Microstructured capacitive sensor with broad detection range and long-term stability for human activity detection

In recent years, flexible stress sensors capable of monitoring diverse body movements and physiological signals have been attracting great attention in the fields of healthcare systems, human–machine interfaces, and wearable electronics. Inspired by the structure of natural eggshell inner membrane (ESIM), we developed a pressure sensor based on MXene (Ti3C2Tx)/Ag NWs (silver nanowires) composite electrodes and the micro-structured dielectric layer to meet the application requirements of wide detection range and long-term stability for the sensors. In the light of the nanoscale-microarray of the dielectric layer and the rough surface of electrode materials, this pressure sensor is expected to allow great and persistent deformation during the loading process. As a result, the device is characterized by an improved sensitivity, fast response (in the millisecond range), wide detection range (0–600 kPa), and long-term stability. The outstanding performance of the proposed sensor makes it possible to detect various human activities, such as speaking, air blowing, clenching, walking, finger/knee/elbow bending, and striking, demonstrating its good application prospects in wearable and flexible electronic devices.

To date, researchers have made lots of efforts to improve the comprehensive performance of the capacitive sensor in various aspects, such as sensitivity, detection limit/range, response time, and stability 12,[22][23][24] . For example, introducing the microstructure has been regarded as one of the effective strategies for remarkably improving the sensor sensitivity due to the distinct deformation of the microstructure and the subsequent increase in the capacitance variation. Ma et al. reported a capacitive pressure sensor consisting of two electrodes and wrinkled polydimethylsiloxane (PDMS) dielectric layer. Compared with the sensor without a microstructure, the fabricated sensor exhibited improved sensitivity from 0.01 to 2.04 kPa −1 in the range of 0-2 kPa 24 . Additionally, bionic microstructures from natural plants, such as aureum leaf 25 , lotus leaf 26 , and rose petals 27 also have attracted much attention because of their delicate and green/cost-effective features. Guo et al. proposed a capacitive sensor consisting of a dried flower petal or a leaf sandwiched between two flexible electrodes. The device worked within a broad-pressure range (0.6-115 kPa), the maximum sensitivity of which was 1.54 kPa −1 . Furthermore, compared with single-component electrode materials, such as carbon nanotubes (CNTs) 28,29 , graphene 30,31 , conductive polymers 32,33 , metal nanostructures [34][35][36] , and MXene (Ti 3 C 2 T x ) [37][38][39][40] , the conducting composite has been widely utilized in the sensor electrode due to its combined advantages [41][42][43][44][45] . For example, the hybrid of one-dimensional (1D) and two-dimensional (2D) conducting materials could effectively prevent the aggregating and stacking of 2D material, thus forming a homogeneous connective network, which leads to good mechanical and electrical properties of the sensor electrode. Liu et al. reported a dendritic-lamellar MXene/carbon nanotube/polyvinylpyrrolidone electrode with high flexibility. The fabricated tactile sensor based on the composite electrode exhibited fast response, demonstrating for voice recognition and pulse measurement 45 .
However, the diverse applications require the pressure sensors to not only respond to a small stimulus but also work under an ultrahigh load 9,46 . For example, the huge impact force coming from a sudden striking usually exceeds hundreds of kilopascals, which is likely to cause the overload or failure of the sensors. Therefore, apart from high sensitivity and fast response, it is highly desirable to ensure a wide detection range and long-term stability of the device.
Herein, we reported a capacitive sensor whose overall performance was improved by designing the microstructure of the dielectric layer and selecting composite materials of the electrode. Inspired by the natural microstructures, we utilized the eggshell inner membrane (ESIM) as the template of the PDMS dielectric layer. Compared with other templates, the unique fibrous ESIM provided the nanoscale-template with a more elaborated microstructure 47 . We could achieve abundant micro-fluctuations from the ESIM, which contributes to the remarkable deformation of the PDMS dielectric layer under the external stimuli. In addition, we proposed a composite electrode based on 2D MXene sheets and 1D Ag NWs. The Ag NWs can effectively decrease the restacking of MXene sheets, forming abundant conducting tunnels to facilitate electron transfer within the electrode. The fabricated MXene sheets and Ag NWs are closely contacted due to the electrostatic adsorption, which makes the conductive network have a stable structure [48][49][50][51][52][53] . In this case, there is no relative slip between two conducting components even under high stress due to the tight combination, leading to a stable electron path within the conductive networks, which is beneficial to the long-term stability of the device under the repeated loading/unloading. As a result, the micro-structured sensor exhibited an improved sensitivity especially in the high-pressure range, broad detection range (0-600 kPa), low detection limit (<16 mg), the fast response time (in the millisecond range), longterm stability, and reproducibility.

