720 for sample 3 at a small strain of 0.9%. The minimum strain that we could detect exactly was up to 0.05%, and the resolution of strain in our sample could reach 0.05%. We fitted our experimental data with the tunneling model in the insets of Figure 2a−c, which will be discussed later. Figure 2d shows the highest GF values of strain sensors based on reported tunneling model and of our samples at different strain under 2%.18,25,30 At a small strain of 0.05%, our GF value could be higher than 3000, whereas at strain 41080
DOI: 10.1021/acsami.7b14501 ACS Appl. Mater. Interfaces 2017, 9, 41078−41086
Research Article
ACS Applied Materials & Interfaces
Figure 3. Evidence of our sensors approaching the percolation threshold: (a) resistance of the sample at different graphite contents. The inset permits the estimation of t = 1.12. (b) Randomly distributed conducting clusters growth according to the classical percolation theory; (I, II) limited clusters period (III) infinite clusters period, and (IV) multiple clusters period. (c−e) Effect of narrow two-dimensional percolative conductive pathway: (c) the thermography diagrams before and after the application of voltage (left part) and the GFs (right part) of the samples with a narrow conductive pathway. (d) The thermography diagrams before and after the application of voltage (left part) and the GFs (right part) of the samples with multiple conductive pathways. (e) The change in the thermography diagrams of a sample with narrow conductive pathway after writing one more trace (left part) and the corresponding change in the GFs (right part). 41081
DOI: 10.1021/acsami.7b14501 ACS Appl. Mater. Interfaces 2017, 9, 41078−41086
Research Article
ACS Applied Materials & Interfaces
increases in Φ correspond to an increase in the average size of the randomly distributed conducting clusters (Figure 3b). Before Φ gets close to the percolation threshold, these clusters are limited and not connected to each other, and the percolation threshold is defined as the point that the first complete connection between clusters from top and bottom (infinite cluster) is formed. Specifically, in the conductiveparticles-filled situation, percolation threshold is the point when the first electrically conductive path connection is formed.41 Thus, if narrow two-dimensional percolative conductive pathway could be formed on the surface, the samples could be considered very close to the percolation threshold and the strain sensors may exhibit a very high sensitivity. To prove the existence of a narrow two-dimensional percolative conductive pathway in our sensors, we made the following experiments. We added a low voltage between the electrodes of the samples (the voltage was low enough to ensure the electric field strength between the electrodes did not cause a breakdown). In this situation, the electric current flowed through the electrically conductive path, whereas no electric current flowed through the other area. So, the temperature of the electrically conductive path would be higher than that of the other area. We used a microscopic thermal imager to detect the temperature changes and differences. Figure 3c gives the thermography diagrams before and after the application of voltage. It could be clearly seen that, although the region with a graphite sheet between the electrodes was wide (several millimeters to more than 1 cm), only one very narrow region showed raised temperatures. These results prove the existence of a narrow two-dimensional percolative conductive pathway on our samples. Then, we need to discuss the importance of the narrow twodimensional percolative conductive pathways for the performance of the strain sensors. To illustrate the significance of a narrow two-dimensional percolative conductive pathway to high sensitivities, we made the following experiments. First, we made a large amount of samples for comparison, and the results demonstrated that almost all of the samples with excellent performance on sensitivities had a narrow two-dimensional percolative conductive pathway (see Figure 3c and also see Figure S3 in the Supporting Information). In contrast, samples with bad performance on sensitivities had several wide conductive paths (see Figure 3e and also see Figure S4 in the Supporting Information). A comparison of the GF values in Figure 3c,d clearly shows that the samples with a narrow twodimensional percolative conductive pathway provide a much higher GF value than the samples with multiple electrically conductive paths. Second, for the highly sensitive sample had been proved to contain narrow two-dimensional percolative conductive pathway, we used graphite rod to write traces on their surfaces again. The comparative results show that these will change the morphology and quantity of electrically conductive paths (see Figure 3e and also see Figure S5 in the Supporting Information). The most critical point was that such changes obviously reduced the GF values of the samples. Thus, the significance of a narrow two-dimensional percolative conductive pathway for high sensitivity was proved. Another interesting aspect is that the gauge factor in terms of initial resistance also showed a clear tendency: a higher initial resistance corresponded to a higher GF, whereas a higher initial resistance corresponded to a lower GF. For example, the GF value of the samples at the strain of 0.2% was 944 for sample 1 (initial resistance: 2.1 MΩ), 614 for sample 2 (initial resistance: 1.2 MΩ), and 367 for sample 3 (initial resistance: 0.57 MΩ).