RESULTS AND DISCUSSION
Micro-structured dielectric layer and composite electrode To improve the detection range and long-term stability of the capacitive sensor, we used the micro-structured PDMS as the dielectric layer and MXene/Ag NWs as the rough electrodes. The microstructure of PDMS was replicated from the structure of the ESIM (Fig. 1a), and MXene sheets was prepared by etching the MAX phase and further delaminating (Fig. 1b, see Methods Section for details). The composite materials consisting of Ag NWs and MXene sheets within the electrode are expected to provide good contact and abundant conducting paths. The flexible capacitive pressure sensor was assembled using the microstructured PDMS dielectric layer sandwiched between two MXene/Ag NWs electrodes (Fig. 1c).
Differently magnified SEM images disclose that the membrane is composed of fibrous protein with an average diameter of about 1 µm, presenting an interconnected network structure. There were a number of open spaces among the protein fibers (Fig. 2a, b and Fig. S1). The inverted microstructure was successfully transferred to the PDMS surface (Fig. 2c, d). These rugged inverse structures and the PDMS with small modulus could cause a wider range of compression deformation, which was expected to greatly improve the sensitivity and detection range of the sensor. The morphology and composition of electrode materials were further investigated. Figure 2e suggests that the MXene after etching features a multilayered structure. The element mapping illustrates that C, Ti, O, and F elements uniformly distribute across the multilayered MXene (Fig. 2f). As confirmed in the AFM image (Fig. 2g), MXene with a few layers was further acquired by ultrasonic treatment. The height distribution of MXene is about 5 nm, which is equal to 4-5 layers of Ti 3 C 2 T x nanosheets stacked together (Fig. 2g, h) 54,55 . The obtained MXene sheets can be completely filtered to form an intact and smooth film due to the stacking (Fig. S2). The inset of Fig. S2 indicates that the MXene sheets uniformly disperse in water without obvious agglomeration. In addition, the average size of the MXene sheets is~100 nm, which is similar to the diameter of the prepared Ag NWs (Figs. S3 and 2i). The 1D Ag NWs were tightly wrapped by the 2D MXene sheets after the combination of the two conductive materials (Fig. 2j), supplying an improved surface roughness of the electrode. XRD Due to the presence of the surface functional groups, the asprepared MXene sheets were negatively charged with a zeta potential of −33.9 mV (Fig. 2l), while the surface of Ag NWs presented a positive charge with a zeta potential of +0.884 mV. Thus, the MXene nanosheets could closely adhere to the surface of Ag NWs owing to their strong electrostatic interaction, which is consistent with the result of SEM in Fig. 2j. The self-assembled composite (Fig. S4) displayed a resultant zeta potential of −18 mV. In the composite system, 1D Ag NWs and 2D MXene sheets can be regarded as the steel bar and cement in the solid buildings, respectively. The tight combination of the two materials could supply definite and stable conductive paths in the networks. The synergy of the two components contributed to the rough surface and high conductivity of the electrode (Fig. S5).