several subsequent samples was usually between 1.0 and 1.4 MΩ, GF higher was than 600 at 0.2% strain; and (3) the initial resistance of the last few samples was usually between 1.9 and 2.3 MΩ, GF was higher than 300 at 0.2% strain. Therefore, we could use these to prepare different samples and sorted them into three main categories, as shown in Figure 2a−c. Evidence of Our Sensors Approaching the Percolation Threshold Using a Statistical Percolation Model. According to the high-definition optical images, we extracted the proportions of graphite sheets (Φ) after every drawing. The dependence of R0 on Φ is shown in Figure 3a. The conductivity of the graphite trace increased by about 4 orders of magnitude when the graphite content was increased from 16.7 to 20.6%, a typical percolation behavior. If the graphite content was larger than Φc, the electrical conductivity of graphite trace could be theoretically predicted by the statistical percolation model40,41 σ = σ0(Φ − Φc)t
(2)
where σ represents the conductivity of composites at a given filler content, Φ is the filler volume content, Φc is the percolation threshold, and t is the critical exponent that depends on the system’s dimension. The typical values for t reported in the literature were between 1.1 and 1.3 for the 2D systems and between 1.6 and 2.1 for the 3D systems.48 As shown in Figure 3a, the percolation threshold Φc is estimated to be 16.9% and the corresponding t value is 1.12, conforming to our preceding 2D systems’ assumptions. As mentioned in our fabrication process, we stopped drawing graphite once we observed a stupendous drop in the resistance and chose these specific samples as our strain sensors, which means that the graphite content of our sensors was very close to the percolation threshold. Area fraction in Figure 3a shows that the percolation threshold is around 17%. However, ideally, according to the percolation theory and simulations,48 the circular disk has a percolation threshold of around 67%. There is a big gap between the two, and the main reason could be the existence of a tunneling conductive effect.30,32,41 So, it is necessary to confirm the existence of a tunneling conductive effect by fitting with a tunneling model. If the tunneling effect exists, then, at the tunneling conductive area, according to the tunneling model, we can figure out the relationship between ln(R/R0) and strain (ε), where R and R0 reflect the resistance before and after strain was applied, respectively ln(R /R 0) = ln(1 + ε) + Xd0ε
(3)
where d0 is the original tunneling distance and X is a physical quantity related to the barrier height between adjacent graphite sheets. Using this equation, we fitted our experimental data as shown in Figure 2a−c, and the experimental data and the fits are in good agreement. From the fitting data, we could deduce that, before the disconnection of an electrically conductive path, the conductive mechanism was consistent with the tunneling model. Although under a certain strain, the electrically conductive path could be broken into disconnection, leading to a sudden increase in resistivity (Figure 2a−c). It can be seen that at the tunneling conductive area, a smaller k leads to a higher GF value. For example, at strain 0.2%, the GF values of the samples was 944 for sample 1, 614 for sample 2, and 367 for sample 3, indicating that a smaller k led to a higher GF value. Evidence of Our Sensors Approaching the Percolation Threshold Using Thermography of Conductive Path/Paths. According to the classical percolation theory,46 41082
DOI: 10.1021/acsami.7b14501 ACS Appl. Mater. Interfaces 2017, 9, 41078−41086
Research Article
ACS Applied Materials & Interfaces
Figure 4. Applications of the strain sensors. (a) Measured characteristics of the resistance difference for the two breathing patterns. The inset is the image of the sensor attached to a person’s chest area for breathing pattern recognition. (b) Measured characteristics of the resistance difference for different words broadcasting when attaching our sensor to a loudspeaker. The inset is the image of the sensor attached to a loudspeaker and analysis of typical signal patterns. (c) The performance of our strain sensors to test quasistatic tensile forces. The inset shows how measurable static force is attached to the sensor.
percolative metal−insulator transition occurs at the critical area fraction (Φc), according to the percolation conductivity formula,40 the ratio of resistance after strain to that before strain is
This is because the sensor with multiple connected paths has a higher initial resistance, and the sensor with a single path has a lower initial resistance. Explanations of the High Sensitivity of Conductive Networks near the Percolation Threshold. Now, we will discuss the conductive mechanisms and theoretically explain the high sensitivity of conductive networks near the percolation threshold. At first, we proceed from the classical percolation theory to show what kind of conditions can be obtained for very sensitive strain sensors (high sensitivity at a small strain). If we define the area occupied by the conductive phase when the sample is laid flat to be Φ0, the area ratio occupied by the conductive phase would become Φ0/(1 + ε) after applying a small strain (ε) to the sample. Because the two-dimensional
R /R 0 = (Φ0 − Φc)t /[Φ0 /(1 + ε) − Φc]t
(4)
Because t is a positive value between 1.1 and 1.3 for the percolation conductance in a two-dimensional plane, it can be deduced that the degree of change in the resistance would reach the maximum after a small strain (e.g.,