Sensing performance of capacitive sensors
We evaluated the performance of the micro-structured and flat capacitive sensors. The sensitivity of the sensor was measured through the slope of the relative change in the capacitance signal versus the applied pressure curve. The change rate of the capacitance of the flat sensor was~50% under 100 kPa, presenting relatively a low sensitivity of 4.37 MPa −1 (Fig. 3a). Furthermore, the change in the capacitance of the flat sensor was almost saturated over 100 kPa. Remarkably, the micro-structured sensor could work with a wider range of up to 600 kPa. The sensitivity value was increased to 10.13 MPa −1 within 100 kPa, and especially by 54.3 and 95.7 times within high-pressure ranges of 100-400 kPa and 400-600 kPa (Fig. 3b).
To identify the contribution of the microstructure to the sensor performance, we simulated the sensor models based on the micro-structured and planar dielectric layer for comparison, and then investigated the variation in the stress and deformation under various pressures by finite element analysis (FEA) (see Methods section for details). According to the definition of capacitance C = εS/4πkd (where ε is the dielectric constant, d refers to the distance between the two plates, k represents the electrostatic constant, and S is the area of the capacitor plate), the device capacitance can be changed conveniently by controlling ε, S, and d. To obtain a higher sensitivity of the sensor, the capacitance variation is expected to increase to the largest extent. Under a given load, the stress and deformation were uniformly distributed across the surface of the flat sensor because of the planar and continuous contact interface between the electrode and dielectric layer (Fig. 3c). Therefore, the deformation and the contact area were hardly changed in this case. The flat sensor exhibited high resistance to the external pressure, resulting in small structural compressibility, as well as few capacitance changes. When the deformation reached saturation, the capacitance turned to be a stable value, which is in accordance with the result in Fig. 3b. Thus, the sensing range of the flat sensor was limited to 100 kPa, characterized by low sensitivity and a relatively narrow sensing range.
The simulation result revealed that the micro-structured sensor experienced a higher level of deformation than the flat one under the same pressure due to the rough surface of the dielectric layer (Fig. 3c). Higher stress was concentrated on the bulges of the microstructure. As the external pressure exerted on the sensor increased, the contact area between the electrode and dielectric layer was gradually increased, while the distance between the two electrodes was decreased. Thus, the micro-structured sensor was able to present a larger capacitance variation under the same pressure, enabling higher sensitivity and a wider detectable range. In addition, the result of the simulated displacement between the two electrodes under various pressures and the structural changes of the capacitive sensors further proved that a larger deformation occurred under the same load in the case of the micro-structured sensor (Fig. 3d, e).
Therefore, the improved performance of the micro-structured sensor can be attributed to the following three aspects. First, the microstructure enables the sensor to experience a higher level of deformation under the same pressure, thus generating more obvious changes in the distance between the two electrodes, as well as the capacitance variations. Second, the high roughness of the electrode due to the combination of 1D and 2D conductive materials provides the improvement of the void space within the electrode. Thus, under a given loading, the stress sensor based on the MXene/Ag NWs electrode is expected to exhibit a large change in the d and C, and then generating a higher sensitivity coefficient (Fig. S6 shows the capacitance changes of the sensor based on smooth ITO/PET electrodes and the micro-structured PDMS dielectric layer, revealing a lower sensitivity within whole working range). Third, there is an air gap between each unit of the bulges in the microstructure. When the load was enhanced, the air gap would be gradually filled by PDMS material. Accordingly, the comprehensive dielectric constant ε of the dielectric layer rose (ε Air = 1, ε PDMS = 2.5-2.7), causing a larger capacitance change.
The instruments for testing the flexible pressure sensors included the push-pull machine, LCR capacitance measurement instrument, and program computer (Fig. 4a), while the important performance parameters for the stress sensors involved the response time and the detection limit. The time-response curve of the loading/unloading pressure indicated that the rise time and fall time of the micro-structured sensor was 280 and 290 ms, respectively (Fig. 4b). It is also worth noting that measuring the response time is limited by the movement rate (2.5 N/s) of the push-pull machine. Thus, the actual response time of the sensor could be less than 50 ms. We also tested the capacitance response of the sensor to rapid tapping using a finger, proving that the fast response within 50 ms (Fig. S7). A repeated signal can be recognized during the loading and unloading process of a grain of rice with a weight of 16 mg (Fig. 4c), denoting that the sensor can be used to reliably detect a subtle stimulus by the capacitive response. Additionally, the sensor showed stable signals in single and multiple cycles under the pressures of 10, 20, 35, and 45 kPa, respectively (Fig. 4d, e). The response to each loading exhibited stepped waves and stable square wave-forms, revealing excellent sensing stability of the sensor within a wide pressure range. Moreover, the micro-structured sensor also demonstrated a precise response independent of the loading speed (0.10, 0.25, and 2.50 N/s) under the application of 2.5 N loading/unloading cycles (Fig. 4f). The rapid change and good recovery of the capacitance at each speed evidenced the fast response and excellent reversibility of the sensor.
To assess the reproducible sensing performance of the device, the sensor was loaded and unloaded with a constant pressure of 40 kPa dynamically for over 20,000 s at least (Fig. 4g). The microstructured sensor exhibited remarkable robustness as the capacitance could recover its initial value even after being subjected to 1800 cycles. The magnified parts clearly showed no obvious amplitude change in the capacitance value at different loading/unloading stages (insets in Fig. 4g), verifying the excellent mechanical stability of the sensor. To confirm the potential for long-term application, the capacitance response of the sensor was further measured after 150 days under the same measuring condition (Fig. 4h, over 2700 cycles was shown in Fig. S8). The sensor still exhibited good stability over the whole period, and the capacitance variation was constant before and after long-term storage and measurement due to the stable conductive network within the electrodes. Furthermore, the SEM image of the dialectic layer after the repeated loading reveals that the PDMS microstructure cannot be destroyed (Fig. S9).
Demonstration of the micro-structured sensor Based on the above characterizations and analyses, the flexible capacitive sensor demonstrated enhanced sensitivity, a wide detection range, low detection limit, rapid response, and longterm stability. Moreover, it was envisioned that the microstructured sensor could have outstanding practicability when detecting diversified body movements. To demonstrate this, the sensor was firmly attached to various body parts of a healthy 26 years old volunteer to collect the capacitance signals in response to the body movements (Fig. 5a). The regular and repetitive output of the electrical signal could be observed when the volunteer repeatedly pronounced the word "studying" (Fig. 5b), caused by the neck muscle movement and the vocal cord vibration. When the sensor was attached to the cheek, the stress of the air blowing on the device were distinctly responded (Fig.  5c). Furthermore, the stable capacitance change-time curves were obtained when the volunteer clenched and released his fist attached by the micro-structured sensor (Fig. 5d). The sensor was also capable of monitoring other body movements, such as the bending of the finger, elbow, and knee. The favorable reversibility during each testing cycle was achieved (Fig. 5e, f, g). Additionally, the angle recognition of the sensor was also demonstrated by attaching the sensor to the index finger (Fig. S10). When the sensor was placed on the sole, the highly stable signal output illustrated that the sensor could be used to monitor the response to walking (Fig. 5h). As a result, the sensor's function of efficiently detecting diverse body movements within a broad detection range was proven.
The sensor was fixed on the surfaces of different objects to measure the sensing capability of the device as a tactile sensor in specific activities (Fig. 6). The stable electrical signals responding to the repeated quick touching and weight pressing on the sensor reflected the rapid response and good mechanical stability of the sensor (Fig. 6a, b). The sensor was also adhered to the beaker and keyboard to monitor the real-time grasping and knocking, which indicated the stable wave-forms and constant capacitance variations in response to the repeated actions (Fig. 6c, d). Figure  6e displays the signal output of the pressure sensor in response to a noncontact force from air blowing of aurilave. The sharp and repeated wave-forms reflected the high sensitivity and fast response of the sensor when detecting the real-time wind pressure. To assess the detection capability under high pressure, the six independent sensors were fabricated into an array to evaluate the volunteer's foot pressure distribution (Fig. 6f). The distinct capacitance values indicated that the sensor array could distinguish different parts of foot pressure when the volunteer was standing. Strike tests of the sensor at different pressures disclosed that the device rapidly responded to the repeatedly strong strikes by the fingertip, stainless cone, and pen point (Fig.  6g-i).
These results demonstrated the outstanding sensing ability of the micro-structured sensor to monitor diverse human activities and physical stimuli, accompanied by the characteristics of high sensitivity, fast response, excellent stability, and broad detection range. The synergy of the rough surface of the conductive materials and the nanoscale-microstructure of the dielectric layer contributed to the improvement of the sensor performance. The proposed sensor shows good prospects in healthcare, soft robots, human-machine interaction, and wearable devices.

Synthesis of Ag NWs and MXene
A hydrothermal method was used for preparing Ag NWs according to our previous work 56,57 . First, 0.788 g silver nitrate, 0.53 g glucose, and 0.23 g ferric sulfate were dissolved in 100 mL distilled water to obtain a lightyellow solution. Second, 4.5 g poly(vinyl pyrrolidone) (K30) was added to the above solution under magnetic stirring for 2 h. Third, the homogeneous mixture was transferred to a Teflon autoclave of 100 mL, and then heat-treated at 180 o C for 9 h. Finally, the gray-green Ag NWs were collected by repeated centrifugation and then dispersed in ethanol. Ti 3 C 2 T x was synthesized by a selective Al etching of MAX phase precursor powder (Ti 3 AlC 2 , 11 Technology Co., Ltd., China) by a minimally intensive layer delamination method. First of all, 1.6 g LiF was dissolved in 9 M HCl of 20 mL, and then 1 g Ti 3 AlC 2 powder was added to LiF/HCl solution with the assistance of magnetic stirring to avoid a sharp increase in the reaction temperature. The etching reaction occurred at 35°C for 48 h with the persistent stir. Subsequently, deionised water was used for washing the etched product until the pH value of the supernatant reached 6. Finally, the Ti 3 C 2 T x dispersion was ultra-sonicated for 2 h under the condition of the ice bath and Ar atmosphere. The delaminated Ti 3 C 2 T x flakes were collected from the upper solution.

Fabrication of micro-structured PDMS dielectric layer
The microstructure of the PDMS (Sylgard 184, Dow Corning) dielectric layer was replicated from the ESIM. First, the ESIM was peeled and then cleaned with deionised water. Second, the mixture of PDMS pre-polymer and a curing agent (10:1 by weight) was dropped on the dried ESIM surface. Finally, after cured at a temperature of 60 o C for 4 h, the ESIM was removed to obtain the PDMS dielectric layer with the inverted microstructure of the ESIM surface.

Construction of capacitive pressure sensor
The flexible capacitive pressure sensor was assembled based on a typical sandwich structure. The mixture of the fabricated Ag NWs and MXene (5:1 by weight) was firstly vacuum-filtrated on a filter membrane, and then we poured the mixture of PDMS pre-polymer and a curing agent (10:1 by weight) onto the MXene/Ag NWs-filter membrane. After cured at a Fig. 5 Demonstration of the micro-structured sensor in response to human activities. a Schematic diagram of the sensor that detects human physiological signals and body motions; b Detection of the vocal cord vibration signal sent from the pronunciation of the word "studying"; c-h The capacitance responses come from the repeated air blowing, clenching and releasing of finger/knee/elbow bending, and walking.
temperature of 60 o C for 4 h, the composite conducting materials are embedded into the PDMS layer. The MXene/Ag NWs was completely transferred to the PDMS surface by removing the filter membrane, acquiring the electrodes. Finally, the electrodes and dielectric layer were cut with an area of 10 mm × 10 mm and then integrated into a sandwiched device. For performance comparison, we fabricated a flat sensor by using the same MXene/Ag NWs composite electrodes and a flat PDMS dielectric layer.

Characterization
A field emission scanning electron microscope (FE-SEM, Zeiss Sigma 500, Germany) was used for observing the morphology of the dielectric layer and electrode materials. The composition of the conductive materials was characterized through an X-ray diffractometer (XRD, X'pert Pro MFD, Panalytical, The Netherlands) with a Cu Kα radiation (λ = 1.54178 Ǻ). The thickness of MXene sheets was determined by an atomic force microscope (AFM, MultiMode 8, Bruker, Germany). The surface potential of the conductive materials was tested using a Malvern Zetasizer (Nano ZS, England). The stress was applied to the sensors by a push-pull force gauge (LE, LiShi, China). The response performance of the stress-dependent sensor was evaluated by measuring the real-time capacitance changes of the device by an LCR (Inductance, Capacitance, Resistance) measurement instrument (IM3526, Rizhi, Japan). The capacitance of the sensor was tested under the frequency of 500 kHz.
In order to more accurately evaluate the advantage of the microstructure in the flexible capacitive sensor, the FEA was conducted to compare the two models based on the micro-structured and planar PDMS dialectic layer by a commercial package Comsol 5.4. The micro-structured model was constructed according to the actual replicated surface of the dielectric layer with the electrodes replaced by the metal plates.

DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request. Fig. 6 Demonstration of the micro-structured sensor in response to exerted pressures. a, b Capacitance response of the sensor to quick touching and weight pressuring; c-e Detection of various repeated actions, including grasping the beaker, knocking the keyboard, and blowing air; f Foot pressure distribution identified by a sensor array; g-i Detection of the repeated strikes of the sensor by the fingertip, stainless cone, and pen point.